IoT, Machine Learning and Blockchain

Technologies for Renewable Energy

and Modern Hybrid Power Systems
RIVER PUBLISHERS SERIES IN INFORMATION
SCIENCE AND TECHNOLOGY
Series Editors:
K.C. CHEN, National Taiwan University, Taipei, Taiwan
and
University of South Florida, USA
SANDEEP SHUKLA, Virginia Tech, USA
and
Indian Institute of Technology Kanpur, India
The “River Publishers Series in Computing and Information Science and Technology” covers research which ushers
the 21st Century into an Internet and multimedia era. Networking suggests transportation of such multimedia contents
among nodes in communication and/or computer networks, to facilitate the ultimate Internet.
Theory, technologies, protocols and standards, applications/services, practice and implementation of wired/wireless
The “River Publishers Series in Computing and Information Science and Technology” covers research which ushers
the 21st Century into an Internet and multimedia era. Networking suggests transportation of such multimedia contents
among nodes in communication and/or computer networks, to facilitate the ultimate Internet.
Theory, technologies, protocols and standards, applications/services, practice and implementation of wired/wireless
networking are all within the scope of this series. Based on network and communication science, we further extend the
scope for 21st Century life through the knowledge in machine learning, embedded systems, cognitive science, pattern
recognition, quantum/biological/molecular computation and information processing, user behaviors and interface, and
applications across healthcare and society. Books published in the series include research monographs, edited volumes,
handbooks and textbooks. The books provide professionals, researchers, educators, and advanced students in the field
with an invaluable insight into the latest research and developments.
Topics included in the series are as follows:-

• Artificial intelligence
• Cognitive Science and Brian Science
• Communication/Computer Networking Technologies and Applications
• Computation and Information Processing
• Computer Architectures
• Computer networks
• Computer Science
• Embedded Systems
• Evolutionary computation
• Information Modelling
• Information Theory
• Machine Intelligence
• Neural computing and machine learning
• Parallel and Distributed Systems
• Programming Languages
• Reconfigurable Computing
• Research Informatics
• Soft computing techniques
• Software Development
• Software Engineering
• Software Maintenance
For a list of other books in this series, visit www.riverpublishers.com
IoT, Machine Learning and Blockchain
Technologies for Renewable Energy
and Modern Hybrid Power Systems

Editors

C. Sharmeela
Anna University, India

P. Sanjeevikumar
Aarhus University, Denmark

P. Sivaraman
Vestas Technology R&D Chennai Pvt. Ltd, India

Meera Joseph
Independent Institute of Education, South Africa

River Publishers
Published, sold and distributed by:
River Publishers
Alsbjergvej 10
9260 Gistrup
Denmark

www.riverpublishers.com

ISBN: 978-87-7022-724-7 (Hardback)

978-87-7022-711-7 (Ebook)


c 2022 River Publishers

All rights reserved. No part of this publication may be reproduced, stored in

a retrieval system, or transmitted in any form or by any means, mechanical,
photocopying, recording or otherwise, without prior written permission of
the publishers.
 Contents

Preface xiii

Acknowledgments xv

List of Figures xvii

List of Tables xxiii

List of Contributors xxv

List of Abbreviations xxix

1 Introduction to IoT 1
Asim Maharjan and Saju Khakurel
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Applications of IoT . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Domestic Applications . . . . . . . . . . . . . . . . 6
1.3.2 Applications in Healthcare . . . . . . . . . . . . . . 7
1.3.3 Applications in E-commerce . . . . . . . . . . . . . 8
1.3.4 Industrial Applications . . . . . . . . . . . . . . . . 9
1.3.5 Applications in Energy . . . . . . . . . . . . . . . . 10
1.4 Technical Details of IoT . . . . . . . . . . . . . . . . . . . 11
1.4.1 Sensors . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.2 Actuators . . . . . . . . . . . . . . . . . . . . . . . 15
1.4.3 Processing Topologies . . . . . . . . . . . . . . . . 16
1.4.4 Communication Technologies . . . . . . . . . . . . 18
1.5 Recent Developments . . . . . . . . . . . . . . . . . . . . . 20
1.6 Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

v
vi Contents

2 IoT and its Requirements for Renewable Energy Resources 29

D. Gunapriya, R. Sivakumar, and K. Sabareeshwaran
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.1.1 IoT and its Necessity . . . . . . . . . . . . . . . . . 30
2.1.2 Challenges in RES . . . . . . . . . . . . . . . . . . 30
2.1.3 Integration of IoT in RES and Benefits . . . . . . . . 32
2.2 Industrial IoT . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.1 Architecture of IoT . . . . . . . . . . . . . . . . . . 33
2.2.2 IoT Components . . . . . . . . . . . . . . . . . . . 34
2.3 RES and IoT . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.3.1 IoT Controls for RES . . . . . . . . . . . . . . . . . 36
2.3.2 Challenges in IoT Implementation . . . . . . . . . . 38
2.4 Challenges of IoT in EMS Post-implementation . . . . . . . 39
2.4.1 Privacy Issues . . . . . . . . . . . . . . . . . . . . . 39
2.4.2 Security Concerns . . . . . . . . . . . . . . . . . . 41
2.4.3 Data Storage Issues . . . . . . . . . . . . . . . . . . 43
2.4.3.1 Challenges in data management . . . . . . 43
2.4.3.2 Challenges in fetching data . . . . . . . . 44
2.4.3.3 Challenges in allocation . . . . . . . . . . 44
2.5 Solution to IoT Challenges . . . . . . . . . . . . . . . . . . 45
2.5.1 Blockchain Methodology . . . . . . . . . . . . . . . 45
2.5.1.1 Blockchain technology infrastructure
features . . . . . . . . . . . . . . . . . . . 47
2.5.1.2 Application domains of blockchain
technology . . . . . . . . . . . . . . . . . 47
2.5.1.3 Challenges of blockchain technology . . . 47
2.5.2 Cloud Computing . . . . . . . . . . . . . . . . . . . 48
2.5.2.1 Reference architecture . . . . . . . . . . . 49
2.5.2.2 Network communication and its
challenge . . . . . . . . . . . . . . . . . . 51
2.5.2.3 Privacy and security . . . . . . . . . . . . 51
2.5.2.4 Background information . . . . . . . . . . 53
2.5.2.5 Big data analytics . . . . . . . . . . . . . 53
2.5.2.6 Provision of program quality . . . . . . . 53
2.5.2.7 IPv4 addressing limit . . . . . . . . . . . 54
2.5.2.8 Legal aspects and social facts . . . . . . . 55
2.5.2.9 Service detection . . . . . . . . . . . . . . 56
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 56
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
 Contents vii

3 Power Quality Monitoring of Low Voltage Distribution System

Toward Smart Distribution Grid Through IoT 61
P. Sivaraman, C. Sharmeela, S. Balaji, and P. Sanjeevikumar
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2 Introduction to Various PQ Characteristics . . . . . . . . . . 63
3.3 Introduction to IoT . . . . . . . . . . . . . . . . . . . . . . 64
3.4 Smart Monitoring using IoT for the Low Voltage Distribution
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.5 Power Quality Monitoring of Low Voltage Distribution
System – Case Study . . . . . . . . . . . . . . . . . . . . . 67
3.5.1 Undervoltage . . . . . . . . . . . . . . . . . . . . . 69
3.5.2 Overvoltage . . . . . . . . . . . . . . . . . . . . . . 69
3.5.3 Interruption . . . . . . . . . . . . . . . . . . . . . . 71
3.5.4 Overload in Branch Circuit . . . . . . . . . . . . . . 72
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 74
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4 Health Monitoring of a Transformer in a Smart Distribution

System using IoT 79
P. Sivaraman, C. Sharmeela, and P. Sanjeevikumar
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2 Introduction to the Transformer . . . . . . . . . . . . . . . . 81
4.3 Failure of the Distribution Transformer . . . . . . . . . . . . 82
4.4 Transformer Health Monitoring System through IoT . . . . . 82
4.4.1 Winding and Oil Temperature Sensor . . . . . . . . 83
4.4.2 Oil Level Monitoring Sensor . . . . . . . . . . . . . 84
4.4.3 Current Sensor and Voltage Sensor . . . . . . . . . . 84
4.4.4 Microcontroller . . . . . . . . . . . . . . . . . . . . 85
4.4.5 LCD or Monitor . . . . . . . . . . . . . . . . . . . 85
4.4.6 Communication System . . . . . . . . . . . . . . . 85
4.4.7 Central Monitoring and Control . . . . . . . . . . . 86
4.5 Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 89
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5 Introduction To Machine Learning Techniques 93

Saniya M. Ansari, Ravindra R. Patil, Rajnish Kaur Calay,
and Mohamad Y. Mustafa
5.1 Why and What is Machine Learning? . . . . . . . . . . . . 93
viii Contents

5.1.1 Phrases in Machine Learning . . . . . . . . . . . . . 94

5.1.2 Steps Involved in Machine Learning Practices . . . . 94
5.1.3 Properties of Data . . . . . . . . . . . . . . . . . . . 94
5.1.4 Real-World Applications of Machine Learning . . . 95
5.2 Classification of Machine Learning Techniques . . . . . . . 96
5.2.1 Supervised Learning . . . . . . . . . . . . . . . . . 96
5.2.1.1 Classification . . . . . . . . . . . . . . . . 97
5.2.1.2 Regression . . . . . . . . . . . . . . . . . 98
5.2.2 Unsupervised Learning . . . . . . . . . . . . . . . . 99
5.2.2.1 Clustering . . . . . . . . . . . . . . . . . 99
5.2.2.2 Association . . . . . . . . . . . . . . . . 100
5.2.3 Reinforcement Learning . . . . . . . . . . . . . . . 100
5.2.3.1 Crucial terms in reinforcement
learning . . . . . . . . . . . . . . . . . . 101
5.2.3.2 Salient features of reinforcement
learning . . . . . . . . . . . . . . . . . . 102
5.2.3.3 Types of reinforcement learning . . . . . . 102
5.2.3.4 Reinforcement learning algorithms . . . . 103
5.3 Some Crucial Algorithmic Mathematical Models in Machine
Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.3.1 Logistic Regression . . . . . . . . . . . . . . . . . . 104
5.3.2 Decision Trees . . . . . . . . . . . . . . . . . . . . 105
5.3.3 Linear Regression . . . . . . . . . . . . . . . . . . . 107
5.3.4 K-Nearest Neighbors . . . . . . . . . . . . . . . . . 108
5.3.5 K-Means Clustering . . . . . . . . . . . . . . . . . 110
5.4 Pre-Eminent Python Libraries Intended for Machine
Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.4.1 Human Detection (OpenCV, HoG, SVM with
Multi-Threading) . . . . . . . . . . . . . . . . . . . 113
5.4.2 Instagram Filters – (OpenCV, Matplotlib, NumPy) . 114
5.5 Machine Learning Techniques in State of Affairs of Power
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 117
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

6 Machine Learning Techniques for Renewable

Energy Resources 121
K. Punitha, S. Anbarasi, and T. Balasubramanian
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 122
 Contents ix

6.2 Overview of Machine Learning . . . . . . . . . . . . . . . . 126

6.3 Deep Learning Architecture . . . . . . . . . . . . . . . . . . 128
6.4 LSTM Network Based Prediction . . . . . . . . . . . . . . . 132
6.5 Concepts of Solar PV and its MPPT Techniques . . . . . . . 134
6.6 Simulation Results and Discussion . . . . . . . . . . . . . . 135
6.6.1 Modeling and Performance Analysis . . . . . . . . . 135
6.6.2 Prediction or Forecasting Methodology . . . . . . . 141
6.6.3 Utilizing Predicted Value in MPPT Technique . . . . 143
6.7 Conclusion and Future Directions . . . . . . . . . . . . . . 145
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

7 Application of Optimization Technique in Modern Hybrid

Power Systems 149
D. Lakshmi, R. Zahira, C. N. Ravi, P. Sivaraman, G. Ezhilarasi,
and C. Sharmeela
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7.2 Modern Power System . . . . . . . . . . . . . . . . . . . . 151
7.2.1 Deregulated Power System . . . . . . . . . . . . . . 152
7.2.2 Components of Deregulation . . . . . . . . . . . . . 152
7.2.3 Types of Transactions . . . . . . . . . . . . . . . . . 154
7.2.3.1 Bilateral transactions . . . . . . . . . . . 154
7.2.3.2 DPM and APF . . . . . . . . . . . . . . . 155
7.2.4 Renewable Energy Sources . . . . . . . . . . . . . . 156
7.2.4.1 Doubly fed induction generator . . . . . . 156
7.2.4.2 DFIG in deregulated power system . . . . 158
7.3 Optimization Techniques and Proposed Technique . . . . . . 161
7.3.1 Controllers . . . . . . . . . . . . . . . . . . . . . . 161
7.3.2 PI Controller . . . . . . . . . . . . . . . . . . . . . 161
7.3.3 Artificial Optimization Algorithm for Tuning PI . . . 162
7.3.3.1 Differential evolution . . . . . . . . . . . 162
7.3.3.2 Flower pollination algorithm . . . . . . . 163
7.3.3.3 Hybrid algorithm . . . . . . . . . . . . . 164
7.3.3.4 Design of a hybrid DE-FPA algorithm
for LFC . . . . . . . . . . . . . . . . . . 165
7.4 Simulation Results and Discussion . . . . . . . . . . . . . . 165
7.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 167
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
x Contents

8 Application of Machine Learning Techniques in Modern

Hybrid Power Systems – A Case Study 173
B. Koti Reddy, Krishna Sandeep Ayyagari,
Raveendra Reddy Medam, and Mohemmed Alhaider
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 174
8.2 Technical Issues in Modern Hybrid Power Systems . . . . . 176
8.2.1 Power Quality . . . . . . . . . . . . . . . . . . . . . 177
8.2.2 Demand-Supply Management . . . . . . . . . . . . 177
8.2.3 Synchronization and Islanding . . . . . . . . . . . . 177
8.2.4 Protective Devices, Safety, and Environment . . . . 177
8.2.5 Human Factor . . . . . . . . . . . . . . . . . . . . . 178
8.3 Application of ML and Optimization Techniques in MHPS . 178
8.4 A Prediction Case Study of ML in MHPS . . . . . . . . . . 179
8.4.1 Forecasting Irradiance of SPP . . . . . . . . . . . . 182
8.4.2 Metrics for Understanding the Performance
of Predictions using ML Methods . . . . . . . . . . 184
8.4.3 Model-Based and Model-Free Regression Techniques 185
8.4.4 Prediction Block . . . . . . . . . . . . . . . . . . . 186
8.4.5 Forecasting of Solar Irradiance with a Model-Based
Regression Approach . . . . . . . . . . . . . . . . . 187
8.4.6 Forecasting of Solar Irradiance with a Model-Free
Regression Approach (ANNs) . . . . . . . . . . . . 190
8.4.7 Normalization, Training, and Testing for Model-Free
Regression . . . . . . . . . . . . . . . . . . . . . . 191
8.5 Optimization Block in MHPS . . . . . . . . . . . . . . . . . 193
8.5.1 Optimization-Assisted ML of MHPS . . . . . . . . 193
8.5.2 Experimental Setup . . . . . . . . . . . . . . . . . . 197
8.5.3 Validation Block . . . . . . . . . . . . . . . . . . . 197
8.5.3.1 Thorough comparisons in voltage-magnitudes
for the actual test day for model-based and
model-free approaches . . . . . . . . . . . 198
8.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 200
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

9 Establishing a Realistic Shunt Capacitor Bank with a Power

System using PSO/ACCS 205
Ali Mohamed Eltamaly, Osama El Sayed Morsy,
Amer Nasr Abd Elghaffar, Yehia Sayed Mohamed,
and Abou-Hashema Ahmed
 Contents xi

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 206

9.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . 208
9.2.1 Power Flow Equations . . . . . . . . . . . . . . . . 209
9.2.2 Mathematical Representation . . . . . . . . . . . . . 210
9.2.3 Sensitivity Calculations . . . . . . . . . . . . . . . . 211
9.3 Capacitor Bank Operation Strategies . . . . . . . . . . . . . 212
9.4 Particle Swarm Optimization . . . . . . . . . . . . . . . . . 214
9.5 Limitation Treatment . . . . . . . . . . . . . . . . . . . . . 216
9.6 PSO Implementation for Offline Capacitor Study . . . . . . 216
9.7 Simulation System for Optimal Capacitor Allocation . . . . 218
9.7.1 Modified System Data . . . . . . . . . . . . . . . . 219
9.7.2 Simulation Study . . . . . . . . . . . . . . . . . . . 221
9.8 Automatic Capacitor Control Scheme . . . . . . . . . . . . 224
9.8.1 ACCS IED Scope . . . . . . . . . . . . . . . . . . . 225
9.8.2 ACCS Operation Logic Steps . . . . . . . . . . . . 225
9.8.3 ACCS Operation Sample . . . . . . . . . . . . . . . 227
9.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 230
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

10 Introduction to Blockchain Technologies 235

Xiaofeng Chen and Xiangjuan Jia
10.1 Introduction and Classification . . . . . . . . . . . . . . . . 236
10.2 Blockchain Technology Characteristics . . . . . . . . . . . . 237
10.2.1 Multi-Centralization . . . . . . . . . . . . . . . . . 237
10.2.2 Tamper-Proof, Traceable, and Transparent . . . . . . 237
10.2.3 High Reliability . . . . . . . . . . . . . . . . . . . . 238
10.3 Blockchain Technology Graph . . . . . . . . . . . . . . . . 238
10.3.1 Core Technology Overview . . . . . . . . . . . . . 239
10.3.2 Expansion Technology Overview . . . . . . . . . . 248
10.3.3 Supporting Technology Overview . . . . . . . . . . 251
10.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 253
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 254

11 Blockchain Technologies for Renewable Energy Resources with

Case Study: SHA–256, 384, and 512 257
Kaung Si Thu, Shubham Tiwari, and Weerakorn Ongsakul
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 258
11.2 Local Energy Trading and Consensus Algorithms . . . . . . 258
11.3 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
xii Contents

11.3.1 Energy Trading Model and Case Study . . . . . . . 260

11.3.2 Performance Result and Evaluation of the Models
at Different Hash Algorithms . . . . . . . . . . . . . 262
11.4 Conclusion and Recommendations . . . . . . . . . . . . . . 266
References . . . . . . . . . . . . . . . . . . . . . . . . . . . 268

Index 271

About the Editors 273

 Preface

Renewable energy resources are alternative to fossil fuels and it always stands
for ‘new findings’ with challenges by the researcher to fulfill the power and
energy demand. The book “IoT, Machine Learning and Blockchain Tech-
nologies for Renewable Energy and Modern Hybrid Power Systems” will
provide an enhanced solution for various aspects of IoT, machine learning,
and blockchain applications from the editors and diverse authors.
This book covers the different sections dealing with fundamentals and
applications of IoT, machine learning, and blockchain technologies in renew-
able energy and hybrid power systems. It includes case studies like power
quality monitoring for low voltage distribution systems through IoT, health
monitoring of distribution transformers through IoT, blockchain with SHA-
256, 384, and 512 application to renewable energy resources, etc.
It well describes topics with theoretical-based analysis and followed
by numerical solutions and simulation results, case studies which make
additional credit to readers for their future research or profession.
The chapters are lucidly covering the significant and bottle-neck chal-
lenges prevailing in the renewable energy and hybrid power systems, enabling
the reader to better understand. The book will be readily available as reference
materials for IoT, machine learning, and blockchain technology applications
to renewable energy & hybrid power systems, and enabling the student
community to create more interest and attention to take up the challenging
renewable energy profession for their endeavors.
It is a unified contribution by international authors from Europe, India,
China, Nepal, the USA, Egypt, Saudi Arabia, and Thailand.

Editors

xiii
 Acknowledgement

Foremost, thanks to the Almighty for his everlasting love throughout this
endeavor.
Acknowledgments are always a phrase to appreciate the resources and
timely solutions either with the digital platform or real-time medium, timely
support and a bond of encouragement is the vital tool for teachers and
researchers from their Institutions. In these regards, we editors express our
sincere thanks to Mr. S. Muthukumaran, Director, TECH Engineering Ser-
vices, Chennai, India; Mr. S. Rajkumar, Executive, JLL, Bengaluru, India;
Mr. K. Sasikumar, Electrical Engineer, Mott MacDonald, Noida, India;
Center for Bioenergy and Green Engineering, Department of Energy Tech-
nology, Aalborg University, Esbjerg, Denmark, Department of Electrical
and Electronics Engineering, KPR Institute of Engineering and Technology,
Tamilnadu, India, Department of Electrical and Electronics Engineering, Col-
lege of Engineering Guindy, Anna University, Chennai, India, Independent
Institute of Education, Johannesburg, South Africa. Editors we got the full
support and executed the task promptly where our Institution devoted the time
and liberty for enhancement with research in particular to make this book a
great success.
I wish One and All for the devoted time frame effort for the grand success
of the book.

Editors

xv
 List of Figures

Figure 1.1 Related concepts of IoT. . . . . . . . . . . . . . . 3

Figure 1.2 Various elements involved in the monitoring
and control of physical systems through IoT. . . . . 4
Figure 1.3 Example of an IoT system – temperature
monitoring. . . . . . . . . . . . . . . . . . . . . . 12
Figure 1.4 A temperature plus humidity sensor. . . . . . . . . 12
Figure 1.5 A soil humidity sensor. . . . . . . . . . . . . . . . 13
Figure 1.6 LDR (left) and photodiodes (right). . . . . . . . . . 13
Figure 1.7 A passive IR (PIR) sensor. . . . . . . . . . . . . . 14
Figure 1.8 An ultrasonic sensor. . . . . . . . . . . . . . . . . 14
Figure 1.9 Example of an automatic door opener. . . . . . . . 16
Figure 1.10 Patient monitoring, an example of on-site
processing. . . . . . . . . . . . . . . . . . . . . . 16
Figure 1.11 Retail stores, an example of an off-site
processing topology. . . . . . . . . . . . . . . . . 18
Figure 1.12 Layers of the OSI model. . . . . . . . . . . . . . . 19
Figure 2.1 Architecture of IoT. . . . . . . . . . . . . . . . . . 34
Figure 2.2 IIoT-ITU reference model. . . . . . . . . . . . . . 35
Figure 2.3 Flow of electrical energy in conventional electric
system. . . . . . . . . . . . . . . . . . . . . . . . 36
Figure 2.4 (a) Exchange of information in client-server
network. (b) Exchange of information in P2P
network. . . . . . . . . . . . . . . . . . . . . . . . 46
Figure 2.5 Structure of IoT in cloud. . . . . . . . . . . . . . . 49
Figure 2.6 Reference architecture of cloud-IoT. . . . . . . . . 50
Figure 2.7 Security challenges in IoT. . . . . . . . . . . . . . 52
Figure 2.8 Interface in industrial platform. . . . . . . . . . . . 54
Figure 3.1 Typical structure of IoT-based monitoring. . . . . . 64
Figure 3.2 Typical block diagram of IoT-based monitoring of
low voltage distribution system. . . . . . . . . . . 66
Figure 3.3 SLD. . . . . . . . . . . . . . . . . . . . . . . . . . 67

xvii
xviii List of Figures

Figure 3.4 Experimental setup. . . . . . . . . . . . . . . . . . 68

Figure 3.5 Block diagram. . . . . . . . . . . . . . . . . . . . 69
Figure 3.6 Flow chart of undervoltage identification. . . . . . 70
Figure 3.7 SMS alert for undervoltage. . . . . . . . . . . . . . 70
Figure 3.8 Flow chart of overvoltage identification. . . . . . . 71
Figure 3.9 SMS alert for overvoltage. . . . . . . . . . . . . . 71
Figure 3.10 Flow chart of interruption identification. . . . . . . 72
Figure 3.11 SMS alert for voltage interruption. . . . . . . . . . 72
Figure 3.12 Flow chart of overvoltage identification. . . . . . . 73
Figure 3.13 SMS alert for overvoltage. . . . . . . . . . . . . . 73
Figure 3.14 Screenshot of real-time monitoring through web. . 74
Figure 4.1 The installation of 100-kVA distribution
transformer. . . . . . . . . . . . . . . . . . . . . . 81
Figure 4.2 Failure and causes of distribution transformer. . . . 83
Figure 4.3 Block diagram of IoT-based health monitoring
of oil-immersed distribution transformer. . . . . . . 84
Figure 4.4 Block diagram of IoT-based transformer health
monitoring. . . . . . . . . . . . . . . . . . . . . . 86
Figure 4.5(a) Hardware experimental setup. . . . . . . . . . . . 87
Figure 4.5(b) Single phase transformer. . . . . . . . . . . . . . . 87
Figure 4.6(a) Online screenshot of the web page – normal
operation. . . . . . . . . . . . . . . . . . . . . . . 88
Figure 4.6(b) Online screenshot of the web page – high winding
temperature. . . . . . . . . . . . . . . . . . . . . . 88
Figure 5.1 Real-world applications of machine learning. . . . 95
Figure 5.2 Classification of machine learning techniques. . . . 96
Figure 5.3 Workflow of supervised learning technique. . . . . 97
Figure 5.4 Workflow of unsupervised learning technique. . . . 99
Figure 5.5 Graphical representation of logistic regression. . . 104
Figure 5.6 Decision tree diagram. . . . . . . . . . . . . . . . 106
Figure 5.7 Graphical representation of linear regression. . . . 107
Figure 5.8 Positive linear relationship. . . . . . . . . . . . . . 108
Figure 5.9 Negative linear relationship. . . . . . . . . . . . . 109
Figure 5.10 Feature space with new arising data point. . . . . . 110
Figure 5.11 Feature space with classification of newer arising
data point. . . . . . . . . . . . . . . . . . . . . . . 110
Figure 5.12 Feature space with all data points. . . . . . . . . . 111
Figure 5.13 Feature space with clusters of all data points. . . . 111
Figure 5.14 Human detection output. . . . . . . . . . . . . . . 114
 List of Figures xix

Figure 5.15 Instagram filtering images. . . . . . . . . . . . . . 116

Figure 5.16 Machine learning in modern hybrid power
system. . . . . . . . . . . . . . . . . . . . . . . . 117
Figure 6.1 Difference between deep learning with machine
learning and artificial intelligence. . . . . . . . . . 126
Figure 6.2 Difference between model (a) development
of traditional programming and (b) development of
machine learning programming. . . . . . . . . . . 127
Figure 6.3 Development of machine learning model. . . . . . 128
Figure 6.4 Simple FFNN. . . . . . . . . . . . . . . . . . . . . 129
Figure 6.5 Simple RNN. . . . . . . . . . . . . . . . . . . . . 130
Figure 6.6 LSTM unit. . . . . . . . . . . . . . . . . . . . . . 131
Figure 6.7 GRU unit. . . . . . . . . . . . . . . . . . . . . . . 131
Figure 6.8 LSTM network model. . . . . . . . . . . . . . . . 132
Figure 6.9 Overall system developed in MATLAB
software. . . . . . . . . . . . . . . . . . . . . . . . 135
Figure 6.10 (a) PV cell single-diode model. (b) PV array
configuration. . . . . . . . . . . . . . . . . . . . . 136
Figure 6.11 (a) I–V characteristics. (b) P–V characteristics. . . 138
Figure 6.12 (a) V–I curve. (b) P–V curve under varying
atmosphere condition. . . . . . . . . . . . . . . . . 138
Figure 6.13 Circuit diagram of cascaded DC–DC converter. . . 139
Figure 6.14 Flowchart of modified P&O MPPT algorithm. . . . 141
Figure 6.15 (a) Influence of irradiation variation. (b) Influence
of temperature variation. (c) Response of the
proposed MPPT algorithm in PV system. . . . . . 142
Figure 6.16 The recorded value of Vmax
for 365 days. . . . . . . . . . . . . . . . . . . . . 143
Figure 6.17 The predicted value of Vmax for another 365 days. . 144
Figure 6.18 Training progress of V max . . . . . . . . . . . . . . 144
Figure 6.19 Converter output voltage with and without
the utilization of LSTM Vref value
in P&O MPPT. . . . . . . . . . . . . . . . . . . . 145
Figure 7.1 Components of deregulation. . . . . . . . . . . . . 152
Figure 7.2 Structure of power system before and after
deregulation. . . . . . . . . . . . . . . . . . . . . 155
Figure 7.3 DFIG-based wind unit model. . . . . . . . . . . . 157
Figure 7.4 Transfer function model of two areas’ deregulated
power system. . . . . . . . . . . . . . . . . . . . . 159
xx List of Figures

Figure 7.5 Transfer function model of PI controller. . . . . . . 161

Figure 7.6 Frequency deviations in area 1. . . . . . . . . . . . 166
Figure 7.7 Frequency deviations in area 2. . . . . . . . . . . . 167
Figure 8.1 Electrical power system. (a) EPS. (b) MHPS. . . . 175
Figure 8.2 Power and inventory flow (traditional at left and
MHPS at right). . . . . . . . . . . . . . . . . . . . 176
Figure 8.3 Single line diagram of the industry. . . . . . . . . . 180
Figure 8.4 Schematic diagram of 6-MW SPP. . . . . . . . . . 183
Figure 8.5 Site weather conditions for solar irradiance. . . . . 183
Figure 8.6 Complete methodology for optimal decision
making. . . . . . . . . . . . . . . . . . . . . . . . 187
Figure 8.7 Final prediction plot. . . . . . . . . . . . . . . . . 189
Figure 8.8 Architecture of artificial neural network. . . . . . . 190
Figure 8.9 Regression analysis of the model-free approach. . . 192
Figure 8.10 Training performance. . . . . . . . . . . . . . . . 193
Figure 8.11 Load, PV, and net injections from the UG. . . . . . 198
Figure 8.12 SLD of industry (left) and power flow result with
ETAP (right). . . . . . . . . . . . . . . . . . . . . 199
Figure 8.13 Nodal voltage magnitude profile. . . . . . . . . . . 199
Figure 9.1 SLD shunt capacitor bank for reactive power
compensation. . . . . . . . . . . . . . . . . . . . . 207
Figure 9.2 Average 24 hours load value during three months
of Thailand. . . . . . . . . . . . . . . . . . . . . . 208
Figure 9.3 SLD n-busbars sample distribution system. . . . . 209
Figure 9.4 Current–voltage characteristic of a capacitor. . . . 212
Figure 9.5 A short transmission power system for the power
flow indication. . . . . . . . . . . . . . . . . . . . 213
Figure 9.6 Flowchart for optimum capacitor allocation
by PSO. . . . . . . . . . . . . . . . . . . . . . . . 217
Figure 9.7 Schematic SLD of the IEEE-15-busbar system. . . 219
Figure 9.8 MVA and MVAR loads in simulation IEEE-15
busbar distribution system. . . . . . . . . . . . . . 221
Figure 9.9 Optimal singly capacitor bank value
of the 15-busbar system. . . . . . . . . . . . . . . 222
Figure 9.10 Value of saving power losses in the 15-busbar
system after using singly capacitor bank. . . . . . . 222
Figure 9.11 Voltage value for IEEE-15 busbars without and with
adding the shunt capacitor bank. . . . . . . . . . . 224
 List of Figures xxi

Figure 9.12 Flowchart operation logic steps for ACCS

module. . . . . . . . . . . . . . . . . . . . . . . . 226
Figure 9.13 Single line diagram for actual busbar voltage
of 11 kV. . . . . . . . . . . . . . . . . . . . . . . . 227
Figure 9.14 Close and open individual logic ACCS/IED module
command. . . . . . . . . . . . . . . . . . . . . . . 228
Figure 9.15 ACCS/IED logic for capacitor bank-1 and capacitor
bank-2 at the close the bus tie circuit breaker. . . . 229
Figure 10.1 Comprehensive graph of blockchain technology. . . 239
Figure 10.2 Core technology graph of blockchain technology. . 240
Figure 10.3 Expansion technology graph of blockchain
technology. . . . . . . . . . . . . . . . . . . . . . 248
Figure 10.4 Supporting technology graph of blockchain
technology. . . . . . . . . . . . . . . . . . . . . . 251
Figure 11.1 Typical structure of blockchain. . . . . . . . . . . 259
Figure 11.2 Demand power from Library building and net
surplus solar power at Energy building. . . . . . . 261
Figure 11.3 Visual model of energy trading. . . . . . . . . . . . 261
Figure 11.4 Heatmap visualization between transaction and
hardware performance. (a) SHA-256. (b) SHA-384.
(c) SHA-512. . . . . . . . . . . . . . . . . . . . . 264
Figure 11.5 Duration of each block index in the transaction. . . 264
Figure 11.6 (a) CPU usage and (b) power consumption
in different SHA scenarios. . . . . . . . . . . . . . 265
Figure 11.7 GPU usage in different SHAs. (a) NVIDIA GeForce
MX330. (b) Intel UHD. . . . . . . . . . . . . . . . 266
Figure 11.8 Trendline models of hardware performance
in SHAs. (a) GPU 1. (b) GPU 2. (c) CPU. . . . . . 267
 List of Tables

Table 1.1 Different types of the protocol used in IoT systems. . 20

Table 2.1 RES challenges as per the root cause analysis. . . . . 31
Table 5.1 Detailing of data properties. . . . . . . . . . . . . . . 94
Table 5.2 Difference between classification algorithm
and regression algorithm. . . . . . . . . . . . . . . . 98
Table 5.3 Advantages and disadvantages of supervised learning
and unsupervised learning. . . . . . . . . . . . . . . . 101
Table 5.4 Difference between supervised learning
and unsupervised learning. . . . . . . . . . . . . . . . 101
Table 5.5 Difference between supervised learning
and reinforcement learning. . . . . . . . . . . . . . . 103
Table 6.1 PV array specification. . . . . . . . . . . . . . . . . . 135
Table 6.2 PV array specification. . . . . . . . . . . . . . . . . . 138
Table 6.3 Cascaded buck and boost converter parameters. . . . . 139
Table 6.4 Comparison of recorded and predicted Vmax . . . . . . 145
Table 7.1 Frequency deviation of each area. . . . . . . . . . . . 167
Table 7.2 System parameters. . . . . . . . . . . . . . . . . . . 168
Table 8.1 ML application in MHPS. . . . . . . . . . . . . . . . 179
Table 8.2 Comparison of parameters of various regression
models. . . . . . . . . . . . . . . . . . . . . . . . . . 188
Table 8.3 Initial regression results for irradiance prediction. . . 189
Table 8.4 Final regression results for irradiance prediction. . . . 189
Table 8.5 Industry technical parameters. . . . . . . . . . . . . . 197
Table 8.6 Optimal value comparison. . . . . . . . . . . . . . . 200
Table 9.1 IEEE-15 busbar simulation system data. . . . . . . . 220
Table 9.2 General data for IEEE-15 busbar system. . . . . . . . 221
Table 9.3 Voltage value at IEEE-15 busbar simulation system
after adding the capacitor bank. . . . . . . . . . . . . 223
Table 9.4 Setting values to operate ACCS/IED at different cases. 228
Table 10.1 Characteristics of common consensus mechanisms. . . 246
Table 11.1 Block information of SHA-256, 384, and 512. . . . . 263

xxiii
 List of Contributors

Ahmed, Abou-Hashema, Electrical Engineering Department. Minia

University, Egypts
Alhaider, Mohemmed, College of Engineering at Wadi Addawaser, Prince
Sattam bin Abdulaziz University, Saudi Arabia;
E-mail: [email protected]
Anbarasi, S., P S R Engineering College, India
Ansari, Saniya M., E & TC Department, Dr D Y Patil School of Engineering
(DYPSOE), India; E-mail: [email protected]
Ayyagari, Krishna Sandeep, Department of Electrical & Computer Engi-
neering, The University of Texas at San Antonio, USA;
E-mail: [email protected]
Balaji, S., IIT Kanpur, India
Balasubramanian, T., P S R Engineering College, India
Calay, Rajnish Kaur, Department of Building, Energy and Material Tech-
nology, UiT The Arctic University of Norway, Norway;
E-mail: [email protected]
Chen, Xiaofeng, Hangzhou Qulian Technology Co., Ltd., China;
E-mail: [email protected]
Elango, S., Coimbatore Institute of Technology, India
Elghaffar, Amer Nasr Abd, CAlfanar Engineering Service, Alfanar Com-
pany, Saudi Arabia, Electrical Engineering Department. Minia University,
Egypt; E-mail: [email protected]
El Sayed Morsy, Osama, Alfanar Engineering Service, Alfanar Company,
Saudi Arabia
Eltamaly, Ali Mohamed, Electrical Engineering Department, Mansoura
University, Egypt; Sustainable Energy Technologies Center, King Saud

xxv
xxvi List of Contributors

University, Saudi Arabia, K.A.CARE Energy Research and Innovation Center,

Saudi Arabia
Ezhilarasi, G., Sri Sairam Institute of Technology, India
Gunapriya, D., Sri Krishna College of Engineering and Technology, India;
E-mail: [email protected]
Jia, Xiangjuan, Hangzhou Qulian Technology Co., Ltd., China;
E-mail: [email protected]
Khakurel, Saju, Electronics and Communication Engineering, Nepal;
E-mail: [email protected]
Lakshmi, D., Academy of Maritime Education and Training (AMET), India;
E-mail: [email protected]
Maharjan, Asim, Electronics and Communication Engineering, Nepal;
E-mail: [email protected]
Medam, Raveendra Reddy, Department of EEE, Maturi Venkata Subba Rao
Engineering (MVSR) College, India; E-mail: [email protected]
Mohamed, Yehia Sayed, Electrical Engineering Department. Minia Univer-
sity, Egypt
Mustafa, Mohamad Y., Department of Building, Energy and Material Tech-
nology, UiT The Arctic University of Norway, Norway;
E-mail: [email protected]
Ongsakul, Weerakorn, Department of Energy, Environment, and Climate,
School of Environment, Resources and Development, Asian Institute of
Technology, Thailand
Patil, Ravindra R., PhD Scholar, Department of Building, Energy and Mate-
rial Technology, UiT The Arctic University of Norway, Norway;
E-mail: [email protected]
Punitha, K., P S R Engineering College, India;
E-mail: [email protected]
Ravi, C.N., Vidya Jyothi Institute of Technology, India
Reddy, B. Koti, Department of Atomic Energy, India;
E-mail: [email protected]
Sabareeshwaran, K., Karpagam Institute of Technology, India;
E-mail: [email protected]
 List of Contributors xxvii

Sanjeevikumar, P., Aarhus University, Denmark;

E-mail: [email protected]
Sharmeela, C., Anna University, India; E-mail: [email protected]
Sivakumar, R., Akshaya College of Engineering and Technology, India;
E-mail: [email protected]
Sivaraman, P., Vestas Technology R&D Chennai Pvt Ltd, India;
E-mail: [email protected]
Thu, Kaung Si, Department of Energy, Environment, and Climate, School
of Environment, Resources and Development, Asian Institute of Technology,
Thailand; E-mail: [email protected]
Tiwari, Shubham, Department of Energy, Environment, and Climate, School
of Environment, Resources and Development, Asian Institute of Technology,
Thailand
Zahira, R., BSA Crescent Institute of Science and Technology, India;
E-mail: [email protected]
 List of Abbreviations

ACCS Automatic capacitor control scheme

AI Artificial intelligence
CPS Cyber-physical systems
DEM De-regulated electricity markets
DFIG Doubly fed induction generator
DG Distributed generation
DN Distribution network
ECDSA Elliptic curve digital signature algorithm
EMM Energy management models
ESS Energy storage systems
ETAP Electrical transient analyzer program
FACTS Flexible AC transmission system
FFNN Feed forward neural network
FPA Flower pollination algorithm
GA Genetic algorithm
GPRS General packet radio service
GPS Global positioning system
GSM Global system for mobile communication
IED Intelligent electronic devices
IoT Internet of things
IR Infrared sensors
IRENA International renewable energy agency
ITU International telecommunication union
KNN K-nearest neighbours
LCD Liquid crystal display
LDR Light-dependent resistors
LFC Load frequency control
LSTM Long-short term memory
M2M Machine to machine
MG Micro grids
MHPS Modern hybrid power system
ML Machine learning

xxix
xxx List of Abbreviations

MPPT Maximum power point tracking

OSI Open systems interconnection model
PI Proportional integral
PID Proportional integral derivative
PoW Proof-of-work
PQ Power quality
PSO Particle swarm optimization
PV Photovoltaic
PV Predictor variable
RNN Recurrent neural network
RTOS Real-time operating system
RV Response
SA Smart appliances
SARSA State action reward state action
SG Smart grids
SHA Secure hash algorithm
SLD Single line diagram
VRES Variable renewable energy resources
WRS Wireless sensor network
 1
Introduction to IoT

Asim Maharjan and Saju Khakurel

Electronics and Communication Engineering, Nepal

E-mail: [email protected]; [email protected]

Abstract
This chapter discusses the field of the Internet of Things, or better known as
IoT, and various concepts related to it along with its vast applications and var-
ious challenges faced by IoT in the present world. A bit of the history of IoT
is also given so that the readers can get a sense of how long ago the concept
of IoT was conceived before becoming the revolution it is today. It goes over
some of the important milestones in the history of IoT from its naming to its
widespread adoption. Some of the applications of IoT, namely, in households,
healthcare, industries, and renewable energies are also discussed to show the
fields that have already begun to integrate IoT technologies in their operation.
A brief description of the enabling technologies of IoT is also given with the
purpose that the readers will be able to grasp the overall working of an IoT
system. It then explores the recent developments and achievements in the
field of IoT to give readers a glimpse of the present-day landscape of IoT.
Finally, the chapter also touches upon the many different kinds of issues and
challenges that present-day IoT systems are facing. These range from the
technical aspects of IoT such as compatibility, interoperability, and security
to the more ethical aspects like privacy.

Keywords: Internet of Things (IoT), sensor, ubiquitous computing, home

automation, Industry 4.0.

1
2 Introduction to IoT

1.1 Introduction
With the exponential evolution of technology in almost every field like
communication technology, digital technology, machinery, robotics, power
and energy, and many more, it is common to have a few of these advancing
technologies combined to form even more complex and advanced techniques
and systems. In recent years, there have been numerous researches for the
expansions in the existing systems to improve their efficiency and application
in various sectors. It was not until a few years back when there were cable line
telephones for communication. But due to the advancement and exploration
in the communication field along with the accelerating development in wire-
less technology and digital technology, today, almost every means of com-
munication is wireless along with the digitized systems embedded in it. And
now with the rapid replacement of smart technologies among all the existing
systems, almost all electronic gadgets are as smart as a human task would be.
Following this, the idea of convergence of multiple technologies gives rise
to various newer ideas for the implementation of the existing technologies for
a better and dynamic system. Internet of Things (IoT) is one of the mixtures
of such multiple existing technologies, which incorporates multiple technical
fields such as sensors, the Internet, software, real-time operating system
(RTOS), embedded systems, etc. The idea of IoT is to act as a bridge that
connects physical things from the real world to the virtual world through the
Internet. This concept of connecting the physical and the virtual worlds was
first initially proposed by Mark Weiser in the early 1900s, where IoT allows
physical objects of the real world to be able to be remotely controlled via
the advancing technologies through the Internet [7]. Later in 2005, the idea
of the IoT was officially introduced by the International Telecommunication
Union (ITU) at the World Summit that was held at the Information Society
in Tunisia. They also released an ITU Internet report which included an in-
depth knowledge about the IoT, its concepts, and its global effect around the
globe [24].
IoT is a large networking platform for interaction and control of multiple
electronic tools via the Internet. In simple words, it is a combination of sen-
sors, processors, and then controllers for actuation, which all communicate
through the use of the Internet. The data streams, from the sensors or any
other smart devices that acknowledge the change in environmental factors,
get stored up in a common platform where the necessary information gets
processed, computed, and analyzed. This filtered data can either be stored for
future reference or can be used immediately for taking actions like controlling
 1.1 Introduction 3

Figure 1.1 Related concepts of IoT.

or switching the other devices sharing the common platform in the network.
It carries the idea of independence of human interventions in monitoring and
control of any digital systems for efficient work through. These systems do
not possess any delay, minimize human effort by expanding the independence
of humans, efficiently utilize the resources, save time, and are transparent.
Figure 1.1 shows the concepts of IoT.
With the expanding knowledge, the unlimited possibility of implementa-
tion areas, and exponential advancement in technology, almost every human
task is replaceable with a monitoring and control system. Few sensors for
observing the changes in respective parameters of a dynamic environment,
a processor for implementing the required algorithm, and controllers or
actuators for maintaining the desired state. This idea expands the horizon of
IoT which now, with few advanced types of equipment, can be implemented
in almost every sector; for example, self-driving cars, medical recordings,
smart home systems, manufacturing to customer care, irrigation systems,
educational institutions, e-commerce, and so on. Figure 1.2 shows the various
elements that are involved in the monitoring and control of the physical
system through IoT.
A smart home can be considered the best example of IoT. The smart
home system incorporates all the features that an IoT has to offer along
with providing a clear concept on “What exactly is an IoT?” As mentioned
4 Introduction to IoT

Figure 1.2 Various elements involved in the monitoring and control of physical systems
through IoT [25].

earlier, IoT is a system where the physical objects are connected to a common
virtual platform for interacting with every other device in the network, which
are programmed to perform certain tasks that require less to no human
interventions.
In a smart home system, multiple sensors are embedded in different
locations of the house that tracks down the changes in a state like temperature,
lighting, air quality, etc. These sensors are the starting point in the system
whose data are then collected to a virtual cloud via the Internet. The data
in the cloud is then processed and filtered after which the processed data is
analyzed in real time. The endpoint to this system can be the APIs or the
controllers in the home itself. This system allows the user to make changes in
their homes when no one is physically present there or just monitor the activ-
ity that is taking place at the home when they are away. Either way, this sys-
tem excludes the human intervention, establishes a virtual connection among
the multiple devices, and takes necessary actions as it is programmed to.
Along with the existing system, there are numerous possibilities to
increase its ability of performance such as a system that connects the user
phone to the home system, calculates its distance from the home and unlocks
the doors, sets the temperature, and turns on the lights on the arrival of the
user. Furthermore, the expandability of these systems is just limited by human
ideas, and with the massive developments that have been occurring in this
field, there soon arrives a future where just the thought of outdoors triggered
in the brain can cause the car to be ready at the front door.
 1.2 History 5

1.2 History
The term Internet of Things was first coined by Kevin Aston in 1999, in a
presentation for Procter & Gamble linking the concepts of radio-frequency
identification (RFID) to the company’s supply chain [1].
However, the first use of the IoT, as we know it now, was even before the
term was coined. The concept first appeared in the mid-1970s at the Com-
puter Science Department of Carnegie Mellon University in a Coke vending
machine. It allowed students to view the status of the vending machine, such
as knowing when the vending machine was full/empty or whether the drinks
inside the vending machine were hot or cold. The vending machines had
micro switches that were used to detect the coke bottles. These switches were
then connected to the university’s mainframe computer based on PDP-10. The
students could then inquire about the status of the vending machine through
an inquiry program written for it.
Back then, the Internet as we know it today did not exist. The communica-
tions were mostly carried out through ARPANET, which would later become
known as the Internet [2].
This is just one of the examples of the concepts of IoT being conceived
before the term was even coined. Many authors have written about devices
interacting with each other through wireless means and have talked about
such concepts in great detail. It is only now that such concepts have taken
concrete forms.
As the years progressed from the mid-1970s, the Internet became more
available to the general public. With large businesses adopting the Internet,
the vast possibilities tied with the Internet rapidly started coming to light.
With computers connected to every part of the world, sharing information
was trivial. Along with the development of the Internet, computers were
becoming smaller, faster, and more available. In 1991, author Mark Weiser
wrote the article “The Computer for the 21st Century” [7]. The concept of
Ubiquitous Computing began with it. The author depicted many real-life
situations illustrating the widespread adoption of computers in performing
day-to-day tasks such as making coffee, reading newspapers, identifying
oneself, and so on. The author also noted how interconnections among these
devices would present many challenges to the networking infrastructure back
then. In the present, however, network infrastructures are very developed and
network bandwidths of up to Gigabits are readily available. Thus, the future
depicted by Mark Weise has certainly turned out to be a reality in today’s
world.
6 Introduction to IoT

Attempts were made in the 1990s to have devices connected by following

a set of standard protocols. Microsoft’s “At Work” and NEST were two big
technologies that attempted to embody the spirit of IoT. However, neither of
them saw any widespread use.
IoT would see its emergence during 2008-2009 [3]. The first international
conference of IoT was held in Zurin in 2008, where 23 out of 92 submissions
were accepted [4]. Following the first international conference, many other
such conferences related to the field of IoT were also held. Thus, IoT started
gaining rapid popularity among scholars and tech enthusiasts alike.
Today, IoT is one the most rapidly growing fields in technology with vast
applications in business, economics, health care, industry, power and energy,
electric vehicles, and so on. With everything being smart and connected to
the Internet, the need to connect them to achieve accessibility of orders of
magnitude greater grows stronger every day.

1.3 Applications of IoT

1.3.1 Domestic Applications
IoT has vast applications in domestic fields. Applications range from the
usual household monitoring to waste management and home security.
Household monitoring is one of the most known examples of IoT
in households. With the help of IoT, people can monitor various equip-
ment in their homes, such as lights and other electronic appliances. With
advancements in technology and several everyday household appliances like
refrigerators, television, etc., becoming smart, they can be connected to the
Internet and can provide us with valuable data about their usage. This data
can be accessed from anywhere in the world through the Internet. Another
one of the major applications is switching, i.e., the act of powering ON/OFF
devices through the internet. With this, we can effectively control our house
from anywhere in the world.
Besides this, IoT can provide us with sophisticated data such as power
usage, temperatures, amount of oxygen, and so on. Analyzing these data
and performing the necessary changes to optimize these variables can help
people dramatically improve their living conditions. Power utilization around
the house can be made more efficient by powering ON/OFF certain appli-
ances based on the data collected through sensors, e.g., light in the room
automatically turns OFF when there is no one in the room.
One of the famous examples of such devices for home automation based
on IoT is the Amazon Echo, which allows users to control many other
 1.3 Applications of IoT 7

devices such as speakers, televisions, and so on [23]. It even has additional

features such as voice commands, web search, temperature sensing, etc.,
which provides immense accessibility to the average person. Thus, the power
of IoT in the household department can be readily seen from this example.
A distinct feature of household automation is centralized control, i.e.,
controlling many appliances using a single device such as your mobile phone,
computer, etc. As an example, web interfaces can be built, through which you
can change the temperature of your smart heating system, control the room
temperature of your smart AC, turn OFF the light, view your electricity usage,
and many more.
With such vast applications in the domestic field and the ease of availabil-
ity of hardware devices, many enthusiasts are also attracted to building such
systems on their own. Thus, domestic applications remain one of the most
prominent areas for deploying IoT-based solutions.

1.3.2 Applications in Healthcare

IoT has a wide range of applications in the health industry. The term Internet
of Medical Things (IoMT) was introduced to specifically refer to the applica-
tion of IoT in healthcare. In an infographic published by Deloitte, over 60% of
EU health organizations have increased their adoption of digital technologies
and have also adopted virtual ways of interacting with patients [10].
The adoption of digital technologies provides doctors and other medical
professionals with new avenues for medical diagnosis and patient monitor-
ing. With IoT-enabled devices, remote health monitoring and emergency
notification systems can be put to use [11]. Many such devices have been
developed and are in a constant state of improvement. Devices for heart rate
monitoring are present today in the form of smartwatches and are capable
of providing sophisticated levels of information such as heart rate, irregular
heartbeat notifications, ECG reports, and even sleep monitoring. The devices
can also be used to provide medical information, such as the dosage and
when to take it, and alerts as well as for monitoring whether a patient has
taken the medicine or not. Through the use of IoT, the data generated by such
devices can be provided to doctors and physicians who can analyze them
and provide better medical diagnoses to the patients. It also allows people to
monitor themselves their own health conditions and improve their lifestyles
to achieve better health.
Another promising application of IoT in the healthcare industry is in
patient record management. All the medical records of patients can be stored
8 Introduction to IoT

and then made to be readily available to the patient and other doctors and
professionals. This way, doctors who are treating the patient will have access
to all the prior records which can lead to a more efficient and accurate
diagnosis [8].
In addition to IoT, machine learning and data analytics have also seen
widespread adoption in healthcare. Machine learning in combination with IoT
can provide a very powerful tool for medical diagnosis. In [12], the authors
have proposed a predictive machine learning algorithm for the detection of
heart diseases by using data such as blood pressure, heart rate, respiratory
rate, and blood sugar. In [13], an IoT system for monitoring the health
condition of patients with high blood pressure is proposed by using HRV
parameters. Many similar studies regarding the field of data analytics in
conjunction with IoT for the medical sector can be found in [14].

1.3.3 Applications in E-commerce

IoT has also found its use in the E-commerce sector. One of the applications
is in supply chain management which deals with handling the flow of goods
and services from production to delivery to the customers. Rapid digitization
has also affected this area very much. With the help of IoT, the information
of each and every item in the warehouses can be made available, and product
identification and tracking in each stage in supply chain management can be
achieved using RFID [26]. Data such as temperature, humidity, etc., are of
much importance when the storage of goods is concerned. Video cameras
have been used for surveillance for a very long time throughout history for
security purposes. Similarly, automated unmanned vehicles also provide a
great productivity boost in the supply chain as this greatly reduces the labor
cost. Consequently, the entire warehouse can be monitored effectively [27].
The survey in [15] highlights the value of IoT-based systems in SCM. In
that survey, 12 participants from various companies belonging to a higher
rank in SCM were interviewed about the effects of adopting an IoT-based
approach. All the participants were very positive about the effects of IoT
in their day-to-day work. Thus, IoT can play a big role in supply chain
management.
Besides this, IoT can also greatly improve the consumer experience in the
retail market. With online shopping/marketing becoming the norm nowadays,
this field has already aggregated IoT in its business model. Information
about the items in a retail shop is readily available through various outlets
 1.3 Applications of IoT 9

such as websites, mobile applications, and such. Food menus in restaurants

along with the vacancy and reservations are only a few clicks away. Many
applications are now available that allow people to compare prices between
various goods and services provided by different providers which can greatly
aid in the decision-making process for a consumer. Online payment systems
are also gaining a lot of traction. With online payment services becoming
more accessible to retail shops along with the convenience of being able
to pay without using physical cash, online payments will likely become the
preferred way of payment in the near future.

1.3.4 Industrial Applications

In the industrial sector, IoT is better known as the industrial Internet of
Things (IIoT). IIoT represents what is known as Industry 4.0 [28], the new
revolution in the industrial sector brought on by the use of IoT, data analytics,
cyber-physical systems (CPS), cloud computing, edge computer, and other
modern technologies. Many technologies related to IoT such as wireless
sensor networks (WSNs) and RFID are heavily used for the identification and
tracking of goods in industries. WSNs enable industries to accurately sense
and monitor their environmental conditions by utilizing a network of sensors
placed throughout the manufacturing plant [29]. It allows them to monitor
the present conditions and make necessary changes to get maximal product
output. Intelligent process control methods with the help of a wide variety
of sensors can be applied in the traditional manufacturing process which can
further boost productivity.
Besides manufacturing, IoT also provides a lot of benefits in the agricul-
tural sector. A large number of sensors connected can provide farmers with
data such as soil humidity, temperature, and pH levels of the entire farm.
This allows the farmers to make proper adjustments to the ever-changing
environment to get better crop yields. RFID is widely used in livestock
farming for tracking animals, their behaviors, and their physical conditions
[16]. The physical conditions of individual animals can be tracked easily and
be made available to the farmers who can take better care of the animals.
Sensors can provide information such as the location of cattle when they
are grazing and their health condition such as temperature, heartbeat, etc.
Any anomaly in their health conditions can now be easily detected and swift
treatment can be done. Thus, productivity in agriculture and farming can be
greatly increased by the use of IoT-enabled technologies.
10 Introduction to IoT

1.3.5 Applications in Energy

IoT has found great applications and huge potential in the field of energy. As
it has been discussed over several applications, one of the major uses of IoT
is for monitoring processes. This works particularly well for increasing the
efficiency of energy utilization.
Various energy monitoring systems have been proposed using IoT, such
as those given in [30–32]. These systems monitor the power consumption
and allow the users to visualize the energy usage of their households and
buildings. This allows people to more efficiently utilize their electricity
consumption by preventing unnecessary usage of electrical appliances. When
released to the general public, small energy savings can greatly add up and
lead to a lower energy consumption overall.
IoT has also seen several applications in the field of managing and
monitoring solar energy. Solar power is greatly influenced by environmental
factors such as the intensity of the sun, cloud movements, etc. [33]. Therefore,
we require constant monitoring for the efficient and reliable operation of
solar systems. A solar facility with IoT capabilities was developed in Arizona
State University park [34, 35]. It allowed users to obtain various information
related to the solar plant operation such as cloud movement and solar panel
analytics. The smart monitoring devices installed were also able to perform
more sophisticated tasks such as fault detection, bypassing of faulty panels,
changing panel connections, etc. Many other solar panel monitoring systems
have also been proposed such as those in [36–38]. These systems allow users
to monitor the voltage, current, and temperature of the solar panels remotely.
Similarly, IoT can also be used to monitor wind turbines. Variables such as
wind speed, rotation of turbines, generated output, etc., can be monitored
from the turbines. Since turbines are generally placed in remote windy areas,
the wireless nature of IoT can provide further benefits in monitoring and
control of the turbines. IoT can also be used for the automated control of solar
as well as wind-powered plants, which can greatly help increase efficiency
and power generation. By monitoring variables such as wind direction, cloud
formations, etc., we can adjust the parameters of solar panels and wind
turbines to get the maximum energy generation from them.
Another application of IoT in the energy field is in the monitoring of smart
grids. The smart grid is a power grid that can deliver power in a controlled
way according to the demand of the consumers [39]. These grids intelligently
integrate the power consumption and supply according to the available as well
as produced energy. Reliable and real-time information are very important
 1.4 Technical Details of IoT 11

in smart grids for robust and efficient delivery of power in smart grids [40].
To this extent, IoT provides a general framework to communicate information
between devices in a smart grid. It can also be used for monitoring and
metering the various power plants, domestic houses, and electric vehicles
which are all a part of the distributed smart grid network. IoT in smart grids
can also be used to monitor and collect consumer data which, when coupled
with data analysis techniques such as machine learning, can be used to create
energy management models (EMMs) [58]. The EMM can then be used to
optimize the performance of the smart grid. They can also be used to predict
and analyze consumer energy demand in the future. Therefore, IoT has a huge
potential in the future in the field of energy.

1.4 Technical Details of IoT

1.4.1 Sensors
Generally speaking, sensors are those devices that convert physical phenom-
ena like temperature, pressure, sounds, etc., into electric signals. Sensors
provide us with the means to observe our world in terms of electrical
signals. These signals are generally converted into a digital form and then
processed.
For example, consider the temperature sensor measuring the temperature
of a greenhouse. In this case, the greenhouse will act as a plant, i.e., a system
which is being monitored by us. The temperature inside the greenhouse is
about 30◦ C. The temperature sensor will give the output in the form of
voltage, say around 4 Volts. This voltage of 4 Volts will then correspond to a
temperature of 30◦ C. Furthermore, these 4 Volts will be converted into digital
form, i.e., in terms of bits by using a special device known as analog-to-digital
converter (ADC). If we assume that the ADC output is 1101 0000 bits or 208
in decimal, then this value of 208 will now correspond to the temperature
of 30◦ C. In most cases, 208 will immediately convert into a numerical
value of 30◦ C to make processing easier. This example only highlights the
major components involved in extracting data from sensors, namely the plant
(greenhouse in our case), the sensor itself, a device to convert the analog
reading into a digital form, and finally a processor to act upon this acquired
data as shown in Figure 1.3.
This is only a relatively simple and a small example of how a sensor
integrates with the IoT environment [59]. The processing device will then
act as an interface for the IoT system to interact with the environment that the
12 Introduction to IoT

     

 


   


     


Figure 1.3 Example of an IoT system – temperature monitoring.

sensor will be monitoring. The complexity of the sensing system depends on a

large number of factors such as the plant being monitored and the phenomena
being monitored. Nowadays, many sensors even come with built-in ADC
and thus only need some form of communication with the processing device
which further simplifies the work done by the processor.
Some of the sensors used to monitor the physical devices and their
description are given below.

A. Temperature Sensor
As the name suggests, these sensors measure the temperature of the environ-
ment. They do so by using thermistors, thermocouples, and other semicon-
ductors. Figure 1.4 shows the typical sensor used to monitor both temperature
as well as humidity.

B. Soil Humidity Sensor

As the name implies, soil humidity sensors are used to monitor soil humidity
in agricultural land. The typical soil humidity sensor is shown in Figure 1.5.

Figure 1.4 A temperature plus humidity sensor.

 1.4 Technical Details of IoT 13

Figure 1.5 A soil humidity sensor [51].

Figure 1.6 LDR (left) and photodiodes (right).

C. Photodetectors
The photodetectors are used to measure the intensity of light in the
environment. Two common light sensors are the light-dependent resistors
(LDRs) and the photodiode. An LDR varies its resistance according to the
amount of light falling on it. Photodiodes, on the other hand, work on the
principle of photoelectric effect where a photon from a light source striking a
semiconductor produces electrons, which, in turn, causes current to flow. The
typical LDR and photodiode are shown in Figure 1.6.

D. Infrared Sensors
Infrared (IR) sensors are used to detect IR radiations that are invisible to
the human eye. They have a wide number of applications such as object
and motion detection. IR can also be used for data transmission. These
sensors can be active or passive. An active sensor has a built-in IR generation
mechanism as well as receiving mechanism, whereas the passive IR sensor
only detects the incoming IR radiations. The typical passive IR sensor is
shown in Figure 1.7.
14 Introduction to IoT

Figure 1.7 A passive IR (PIR) sensor [53].

Figure 1.8 An ultrasonic sensor [51].

E. Ultrasonic Sensors
Ultrasonic sensors are used for measuring distance. They work by measuring
the time difference between the transmitted and the reflected ultrasonic waves
which can then be converted into the distance by using the speed of the
ultrasonic wave. A typical ultrasonic sensor is shown in Figure 1.8.

F. Gas Sensors
Gas sensors are used to detect the presence and concentration of various
kinds of gases in the atmosphere. One of their major uses is in air pollution
monitoring, where the sensor is used to measure the harmful gases present in
the air such as CO2 , carbon monoxide, etc.
 1.4 Technical Details of IoT 15

G. Atmospheric Pressure Sensor

Atmospheric pressure sensors are used to measure the atmospheric pressure
and also sometimes as an altimeter for measuring the altitude of aircraft,
spaceships, etc.

1.4.2 Actuators
Actuators are those components that convert an incoming signal such as
an electrical voltage into a physical form such as a force, velocity, light,
temperature, etc., or even into a non-mechanical form such as voltage and
current itself [42]. They receive an input signal from a controller and produce
an output that depends on the input signal. The output of the actuator interacts
with its surroundings to bring some kind of change to it. As an example, a
DC motor can be thought of as an actuator with an electrical voltage as its
input signal and the rotation of its shaft as the output. The rotation of the
shaft can interact with the physical world in several different ways, whether
it is changing the position of the system by driving some wheels, opening or
closing a door, pumping water, etc. Depending on the type of input signals
and the corresponding output signal, actuators can be classified into several
classes such as hydraulic, pneumatic, electric, mechanical, and so on [41].
In IoT-based systems, electric actuators are dominant primarily because
they can be controlled using electrical signals. Some examples of electric
actuators are DC motors, servo motors, electromechanical relays, etc. Electric
actuators are also versatile and flexible in the sense that input signals can be
controlled by the software running in the processor which allows for much
better control over their output. For example, the speed of a DC motor and
the position of a servo motor can be easily controlled through software. Thus,
by including programming logic, a single electric actuator can be used for
a vast number of applications. Furthermore, a processor can use the input
data from the sensors to generate the input signals for actuators. This is
a fundamental aspect of smart devices where devices can produce outputs
according to the inputs without any human interference [42]. The typical
example of the automatic door opening and closing is shown in Figure 1.9.
Figure 1.10 shows the simple automatic door opening system in which a
DC motor is used as an actuator whose goal is to simply open and close the
door. The camera acts as a sensor that gives pictures as input to the processor.
The processor is then tasked with person detection. When it detects someone
approaching the door, it gives an electrical signal to the DC motor to open
the door. Obviously, in an IoT-based system, we would be able to do much
16 Introduction to IoT

 
  




Figure 1.9 Example of an automatic door opener [49], [50].




   
 
  

Figure 1.10 Patient monitoring, an example of on-site processing [50].

more than just opening and closing of doors based on the camera image.
With multiple of these systems placed at various locations along with popper
person identification, we can track people in buildings solely based on which
doors they have passed through.

1.4.3 Processing Topologies

Data processing is a very important part of IoT systems. With vast amounts
of data collected from the physical system, it is of utmost importance to have
processing capabilities that can work with huge amounts of data efficiently.
Several things have to be considered for choosing the right processing topol-
ogy. Data considerations are some of the most important factors. For example,
IoT-based systems can be real time, where data is collected and processed in
real time. These systems are generally process monitoring systems such as
monitoring the temperature in a room, the pressure inside a chemical reaction
chamber, monitoring the heart rate and blood pressure of a medical patient,
 1.4 Technical Details of IoT 17

etc. These systems involve a comparatively smaller workload; however, the

latency must also be very small. Here, latency would mean the time between
when the observation was taken (i.e., when the heart rate of a patient is
measured) and when the processor would produce the corresponding output.
On the other hand, IoT systems also have to deal with non-time critical data
such as soil/humidity moisture from agricultural farms. These data can be
stored in a database, possibly in a data center, and can be processed later
on. Thus, latency is not a very important factor for these types of systems.
The level of sophistication required is also very less compared to real-time
systems in which every component in the IoT system must run with minimal
latency for proper functioning [43].
Because of the nature of real-time systems, it has become necessary to
add computing power near the source of data so that the latency is minimized
as much as possible. Thus, two topologies are naturally born from this: on-site
processing and off-site processing [17]. On-site processing refers to the data
processing that occurs at the source itself. Here, the sensors are coupled with
devices that have significant processing capabilities. These devices may be
microprocessors, microcontrollers, or even large mainframe or supercomput-
ers depending on the need. In this scenario, it is no longer dependent on the
network for computing and, therefore, in some ways, more reliable for real-
time systems. It has the added benefit of needing smaller bandwidth. The data
collected can be shared in the IoT system at a later time. Here, the processing
of data takes a higher priority than the data being available to the end-users.
Figure 1.10 shows the typical on-site data processing system.
On a similar note, we have off-site processing. Here, the processing is
done remotely, i.e., not at the source of the data. The data generated at the
source has to be transmitted somehow to the processing site where it is
processed and the processed results are then transferred to the users. Thus,
it will generally have higher latency since it involves the transmission and
reception of data over a network. It is generally used when the processing
of data can be delayed with no repercussions. One of the main advantages
of off-site processing is that it can greatly reduce implementation costs.
Multiple sensor nodes can utilize a single processing unit compared to on-
site processing where every sensor node needs to be paired with a processing
unit. They can also be used for heavier workloads on the data since off-site
processing gives us more flexibility in choosing more powerful hardware. As
an example, consider a retail store that changes its goods based on customers’
preferences. For this, the data of customers’ preference of items are logged.
Off-site processing is preferable for this since the store needs to collect a large
18 Introduction to IoT

 

 




    


  

 


  
   




 

 

 



Figure 1.11 Retail stores, an example of an off-site processing topology [50].

amount of data from the customers before ordering new goods in the store.
Also, the store can use pre-existing systems for this task. The data logged
in a day can be sent to a cloud service, such as an item recommendation
system, which can perform data analysis on the customer data to figure out
the customer preferences and the kinds of items that customers are potentially
looking for. The cloud service can also be made available to be used by
other retail stores. This way, every store does not need to deploy its item
recommendation system and can use pre-existing ones, thus saving a large
amount of money and time. An illustration of an off-site processing system is
shown in Figure 1.11 [17].

1.4.4 Communication Technologies

Communication forms the basis for devices to transfer and share their data in
IoT. To this end, there must be a connection between the devices and certain
protocols have to be followed. The open systems interconnection (OSI) model
is a popular model on which many different communication systems are
based [44]. The OSI model defined seven layers [45], with each layer having
its unique functionality and also utilizing the functions of the layer as listed
in Figure 1.12.
The seven layers combined provide the full communication service from
one device to another. The physical and the datalink layers are responsible
 1.4 Technical Details of IoT 19

Figure 1.12 Layers of the OSI model.

for data transmission from one device to another through a single connection;
for example, from the sensor to a microprocessor or your phone to the nearby
Wi-Fi router. The network layer is responsible for data transmission through
the network, i.e., from the sensor node to your computer. The transport layer
is responsible for data transmission between processes, i.e., from the sensor
node to the program that is using the sensor node. The above layers deal with
data presentation and processing. These layers are generally present in every
communication system in one way or another.
The bottom four layers of the OSI layer are of particular importance.
Because of the wide variety of devices that are used, there is also a large
variation in the technology used in each of the layers. The physical and the
data link layers consist of various wired and wireless communication pro-
tocols used to connect two devices through a single link. Generally, sensors
are connected to microprocessors through a wired channel. The sensor can
communicate with the processor using various protocols such as I2C, SPI,
USART, and so on. The processors, being more powerful devices, are then
tasked with sending the data over the network which can be done through a
wired channel such as Ethernet or a wireless channel. IEEE 802.15.4, Zigbee,
LoRa, Bluetooth, Wi-Fi, and RFID are some of the popular technologies used
for wireless communication [17, 46].
20 Introduction to IoT

Table 1.1 Different types of the protocol used in IoT systems.

Functionality Example protocols
Infrastructure They handle the routing of data so IPv4, IPv6, QUIC,
protocols that it reaches the correct nanoIP, etc.
destination. These are very
dependent on the type of network
used, e.g., Internet.
Discovery protocols They are used to discover whether mDNS, UPnP, etc.
a certain device or a certain
service is available in a network
or not. Besides this, they are also
used when a new device is
connected to a network to
establish proper communication
with the device.
Data protocols They deal with data transfer, MQTT, REST,
storage, and distribution. Data WebSockets, etc.
transmission can occur between
one source and multiple
destinations. Thus, these
protocols have to be adept at
handling large amounts of data
transfer.
Identification They are used to identify devices EPC, uCode, URI, etc.
protocols in the network uniquely.

The other layers then deal with the transmission of data from the source
endpoint to the destination endpoint. Besides this, they also provide other
important functions such as device discovery, identification, and manage-
ment. Thus, we can classify the various protocols used in terms of their usage
into the groups as listed in Table 1.1 [17].

1.5 Recent Developments

With the advancements in extensive digital technology and progressive opti-
mal algorithms, IoT is emerging in almost every sector as possible and with its
capacity expanded to virtual memories as well as the scalability factor. IoT is
now on the verge of taking over the services by Internet and electronic gadgets
to a newer level. For a smart future, IoT is the key source impacting the larger
fields like big data and smart cities. Furthermore, due to its quick adaptability,
increasing capability, better security, larger scalability, and reusability, it is
safe to say that IoT is going to have a touch on everyone’s day-to-day life
 1.5 Recent Developments 21

activities. The whole idea of IoT stands on data entry, its manipulation, and
necessary actions, but with the wide application area and increasing data
levels including some confidential information too, the future of IoT leans
toward complicated security and privacy issues. This concern is captivating
the concentration from both industry and academia for performing research
on IoT’s security and privacy. Along with the security issues, the future of
IoT also concerns the management issues like monitoring, updates, diagnos-
tics, crash analysis and reporting, processor speed, power consumption, OS,
standards and units, platforms, and so on. Hence, the recent development in
IoT is concerned with these factors, especially prioritizing the security and
privacy issues.
Since IoT connects multiple devices into a common network, the safest
and secure point-to-point connection is preferred. Various existing standards
have been used in attempts to solve the interconnection issues between the
devices in the network. Also, various alliances and standards such as ITU-
T, IEEE P2413, Allen Alliance, Open Connectivity Foundation, and IPSO
Alliances have suggested a framework for IoT and connectivity of local IoT
devices. OneM2M is one such proposed standard that aims to provide inte-
gration and interconnection between various smart IoT devices. The oneM2M
project was founded by the Standards Developing Organizations (SDOs) of
many countries such as Korea, USA, Japan, and so on [18].
The integrated circuits designed for IoT systems are focused on specific
purposes and thus mainly utilize a robust and powerful system on chips
(SoCs) for their operation. With security always being one of the most
important aspects of an IoT-based system, IoT systems need to have security
measures to protect the user data. Many different approaches for security in
IoT have been proposed in recent years using new technologies like artificial
intelligence (AI), machine learning, big data, and blockchain [55]. Some
approaches using blockchain to improve security and privacy have been
discussed in [53] and [54]. Similarly, a discussion on other approaches using
machine learning can be found in [56] and [57]. Utilizing the SoC nature
of IoT devices, a physically unclonable function (PUF) based cryptographic
security solution for IoT SoCs has been proposed which deploys a mechanism
using PUF and symmetric cryptography to provide a secure means for data
communication [47].
Besides the advancement that is dedicated solely to IoT, advancements in
other fields are also showing their impacts on bettering the IoT. The upcoming
generation of mobile technology is predicted to highly benefit the IoT field.
This is because of the invention of 5G technology though its specifications are
22 Introduction to IoT

yet to be finalized. With the greater speed of connectivity and lower latency,
it will be aiding in speeding up IoT devices and help in better remote access.
Another major emerging field is that of artificial intelligence (AI). Integration
of AI with IoT can be a boon to the world of innovation. Using the data
obtained from smart IoT devices in order to train themselves as smarter as a
human brain can be a game-changer to the world of automation. Self-driving
cars are a great example of this combination. Cloud computing is another
major area that is immensely alternating with every other data storage and
manipulation system available. The field of robotics has also seen the use of
IoT and other cloud technologies. This is referred to as cloud robotics that
aims to utilize IoT and other cloud services [48]. It allows robots to share
information and utilize many powerful resources that are distributed in the
cloud.

1.6 Challenges
Like many modern technologies, IoT also faces many challenges that prevent
it from widespread adoption. IoT faces challenges at both the technological
levels and at the ethical level. From a technological perspective, IoT faces
challenges such as interoperability, scalability, security and privacy concerns,
data concerns, etc.
With a large number of IoT systems being deployed, there is a higher
chance of interoperability issues, i.e., integration of different systems into a
single one. Many systems are proprietary and thus incompatible with each
other. Making them compatible would require compatibility at all levels such
as in hardware, protocols, etc. [19]. In order to overcome these issues, stan-
dards need to be developed which cover the technical requirements including
hardware and software requirements for all the devices for better interoper-
ability. Various efforts have been made by standards organizations like ITU,
ISO, IEEE, IEC, etc., to build standards for fundamental IoT technologies like
RFID, WSNs, etc. [20]. Scalability, on the other hand, brings along a lot of
other major issues. Most IoT systems have to be designed and used on a large
scale. This brings about the concerns of security and privacy among many
others. IoT is an amalgamation of all kinds of technologies that raise security
concerns at all levels. Each part of an IoT ecosystem must be secure against
vulnerabilities and be robust against outer disturbances and interference. A lot
of devices used in IoT have limited computational capabilities, which makes
it very difficult to implement robust security measures. This can open up
the entire ecosystem to vulnerabilities. One famous example of IoT security
 References 23

being compromised is the Mirai botnet [21]. Mirai is a malware that targeted
IoT devices to create a very large botnet that was used for many distributed
denial of service (DDoS) attacks in 2016. It did so by guessing the credentials
of the devices. Many users did not change the credentials of the device
from the factory defaults due to which it was very easy for the malware to
infect hundreds of thousands of devices. An army of such devices was used
to conduct DDoS attacks on companies like Krebs on Security, OVH, and
Dyn. This alone highlights the disastrous risks of having vulnerabilities on
IoT-based systems.
In addition to this, privacy concerns also arise when we have devices that
can monitor our daily activities and transfer that data through a large network.
Any device that is connected to the Internet is susceptible to many different
forms of attack that involve stealing one’s identity, accessing other people’s
data, and so on. Websites like Shodan (shodan.io) allow people all over the
Internet to legitimately access other people’s devices like IP cameras [22]. In
this age, where privacy is of utmost importance, IoT devices have to maintain
the highest level of privacy possible.

1.7 Conclusion
In this chapter, a brief overview of the entire IoT ecosystem was presented
along with its history, applications, enabling technologies, and recent devel-
opments and challenges. IoT is a technology with vast applications in many
fields that affect our day-to-day lives. Improvements brought in the fields of
healthcare, industries, energy management, etc., by the sensing, monitoring,
and communicating capabilities of IoT have greatly upgraded our lifestyle
and it will keep on enhancing our lifestyle in the future as the IoT technology
further evolves. However, IoT is still not a technology that is without faults.
Many technical and ethical issues arising from the use of IoT have hindered
it from its widespread adoption. However, IoT remains one of the most
influential pieces of technology in the present world, and, thus, with ongoing
research and developments, these limitations will hopefully be overcome in
the near future. Therefore, IoT is a field with nearly limitless potential in the
foreseeable future.

References
[1] K. Ashton, “That ‘Internet of Things’ thing,” RFID Journal, 22-Jun-
2009. [Online]. Available: https://www.rfidjournal.com/that-internet-of
-things-thing. [Accessed: 07-Sep-2021].
24 Introduction to IoT

[2] “The ‘Only’ Coke Machine on the Internet,” Carnegie Mellon School of
Computer Science. [Online]. Available: https://www.cs.cmu.edu/~{}c
oke/history_long.txt. [Accessed: 07-Sep-2021].
[3] E. Dave, “The Internet of things: How the next evolution of the Internet
is changing everything,” 2011. [Online]. Available: https://www.cisco.
com/c/dam/en_us/about/ac79/docs/innov/IoT_IBSG_0411FINAL.pdf
[Accessed: 23-Oct-2021].
[4] C. Floerkemeier, E. Fleisch, M. Langheinrich, F. Mattern, and S.E.
Sarma, in Proceedings of the Internet of Things First International
Conference, IoT 2008, Zurich, Switzerland, March 26–28, 2008. Berlin,
Heidelberg: Springer-Verlag Berlin Heidelberg, 2008.
[5] E. Irmak and M. Bozdal, “Internet of Things (IoT): The most up-to-date
challenges, architectures, emerging trends and potential opportunities,”
International Journal of Computer Applications, vol. 179, pp. 20–27,
May 2018.
[6] N. Sharma, M. Shamkuwar, and I. Singh, “The history, present and
future with IoT,” Intelligent Systems Reference Library, pp. 27–51,
2018.
[7] M. Weiser, “The computer for the 21st century,” ACM Sigmobile Mobile
Computing and Communications Review, vol. 3, no. 3, pp. 3–11, 1999.
[8] R. De Michele and M. Furini, “IoT healthcare,” in Proceedings of the
5th EAI International Conference on Smart Objects and Technologies
for Social Good, 2019.
[9] K. Saleem, I.S. Bajwa, N. Sarwar, W. Anwar, and A. Ashraf, “IoT
healthcare: Design of smart and cost-effective sleep quality monitoring
system,” Journal of Sensors, vol. 2020, pp. 1–17, 2020.
[10] S. Allen, “2021 global health care sector outlook,” Deloitte, 25-Jun-
2021. [Online]. Available: https://www2.deloitte.com/global/en/pa
ges/life-sciences-and-healthcare/articles/global-health-care-sector-outl
ook.html. [Accessed: 07-Sep-2021].
[11] M. Ersue, D. Romascanu, J. Schönwälder, and A. Sehgal, “Management
of networks with constrained devices: Use cases,” no. 7548. RFC Editor,
May 2015.
[12] P.M. Kumar and U. Devi Gandhi, “A three-tier Internet of Things
architecture with machine learning algorithm for early detection of heart
diseases,” Computers & Electrical Engineering, vol. 65, pp. 222–235,
2018.
 References 25

[13] R.N. Kirtana and Y.V. Lokeswari, “An IoT based remote HRV monitor-
ing system for hypertensive patients,” in Proceedings of 2017 Interna-
tional Conference on Computer, Communication and Signal Processing
(ICCCSP), 2017, pp. 1–6, doi: 10.1109/ICCCSP.2017.7944086.
[14] W. Li, Y. Chai, F. Khan, S.R. Jan, S. Verma, V.G. Menon, Kavita, and
X. Li, “A comprehensive survey on machine learning-based big data
analytics for IoT-enabled smart healthcare system,” Mobile Networks
and Applications, vol. 26, no. 1, pp. 234–252, 2021.
[15] T. de Vass, H. Shee, and S. Miah, “IoT in supply chain management:
Opportunities and challenges for businesses in early Industry 4.0 con-
text,” Operations and Supply Chain Management: An International
Journal, pp. 148–161, 2021.
[16] T.M. Brown-Brandl, “1 using RFID in animal management and more,”
Journal of Animal Science, vol. 97, no. Supplement_2, pp. 1–2, 2019.
[17] S. Misra, A. Mukherjee, and A. Roy, Introduction to IoT. Cambridge:
Cambridge University Press, 2021.
[18] H. Park, H. Kim, H. Joo, and J.S. Song, “Recent advancements in
the Internet-of-Things related standards: A onem2m perspective,” ICT
Express, vol. 2, no. 3, pp. 126–129, 2016.
[19] M. Noura, M. Atiquzzaman, and M. Gaedke, “Interoperability in inter-
net of things: Taxonomies and open challenges,” Mobile Networks and
Applications, vol. 24, no. 3, pp. 796–809, 2018.
[20] M. Aly, F. Khomh, Y.-G. Gueheneuc, H. Washizaki, and S. Yacout, “Is
fragmentation a threat to the success of the internet of things?,” IEEE
Internet of Things Journal, vol. 6, no. 1, pp. 472–487, 2019.
[21] M. Antonakakis et al., “Understanding the Mirai Botnet,” in Proceed-
ings of the 26th USENIX Conference on Security Symposium, 2017,
pp. 1093–1110.
[22] H. Lin and N. Bergmann, “IoT privacy and security challenges for smart
home environments,” Information, vol. 7, no. 3, pp. 44, 2016.
[23] Alexa Smart Home-Learn About Home Automation. [Online]. Available:
https://www.amazon.com/alexa-smart-home [Accessed: 23-Oct-2021].
[24] F. Wang, L. Hu, J. Hu, J. Zhou, and K. Zhao, “Recent advances in the
internet of things: Multiple perspectives,” IETE Technical Review, 2016,
doi: 10.1080/02564602.2016.1155419.
[25] Available: https://pixabay.com/illustrations/iot-internet-of-things-net
work-3337536/ [Accessed: 23-Oct-2021].
26 Introduction to IoT

[26] A. Sarac, N. Absi, and S. Dauzère-Pérès, “A literature review on the

impact of RFID technologies on supply chain management,” Interna-
tional Journal of Production Economics, vol. 128, no. 1, pp. 77–95,
2010.
[27] M. Attaran, “Digital technology enablers and their implications for sup-
ply chain management,” Supply Chain Forum: An International Journal,
vol. 21, no. 3, pp. 158–172, 2020.
[28] M.T. Okano, “IoT and industry 4.0: The industrial new revolution,” in
Proceedings of International Conference on Management and Informa-
tion Systems, vol. 25, Sep. 2017, pp. 26.
[29] L.D. Xu, W. He, and S. Li, “Internet of things in industries: A sur-
vey,” IEEE Transactions on Industrial Informatics, vol. 10, no. 4,
pp. 2233–2243, Nov. 2014, doi: 10.1109/TII.2014.2300753.
[30] V. Marinakis and H. Doukas, “An advanced IoT-based system for intel-
ligent energy management in buildings,” Sensors, vol. 18, no. 2, pp. 610,
Feb. 2018. [Online]. Available: http://dx.doi.org/10.3390/s18020610
[31] J. Yun, S.-S. Lee, I.-Y. Ahn, M.-H. Song, and M.-W. Ryu, “Mon-
itoring and control of energy consumption using smart sockets and
smartphones,” Communications in Computer and Information Science,
pp. 284–290, 2012.
[32] K. Chooruang and K. Meekul, “Design of an IoT energy monitoring
system,” in Proceedings of 2018 16th International Conference on ICT
and Knowledge Engineering (ICT&KE), 2018, pp. 1–4, doi: 10.1109/IC
TKE.2018.8612412.
[33] R.R. Hernandez, S.B. Easter, M.L. Murphy-Mariscal, F.T. Maestre,
M. Tavassoli, E.B. Allen, C.W. Barrows, J. Belnap, R. Ochoa-Hueso,
S. Ravi, and M.F. Allen, “Environmental impacts of utility-scale
solar energy,” Renewable and Sustainable Energy Reviews, vol. 29,
pp. 766–779, 2014.
[34] A.S. Spanias, “Solar energy management as an internet of things (IoT)
application,” in Proceedings of 2017 8th International Conference on
Information, Intelligence, Systems & Applications (IISA), 2017.
[35] S. Rao et al., “An 18 kW solar array research facility for fault detection
experiments,” in Proceedings of 2016 18th Mediterranean Electrotech-
nical Conference (MELECON), 2016, pp. 1-5, doi: 10.1109/MELCON
.2016.7495369.
[36] M. Sharma, M.K. Singla, P. Nijhawan, S. Ganguli, and S.S.
Rajest, “An application of IoT to develop concept of smart remote
 References 27

monitoring system,” Business Intelligence for Enterprise Internet of

Things, pp. 233–239, 2020.
[37] D.D. Prasanna Rani, D. Suresh, P. Rao Kapula, C.H. Mohammad
Akram, N. Hemalatha, and P. Kumar Soni, “IoT based smart solar
energy monitoring systems,” Materials Today: Proceedings, 2021.
[38] M. Ali and M.K. Paracha, “An IoT based approach for monitoring solar
power consumption with Adafruit cloud,” International Journal of Engi-
neering Applied Sciences and Technology, vol. 04, no. 09, pp. 335–341,
2020.
[39] P. Siano, “Demand response and smart grids—A survey,” Renewable
and Sustainable Energy Reviews, vol. 30, pp. 461–478, 2014.
[40] V.C. Gungor, B. Lu, and G.P. Hancke, “Opportunities and challenges
of wireless sensor networks in smart grid,” IEEE Transactions on
Industrial Electronics, vol. 57, no. 10, pp. 3557-3564, Oct. 2010, doi:
10.1109/TIE.2009.2039455.
[41] M. Anjanappa, K. Datta, T. Song, R. Angara, and S. Li, “Introduction to
sensors and actuation,” in The Mechatronics Handbook, 2nd ed., R. H.
Bishop, Ed. Hoboken: CRC Press, 2008.
[42] H. Janocha, Ed., Actuators, 1st ed. Berlin: Springer, 2004.
[43] P.K.D. Pramanik and P. Choudhury, “IoT data processing: The different
archetypes and their security and privacy assessment,” in Internet of
Things Security: Fundamentals, Techniques and Applications, S. K.
Shandilya, S. A. Chun, S. Shandilya, and E. Weippl, Eds. Gistrup: River
Publishers, 2018, pp. 37–54.
[44] A. Rayes and S. Salam, “The Internet in IoT—OSI, TCP/IP, IPv4, IPv6
and Internet routing,” in Internet of Things from Hype to Reality. Berlin:
Springer, 2017, pp. 35–56.
[45] H. Zimmermann, “OSI reference model—The ISO model of architec-
ture for open systems interconnection,” IEEE Transactions on Commu-
nications, vol. 28, no. 4, pp. 425-432, Apr. 1980, doi: 10.1109/TCOM.1
980.1094702.
[46] S. Al-Sarawi, M. Anbar, K. Alieyan, and M. Alzubaidi, “Internet of
Things (IoT) communication protocols: Review,” in Proceedings of
2017 8th International Conference on Information Technology (ICIT),
2017, pp. 685–690, doi: 10.1109/ICITECH.2017.8079928.
[47] A. Balan, T. Balan, M. Cirstea, and F. Sandu, “A PUF-based crypto-
graphic security solution for IoT systems on chip,” EURASIP Journal
on Wireless Communications and Networking, vol. 2020, no. 1, 2020.
28 Introduction to IoT

[48] G. Hu, W.P. Tay, and Y. Wen, “Cloud robotics: architecture, challenges
and applications,” IEEE Network, vol. 26, no. 3, pp. 21-28, May-Jun.
2012, doi: 10.1109/MNET.2012.6201212.
[49] Available: https://www.freesvg.org [Accessed: 23-Oct-2021].
[50] D.C Motor image by Dcaldero8983, CC BY-SA 3.0. Available: https:
//creativecommons.org/licenses/by-sa/3.0, via Wikimedia Commons.
[51] Images by SparkFun, CC BY 2.0. Available: https://creativecommons.
org/licenses/by/2.0
[52] PIR image by Nowforever CC BY-SA 4.0. Available: https://creativeco
mmons.org/licenses/by-sa/4.0 via Wikimedia Commons.
[53] O. Alfandi, S. Khanji, L. Ahmad, and A. Khattak, “A survey on boosting
IoT security and privacy through blockchain,” Cluster Computing, vol.
24, no. 1, pp. 37–55, 2020.
[54] S.N. Mohanty, K.C. Ramya, S.S. Rani, D. Gupta, K. Shankar, S
K. Lakshmanaprabu, and A. Khanna, “An efficient lightweight inte-
grated blockchain (ELIB) model for IoT security and privacy,” Future
Generation Computer Systems, vol. 102, pp. 1027–1037, 2020.
[55] B.K. Mohanta, D. Jena, U. Satapathy, and S. Patnaik, “Survey on IoT
security: Challenges and solution using machine learning, artificial intel-
ligence and blockchain technology,” Internet of Things, vol. 11, 2020,
Art. no. 100227.
[56] R. Ahmad and I. Alsmadi, “Machine learning approaches to IoT secu-
rity: A systematic literature review,” Internet of Things, vol. 14, 2021,
Art. no. 100365.
[57] M.A. Amanullah, R.A. Habeeb, F.H. Nasaruddin, A. Gani, E. Ahmed,
A.S. Nainar, N.M. Akim, and M. Imran, “Deep learning and big data
technologies for IoT security,” Computer Communications, vol. 151,
pp. 495–517, 2020.
[58] W. Ahmed et al., “Machine learning based energy management model
for smart grid and renewable energy districts,” IEEE Access, vol. 8,
pp. 185059–185078, 2020, doi: 10.1109/ACCESS.2020.3029943.
[59] P. Sivaraman and C. Sharmeela, IoT-Based Battery Management Sys-
tem for Hybrid Electric Vehicle, in P. Sanjeevikumar (eds), Artificial
Intelligent Techniques for Electric and Hybrid Electric Vehicles, John
Wiley-Scrivener Publishing, 2020.
 2
IoT and its Requirements for Renewable
Energy Resources

D. Gunapriya1 , R. Sivakumar2 , K. Sabareeshwaran3 ,

and C. Sharmeela4
1 SriKrishna College of Engineering and Technology, India
2 Akshaya College of Engineering and Technology, India
3 Karpagam Institute of Technology, India
4 Anna University, India

E-mail: [email protected]; [email protected];

[email protected]

Abstract
The worldwide energy production is majorly based on fossil fuels, and it cov-
ers 84% of world key energy consumption as on 2020 statistics. To support
global energy demand, both government and private sectors have started to
focus the problems on fossil fuel depletion and the impact of Greenhouse
gas emissions in climatic change. To reduce the emission of Greenhouse
gas and to obtain energy transition of sustainable range, it is necessary to
integrate renewable energy in the power grid as per recent studies. Though
the renewable energy sector kept its phenomenal growth in the current sce-
nario and the prediction of its supply to fulfill two-thirds of global energy
demand in 2050, there are challenges in power generation, transmission, and
distribution where the economic feasibility plays a role. Modern technology,
such as the Internet of Things (IoT), plays an important role in overcoming
these problems and optimizing renewable energy supplies. Understanding the
use of IoT technology fosters trust in this sector to give a better solution
for renewable energy management such as generation, transmission, and
distribution, as well as effective energy sharing to the grid. This chapter
discusses IoT technologies and their integration in the renewable energy
sector to increase their efficiency. Furthermore, information transmission and

29
30 IoT and its Requirements for Renewable Energy Resources

storage for data analytics to provide optimum energy supply, and issues in
IoT implementation with solutions to security and privacy maintenance are
highlighted.

2.1 Introduction
2.1.1 IoT and its Necessity
The modern equipment offers effective communication among themselves
besides carrying out their functions, which supports effective system opera-
tion. In the fourth industrial revolution, robotics systems, wireless communi-
cation, the Internet of Things (IoT), cloud computing, and so on dominated
the functions [1]. Now the issue arises as to “Why should the equipment
in the system, or system to system, interact with each other?” The solution
is that by using effective communication, superfluous functions in the sys-
tem may be turned off to conserve energy and optimize system operations.
Furthermore, the gathering of operational information (equipment data and
environmental data) in real time via communication is utilized for data ana-
lytics to predict/measure equipment performance to avoid unexpected system
failure [2]. In addition, data analysis is utilized to make decisions on system
enhancements and adjustments depending on future demand. For more than
a decade, machines interacted using machine-to-machine (M2M) commu-
nication, a one-to-one communication method used to link one machine to
another [3]. However, when sensors and actuators are utilized with devices
for sending and receiving data to monitor the devices, transmission control
protocol/Internet protocol (TCP/IP) is used for large information transmission
between the communicating devices. Kevin Ashton invented the term “Inter-
net of Things” (IoT) in 1999, and his concept of IoT technology was based on
radio frequency identification (RFID) device communication, which differs
greatly from today’s IP-based connectivity. After 2011, the IoT boom was
happening with the network layer of IPv4 and IPv6, which became essential
to IoT with the aid of developing technologies. In this rapid development,
there are already 6.4 billion linked devices and sensors in IoT, with 5.5 million
devices connecting to IoT on a daily basis [4].

2.1.2 Challenges in RES

The current fluctuating utility power demand in the distribution system world-
wide is managed by incorporating the variable renewable energy resources
(VRES) into the distribution system with the help of decarbonizing effect.
 2.1 Introduction 31

Table 2.1 RES challenges as per the root cause analysis.

Factors Cause
Stability Stability violations and re-dispatch allocation
Steadiness The mismatch between source and load
Eminence Safety hazards and aging of equipment

VRES, such as solar power grid systems and wind power grid systems,
plays a significant role in accomplishing the power sector decarbonization.
However, they are not the same as traditional power production systems. By
integrating the VRES in the distribution system, there will be possibilities for
the formation of difficulties and challenges in the operation of the distribution
network (DN) [5]. Furthermore, in the context of the growing role of VRES
in meeting current power demand, it is, therefore, imperative to address the
new problems that arise in power systems. If these problems and issues are
not addressed, they may endanger the reliability of the power supply system
as well as the decarbonization targets [6].
Table 2.1 lists RES challenges as per the root cause analysis and offers an
overview of the signs of growing VRE penetration discovered by a literature
study. The symptoms may be classified that correspond to the power system’s
essential performance criteria. The categories are briefly described in the
sections that follow. The end user’s primary performance requirement is
adequate power quality. The power quality factor includes the conditions
for uninterrupted power sources, consistent voltage and current parameters,
and safe conditions in the event of disruptions. The uniqueness and asyn-
chronous facts of VRES are two basic features that are primarily responsible
for power quality issues. The current factor is concerned with the effective
DN of electricity. In comparison to the other categories, the root causes
of problems in the flow category are numerous. The greatest proportion of
flow problems is caused by VRE unpredictability, modularity, and location
restrictions. The stability factor is concerned with power rate modulations and
voltage fluctuation, together with restart and recovery of the system after the
power distribution interruptions. The problems that deal with the balancing
of the power supply with short and long terms in the distributions systems
as well as managing the demand are the concerns of the eminence factor in
the VRES power generation system. This includes the management of the
power systems’ upgradation competency and maintaining low power gen-
eration levels among the systems. But VRES power generation fluctuations
and unpredictability create balancing problems. To summarize, the analysis
of causes in VRES stipulates a reliable foundation for the classification of
problems in power systems with its growth and adaptations.
32 IoT and its Requirements for Renewable Energy Resources

2.1.3 Integration of IoT in RES and Benefits

Due to the depletion of fossil fuels and the growth in environmental contami-
nation, contemporary technological development and research have mostly
focused on the adoption of green energy production sources and the best
use of existing energy resources [7]. Smart grid and microgrid technology
innovation is highly useful in achieving these aims through power suppliers.
The purpose of smart grids is to effectively deal with energy distribution
management which involves the generation, transmission, and distribution
operations. To effectively manufacture, use, and distribute energy, all par-
ticipants, such as distributors, customers, and producers, are authorized to
engage in two-way communication. As a result, energy efficiency may be
achieved by effectively using the available energy sources. Since smart and
microgrid technologies are becoming a preferred method to power generation
due to the focus on green energy production globally, there is also an increase
in demand for cleaner energy manufacturers and suppliers.
The recent research works depict that the recent developments of sensors,
data storage systems, and data analytical tools are majorly used to achieve the
balancing of power supply against the power requirement effectively to the
consumers who are connected to the smart grid system. The advancements of
smart devices and their connectivity through the Internet bring the meaning of
the Industrial Internet of Things (IIoT) and the purpose of use in power grid
system. Smart devices aid in increasing operational efficiency, optimizing
company operations, and safeguarding the system. Power generation plan-
ning is essential to optimize the usage of power and its costs. The energy
planning must include the numerous energy resources available and then
optimize depending on feedback from many associated elements. The energy
consumption needs are conveyed to the framework via the IoT, which aids in
communication with various IoT-based smart devices and provides a response
to smart grids to achieve the optimal decision.

2.2 Industrial IoT

In today’s world, the rapid development of IoT is radically transforming
people’s lives. Currently, IoT manufacturers are concentrating their efforts
on offering IoT solutions to specific industries rather than broad applications.
There is also a significant opportunity in the rising demand for IoT devices
in the industrial sector to deliver industry-specific solutions. The IIoT is
distinguished by the interconnection of machines, people, and computers to
enable smart industrial processes that use advanced data analytics to alter
 2.2 Industrial IoT 33

business outcomes. The IIoT is an extension of the IoT that has evolved as a
broad notion of applying the IoT to the industrial sector [8].
However, because of the integration requirement, it confronts specific
problems that distinguish it from other IoT systems and services. Although
the fundamental concepts of IIoT and IoT are similar, i.e., smart device
connections that enable remote operation of the equipment, receive and store
data, analyze the data, and supervise and control the variables that distinguish
the IIoT as the subcategory of IoT which develops the strict requirements for
uninterrupted services and security with efficient operational technology used
in the industrial fields. The industrial sector’s particular qualities - technology
and requirements – result in specialized solutions and services to support the
industrial division’s concentration on a customized IoT concept. As a result,
the industrial sector has taken a keen interest in the creation of specialized
ideas ranging from policy to technology services.

2.2.1 Architecture of IoT

The first stage in building an IoT system for a given application is to
select acceptable and dependable IoT devices, communication protocols, data
storage devices, and data analytic methodologies for application performance
prediction [9]. The advancement and distribution of IIoT systems and services
require the creation of designs that allow for competent and active operations,
including communication, predicted endway services, and the major investors
tangled for devices, cyber–physical systems, communication networks (CNs),
service providers, and professional promoters. The International Telecommu-
nication Union (ITU) addressed the issues in the development of standards
and architecture for IoT by publishing the ITU-T Y.2060 recommendation
in 2012, which presents a reference architecture for IoT in general as well as
applications that come under the IIoT umbrella, such as smart grid system and
intelligent transportation systems, e-wellbeing, and so on. The ITU initiative
has broadened the communications vision to encompass communication of
“everything” as well as communicate ideas of “any time” and “any place.”
It is vital to highlight that it covers all anticipated applications, including
industrial ones like smart grids and smart transportation systems, among
others. ITU considers physical and virtual items that are recognizable and
capable of connecting to CNs to be “things.” Importantly, because com-
munication is such a key component of the IoT concept, physical items
that fall under the “devices” category must be connected to networks for
any analog data to be converted to digital and communicated via Internet
connections. Devices can be simple data transmitting devices that transfer
34 IoT and its Requirements for Renewable Energy Resources

Figure 2.1 Architecture of IoT.

and store data, data acquisition devices that interact with the equipment
via encoders and decoders, sensors and actuators or machines, end service
devices, and consumer electronic devices that have embedded processing and
communication resources.
The connectivity of devices is the fundamental idea of IoT architecture
based on the ITU reference model. Figure 2.1 shows the architecture of
IoT wherein the model analyzes three communication techniques based on
the usage of gateways (G), local network (LN), and the CN. The smart
devices can connect openly, over LNs, and/or over the CN, without the
need for gateways, or they can communicate via gateways. The IIoT has
the key features of interconnectivity, scalability, diversity, service facility, the
dynamic information of the devices, and connectivity. Development becomes
a substantial factor that needs to be included at all levels of IIoT as the number
of connected devices increases dramatically. The development issue is present
not only to communication points and the number of smart devices but also
to the size of collected and transmitted data including its administration
in terms of huge storage and analyzing. The dynamic performances of the
devices, which turn on and off automatically or join and detach from CNs,
will complicate and demand more from the landscape.

2.2.2 IoT Components

Figure 2.2 illustrates the IIoT-ITU reference model, which was presented to
satisfy the aforementioned needs. It is a normal layer construction model
 2.2 Industrial IoT 35

Figure 2.2 IIoT-ITU reference model.

which consists of four tiered layers, particularly device layer, network layer,
software and service assistance layer, and lastly application layer, and two
vertically integrated layers which are crosscutting the four-tiered layers,
defining managing as well as security capabilities and attributes to all hierar-
chical levels. The first layer that presents the device information, probably the
bottom in the hierarchy, consists of the performance of equipment and CNs.
Thinking about the primary interest of ITU in marketing communications,
the level describes communication-centered performance for the devices:
Importantly, the unit level includes protocol transformation because products
might put into action different protocols and, therefore, requires process
transformation for interoperability.
The network layer provides a summary of related protocols and the
conversion of the data from the devices to network-level protocols. This
layer contains information about how the network works, including how
the data in the model is transported. For networking, they may include
control options for accounting, authorization, authentication, mobility, and
network connectivity, while for transportation, they foresee user traffic trans-
portation along with the transport of management and management info for
(I)IoT service as well as applications. The service assistance and application
assistance layers include both service/application-specific and generic func-
tionality (capabilities) which allows (I)IoT applications as well as services.
Looking at the distributed character of (I)IoT services as well as programs,
there exists generic efficiency, like information processing and storage and
36 IoT and its Requirements for Renewable Energy Resources

Figure 2.3 Flow of electrical energy in conventional electric system.

particular functionality per service and application since emerging solutions

have requirements that are different; for instance, smart grid operation puts
various privacy needs than a smart toll management process for transporta-
tion services. Lastly, the application level, probably the highest hierarchical
level, consists of the (I)IoT applications as well as services. The managing
vertical and crosscutting level involves both generic and program domain-
specific functionalities, account management, security, fault, performance,
and resource.

2.3 RES and IoT

For a long time, the current electrical grid design has provided excellent
service to anyone who has access to energy. Even though the unit is evolving,
it is critical to comprehend the devices that have brought us to this point. The
facility of electrical energy typically alters on three important segments: dis-
tribution, transmission, and generation. Figure 2.3 shows the ?ow of electrical
energy in the conventional electric system.
Approximately 5% of electrical energy is lost as line losses in long-
distance transmission lines. The existing power grid was meant to satisfy the
twentieth-century demand.

2.3.1 IoT Controls for RES

The IoT is used to transmit the energy usage requirements to the framework,
which allows it to interface with various IoT-based devices and send feedback
 2.3 RES and IoT 37

to microgrids to carry out the optimal option. Four goals could be achieved
using the projected look as follows.
• Initially, it creates a self-contained IoT for transmitting and receiving
data from end-users.
• As a result of the foregoing, it regulates the power usage of products
based on the goals and constraints established by customers.
• Third, it provides buyers with feedback on their power consumption
patterns to help them save money and energy.
• Finally, it has inexhaustible power sources in the network.
It provides applications like remote monitoring, prognostic maintenance, and
cybersecurity and enhances the professional process, laborer well-being, and
then advanced distributed management.
To enhance the consumption of distributed power resources (DERs), the
grid must become more intelligent to improve traditional grid infrastructure
while also promising IoT coordination. Transmission operators can account
for the distribution and generation elements of the grid’s resources and
assets. Operators can regulate crucial aspects of the grid while also including
distribution. A decentralized strategy brings the generation near to the needed
load, reduces transmission losses as well as vulnerabilities, and expands
the overall dependability, versatility, efficiency, and sturdiness of the grid.
Communication is closer and bidirectional to near real-time, empowering
customers to be apt to oversee expenses and loads. Additionally, power fees
may be progressively unique. Besides, smart building products empower
intelligent equipment through the web. The smart grid bolsters the use of
even more nuanced and effective demand management plans as well as the
delivery of progressively educated methods by the customers. Intelligent
digital meters, also referred to as “smart meters,” are basic to the smart grid.
To enhance the utilization of unlimited energy resources, these grids improve
customer involvement infrastructure as well as assurance to sign up with IoT
at the community level. On a town level, a reasonable microgrid allows for
a bidirectional stream of communication and electricity between electricity
suppliers and customers. IIoT plays a critical role in addressing the issue
of non-accessibility of specific renewable energy solutions by monitoring
power use, energy generation, and integration, particularly for intelligent
microgrids. Smart microgrids improve nearby dependability by utilizing an
explicit dependability enhancement program that includes surplus distribu-
tion, clever switches, energy generation, energy storage, automation, and
other smart technologies. With the advances and switches in the electricity
38 IoT and its Requirements for Renewable Energy Resources

sector, utilities, as well as the power sector, constantly change for the long
term. Nevertheless, the grid is going to keep on being essential to the
electricity industry.

2.3.2 Challenges in IoT Implementation

The basic functions of smart grid (SG) advancement consist of the devel-
opment of power grid systems, standards and communication, analytical
intelligence, economic and environmental factors, and testbed. Several of
the difficulties for sustainable growth include electricity issues, manpower
capabilities, political policies, and technology [10]. The computational intel-
ligence attribute consists of innovative analytical equipment that will enhance
the majority power network by using heuristics, evolution programming,
choice support equipment, and adaptive optimization methods, which are
promising resources for the layout and computation demand of SG. Renew-
able energy development makes use of variability, and its consumption is
also commercially and economically feasible to alleviate demand anxiety
while also improving reliability, minimizing losses, and reducing the carbon
footprint caused by winter and gas-based energy solutions. The improve-
ment of automation, interaction, and requirements is needed to make sure
that rapid decision-making promotes effectiveness and dependable operation.
These options are determined by the energy and the customers. Human
development, homegrown know-how, the work of scholars and government,
interconnection challenges, engineering challenges, technologies in plant
layout, maintenance, and operation, and improvement of SG with notable
architecture for supporting energy efficiency as well as need for supply
control are pathways for the renewable development of SG.
Generally, the primary technical challenge in the strength process func-
tioning in a deregulated or competitive atmosphere is increasing the power
transfer power of current transmission methods to stay away from the con-
gestion in the method. Many techniques have been recommended to manage
the problem. Several of them include maximum energy flow, grounded model
scheduling, use of complex systems like FACTS, and sending out decades,
and they will help you mitigate congestive network situations on the con-
strained transmission track. The issues for consideration in computational
complications for the improvement of the SG feature are the following:
• penetration of unlimited energy resources (RERs); participant bidding
methods prevent businesses from providing answers with environmental
objectives;
 2.4 Challenges of IoT in EMS Post-implementation 39

• the high reliance on energy method versions on intelligent operation and

influence, power method planning, and management prevents businesses
from providing solutions with complex goals;
• lack of comprehensive knowledge by engineers and operators of com-
putational equipment that is actually user-friendly and immediately
interpretable;
• the control and operation are complex power systems because of
the complexity of computational resources used for uncertainties and
modeling;
• forecasting tonne demand, ancillary services, and prices avoids busi-
nesses from providing answers with economic objectives;
• risk minimization in the power sector that is electric with buy-in compu-
tational resources seeks to establish a tradeoff between maximizing the
anticipated return shipping.
The main complex issues of SG are interoperability, networking communica-
tions, demand response, power storage, and distribution grid control.

2.4 Challenges of IoT in EMS Post-implementation

The challenges of IoT are classified into various categories as stated in the
following:
• privacy issues;
• security issues;
• data storage issues.
2.4.1 Privacy Issues
The usage of IoT is widespread as it offers a unique operation for various
applications such as wellness and transportation systems. IoT also provides a
better possibility for the interconnection of various sensors. This system has
some privacy issues, such as the following.
Authentication:
Authorization to access the database of IoT is the main privacy issue. This
challenge arises as many rely on weak passwords for authenticating the IoT.
Encryption:
Encryption is the process of hiding secure data. It has been noted that many
IoT-empowered gadgets do not encrypt the data while transferring over the
Internet.
40 IoT and its Requirements for Renewable Energy Resources

Insecure Interface:
IoT devices mostly have a screen for controlling or assigning the schedule.
This interface needs proper authentication to alter the operation.
Credentials:
Many people incorporate and use the default credentials due to a lack of
knowledge in changing the username and password. This is vulnerable to
hackers accessing and re-altering the operation.
Coding Practice:
The IoT devices are designed to be user-friendly so that the machine coding
can be changed as needed. But this has also been a challenge in privacy
concerns since the coding is done or altered without the proper coding ethics.
User Privacy:
When transitioning to advancement, the privacy of the customer/user
becomes much more important. For example, a grocery shop collects the
user’s information, particularly their location and the frequency of products
purchased. Unless the data is securely handled or encrypted, the health of the
user is exposed, which is a privacy thread concerning IoT.
Reducing Privacy Issues and Security Challenges:
IoT products are designed in such a way to be secure if the proper protocols
are followed.
Network Validation:
Proper scheduled validation needs to be done, such as network traffic and any
modifications in encryption. If the encryption is properly done, the security
issues can be reduced to some extent.
Edge Channel Validation:
With the aid of penetration assessment, the edge channel defense can be
validated for both software and hardware modules. This penetration will
reduce the recent threads on privacy issues.
Coding Ethics:
The codes of various products are kept secure to be utilized for marketing
purposes. Early safe code feedback is recommended to reduce the upcoming
challenges. On the other hand, the financial cost of rectifying the security
issues will be reduced if proper handling of the code is done at proper
intervals.
 2.4 Challenges of IoT in EMS Post-implementation 41

2.4.2 Security Concerns

IoT-based systems are predicted to face various security threats, such as the
following.
Spoofing:
It is like a misusage of a person’s or company’s identity. Spoofing aims to
represent the identity in a different manner and to spoil the reputation.
Tampering:
Data tampering is mostly seen in the marketing category. This challenge is
put forth by many leading organizations, such as obtaining and analyzing the
usage of a particular product by consumers and increasing the price of that
product. The tampering is not limited to organizations; it is done in residential
too by obtaining the power usage from smart meters and imposing additional
prices for the same usage.
Compromising:
It has been noted that only some products are tamper-resistant, whereas many
fall under the non-tamper criteria. As an example, the meters are tamper-
resistant, whereas the connected sensors or actuators are not tamper-resistant.
Compromising the resistance will also lead to the security thread. Most of
the IoT devices are IP enabled; so it is easier to incorporate into a common
Internet. This methodology gives a key for attackers to access the data,
leading to a security vulnerability. This cyber-attack not only damages the
software part but also damages the physical parts of the IoT.
Scalable Issue:
The protection of the IoT is not scaled down to certain regions; it has been
designed to cover a wide range of products. This is the key for attackers to
easily obtain the credentials of the user.
IoT Communication:
Nowadays, the use of driverless cars is common. The technology behind this
is IoT. If proper security is not provided, it may lead to unpredictable damage
to both humans and the product.
Deployment:
The SG is intended to cover a wide area, as a single SG covers the entire
nation, which leads to the work being unsupervised. If adequate SGs are
42 IoT and its Requirements for Renewable Energy Resources

provided at distant locations without any physical boundary, monitoring

security is easily accessible and tampering of data can also be mitigated to
some extent.
Outdated Systems:
The integration or connection of IoT security monitoring is done in the old
available system and also through some private bodies. Usage of IoT-based
SG in the existing system is a noteworthy issue since there is no possibility to
connect the existing system with the new upgraded modules.
Resources:
Most of the SG devices are restricted in accessing the resources, which are
predominantly widely connected. For enhancement of the security concerns,
particular attention is needed to accommodate the solutions.
Integration:
There is still usage of legacy systems that do not support the TCP/IP protocol
and cannot connect with IP-based systems without the aid of third-party
gateways, leading to a lack of end-to-end encryption. And also, there is a
notable issue of security-enhanced device interconnection with unsecured
devices, which makes both the devices vulnerable to security issues.
Trust Management:
Devices that are from different entities do not acquire trust till a minimum
level of security enhancement is provided. Nonetheless, the question of estab-
lishing trust between various entity device connections remains unanswered,
particularly in long-distance connections.
Bandwidth:
The SG devices are to be made to respond faster and more accurately
to time-to-time variations. SCADA, for example, is a computer-controlled
system used in transmission and distribution systems that must be capable of
responding to time-to-time variations in electrical parameters such as voltage,
current, power, and so on. The devices employed in weather prediction should
be very fast in updating the meteorological changes by obtaining data from
various sensors.
Enhancement of IoT Security:
The key security that needs to be incorporated for the IoT used in the smart
grid is as listed in the following.
 2.4 Challenges of IoT in EMS Post-implementation 43

Validation:
The competence to check and verify the actual devices connected and detect
if there is any new unknown device connected or accessing the data transmis-
sion and distribution. The energy supplier needs to validate the user’s smart
meter for consolidating the bill of use.
Genuineness:
The supplier needs to provide a security enhancement to the database so
that the accessed data or bill does not tamper illegally. Periodic software
enhancement is needed for smart meters, as the whole process relies on the
connected meter.
Inscrutability:
The secured data of the user should be made accessible only by the allocated
SG operator and the energy provider for that particular user. And, addition-
ally, the user is restricted to knowing only the particular data accessed and
not the whole accessed data of all consumers.
Privacy Protection:
Even though the SG operator and energy provider have the data of the user,
irrespective of the units consumed or energy price, the data should not be
used for any other purpose than the calculation procedure of the bill. Simply,
the data collected for billing should not be used for another purpose without
proper approval from the user.
Authentication:
Ensure that the allocated person from the SG is accessing the data on both
software and hardware. Periodic validation is required to ensure that the
allocation of rights for data access is needed. If any mismatch is noted in
the usage of power, the authorized person needs to make a field visit and
investigate whether there are any configuration issues in the smart meter or
any physical alterations have been made.

2.4.3 Data Storage Issues

2.4.3.1 Challenges in data management
Every IoT device is capable of collecting a large amount of data. The issue
occurs in the storage of all the data at a single point. As per the research, it has
been clear that, at present, there is no adequate storage available to store the
44 IoT and its Requirements for Renewable Energy Resources

entire sensor and actuator collected data. Very few entities have invested in
developing an adequate storage space for housing all their device data. Many
have compromised on storing all the data by storing only the data based on
the need and its value. But this causes an issue when we need data, which, at
that time, may be considered unneeded. Increasing the number of data centers
will only be a part of the solution to deal with the daily increase in IoT device
usage. The proper segregation of data from the device is needed so that it is
only easily accessible with available bandwidth. For instance, if we need to
calculate the usage of power by a user, the accessing of smart meter details
should be simplified, and, most of the time, they need to be quickly accessible.

2.4.3.2 Challenges in fetching data

Even though all the data is stored in data management, for easy access, we
rely on data fetching tools. The data stored is not only the initial cases but also
needs periodic updating. For example, the billing details of the user need to
be updated at every instance of calculating the energy usage. The streaming
of weather data is based on the data obtained from the sensor and stored
in a particular location/allocation of the dataset, with the data being fetched
from a particular set from time to time, for instance. Data fetching tools play
a significant role in updating the proper data to the allocated location and
fetching the required data from the correct dataset. This fetching also needs
to be done as per the individual user’s needs. The streaming of digital content
is made simple with these data mining tools. They have been enhanced in
terms of security to detect tampering and illegal data retrieval. The digital
media streams are completely reliant on these tools, as significant progress
has been made in the data retrieval process.

2.4.3.3 Challenges in allocation

When comparing the development in consumer-related products, the develop-
ment of IoT has a significant mark, such as innovation in sensors and wireless
technologies. The day-to-day enhancement also needs to be under the security
protocol. New IoT innovations will be unmotivated if they deviate from
security concerns and standards. A small mistake in data handling can lead
to unpredictable damage. For example, when handling the SCADA system,
if any wrong data has been put forward based on that data, modifications are
made to the physical devices, which leads to damage to the user-connected
devices. Another instance, such as health monitoring, where wrong data about
health is interpreted and the user consumes medicine concerning the IoT data,
leads to unpredictable damage to humans. If even one of these incidents is
 2.5 Solution to IoT Challenges 45

reported and made public, users’ trust in IoT devices will suffer. The data
management needs to fetch and provide the user-required data irrespective
of the available bandwidth. The issue occurs when there is saturation on the
Internet since most of the IoT devices are operated on IP. The usage of IoT
is fulfilled only when every user has trust and enhancement in the security of
the system.

2.5 Solution to IoT Challenges

Through the implementation of IoT-based technology in the smart grid, it has
become more reliable and efficient in terms of operation. Even though IoT
provides good support in the operation of the smart grid system, it has certain
security flaws, as discussed in previous topics. Those security issues can be
overcome with the aid of various methodologies such as blockchain, artificial
intelligence based operations, and distributed frameworks. In the upcoming
sections, certain opportunities available in the IoT to mitigate security issues
are discussed.

2.5.1 Blockchain Methodology

The blockchain methodology is an upcoming methodology concerned with
the distribution of information and the management of storage. The major
advantage of blockchain is that there is no centralized administration. In the
context of IoT with blockchain, it has the capability of interconnection as
well as information exchange in a peer-to-peer (P2P) community, as shown in
Figures 2.4(a) and (b), which depict information exchange in a client-server
network and a P2P network, respectively. Blockchain technology operates
based on a centralized algorithm; so there is less possibility of data access
by any third person. If any information on the server is accessed or modified,
that info is shared with all the community users, which enhances the security
of the smart grid data.
Blockchain technology has a periodic update in terms of enhancing
security, which makes it a failure for cybercriminals to access centralized
data. And also, blockchain integration with IoT is done in a manner such
that the data can be backtracked and also notified if there is any conflict in
the provided information. The data and information in the IoT are encrypted
through blockchain engineering. In the P2P communication process, the
interactions are built with the aid of a key exchange mechanism [11].
Through the cryptographic key exchange mechanism, the blockchain-
protected data is assured to be safe and no intruder can interface with and
46 IoT and its Requirements for Renewable Energy Resources

Figure 2.4 (a) Exchange of information in client-server network. (b) Exchange of informa-
tion in P2P network.

access the data. The IoT-based smart grid data can be easily backtracked
consequently. In the next part, the major function of blockchain, which
benefits the IoT-implemented smart grid, is discussed.
• Nature of Distribution:
The distribution process is faster, reliable, and efficient due to the
elimination of various centralized architectural issues.
• Security Protection:
Since all the data transmission is encrypted cryptographically, the
reliability in the privacy protection is enhanced [12].
• Eminence:
The possibility of data backtracking helps in analyzing all the changes or
modifications made if any in the system. Easy traceability of blockchain
topology helps in tracing the changes that occurred in IoT devices [13].
• Credentials:
All the IoT devices have unique credentials so that the backtracking of
every IoT device is possible with the help of blockchain methodology.
Because of the shift from centralized to distributed generation, blockchain
technology has emerged as a promising solution for establishing a dependable
consensus mechanism. Blockchain allows P2P electronic payments and is
categorized into two classes, namely public chains and generalized private
chains. The public chain allows every individual or institution to join and
leave freely and is a fully open network. A private chain is a network in
which participants are subject to restrictions. Blockchain’s key characteris-
tics include its decentralized structure, distributed data storage mechanism,
consensus agreement, and asymmetric encryption for network security.
 2.5 Solution to IoT Challenges 47

2.5.1.1 Blockchain technology infrastructure features

The backbone of infrastructure includes distributed data storage and verifica-
tion platform based on blockchain, a sensing environment based on IoT, and
a service delivery system that is cloud-based [14]. The above cyberinfras-
tructure supports physical energy entities for generation, transmission, and
DNs. In physical infrastructure energy flow, the segment includes generation
from power plants and distributed generation resources. Both renewable and
nonrenewable power plants are included in the generation segment and their
transmission at different levels is considered. Energy entities like energy
market systems, load aggregators, and energy retailers are also involved in
physical infrastructure.
The cyberinfrastructure has to support the entities being monitored, the
need to control equipment, and its coordination with physical infrastructure.
In the cyberinfrastructure, the IoT devices are equipped with protocol stacks
and communication interfaces to communicate with one another and their
users. The blockchain and cloud infra will collect data from multi-domain
and stakeholders’ decisions are made. The decentralized and distributed
processing environments [15] are provided to interface with the IoT so that
it provides services like data storage, a platform for smart contracts, and also
authorization for agreements and assets. Depending on the application and
its requirements, the data collected by IoT can be stored in blockchain or
uploaded to the cloud.

2.5.1.2 Application domains of blockchain technology

The application of blockchain technology in future energy systems has huge
potential. In a data management system, a blockchain structure can support
grid data protection and smart meter data aggregation. Blockchain technology
has the potential to reconstruct the energy market so that a centralized trading
structure can be adopted. For the above applications, the limitations are cyber-
attack vulnerability, difficulty in establishing an open, cross-border energy
market system [16], and the fear of a centralized market structure. The
blockchain also provides a data synchronization mechanism and an infor-
mation recording mechanism. The technology also provides authentication
services for its energy stakeholders.

2.5.1.3 Challenges of blockchain technology

Though blockchain technology has the potential to provide a solution for
future energy systems when applied to practical cases, certain challenges
have to be overcome, and a few are highlighted here. Information redundancy
48 IoT and its Requirements for Renewable Energy Resources

occurs as multiple data copies are created on networked nodes, and, also,
it would take extra storage space and more power consumption [17]. A
performance scalability problem exists in the blockchain system where the
consensus protocol can slow down the system when nodes are added. The
design flaws and bugs that exist in programs written by a human cannot be
fixed or reversed as blockchain techniques are irreversible. Verification tools
are to be designed for smart contracts or user-defined programs. Computer
network traffic occurs when the blockchain network is coordinated with other
infrastructures. Therefore, the differential functional energy organizations
have to co-operate for its implementation. The system may suffer due to
congestion in the network, deviation in voltage levels, and overloading when
integrated with physical grids.

2.5.2 Cloud Computing

Cloud computation systems have developed in recent decades, and the pro-
cessing of information gathered from systems is the major strategy followed
in cloud computing. It mainly offers a solution for the storage, applica-
tion, and computation of data online, which flows from IoT products. The
cloud embodies the “Internet” and computing embodies the computation
and processing solutions provided by this strategy [18]. Cloud computing
encompasses both Internet application program solutions and data center
based hardware systems [19]. Based on the characteristics, fundamental
data and structures of multifaceted computation capabilities are processed
in cloud computation [20]. The major advantages of using cloud computing
systems are that (1) hardware prices could be lowered; (2) computational
power and storage ability could be improved; and (3) ease of operation in
managing information due to multi-core architectural systems. Additionally,
a cloud computational system is a protected system that offers information,
computation energy, and storage space that is required from a terrestrial
location [21]. Figure 2.5 shows the structure of the IoT in the cloud, and,
here, the function of cloud computing is to allow the fundamental data from
IoT components to be quickly examined, managed, and sorted resourcefully.
Moreover, cloud computing eliminates the costs associated with purchasing
software and hardware and operating the algorithms for processing the IoT
information, significantly reducing the amount of electrical energy required
for neighborhood detail computation.
It is not easy to make each item and tool within the smart grid component
of the IoT system and, after that, make all options manageable, controllable,
and accessible through cloud computing. Simply said, several issues have
 2.5 Solution to IoT Challenges 49

Figure 2.5 Structure of IoT in cloud.

been identified as requiring access to the cloud-IoT to improve the usage of

power DNs [22]. The current situations demand that we resolve the cloud
computation system from the perspective of the energy industry apart from
resources and data. Though the cloud-IoT generates greater opportunities
economically in the energy industry, it raises the risk in the system also.
Discretion, safety measures, and, predominantly, individuality protection
are becoming important aspects within the recurrent and extremely shifting
cloud environments, which would be the foundation of public and private
clouds integrated with undertakings by companies. In cloud-IoT, diversified
networks and solutions might be required and unified to allow for different
services and diverse data. The system should have the versatility and adapt-
ability to help all data types based on the needs of program quality support. It
can be declared that although the blend of IoT as well as cloud companions
might get over several of the present issues of these two solutions, which
should be solved distinctly together, brand new tasks might be produced in
such a synergistic and joint connection that the following portion intends to
talk about several of these difficulties [23].

2.5.2.1 Reference architecture

The IoT plus cloud area is highly heterogeneous, and there are no appropriate
requirements for services and data. The standing options have fundamental
50 IoT and its Requirements for Renewable Energy Resources

Figure 2.6 Reference architecture of cloud-IoT.

differences and varied ranges concerning the union of IoT and cloud com-
puting systems. This difference is, to a significant degree, because of the
absence of regular tools to allow for the style of these remedies, and that
deteriorates the intricacy and also necessitates major work for their design
and architecture. This issue is aggravated by the absence of proper direction
and framework to meet non-functional and functional requirements as these
results hinge on their deployment and implementation. Additionally, data
produced by various IoT products does not obey any normal format, meaning
they are frequently signified in formats that are different, different devices,
and so on. Presently, you will find pervasive platforms and no tools for
describing IoT equipment and their distinctive capabilities in a uniform and
a regular means; so the program agents cannot just do the tasks of theirs,
for instance, automated detection and control of products (services, data,
and orchestration) [24]. One of several solutions to ease these problems is
certainly the reference architecture. The reference architectures can have fun
with an important part of the meaning of the structure blocks needed to
create architectural methods that mix IoT as well as cloud computing prin-
ciples. Referring architectures could be observed as intellectual architectures
integrating experience and science in a specific program domain, therefore
enabling the expansion, interoperability, tuning, and progress of program
systems within the specified domain.
Therefore, the recommendations indicated by reference architecture may
be viewed as important components in creating cloud-IoT besides their
increasing complexity and size. Additionally, because of the benefits of
 2.5 Solution to IoT Challenges 51

compatibility in acquiring results, this interoperability could be looked after

by reformation and producing architectural remedies that are based upon
a reference architecture. Among the readily accessible remedies, the open-
IoT wedge, as an innovator, appears to be on the road to being realized
as the guide structure to combine IoT as well as cloud computing systems
by presenting comprehensive explanations and procedures of comprehensive
devices [25].

2.5.2.2 Network communication and its challenge

Smart grid equipment and parts that are connected to the Internet gener-
ate and send out a vast amount of data that must be categorized, sorted,
and enhanced in the cloud. If every piece of information produced by all
machines were delivered to the cloud, a substantial stream of information
would flow throughout the networks, causing network congestion. As a result,
it leads to communication delays, which could be particularly unbearable
for apps that are running in real time (or have quasi-real-time performance).
Although current cloud systems appear to be capable of providing multiple
sources for significant data analytics, widespread usage of these materials
can be prohibitively expensive. As a result, it is critical to develop innovative
approaches to limit the number of cloud methods used in IoT programs as
well as how cloud resources should perform effective processing on the data
generated by-products and how to successfully engage computing resources.
Cloud computing technology, which enables cloud computing on the system’s
borders to offer different solutions like processing and analyzing storage
space facilities, as well as community services between the platform and
connected devices on the cloud, is one of the suggested techniques for dealing
with this challenge. A cloud computing system is a more advanced version of
traditional cloud computing designed to assist IoT applications with response
time constraints, geographical distribution, and mobility requirements. Smart
gateways and regional clouds would be required for the notion to be imple-
mented, as they are likely to be used for communication interceding between
IoT systems and cloud sources as well as pre-processing and short-term
storage of data on smart grids [26].

2.5.2.3 Privacy and security

The adoption of IoT infrastructure on impending smart grids faces signif-
icant security challenges. Figure 2.7 shows the security challenges in the
IoT. Although security is a huge concern on the current Internet platform,
when IoT and cloud computing are combined in smart grids, it becomes a
52 IoT and its Requirements for Renewable Energy Resources

Figure 2.7 Security challenges in IoT.

much larger and more vulnerable issue that may be considered the biggest
challenge in IoT. This assumption is supported by the outstanding community
redistribution and, as a result, additional access points in the process, as well
as the inescapable importance of reliable power service. Furthermore, the
items are hooked up to have a simpler design than computer systems, which
makes it hard to implement security equipment online ordinarily. The IoT is a
great deal closer to actual existence than today’s Internet, and the private and
important info of customers and energy manufacturers is discussed on the
information system. Assaulting such a system will be the same as interfering
with the day-to-day operations of smart grid owners in the distribution,
transmission, and generation sectors.
The network and protocol security provide an Internet-protected com-
munication mechanism with identity management for authorization and
authentication to ensure that information is produced by a specific provider,
as well as limiting access, maintaining privacy, and establishing trust between
subscribers and entities to aid political aims. Finally, the most serious security
problems are IoT attack types such as denial of service (DoS), physical
 2.5 Solution to IoT Challenges 53

damage, eavesdropping, and node capturing for data extraction by control

organizations [27].

2.5.2.4 Background information

Not only is the most advanced sensor data managed delicately in the IoT,
but the context of that data, which includes some type of information that
could be expected to separate an individual’s identity, location, or the state of
an item, also necessitates a complex administration process. Contextual ele-
ments such as metrics, component status, location, and information precision
could be translated into raw data from sensor devices for further processing
or retrieval. Even though adding contextual information increases the value of
sensed data, it also increases the variability of the data in terms of the format
utilized. Combining several screen models creates a barrier in the informa-
tion sharing and retrieval process. To address the problem, data generated
by a variety of IoT-enabled devices must be collected and refined to raise
awareness, and access to and authorization for investigation and extraction
should be granted to a wide range of end-users and applications [28].

2.5.2.5 Big data analytics

The simple fact that 50 billion products will be connected to the system
by 2020 needs a special focus on the conveyance, access, storage, and
processing of the massive amounts of data that these devices will gener-
ate. IoT technological innovation will undoubtedly be on the list of major
components of serious data analytics because of recent scientific growth, and
the cloud permits complex analytics and long-term storage. The widespread
use of mobile devices and receptors is, in fact, a signal and a request for
a scalable computing wedge (approximately 2.5 quadrillion bytes of data is
produced every day). Handling this volume of data is a significant challenge,
as the program’s overall operation closely resembles the functions of the
information management services. As an example, cloud-based systems are
developing and improving ways to summarize important details based on
semantic feature selection algorithms. As a result, after the NoSQL advances,
both proprietary and open-source database technology therapies are used to
tackle big data [29].

2.5.2.6 Provision of program quality

As the number of datasets grows and the range of unpredictability, uncer-
tainty, and sorts of details in the product expands, the quality of support
becomes a contentious issue. Any amount or type of information could be
54 IoT and its Requirements for Renewable Energy Resources

used at any time. Some of the information can also be classified as critical
information. As a result, queries must be dynamically prioritized on the cloud
side. The bandwidth, noise, delay time, and program missing ratio are com-
monly used to calculate service quality. The level of service quality must be
determined by the data type and the urgency with which the information must
be delivered on the sync node (access point) [30]. In protocol supporting vari-
ous protocols for joining various objects on the Internet, even in the case, they
have homogeneous nature are found. For instance, an IoT sensor could use
unique protocols, like Wireless HART, IEEE 1451, Zig Bee, and also 6LOW-
PAN [31]. The gateway Hojo motor magnetic generator is required for a few
protocols, while others cannot be supported. The kind and gate of the sensor
utilized decide this. A sensor that is easier or cheaper to obtain has a higher
priority than others from the subscriber’s and user’s perspective. As a result,
there is no way to guarantee that the newly added sensor will be profitable.
The chart of typical gateway protocols could be used to solve this problem.

2.5.2.7 IPv4 addressing limit

The web is a significant element of the IoT, though the Internet address is
restricted to IPv4. Furthermore, numerous solutions, such as CoAP, enable
interoperability with embedded devices and elements from the Internet,
and new product development activities highlight the necessity to eliminate
network address translation (NAT) systems and address these issues with
a unique IP address. Simply put, when items have become a part of the
Internet (IoT), they need a unique and exclusive identification, and Figure 2.8
shows the interface in the industrial platform. Furthermore, mobile items,
such as moveable sensor nodes on all types of cars (especially electric

Figure 2.8 Interface in industrial platform.

 2.5 Solution to IoT Challenges 55

vehicles) and other moving objects, must map their real-time identification
to the new community they have entered. Assigning IPv6 addresses to this
pervasive network is wise since the IPv6 address space is deemed to be
ample to handle IoT systems. To overcome this issue, IPv6 employs the IP128
short address, which offers benefits such as Internet integration, continuous
connectivity, and rest compliance. SLoWPAN and Zig Bee IP protocols have
good compatibility with IPv6 in IoT-enabled embedded goods.
However, there are currently very few industrial platforms that have these
capabilities. Because of IoT adoption, devices will transition from human-
centered pursuits to networks featuring machine-to-human and machine-
to-machine connections, and IPv6 will pave the way. If IPv6 is used to
recognize the communications of objects, the widespread deployment of IPv6
becomes a contentious issue. Unless a proper, regular, and effective system
for cohabitation of IPv6 and IPv4 is implemented, the IPv6 to tools (objects)
project will not be successful [32]. According to research focused on the
coexistence of IPv6 and IPv4, the smooth transition to IPv6 should be further
developed.

2.5.2.8 Legal aspects and social facts

These two critical concerns are linked to one another in some way. Consider
a cloud-based, data-driven service provided by a single person. On the one
hand, the main service provider must adhere to various international regula-
tions, while, on the other hand, consumers must be motivated to participate in
data collection. It is easier to provide people with the option of collaborating
on information-gathering items that represent a specific concept. Recently,
several issues have been identified that may be found in almost any system,
such as the use of people as sensor information sources. The response
also contributes to the users’ continued loyalty (such as combining human-
made qualitative observations with quantitative observations produced by
machines; or maybe they want to determine and control reliability, reputation,
data quality, and belief in) [33]. Subscribers and users may be enabled
by up-to-date customized applications as well as blocks such as toolkits,
frameworks, and accelerators that allow users to engage with IoT in the same
way that they would with the Internet. Such gadgets must allow design experts
to learn about consumer performance while also providing owners with an
engaged role in the engineering design system. As a result, to achieve this
goal, the resources should make it possible for specific people to see different
design characteristics simply and cost-effectively. To address this problem,
scientists and experts are working to develop appropriate tools for using a
56 IoT and its Requirements for Renewable Energy Resources

collaborative sampling technique in which both designers and specific users

uncover applications and connections.

2.5.2.9 Service detection

The cloud supervisor, or possibly agent in cloud-IoT, oversees the identifi-
cation of new services and goods for individual consumers. On the IoT, any
item can become a component of the system at any time, and the system can
become a component at any time. Some IoT nodes are capable of mobility.
Finding new solutions, as well as their instantaneous status and updating the
currently established services, remains a contentious issue. For a larger and
more complicated IoT, management equipment may be necessary, which is
responsible for monitoring the condition of IoT nodes, keeping an eye on
cellular nodes and maintaining the current state of both the present IoT and
recently added IoT nodes. For this work, a consistent means of obtaining
assistance is required [34].

2.6 Conclusion
With the rise in global demand for energy and the numerous barriers to its
extraction, like fossil fuel depletion, attention toward climate change prob-
lems, etc., the world is facing a new shift in energy generation. The advent of
IoT and its smart integration with current power distribution systems have
tremendously played a role in reducing the socioeconomic-environmental
impacts of the current energy distribution system, thus fulfilling the needs
of energy use. Faced with several challenges in this type of integrated power
distribution system, the state-of-the-art IoT power sector is poised to move
from the latest hierarchy to build a smarter, decentralized system. In this
study, we discuss the role of the IoT in the energy supply chain and the
application of smart grids in the power distribution system. Furthermore, this
article discusses the role of the IoT and its structure and challenges in inte-
grating it into the energy chain as well as tackling these challenges with IoT
support and its benefits. This study reviews the many components of an IoT
system, such as supporting communication and sensor technologies, and how
they might be used in the energy sector, such as temperature, humidity, light,
speed, passive infrared, and proximity sensors. Also, this study discusses
cloud computing and data analytic platforms, which are data analysis and
visualization tools that may be used for a variety of smart energy applications,
from buildings to smart cities. Moreover, this chapter highlights the after-
implementation challenges of IoT in energy management systems, such as
 References 57

privacy problems and challenges, security problems and their challenges,

and data storage problems and their challenges. This chapter discusses the
IoT challenges and their solutions through blockchain technology and cloud
computing briefly like the future scope of the research.

References
[1] Friess, D. A. (2016). Ecosystem services and disservices of Mangrove
forests: Insights from historical colonial observations. Forests, 7(9), 183.
https://doi.org/10.3390/f7090183.
[2] Gardoni, P., & Murphy, C. (2020). Society-based design: Promoting
societal well-being by designing sustainable and resilient infrastructure.
Sustainable and Resilient Infrastructure, 5(1-2), 4–19.
[3] Li, Z. Q., Xu, H., & Zhang, Y. (2014). Remote sensing of haze pollu-
tion based on satellite data: Method and system design. Environmental
Monitoring in China, 30(03), 159–165.
[4] Parker, D. S. (2020). The implementation of the Internet of Things
(IoT): A case study of the barriers that prevent implementation of
IoT within small and medium enterprises (SME), Doctoral dissertation,
Northcentral University.
[5] Muruganantham, B., Gnanadass, R., & Padhy, N. P. (2017). Challenges
with renewable energy sources and storage in practical distribution
systems. Renewable and Sustainable Energy Reviews, 73, 125–134.
[6] Sinsel, S. R., Riemke, R. L., & Hoffmann, V. H. (2020). Challenges and
solution technologies for the integration of variable renewable energy
sources—A review. Renewable Energy, 145, 2271–2285.
[7] Bhattacharjee, S., & Nandi, C. (2019). Implementation of industrial
internet of things in the renewable energy sector. In The Internet of
Things in the Industrial Sector, pp. 223–259, Springer, Cham.
[8] Serpanos, D., & Wolf, M. (2018). Industrial Internet of Things. In
Internet-of-Things (IoT) Systems, pp. 37–54, Springer, Cham.
[9] Al-Ali, A. R., Zualkernan, I. A., Rashid, M., Gupta, R., & AliKarar, M.
(2017). A smart home energy management system using IoT and big
data analytics approach. IEEE Transactions on Consumer Electronics,
63(4), 426–434.
[10] Salkuti, S. R. (2020). Challenges, issues and opportunities for the devel-
opment of smart grid. International Journal of Electrical and Computer
Engineering (IJECE), 10, 1179–1186.
58 IoT and its Requirements for Renewable Energy Resources

[11] Pilkington, M., (2016). Blockchain technology: Principles and appli-

cations. In Research Handbook on Digital Transformations, vol. 225,
Edward Elgar Publishing, Cheltenham, U.K.
[12] .Veena, P., Panikkar, S., Nair, S. & Brody, P. (2015). Empowering the
edge-practical insights on a decentralized Internet of Things. Empow-
ering Edge Practical Insights: A Decentralized Internet Things, 17,
2015.
[13] Prisco, G. (Nov. 2015). Slock.it to introduce smart locks linked to
smart Ethereum contracts decentralize the sharing economy. [Online]
Available: https://bitcoinmagazine.com/articles/sloc-it-tointroduce-sma
rt-locs-lined-to-smart-ethereumcontractsdecentralizethe-sharing-econ
omy-1446746719.
[14] Stergiou, C., Psannis, K. E., Kim, B. G., & Gupta, B. (2018). Secure
integration of IoT and cloud computing. Future Generation Computer
Systems, 78, 964-975.
[15] Ji, C., Li, Y., Qiu, W., Awada, U., & Li, K. (Dec. 13–15, 2012). Big
data processing in cloud computing environments. In Proceedings of the
2012 12th International Symposium on Pervasive Systems, Algorithms
and Networks, San Marcos, TX, USA, pp. 17–23, vol. 107.
[16] Vukolic, M. (2015). The quest for scalable blockchain fabric: Proof-of-
work vs. BFT re-plication. In Proceedings of the International Workshop
on Open Problems in Network Security, Springer, Berlin, pp. 112–125.
[17] Blockchain. (2017). Transactions per block. [Accessed 2-Nov-2017].
Available: https://blockchain.info/charts/n-transactions-per-block
[18] Vytelingum, P., Ramchurn, S. D., Voice, T. D., Rogers, A., & Jennings,
N. R. (2010). Trading agents for the smart electricity grid. In Proceed-
ings of the 9th International Conference on Autonomous Agents and
Multi-Agent Systems (IFAAMAS), pp. 897–904.
[19] Block, C., Neumann, D., & Weinhardt, C. (2008). A market mechanism
for energy allocation in micro-CHP grids. In Proceedings of the 41st
Annual Hawaii International Conference on System Sciences, IEEE,
p. 172.
[20] Block, C. A., Collins J., Ketter W., & Weinhardt, C. (2009). A multi-
agent energy trading com- petition. [Accessed 22-Nov-2017]. Available:
www.erim.eur.nl
[21] Ketter, W., Collins, J., & Reddy, P. (2013). Power TAC: A compet-
itive economic simulation of the smart grid. Energy Economics, 39,
262–270.
 References 59

[22] Yassine, A., Singh, S., Hossain, M. S., & Muhammad, G. (2019). IoT
big data analytics for smart homes with fog and cloud computing. Future
Generation Computer Systems, 91, 563–573.
[23] Carvalho, O., Garcia, M., Roloff, E., Carreño, E. D., & Navaux, P.
O. (Sep. 2017). IoT workload distribution impact between edge and
cloud computing in a smart grid application. In Proceedings of the Latin
American High Performance Computing Conference, Springer, Cham,
pp. 203–217.
[24] Forcan, M., & Maksimoviæ, M. (2019). Cloud-fog-based approach
for smart grid monitoring simulation. Modelling Practice and Theory,
101988.
[25] Jamborsalamati, P., Fernandez, E., Hossain, M. J., & Rafi, F. H. M. (Nov.
2017). Design and implementation of a cloud-based IoT platform for
data acquisition and device supply management in smart buildings. In
Proceedings of the 2017 Australasian Universities Power Engineering
Conference (AUPEC), IEEE, pp. 1–6.
[26] Stergiou, C., Psannis, K. E., Kim, B. G., & Gupta, B. (2018). Secure
integration of IoT and cloud computing. Future Generation Computer
Systems, 78, 964–975.
[27] Chiang, M., & Zhang, T. (2016). Fog and IoT: An overview of research
opportunities. IEEE Internet of Things Journal, 3(6), 854–864.
[28] Cai, H., Xu, B., Jiang, L., & Vasilakos, A. V. (2016). IoT-based big data
storage systems in cloud computing: perspectives and challenges. IEEE
Internet of Things Journal, 4(1), 75–87.
[29] Zahoor, S., Javaid, S., Javaid, N., Ashraf, M., Ishmanov, F., & Afzal,
M. (2018). Cloud–fog–based smart grid model for efficient resource
management. Sustainability, 10(6), 2079.
[30] Mekala, M. S., & Viswanathan, P. (Aug. 2017). A survey: smart agricul-
ture IoT with cloud computing. In Proceedings of the 2017 international
conference on microelectronic devices, circuits and systems (ICMDCS),
IEEE, pp. 1–7.
[31] Aazam, M., Huh, E. N., St-Hilaire, M., Lung, C. H., & Lambadaris,
I. (2016). Cloud of things: Integration of IoT with cloud computing.
In Proceedings of the Robots and Sensor Clouds, Springer, Cham,
pp. 77–94. [45]. Stojmenovic, I. (Nov. 2014). Fog computing: A cloud to
the ground support for smart things and machine-to-machine networks.
In Proceedings of the 2014 Australasian Telecommunication Networks
and Applications Conference (ATNAC), IEEE, pp. 117–122.
60 IoT and its Requirements for Renewable Energy Resources

[32] Kaur, M. J., & Maheshwari, P. (Mar. 2016). Building smart cities appli-
cations using IoT and cloud-based architectures. In Proceedings of the
2016 International Conference on Industrial Informatics and Computer
Systems (CIICS), IEEE, pp. 1–5.
[33] Tom, R. J., & Sankaranarayanan, S. (Jun. 2017). IoT based SCADA inte-
grated with fog for power distribution automation. In Proceedings of the
2017 12th Iberian Conference on Information Systems and Technologies
(CISTI), IEEE, pp. 1–4.
[34] Khan, Z., Anjum, A., & Kiani, S. L. (Dec. 9–12, 2013). Cloud based
big data analytics for smart future cities. In Proceedings of the 2013
IEEE/ACM 6th International Conference on Utility and Cloud Comput-
ing, Dresden, Germany, pp. 381–386.
 3
Power Quality Monitoring of Low Voltage
Distribution System Toward Smart
Distribution Grid Through IoT

P. Sivaraman1 , C. Sharmeela2 , S. Balaji3 , P. Sanjeevikumar4 ,

and S. Elango5
1 VestasTechnology R&D Chennai Pvt Ltd, India
2 Anna University, India
3 IIT Kanpur, India
4 Aarhus University, Denmark
5 Coimbatore Institute of Technology, India

E-mail: [email protected]; [email protected];

[email protected]

Abstract
Smart distribution grid allows bidirectional power flow in the distribution
system with a mixture of multiple renewable energy sources. The distribution
system is highly affected by various power quality (PQ) disturbances like
harmonics, transients, undervoltage, overvoltage, unbalance, etc. Ensuring
the quality of power is supplied to the consumers is one of the major concerns
of the distribution company. The distribution company has to monitor the PQ
parameters at various locations of the smart distribution grid. This chapter
discusses the monitoring of various PQ problems, such as undervoltage,
overvoltage, interruption, and overload of a smart distribution system. It
employs the remote communication stations (RCS) at the various locations
of a smart distribution system. RCS monitors the various PQ parameters con-
tinuously and compares them with the reference value. It gives notifications
whenever the PQ parameters exceed the reference values. This Internet of
Things based monitoring system detects the PQ events such as undervoltage
and overvoltage, unbalanced voltage, and overload. It sends the details to the
distribution company/users using alert SMS and also on the web.

61
62 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

Keywords: Internet of Things (IoT), smart grid, distribution system, power

quality (PQ), GSM.

3.1 Introduction
The distribution system is the last part of the power system, distributing
electric power to the end-users. The distribution system is the electrical
system between the sub-station and the consumer’s meters fed by the dis-
tribution system [1]. It comprises feeders, distributors, and service mains.
An AC distribution system is classified into a primary distribution system
and secondary distribution system. The primary distribution system voltages
vary in different countries. The typical primary distribution voltages are 33,
22, 11, kV carried by 3Φ, 3-W system and secondary distribution system
employs 415 V carried by 3Φ, 4-W system [2–5, 26]. Distribution systems
comprise residential, commercial, and industrial systems, and it distributes
the electric power supply to the end-users [6–8]. The distribution system
shall have high supply reliability, efficiency, and power supply quality to
its end-users [3]. To have such a distribution system, certain areas need
to be monitored and controlled. They are, location of fault identification,
clearing the fault in lesser time, effectively monitoring the power quality (PQ;
voltage and current variation), maintaining the voltage drop and PQ variation
within the acceptable tolerance limits, and maintaining the proper database of
failure/outage events [9, 10].
The distribution system is highly affected by various PQ disturbances,
such as harmonics, transients, undervoltage, overvoltage, unbalance, inter-
ruption, flicker, etc. [11–14]. The poor PQ results in premature equipment
failure, mal-operations, tripping of sensitive critical equipment, and financial
losses to end-users as well as distribution companies [15, 22]. Hence, it is
essential to monitor the PQ parameters at various locations of the distribution
grid to find out quality of the power supply is delivered to the end-users.
The equipment shall monitor the PQ parameters, such as voltage and
current transformer, PQ meter, or analyzer. With this equipment, PQ param-
eters like voltage sag, swell, transients, harmonics, flicker, etc., shall be
monitored/measured at the measurement point [16, 17]. There are two types
of PQ measurement carried out at the site. The first one is doing the mea-
surement after the failure of any devices/equipment, i.e., finding the reason
why the device/equipment failed. This method is widely used to find the
reason for failure and then propose a mitigation plan to avoid equipment
failure in the future. The second one is doing the measurement periodically
 3.2 Introduction to Various PQ Characteristics 63

or regularly as preventive maintenance to identify any existing PQ problem

and mitigate it before it leads to equipment failure, mal-operation, etc. Either
by permanent PQ monitoring (by means of permanent current transformer,
voltage transformer, and PQ analyzer at the site installation) or temporary
PQ monitoring through portable equipment (portable current transformer,
voltage transformer, and PQ analyzer at the site installation). In permanent-
type monitoring, the capital investment cost is high. Hence, in many places,
portable PQ instruments are used for temporary PQ monitoring.
In recent years, the Internet of Things (IoT) has been widely used to
monitor and control equipment from a remote location or centralized loca-
tion [21]. It acts as a mediator between the physical equipment (hardware)
and application (software). IoT-based technology is used to monitor the PQ
parameters from the remote location or centralized location and the actual
measurement site as well. It is using various network technologies (wireless
communication technologies) as a communication platform for remote mon-
itoring [18–20, 23–25]. This chapter discusses IoT-based PQ monitoring in a
smart distribution system.

3.2 Introduction to Various PQ Characteristics

As per IEEE Std 1100-2005, the PQ is defined as powering and grounding
electronic equipment in a manner suitable to the operation of that equip-
ment and compatible with the premise wiring system and other connected
equipment [9]. The PQ characteristics as per IEEE Std 1159-2019 are as
follows [10]:
• short duration root mean square (RMS) variation;
• long duration RMS variation;
• transients;
• unbalance or imbalance;
• voltage fluctuations;
• waveform distortion;
• power frequency variations.
Transients are momentary changes/variations in voltage or current or both
due to lightning, faults, switching operations, etc.
Short duration RMS variation is called a variation in RMS voltage for
between 0.5 cycles and 60 seconds.
Long duration RMS variation is called a variation in RMS voltage of over
60 seconds.
64 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

Imbalance or unbalance is the ratio of the magnitude of the nega-

tive sequence component to the positive sequence component expressed in
percentage in a three-phase system.
Voltage fluctuation is called a continuous change in the voltage (instanta-
neous cycle-to-cycle voltage) because of connected load resistance change in
every cycle.
Undervoltage is called long-duration RMS voltage magnitude reduction
between 0.9 and 0.8 pu for over 1 minute. The causes of undervoltages
are continuous energization of large capacity loads, overloaded circuits,
switching OFF capacitor banks, and undervoltage.
Overvoltage is called long-duration RMS voltage magnitude that
increases above 1.1 pu for over 1 minute. The causes of overvoltage are
switching OFF large capacity loads, switching ON capacitor bank, and
incorrect tap operation of transformers.
Sustained interruptions are called long-duration RMS voltage magnitude
reduction of less than 0.1 pu for over 1 minute.
Waveform distortion is called steady-state distortion or deviation from the
ideal sinusoidal characteristics [11].

3.3 Introduction to IoT

IoT is a medium of communication or mediator between hardware equipment
and software. The typical structure of IoT-based monitoring is shown in
Figure 3.1. In this method, all the metering and controlling equipment are
connected to a centralized or common platform, i.e., IoT. It will collect the
data from the connected hardware equipment and transfer it to the customer
application’s storage and data analytics.

Figure 3.1 Typical structure of IoT-based monitoring.

 3.4 Smart Monitoring using IoT for the Low Voltage Distribution System 65

It comprises three layers, namely hardware layer (sensors or measuring

equipment), middle layer (IoT platform), and application layer (customer
application, data analytics, and storage) [29, 30]. The hardware layer com-
prises measuring types of equipment like a current transformer, voltage
transformer, temperature sensor, humidity sensor, infrared sensor, photode-
tectors, gas sensors, pressure sensors, ultrasonic sensors, etc. The middle
layer receives the signals from the hardware layer and transfers these signals
to the application layer through communication mediums like GSM, GPRS,
etc. Finally, the application layer receives the input from the middle layer,
analyzes the received input (comparing the value with pre-defined/reference
value), stores it, and communicates to the users using human-machine
interface (HMI).

3.4 Smart Monitoring using IoT for the Low Voltage

Distribution System
Electricity is one of the basic needs of people in the modern world. It is
very important to have quality and reliability of supply distributed to the
consumer’s equipment in the distribution system. To maintain the quality
and reliability of the supply, the following areas should be improved in a
distribution system:
• identifying the faults that occurred at the instant;
• clearing the fault in lesser time;
• maintaining the proper database of fault and outage event;
• monitoring the quality of power;
• maintaining the voltage drop and PQ variation within the acceptable
tolerance limits.
The typical block diagram of IoT-based monitoring of low voltage distribu-
tion systems is shown in Figure 3.2.
It comprises voltage and current transformers/sensors, microcontroller,
and communication medium (transmitter) at the remote site location where
the actual monitoring has been carried out. The receiver receives the input
signal from the remote site location at the central monitoring and control
center, processes and analyzes it, stores it, and communicates to the users
through the monitor, alarm signal, warning signal, etc.
From the literature, many wireless communication technologies are used
to communicate/transfer the data between remote sites and central monitoring
locations [28]. The following wireless communication technologies are
widely used for communication:
66 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

Figure 3.2 Typical block diagram of IoT-based monitoring of low voltage distribution
system.

1. Bluetooth communication;
2. Wi-Fi communication;
3. ZigBee communication;
4. GPS communication;
5. GPRS communication;
6. GSM communication.
1. Bluetooth Communication:
The Bluetooth communication technology is mostly used to establish com-
munication between mobile phones for data transfer in a short distance. This
method has a limited connectivity coverage area and is insecure, and free
from interference [27].
2. Wi-Fi Communication:
Like Bluetooth, Wi-Fi communication technology is also used for connection
between multiple equipment at lesser coverage distance. This method is
widely used for providing the Internet connection to the devices.
3. ZigBee Communication:
The wireless ZigBee communication is used for networking and connectivity
between the devices based on IEEE Std 802.15.4. This method is more
suitable for connecting many devices due to its network structure flexibility.
3.5 Power Quality Monitoring of Low Voltage Distribution System – Case Study 67

4. GPS Communication:
The global positioning system, in short form GPS, is satellite-based wireless
communication technology. This method offers a wider coverage range with
an accuracy range of +/-10 m.
5. GPRS Communication:
The general packet radio service, in short form GPRS, is widely used in
mobile phone communication across the globe. This method enables the
wireless connection from anywhere within the coverage range.
6. GSM Communication:
The global system for mobile, in short form GSM, is a widely used
mobile communication method. It has the frequency range of either 900 or
1800 MHz.

3.5 Power Quality Monitoring of Low Voltage Distribution

System – Case Study
The single line diagram (SLD) of the circuit is considered for the PQ
measurement as shown in Figure 3.3.

Figure 3.3 SLD.

68 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

Figure 3.4 Experimental setup.

The experimental setup for the SLD shown in Figure 3.3 with the smart
monitoring system is shown in Figure 3.4. It consists of three-phase, 415-V,
50-Hz power source, loads (three numbers of single-phase lamps), microcon-
troller, GSM modem, current transformer, and potential transformer.
The control block diagram of the experimental setup shown in Figure 3.4
is shown in Figure 3.5. The PIC16F877A receives the load (lamps) voltage
and the current using the potential transformer and the current transformer,
respectively. It compares the voltages and current magnitudes with respect to
reference values (limits). If the limits are violated, an alarm is triggered in
addition to a pop-up on the LCD screen. The LCD and buzzer are used to
communicate the deviations in the measured voltage and current to the users
locally, while GSM-based communication is used for remote monitoring.
3.5 Power Quality Monitoring of Low Voltage Distribution System – Case Study 69

Figure 3.5 Block diagram.

GSM modem can accept any GSM network as a SIM card and just like a
mobile phone with its unique phone number. This modem uses the RS232
port for communication with embedded applications. The SIM900A is a
complete dual-band GSM/GPRS solution in an SMT module featuring an
industry-standard interface; the SIM800 delivers GSM/GPRS 900/1800 MHz
performance for voice, SMS, data, and fax in a small form factor and with
low power consumption. The LCD can perform the local monitoring; remote
monitoring can be done through the website and mobile via SMS.

3.5.1 Undervoltage
An undervoltage is defined as a reduction of RMS voltage magnitude for over
a 1-minute time duration below specified limits. The nominal voltage is 240
V with a tolerance of ±10%. Hence, the reference voltage (lower limit) is
set as 216 V. If the monitored voltage is less than 216 V for over 1 minute,
the microcontroller sets off an alarm using the buzzer in addition to a pop-
up message on the LCD as “undervoltage.” The flow chart of undervoltage
identification is shown in Figure 3.6.
If the consumer’s phase voltage reduces below the reference value, an
alert notification is sent via SMS using the GSM module. Additionally,
the same information is recorded and communicated through a webpage.
Figure 3.7 illustrates the notifications received by the customer through
SMS text.

3.5.2 Overvoltage
An overvoltage is defined as the increase in RMS voltage magnitude over
a 1-minute time duration. The nominal voltage is 240 V with a tolerance
of ±10%. Hence, the reference voltage is set as 264 V. If the measured
70 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

Figure 3.6 Flow chart of undervoltage identification.

Figure 3.7 SMS alert for undervoltage.

voltage is higher than 264 V for over 1 minute, then the microcontroller sets
off an alarm using the buzzer in addition to a pop-up message on the LCD
as “overvoltage.” The flow chart of overvoltage identification is shown in
Figure 3.8.
Figure 3.9 illustrates the SMS received by the consumer during the advent
of an upper limit violation on the voltage.
3.5 Power Quality Monitoring of Low Voltage Distribution System – Case Study 71

Figure 3.8 Flow chart of overvoltage identification.

Figure 3.9 SMS alert for overvoltage.

3.5.3 Interruption
An interruption is defined as a decrease in RMS voltage magnitude less than
0.1 pu. If the time duration is less than 1 minute, then it is called a momentary
interruption, and if it is 1 minute, it is called a sustained interruption. The
nominal voltage is 240 V with a tolerance of ±10%. Hence, the reference
voltage is set as 24 V. If the measured voltage is lesser than 24 V for less than
1 minute which is a momentary interruption and over 1-minute time duration
which is sustained interruption, then the microcontroller will send an alarm
through the buzzer and pop-up through LCD as an interruption. The flow
chart of interruption identification is shown in Figure 3.10.
72 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

Figure 3.10 Flow chart of interruption identification.

Figure 3.11 SMS alert for voltage interruption.

The end-user will receive the SMS in mobile phone for voltage interrup-
tion alert as shown in Figure 3.11.

3.5.4 Overload in Branch Circuit

An overload may be termed as the increase in RMS current drawn by the load
exceeding its nominal value. As a result of the higher load current, the current
flow in the branch circuits is increased. This increase in current is known as
3.5 Power Quality Monitoring of Low Voltage Distribution System – Case Study 73

overloading of branch circuits. For example, the nominal current of the load is
100 A. If the current flow in this branch is over 100 A, it is called overloading.
Up to 10% of the overloading (on the nominal value) is acceptable. To
identify the branch circuit’s overloading, the reference current is set as 0.2
A. If the measured current is higher than 0.2 A, the microcontroller will send
an alarm through the buzzer and pop-up through the LCD as overcurrent. The
flow chart of overvoltage identification is shown in Figure 3.12.
The end-user will receive the SMS in mobile phone for overcurrent alert
as shown in Figure 3.13.

Figure 3.12 Flow chart of overvoltage identification.

Figure 3.13 SMS alert for overvoltage.

74 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

Figure 3.14 Screenshot of real-time monitoring through web.

Through this IoT-based monitoring system, consumers will immediately

be notified about the current PQ parameters through the mobile communi-
cation system via SMS. This enables supervision/monitoring of the real-time
PQ parameters. The real-time web-based monitoring of low voltage smart dis-
tribution can be monitored anywhere in the world. The data can be accessed
through the web link http://vehiccleiot.in/distribution/view.php. Figure 3.14
enumerates the real-time parameters displayed on the webpage.

3.6 Conclusion
In the modern world, electricity is one of the basic needs of human beings.
Supplying the quality and high reliable power supply is the most impor-
tant concern for all the distribution companies. This chapter discusses the
PQ monitoring at a low voltage smart distribution system through contin-
uous monitoring with advanced features like analyzing the PQ parameter,
storing/keeping the proper database, etc. It also sends the alert notification
to the users via SMS and real-time monitoring on the web page. Power
distribution companies can supply a more reliable power supply to their
customer by using this monitoring system.
 References 75

References
[1] P. Sivaraman and C. Sharmeela, Introduction to electric distribution
system. In B. Kahn, H. H. Alhelou, and Ghassan (Eds.), Handbook of
Research on New Solutions and Technologies in Electrical Distribution
Networks. Hershey, PA: IGI Global, 2020, pp. 1–31.
[2] P. Sivaraman and C. Sharmeela, Existing issues associated with electric
distribution system. In B. Kahn, H. H. Alhelou, and Ghassan (Eds.),
Handbook of Research on New Solutions and Technologies in Electrical
Distribution Networks. Hershey, PA: IGI Global, 2020, pp. 1–31.
[3] T.S. Somkuwar and M.G. Panjwani, Review paper on electrical distri-
bution line monitoring, International Journal of Advance Research in
Computer and Communication Engineering, 2015.
[4] P. Sivaraman, D. Gunapriya, K. Parthiban, and S. Manimaran, Hybrid
fuzzy PSO algorithm for dynamic economic load dispatch, Journal
of Theoretical and Applied Information Technology, vol. 62, no. 3,
pp. 794–799, April 2014.
[5] P. Sivaraman, S. Manimaran, K. Parthiban, and D. Gunapriya, PSO
approach for dynamic economic load dispatch problem, International
Journal of Innovative Research in Science, Engineering and Technology,
vol. 3, no. 4, pp. 11905–11910, April 2014.
[6] P. Sivaraman, C. Sharmeela, R. Mahendran, and A. Thaiyal Nayagi,
Basic Electrical and Instrumentation Engineering. Hoboken, NJ: John
Wiley & Sons, 2020.
[7] P. Sivaraman, C. Sharmeela, and D.P. Kothari, Enhancing the voltage
profile in distribution system with 40 GW of solar PV rooftop in Indian
grid by 2022: A review, In Proceedings of 1st International Confer-
ence on Large Scale Grid Integration of Renewable Energy in India,
September 2017.
[8] IEEE Std 141-1993, IEEE Recommended Practice for Electrical Power
Distribution for Industrial Plants.
[9] IEEE Std 1100-2005, IEEE Recommended Practice for Powering and
Grounding Electronic Equipment.
[10] IEEE Std 1159-2019, IEEE Recommended Practice for Monitoring
Electric Power Quality.
[11] P. Sivaraman and C. Sharmeela, Power quality problems associated
with electric vehicle charging infrastructure. In P. Sanjeevikumar, C.
Sharmeela, and J. B. Holm-Nielsen (Eds.), Power Quality in Modern
Power Systems. Amsterdam: Elsevier, 2021.
76 Power Quality Monitoring of Low Voltage Distribution System Toward Smart

[12] R. Mahendran, P. Sivaraman, and C. Sharmeela, Three phase grid inter-

faced renewable energy source using active power filter, In Proceedings
5th International Exhibition & Conference, GRIDTECH 2015, New
Delhi, India, April 2015, pp. 77–85.
[13] P. Sivaraman and C. Sharmeela, Power quality and its characteristics, In
P. Sanjeevikumar, C. Sharmeela, and J.B. Holm-Nielsen (Eds.), Power
Quality in Modern Power Systems. Cambridge, MA: Academic Press,
2021.
[14] P. Sivaraman and C. Sharmeela, Power system harmonics, In P.
Sanjeevikumar, C. Sharmeela, and J.B. Holm-Nielsen (Eds.), Power
Quality in Modern Power Systems. Cambridge, MA: Academic Press,
2021.
[15] P. Sivaraman and C. Sharmeela, Battery energy storage system address-
ing the power quality issue in grid connected wind energy conversion
system, In Proceedings of the 1st International Conference on Large-
Scale Grid Integration Renewable Energy in India, New Delhi, India,
September 2017.
[16] P. Sanjeevikumar, C. Sharmeela, J.B. Holm-Nielsen, and P. Sivaraman,
Power Quality in Modern Power Systems. Cambridge, MA: Academic
Press, 2021.
[17] C. Sharmeela, P. Sivaraman, P. Sanjeevikumar, and J.B. Holm-Nielsen,
Microgrid Technologies. Hoboken, NJ: John Wiley & Sons, 2021.
[18] B. Li, B.Z.J. Guo, and H. Yao, Study of cognitive radio based wireless
access communications of power line and substation monitoring sys-
tem of smart grid, In Proceedings of the International Conference On
Computer Science and Service System, IEEE, 2012.
[19] V.C. Gungor, B. Lu, and G.P. Hancke, Opportunities and challenges
of WSN in smart grid, IEEE Transactions on Industrial Electronics,
vol. 57, no. 10, October 2010.
[20] J. Gutiérrez, J.F. Villa-Medina, A. Nieto-Garibay, and M.. Porta-
Gándara, Automated irrigation system using a wireless sensor net-
work and GPRS module, IEEE Transactions on Instrumentation and
Measurement, vol. 63, no. 1, January 2014.
[21] P. Sivaraman and C. Sharmeela, IoT-based battery management sys-
tem for hybrid electric vehicle, In A. Chitra, P. Sanjeevikumar, J. B.
Holm-Nielsen, and S. Himavathi (Eds.), Artificial Intelligent Techniques
for Electric and Hybrid Electric Vehicles. Beverly, MA: Scrivener
Publishing LLC, 2020, pp. 1-16.
 References 77

[22] C. Sharmeela, P. Sivaraman, and S. Balaji, Design of hybrid DC mini-

grid for educational institution: Case study, Lecture Notes in Electrical
Engineering, vol. 580, pp. 125-134, 2020.
[23] L. Atzori, A. Iera, and G. Morabito, The Internet of Things: A survey,
Computer Networks, vol. 54, pp. 2787–2805, 2010.
[24] N. Harish, V. Prashal, and D. Sivakumar, IoT based battery management
system, International Journal of Applied Engineering and Research,
vol. 13, pp. 5711–5714, 2018.
[25] S. Balaji and S. Usa, Characterization of polypropylene under DC-
polarity reversal, In Proceedings of the 2019 IEEE 4th International
Conference on Condition Assessment Techniques in Electrical Systems
(CATCON), 2019, pp. 1-5, doi: 10.1109/CATCON47128.2019.CN0064.
[26] N. Sharma, M. Shamkuwar, and I. Singh, The history, present, and future
with IoT, Intelligent Systems Reference Library, pp. 27–51, 2018.
[27] P. Sivaraman and C. Sharmeela, Solar micro-inverter, In J. Zbitou, C.
Pruncu, and A. Errkik (Eds.), Handbook of Research on Recent Devel-
opments in Electrical and Mechanical Engineering. Hershey, PA: IGI
Global, 2020, pp. 283–303.
[28] M. Weiser, The computer for the 21st century, ACM SIGMOBILE
Mobile Computing and Communications Review, vol. 3, no. 3, pp. 3–11,
1999.
[29] M. Ersue, D. Romascanu, J. Schnwlder, and A. Sehgal, Management of
networks with constrained devices: Use cases, no. 7548. RFC Editor,
May 2015.
[30] S. Balaji, V.S. Devendran, and S. Chenniappan, Electromagnetic anal-
ysis of underground cables for upcoming smart city: Case study, In
Proceedings of the 2020 IEEE International Conference on Power and
Energy (PECon), 2020, pp. 259-263, doi: 10.1109/PECon48942.2020.
9314306.
[31] T. de Vass, H. Shee, and S. Miah, IoT in supply chain management.
Opportunities and challenges for businesses in early Industry 4.0 con-
text, Operations and Supply Chain Management: An International
Journal, pp. 148–161, 2021.
[32] H. Tang, S. Lu, G. Qian, J. Ding, Y. Liu, and Q. Wang, IoT-based signal
enhancement and compression method for efficient motor bearing fault
diagnosis, IEEE Sensors Journal, vol. 21, pp. 1820-1828, 2021.
[33] S. Balaji and T. Ayush, Placement of energy exchange centers and
bidding strategies for smartgrid environment, 2020 Wiley, Smart Grid
Technologies. ISBN: 978-1-119-71087-5.
 4
Health Monitoring of a Transformer
in a Smart Distribution System using IoT

P. Sivaraman1 , C. Sharmeela2 , and P. Sanjeevikumar3

1 VestasTechnology R&D Chennai Pvt Ltd, India
2 Anna University, India
3 Aarhus University, Denmark

E-mail: [email protected]; [email protected];

[email protected]

Abstract
Electricity places an important role in the modern world. Transformers are
one of the key piece of equipment in modern power systems. Because of
various reasons, the failure of transformers can happen unexpectedly, result-
ing in a power supply outage to the end-users. Distributed generation (DG)
allows the bidirectional power flow in the system with higher integration of
distributed renewable energy sources (RES), especially solar photovoltaics.
It is possible to install the solar PV system in the range between 10%
and 120% of the distribution transformer rating under the DG. Failures of
distribution transformers affect the power supply to end-users and the DG
integration into the distribution grid. Consistent monitoring of distribution
transformers by means of physical inspection is one of the major concerns
of distribution companies/end-users to prevent unexpected failure. Currently,
distribution transformer parameters like state of working (i.e., online/offline)
is monitored, and other parameters like oil level, oil temperatures, winding
temperature, etc., are not monitored from the remote end. This chapter
discusses the monitoring of health conditions of a laboratory prototype
250-VA single phase oil-immersed transformer using Internet of Things (IoT)
technology to achieve the conventional distribution systems toward a smart
distribution system.

79
80 Health Monitoring of a Transformer in a Smart Distribution System using IoT

Keywords: Distribution transformer, distribution system, distributed genera-

tion (DG), Internet of Things (IoT), smart grid.

4.1 Introduction
Conventionally, electricity is generated by large power plants at remote loca-
tions and the power is then transferred to the load center [3]. Power flow in the
conventional power system is unidirectional, and transformers here play an
important role. The modern electric power systems allow bidirectional power
flows with smart communication devices. The concept of distributed gener-
ation (DG) introduced localized power generation closer to the load center
and it allows bidirectional power flow into the system. It is possible to install
the solar PV system in the range between 10% and 120% of the distribution
transformer rating under the DG [22]. Failures of distribution transformers
affect the power supply to end-users and the DG integration into the distri-
bution grid. The end-users always need high reliability in input power supply
from the distribution company for trouble-free operation of their loads. The
supply’s reliability depends on the performance of the equipment in the sys-
tem, particularly transformers [23–25]. An unexpected failure of distribution
transformers results in power outages to the end-users, reducing the reliability
of the power supply. Usually, power or distribution transformers have a
20–35-year design lifetime, and this can be extended up to 60 years with
good maintenance in practice. Failure of distribution transformers is common
because of poor maintenance, overloading and harmonics, lightning over cur-
rent/voltages, internal/external loose connections, and internal/external short
circuits. Hence, monitoring transformers are essential for high supply relia-
bility to the end-users in the distribution system towards the smart distribution
system. Presently, PLC-based systems are used for monitoring transformers
locally at the site location. In a distribution system in metro cities, many dis-
tribution transformers are in place/operation, and it is difficult to monitor all
of them locally. Hence, it is essential to monitor all the distribution transform-
ers in a common/centralized location. An Internet of Things (IoT) based tech-
nology enables the monitoring of many distribution transformers used in the
distribution system at a commonplace, i.e., central monitoring control center
or main control center. An IoT-based system has been designed so that it will
continuously monitor the essential parameters of the transformers through-
out its day-to-day operation. The controller is used to compare the various
measured parameters (like the voltage, current, oil level, or temperature) with
the reference values. If any values exceed the reference value, it will give the
notification and/or alarm to prevent damage/failure/tripping.
 4.2 Introduction to the Transformer 81

4.2 Introduction to the Transformer

A transformer is static electrical equipment, and it is used to transfer energy
between two or more circuits using electromagnetic induction. The trans-
formers transform the voltage level from one voltage to another voltage at
the same operating frequency [2, 3]. It comprises two windings, namely
primary windings and secondary windings [7]. These two windings are not
electrically connected but are connected through the magnetic core. It works
on the principles of Faraday’s law of electromagnetic induction.
The main parts of the transformer are as follows:
1. core (made up of silicon steel);
2. the primary winding (made up of copper or aluminum);
3. secondary winding (made up of copper or aluminum.
The installation of a distribution transformer of 100 kVA, 11 kV/0.433 kV,
50 Hz used to power the residential and agriculture loads in a rural village in
India is shown in Figure 4.1.

Figure 4.1 The installation of 100-kVA distribution transformer.

82 Health Monitoring of a Transformer in a Smart Distribution System using IoT

4.3 Failure of the Distribution Transformer

The life of the distribution transformer is significantly affected/reduced due
to overloading [6]. This results in unexpected tripping and/or failures of
the distribution transformer. Hence, it leads to power supply interruption to
various end-users connected to it, and the system’s reliability is questionable
[8]. The most common causes for failure of distribution transformer are
[13–15] as follows:
1. overloading;
2. improper maintenance;
3. electrical faults (both internal and external);
4. overvoltage is due to lightning surge;
5. loose connections;
6. insulation degradation;
7. environmental factors like moisture, etc.;
8. overheating because of harmonics and unbalanced loading;
9. accident to its physical infrastructure by vehicles, etc.;
10. accident to its physical infrastructure by vehicles, etc.;
11. presence of oxygen;
12. solids in the insulating oil.
The failure and causes of distribution transformer in Punjab State Power
Corporation Limited in India between 2010 and 2015 are given in Figure 4.2.
Three hundred and forty-eight distribution transformers failed in this
period [1].
In order to avoid the unexpected failure or outage of the distribution trans-
former, it is essential to monitor the distribution transformer and maintain
them with care.

4.4 Transformer Health Monitoring System through IoT

The advantages of monitoring the transformer through IoT are as follows
[4, 5].
• Monitoring many distribution transformers manually is difficult, and
IoT-based monitoring is more convenient for distribution operators or
companies.
• More reliable as compared to local monitoring.
• Economically, IoT-based monitoring is cheaper as compared with local
monitoring.
 4.4 Transformer Health Monitoring System through IoT 83

Figure 4.2 Failure and causes of distribution transformer.

Figure 4.3 shows the typical block diagram of IoT-based health monitor-
ing of oil-immersed distribution transformer.
The IoT-based system monitors the oil-immersed transformer essential
parameters throughout its operation continuously. They are oil temperature,
winding temperature, oil level, voltage, and current flow. The sensors are used
to measure these parameters and communicate the measured values to the
microcontroller.

4.4.1 Winding and Oil Temperature Sensor

Due to the overloading of the transformer, the winding temperature and oil
temperature get increased, resulting in transformer failure or outage. Hence,
it is essential to monitor the temperature inside the transformer [20]. If
winding or oil temperature increases above the allowed temperature, it is
mandatory to reduce the temperature to avoid transformer failure or outage.
The temperature sensor is used to measure the temperature inside the trans-
former cubicle, i.e., winding temperature and oil temperature. This sensor
will measure the transformer winding and oil temperature and communicate
it to the microcontroller.
84 Health Monitoring of a Transformer in a Smart Distribution System using IoT

Figure 4.3 Block diagram of IoT-based health monitoring of oil-immersed distribution

transformer.

4.4.2 Oil Level Monitoring Sensor

The transformer oil is generally used for two main reasons. They are insu-
lation and cooling. Transformer oil levels will be reduced due to leakages
and evaporation due to overheating. It is essential to maintain the transformer
oil level at the required level during the normal operation. In case the oil
level is reduced below the required level, it affects both cooling as well as
insulation. So, it is not safe to operate the transformer with a lesser oil level on
it. Hence, it is essential to monitor the oil inside the transformer. An ultrasonic
sensor is used to measure the oil level and communicate the oil level to the
microcontroller [21].

4.4.3 Current Sensor and Voltage Sensor

The current is drawn from the source through the transformer during the oper-
ation of electrical equipment (i.e., loads). Whenever there is an overloading
 4.4 Transformer Health Monitoring System through IoT 85

condition, the transformer also gets overloaded. Due to overloading, the

temperature inside the transformer is also increasing. Hence, it is essential
to monitor the transformer loading condition. Current sensors or current
transformers measure the current flowing through the transformer and give
the output. Based on the current flow through the transformer, the output
signal is proportional to it.
Similarly, a voltage sensor is used to measure the voltage at which
the transformer operates and communicate the output signal to the
microcontroller.

4.4.4 Microcontroller
A microcontroller is a device used to compare the various measured quan-
tities, such as oil temperature, winding temperature, oil level, voltage, and
current with the pre-defined or reference values. It will communicate with a
local LCD for local monitoring and remote monitoring through IoT. If any
of the measured values exceeds the reference value, it will initiate the alarm
and/or notification.

4.4.5 LCD or Monitor

The LCD or monitor is used as a human-machine interface for displaying the
various measured quantities for human understanding. The LCD or monitor
is used for local monitoring at the remote site location.

4.4.6 Communication System

A communication system transfers the various measured quantities, like oil
level, winding temperature, etc., to a central monitoring location from a
remote site location. Many proven wireless communication technologies are
available in the literature to transfer the data between two or more devices
[9–12]. Some of the wireless communication technologies for long-distance
communications are as follows:
1. Global system for mobile communication (GSM)
2. General packet radio service (GPRS)
3. Global positioning system (GPS)
The GSM is a mobile phone communication system used to communicate
between multiple devices [16–19]. It is used to communicate the remote
site location measuring devices with central monitoring and control centers.
GSM acts as a transmitter at remote site locations and transfers the measured
quantities to central monitoring. In central monitoring and control location,
86 Health Monitoring of a Transformer in a Smart Distribution System using IoT

GSM acts as a receiver and will receive the signal from the remote site
location.

4.4.7 Central Monitoring and Control

The central monitoring and control center are commonplace to monitor the
various equipment conditions and their parameters, as shown in Figure 4.3. It
received the measured quantities from various equipment in multiple remote
site locations. It displays the various measured parameters of each equip-
ment per remote site location for easy user understanding. If any measured
quantities exceed the reference value, it will initiate the alarm and/or notifi-
cation. The central monitoring and control center operator must acknowledge
the alarm/notification and necessary action should be taken to prevent the
transformer failure.

4.5 Case Study

It showed the block diagram of the laboratory setup of IoT-based transformer
monitoring system in Figure 4.4.

Figure 4.4 Block diagram of IoT-based transformer health monitoring.

 4.5 Case Study 87

Figure 4.5(a) shows the experimental hardware setup, and the single-
phase transformer is shown in Figure 4.5(b).

Figure 4.5(a) Hardware experimental setup.

Figure 4.5(b) Single phase transformer.

88 Health Monitoring of a Transformer in a Smart Distribution System using IoT

The monitoring of various transformer parameters through IoT for nor-

mal operation is shown in Figure 4.6(a) and that of high winding and oil
temperature is shown in Figure 4.6(b).

Figure 4.6(a) Online screenshot of the web page – normal operation.

Figure 4.6(b) Online screenshot of the web page – high winding temperature.
 References 89

The system operator must acknowledge the alarm/notification; necessary

preventive action must be taken to avoid the transformer failure.

4.6 Conclusion
The IoT-based health monitoring of the transformer is presented in this
chapter. The essential parameters such as oil level, winding temperature,
oil temperature, ambient temperature rise, and transformer loading can be
monitored continuously throughout its operation. If any parameter exceeds
the reference value, it will give the abnormality alarm/notification. Necessary
preventive action needs to be taken immediately to prevent the transformer
from failing. The IoT-based method is very convenient for monitoring many
distribution transformers in urban distribution systems in metro cities like
Chennai. As compared with manual monitoring, this method is much useful
and reliable.

References
[1] J. Singh, S. Singh, and A. Singh, Distribution transformer failure modes,
effects, and criticality analysis (FMECA), Engineering Failure Analysis,
pp. 180–191, vol 99, 2019.
[2] R. Murugan and R. Ramasamy, Failure analysis of power transformer
for effective maintenance planning in electric utilities, Engineering
Failure Analysis, pp. 182–192, vol 55, 2015.
[3] P. Sivaraman, C. Sharmeela, R. Mahendran, and A. T. Nayagi, Basic
Electrical and Instrumentation Engineering, John Wiley & Sons, 2020.
[4] A.-R. AI-Ali, A. Khaliq, and M. Arshad, GSM-based distribution
transformer monitoring system, IEEE MELECON 2004, Croatia, May
12–15, 2004, pp. 999–1002, vol. 3.
[5] D. Duttachowdhury, V. Patil, A. Parab, R. Patel, and K. Niikum, Trans-
former monitoring and control using IoT, IOSR Journal of Engineering,
pp. 40–43, vol. 10, Mar. 2018.
[6] R. Singh and A. Singh, Causes of failure of distribution transformers
in India, 9th International Conference on Environment and Electrical
Engineering (EEEIC 2010), 2010, pp. 388–391.
[7] P. Sivaraman and C. Sharmeela, Introduction to electric distribution
system. In B. Kahn, H.H. Alhelou, & Ghassan (Eds.), Handbook of
Research on New Solutions and Technologies in Electrical Distribution
Networks, pp. 1–31. Hershey, PA: IGI Global, 2020.
90 Health Monitoring of a Transformer in a Smart Distribution System using IoT

[8] P. Sivaraman and C. Sharmeela, Existing issues associated with electric

distribution system. In B. Kahn, H. H. Alhelou, & Ghassan (Eds.),
Handbook of Research on New Solutions and Technologies in Electrical
Distribution Networks, pp. 1–31, Hershey, PA: IGI Global, 2020.
[9] B. Li, B. Zhang J. Guo, and H. Yao, Study of cognitive radio based
wirelesss access communication of power line and substation monitor-
ing system of smart grid, International Conference on Computer Science
and Service System, IEEE, 2012.
[10] P. Sivaraman and C. Sharmeela, IoT-based battery management system
for hybrid electric vehicle, In A. Chitra, P. Sanjeevikumar, J.B. Holm-
Nielsen, & S. Himavathi (Eds.), Artificial Intelligent Techniques for
Electric and Hybrid Electric Vehicles, pp. 1–16, Beverly, MA: Scrivener
Publishing LLC, 2020.
[11] L. Atzori, A. Iera, and G. Morabito, The Internet of Things: A survey,
Computer Networks, pp. 2787–2805, vol 54, 2010.
[12] N. Harish, V. Prashal, and D. Sivakumar, IOT based battery management
system, International Journal of Applied Engineering and Research,
pp. 5711–5714, vol 13, 2018.
[13] P. Rawal, J. Pratipalsinh, and V. Devdhar, Distribution transformer fail-
ure analysis in Gujarat DISCOM, International Journal of Latest Tech-
nology in Engineering, Management & Applied Science, pp. 48–51,
vol 6, Jun. 2017.
[14] R. Gayathri, S. Nanthini, M. Maheshwari, and S. Akila, Transformer
health monitoring system based Internet of Things, International Journal
of Pure and Applied Mathematics, pp. 959–964, vol 119, no. 15, 2018.
[15] S. Karthik, B. Ezhilkumaran, V. Vaibhav, and S. Kishorekumar, Real
time health monitoring system of transformer using IoT, International
Journal of Pure and Applied Mathematics, pp. 921–925, vol 119, no. 15,
2018.
[16] N. Sharma, M. Shamkuwar, and I. Singh, The history, present and future
with IoT, Intelligent Systems Reference Library, pp. 27–51, 2018.
[17] A.A. Sonune, M.S. Talole, S.V. Sonkusale, A.A. Akotkar, P.S. Jaiswal,
and V.N. Gayki, Condition monitoring of distribution transformer using
IoT, International Journal of Engineering Research & Technology,
pp. 335–338, vol 9, no. 06, Jun. 2020.
[18] A.K. Gautam, P. Gharpure, and P.V. Gawande, GSM based remote
monitoring of parameters of transformer, International Journal of Mod-
ern Trends in Engineering and Research, pp. 409–412, vol 03, no. 03,
Mar. 2016.
 References 91

[19] C.D. Oancea, GSM infrastructure used for data transmission, 7th Inter-
national Symposium on Advanced Topics in Electrical Engineering
(ATEE), May 2011, pp. 1–4.
[20] P. Kore, V. Ambare, A. Dalne, G. Amane, S. Kapse, and S. Bhavarkar,
IoT based distribution transformer monitoring and controlling system,
International Journal of Advance Research and Innovative Ideas in
Education, pp. 122–126, vol 5, no. 2, 2019.
[21] P. Dimpal, P. Mayank, G. Pratik, K. Pramod, and D. Kshitija, Monitor-
ing and controlling of transformer using IoT, International Journal of
Advance Research and Innovative Ideas in Education, pp. 1250–1256,
vol 6, no. 5, 2020.
[22] P. Sivaraman, C. Sharmeela, and D.P. Kothari, Enhancing the voltage
profile in distribution system with 40 GW of solar PV rooftop in Indian
grid by 2022: A review, 1st International Conference on Large Scale
Grid Integration of Renewable Energy in India, Sep. 2017.
[23] P. Sivaraman and C. Sharmeela, Power quality and its characteristics, In
P. Sanjeevikumar, C. Sharmeela, & J. B. Holm-Nielsen (Eds.), Power
Quality in Modern Power Systems, Cambridge, MA: Academic Press,
2021.
[24] P. Sivaraman and C. Sharmeela, Power system harmonics, In P. Sanjee-
vikumar, C. Sharmeela, & J. B. Holm-Nielsen (Eds.), Power Quality in
Modern Power Systems, Cambridge, MA: Academic Press, 2021.
 5
Introduction To Machine Learning
Techniques

Saniya M. Ansari1 , Ravindra R. Patil2 , Rajnish Kaur Calay3 ,

and Mohamad Y. Mustafa3
1E & TC Department, Dr D Y Patil School of Engineering (DYPSOE), India
2 PhD Scholar, Department of Building, Energy and Material Technology,
UiT The Arctic University of Norway, Norway
3 Department of Building, Energy and Material Technology, UiT The Arctic

University of Norway, Norway

E-mail: [email protected]; [email protected];
[email protected]; [email protected]

5.1 Why and What is Machine Learning?

Machine learning (ML), an integral part of artificial intelligence (AI), has
emerged as one of the most important technologies in today’s world. So what
do these terms mean? In this chapter, the concept and involved techniques in
ML have been elucidated.
AI has the potential to make decisions like humans and consists of
standard rules encoded in the form of computer programs. Computers and
computational techniques are included in the central field of computer sci-
ence. The ML algorithms learn from the processed data and build predictions
on that data. It is possible to alter the action and reaction of large data in ML
to achieve higher adaptability, efficiency, and scalability. Next, deep learning
is also a specific kind of advanced ML which uses deep neural networks
to solve distinct problems. The deep neural network implements minuscule
computations on numerous layers for performing duties like humans [1].
The American pioneer in the AI and computer gaming area, Arthur Samuel,
first introduced the ML phrase in 1959 and defined it as follows: “it gives
computers the ability to learn without being explicitly programmed.”

93
94 Introduction To Machine Learning Techniques

Then, Tom Mitchell in 1997 defined ML as, “A computer program is

said to learn from experience E concerning some task T and some perfor-
mance measure P, if its performance on T, as measured by P, improves with
experience E.”

5.1.1 Phrases in Machine Learning

• Model – It is the key element of ML technology and is trained by
utilizing an ML algorithm to generate outturns.
• Algorithm – It is a bunch of rules and computational techniques to
return meaningful details.
• Training data – It is the crucial aspect in the ML algorithms which
comprises features, patterns, and key trends.
• Testing data (TD) – It helps to assess the accuracy of the trained model.
• Predictor variable (PV) – It is an attribute of the data which can
forecast the outturn.
• Response variable (RV) – It is an attribute of the output variable and
the PV(s) should predict it.

5.1.2 Steps Involved in Machine Learning Practices

The ML practices consist of the following steps:
• defining the problem statement and following the objectives;
• collecting the data;
• processing the data to return meaningful information;
• selecting algorithms;
• training of the model;
• assessment and optimization of the model;
• tuning of the algorithm;
• generating outturns.

5.1.3 Properties of Data

The detailing of data properties is listed in Table 5.1.

Table 5.1 Detailing of data properties.

Data Property Detailing
Velocity Data creation and streaming rate
Volume Data scale
Veracity Assurance and truthiness in data
Variety Contrasting types of data images, audios, text, videos
Value Significance of data in the form of details
 5.1 Why and What is Machine Learning? 95

5.1.4 Real-World Applications of Machine Learning

Figure 5.1 presents the real-world applications of machine learning
techniques.

Figure 5.1 Real-world applications of machine learning.

96 Introduction To Machine Learning Techniques

5.2 Classification of Machine Learning Techniques

ML techniques are classified into three significant types to be contingent on
the mode of the learning [5]. The ML techniques are as follows:
• supervised learning;
• unsupervised learning;
• reinforcement learning (RL).
Let us dive deep into the classified techniques of ML as shown in Figure 5.2.

5.2.1 Supervised Learning

It is the classified learning technique of ML which learns from the proficiently
“Labeled” data to yield the output. It means that input data have beforehand
labeled with intended output. It comes to know that labeled data given to the
algorithms play a crucial role in training and contain all features to generate
the final output [8]. This data is nothing to a supervisor in the learning
process and similar to the student who acquires knowledge from the teacher’s
supervision.

Figure 5.2 Classification of machine learning techniques.

 5.2 Classification of Machine Learning Techniques 97

Figure 5.3 Workflow of supervised learning technique.

The supervised learning algorithms have a goal of coming upon a map-

ping function to map the input variable (x) and the output variable (Y) [3].
The following equation shows the correlation:

Y = f (x). (5.1)

In the realistic scenario, supervised learning is applied for different

applications such as classification of images, fraud detection, assessment of
risk, prediction of the score, spam email detection, etc. The workflow of the
supervised learning is given in Figure 5.3.
In an example, consider we have an image of distinct types of flowers and
the supervised learning model has a goal to recognize the flowers and classify
them appropriately. For this, we will train the ML model for different features
of flowers and will test on the new dataset (testing dataset) of flowers. So, ML
trained model by a suitable algorithm will identify the flower and predict the
outturn.
There are two types of supervised ML algorithms.

5.2.1.1 Classification
If an output variable is categorical, then a classification algorithm is applied
to get intended results. This means it consists of classes for respective data
[9, 10].
98 Introduction To Machine Learning Techniques

In classification, take a look at the following equation:

y = f (x). (5.2)

Here, y is the categorical output.

This means y (discrete output function) is mapped to x (input variable).
Following are some crucial classification algorithms:
• logistic regression;
• random forest;
• support vector machines;
• decision trees.

5.2.1.2 Regression
For relating input variable and output variable, the regression algorithms
come into services and are applied for the prediction of continuous output
variables (output is always real value) [11]. Following are some favored
regression algorithms:
• linear regression;
• non-linear regression;
• Bayesian linear regression;
• regression trees.
The difference between the classification algorithm and the regression algo-
rithm is given in Table 5.2.

Table 5.2 Difference between classification algorithm and regression algorithm.

Classification algorithm Regression algorithm
• These algorithms are utilized • These algorithms are utilized
with discrete data. with continuous data.
• It records the input value (a) • It records the input value (a)
per the discrete output per the continuous output
variable (b). variable (b).
• In this, it is tried to find out • In this, it is tried to find out
the decision boundary to the best fit line to predict the
classify the dataset. right output.
• It is divided into binary class • It is divided into linear and
classifiers. non-linear regression.
• In this, the output variable • In this, the output variable
must seem to be a discrete must seem to be real value.
value.
 5.2 Classification of Machine Learning Techniques 99

5.2.2 Unsupervised Learning

In an unsupervised learning technique, algorithms learn from simple exam-
ples and do not depend on certain related responses. An algorithm itself
determines the insight and hidden patterns from data. It is the same as a
human control system to learn newer things. In this way, it can be said that an
unlabeled dataset is used to train the models and permitted to proceed with
no supervision on data [12]. The regression and classification problem cannot
be solved plainly and it is only due to the unavailability of output data.
Overall, the unsupervised learning technique has an objective to search
out the basic pattern of the dataset, group the data based on equality,
and record the respective dataset in compressed format. The workflow of
unsupervised learning techniques is given in Figure 5.4.
In an example, consider we have an image of different types of flowers
and the objective is to identify flowers appropriately. For this, an unsupervised
learning algorithm will solve this problem by clustering the imagery dataset
of flowers into groups as per image similarities.
Unsupervised learning is used due to the following points.
• It helps to find important insights and hidden patterns from data.
• It is similar to the critical thinking of humans to learn newer things from
experiences.
• It works on non-labeled and non-categorized data and it is the function
why it is called an unsupervised learning technique.
• In a realistic scenario, it is not possible to have input data always with
correlating output. So, unsupervised learning helps to solve such tasks.
There are also two types of unsupervised ML algorithms.

5.2.2.1 Clustering
In clustering, the objects are grouped into clusters such as one group of
objects with higher feature similarities and one new group of objects with
lower or no feature similarities.

Figure 5.4 Workflow of unsupervised learning technique.

100 Introduction To Machine Learning Techniques

Simply, clustering analyzes the feature similarities between the objects

and makes a category of inherent data depending on it. There are differ-
ent types of clustering based on soft and hard clustering methods such
as density-based, fuzzy, distribution model-based, hierarchical, partitioning
type, etc. [13].

5.2.2.2 Association
Association is used to find correlation within dataset variables. This means
the bunch of things that seem jointly in the dataset is determined by associa-
tion [14]. It is nothing but finding a rule which reports about the dataset.
The analysis of the market basket (a person wants to purchase A as well
as B) is the best example of association.
Following are some favored unsupervised learning algorithms.
• KNN (k-nearest neighbors);
• anomaly detection;
• K-means clustering;
• principle component analysis;
• neural networks.
Table 5.3 presents the advantages and disadvantages of both supervised and
unsupervised learning, as well as the difference between them, which is given
in Table 5.4.

5.2.3 Reinforcement Learning

This ML technique depends on feedback and its smart factor or agent
(computer program) learns from prior experiences obtained by performing
movement to deal with the surroundings. For good movement, the factor
obtains positive feedback, whereas, for worst movement, it obtains negative
feedback [2].
Overall, it learns from trial-and-error experiences acquired from realistic
scenarios to perform excellently. It can be said that the smart factor (computer
program) relates to the surroundings and learns from it to move.
The AI smart factor or agent is concerned with RL where it learns
automatically from earlier experiences without any human perception [15].
The AlphaGo and self-driving cars are the best state-of-the-art ML
examples of RL and comprise decision-making sense.
The smart factor or agent consists of three tasks such as acting, changing,
or remaining in the same state and obtaining feedback. These tasks are used to
inspect and learn about the surroundings. For positive feedback and negative
 5.2 Classification of Machine Learning Techniques 101

Table 5.3 Advantages and disadvantages of supervised learning and unsupervised learning.
Supervised Advantages • It helps the model to predict the outruns
learning based on earlier experiences.
• It solves mostly occurring real-world issues.
• In this, we get a precise idea regarding
object classes.
Disadvantages • It cannot handle critical tasks.
• The computation time is large for model
training.
• It is not able to predict right outruns if the
test and train data are different.
Unsupervised Advantages • It handles critical tasks because it does not
learning comprise input labeled data.
• It is favored mostly because unlabeled data
is available simply as compared to labeled
data.
Disadvantages • It is a hard technique as compared to
supervised learning as it does not comprise
consistent outturns.
• The results might not seem so much correct
due to no pre-idea about outturns.

Table 5.4 Difference between supervised learning and unsupervised learning.

Supervised learning Unsupervised learning
• It is computationally very complex. • It is computationally less complex.
• The input data is known and labeled. • The input data is unknown and unlabeled.
• It is classified into two types of • The clustering and associations are two
problems as classification and types of problems in unsupervised
regression. learning.
• It requires supervision to learn the • It does not require supervision to learn the
model. model.

feedback, the smart factor or agent obtains a positive point and negative point,
respectively.

5.2.3.1 Crucial terms in reinforcement learning

a) Smart Factor or Agent () –
The unit discerns or inspects the surrounding and make movement on
them.
b) Environment () –
A place inside which smart factor or agent is available or near.
102 Introduction To Machine Learning Techniques

c) State () –
After the movement of smart factor or agent, the environment gives back
its place.
d) Action () –
The movements of the smart factor or agent inside the environment.
e) Policy () –
It is the plan of the action appealed by the smart factor or agent for the
following action or movement on the present state.
f) Reward () –
It is the feedback given to the smart factor or agent by the environment
or surrounding to assess the movement of smart factor or agent.
g) Q-value () –
It is similar to the value; even so, it catches one extra parameter like
current action or movement (a).
h) Value () –
It is inverse to the short-term reward and looked for long-term return.

5.2.3.2 Salient features of reinforcement learning

• It is a trial and error based procedure.
• The smart factor or agent may obtain a setback reward.
• There is no need to teach the smart factor or agent about surroundings
or the environment as well as what movement is required to proceed.
• In this, the smart factor or agent requires to automatically inspect the
surrounding or the environment to obtain the greatest positive rewards.
• The value-based, policy-based, and model-based are three methods to
execute RL.

5.2.3.3 Types of reinforcement learning

Primarily, there are two classes of RL as given below.
a) Positive Reinforcement
Positive reinforcement means to join some kind to enhance a propensity
so that an expected behavior will be taking place anew. It affects the
smart factor’s or agent’s behavior conclusively as well as enhances its
robustness.
This positive reinforcement can continue the alterations for large spam;
even so, maximum positive reinforcement may require a burden of
places which can minimize the outturns.
 5.2 Classification of Machine Learning Techniques 103

Table 5.5 Difference between supervised learning and reinforcement learning.

Supervised learning Reinforcement learning
• It functions on the occurring dataset. • It functions by interrelating with the
surrounding or environment.
• There is the availability of a labeled • No labeled dataset is available in it.
dataset.
• When the input is present, then • In this, decisions are reserved
decisions are produced. sequentially.
• In this, the algorithm is trained to • There is no earlier training given to the
predict the outturns. smart factor or agent.

b) Negative Reinforcement
It is inverse to the positive reinforcement and newly enhances the
tendency of coming behavior by keeping negative terms away.
It is highly productive than positive reinforcement and based on behavior
as well as places.
The difference between supervised learning and RL is given in
Table 5.5.

5.2.3.4 Reinforcement learning algorithms

There are some key RL algorithms as given below [16].
a) Q-Learning (Q Stands for Quality)
It is an off-policy RL algorithm and is utilized for the learning of
temporal differences (TDs).
In this, Q (s, a) is a value function that specifies in what way the good
action “a” should be taken at a specific “s” state.
b) Deep Q Neural Network (DQN)
It is Q-learning by comprising neural networks. It is utilized for the
surrounding of big state space.
c) State Action Reward State Action
It is an on-policy RL algorithm and is utilized for the learning of TDs.
The state action reward state action (SARSA) is applied to compute Q π
(s, a) intended for recently proposed policy π and all (s-a) pairs.
The name SARSA is given due to using of Q (s, a, r, s’, a’).
Here, s = initial state, a = initial action, s’ = new state, a’ = new action,
and r = reward.
d) Temporal Difference
It is a method to learn about the prediction of an extent that rests on
imminent signal values.
104 Introduction To Machine Learning Techniques

To forecast an amount of the entire quantity of reward predictable over

the imminent signal, the TD algorithm is utilized.

5.3 Some Crucial Algorithmic Mathematical Models in

Machine Learning
The crucial mathematical models of ML are specified below [4].

5.3.1 Logistic Regression

It is consumed to forecast the variable reliant on category by utilizing a
specified set of self-regulating variables. The values of categorical variables
are predicted by a logistic regression algorithm [17]. The S-curve is used to
classify the distinct samples. The logistic regression has three types based on
categories such as binomial, multinomial, and ordinal.
The outturn of logistic regression lies between 0 and 1 as shown in
Figure 5.5, and this algorithm is applied in the state of probability requirement
within two classes.

Figure 5.5 Graphical representation of logistic regression.

 5.3 Some Crucial Algorithmic Mathematical Models in Machine Learning 105

• Sigmoid Function:
This mathematical function is applied for mapping the predicted values to
probabilities. In this, assessment is done for real value into a different value
on a scale of 0 and 1. So, it is said that this algorithm is based on the threshold
value. The logistic regression values do not go across the boundary, and, due
to this, the S-form of the curve is structured and known as the logistic function
or sigmoid function.
• Theories:
The dependent variable must remain a categorical kind. The multi-
collinearity should not be applicable for the independent variable.
• Mathematical Equation:
This equation is gained from the linear regression equation. We have the
straight-line equation as follows:

y = b0 + b1 × 1 + b2 × 2 + · · · + bn × n. (5.3)
Now divide the above equation by 1 - y, and we get the following:

y / (1 − y) ; ; 0 for y = 0 and infinity for y = 1. (5.4)

Taking logarithm on the above equation, we get the resulting equation as

follows:

log [ y / (1 − y) ] = b0 + b1 × 1 + b2 × 2 + · · · + bn × n. (5.5)

5.3.2 Decision Trees

It is utilized for both regression and classification problems but is typically
favored for classification problems. It contains two nodes as the decision node
and leaf node.
These inner nodes signify the dataset’s features in which the decision
rules are denoted by branches and outturns by leaf node [18]. Figure 5.6
shows the interconnection inside the decision tree technique.
This workflow of the algorithm is as given below.
• Start tree through S as a root node covering the comprehensive dataset.
• By making use of the attribute selection measure (ASM), search for the
finest dataset attribute.
• S is divided into subsets comprising probable values intended for the
finest attributes.
106 Introduction To Machine Learning Techniques

Figure 5.6 Decision tree diagram.

• Make a node of the decision tree for covering the greatest attribute.
• Create newer decision trees by utilizing an earlier initiated subset of a
given dataset and repeat this process till getting the final node which
cannot be classified further which is nothing but leaf node.
How to select attribute for root and sub-nodes?
The attribute selection technique (ASM) is utilized for a selection of the
finest attribute for root and sub-nodes. It has two types as given below with
mathematical equations.
• Information Gain:
It computes the total of information about the class provided by the feature.
The formula for calculating information gain is as follows:

Information Gain = Entropy −

[(Weighted Average) ∗ Each Features Entropy ] . (5.6)

Here, entropy is given by

Entropy = [(− P (Yes) log2 P (Yes)) − (P (No) log2 P (No))]. (5.7)

Entropy is denoted by S - total number of samples; P (Yes) – the probability

of Yes; P (No) – the probability of No.
• Gini Index:
It measures purity or impurity utilized through generating a decision tree in
the algorithm of classification and regression tree (CART).
 5.3 Some Crucial Algorithmic Mathematical Models in Machine Learning 107

It is computed by the below formula as


Gini Index = 1 − j P j 2. (5.8)

This algorithm is easy and valuable for decision-related issues and involves
less data cleaning.

5.3.3 Linear Regression

It is consumed to forecast the variable reliant on a continuous state by
utilizing a specified set of self-regulating variables. The values of continuous
variables are predicted by a linear regression algorithm [11].
Simply, it clarifies the connection within variable which generally seems
in the format of one dependent and one or more independent. The graphical
representation of linear regression is given in Figure 5.7.

Figure 5.7 Graphical representation of linear regression.

108 Introduction To Machine Learning Techniques

• Mathematical Equation:
The basic equation of regression is given as follows:
y = c + b ∗ x. (5.9)
Here, b = coefficient of regression, c = constant, x = score of proceeding
independent variable, and y = score of projected dependent variable.
It is of two types, namely simple linear regression and multiple linear
regression.
The linear regression line represents the correlation between the depen-
dent and independent variables. Figure 5.8 signifies about positive linear
relationship, whereas Figure 5.9 depicts a negative linear relationship.

5.3.4 K-Nearest Neighbors

It is a non-parametric algorithm and it reflects that there are no assumptions
completed on primary data. The lazy learner algorithm is also another name
of the KNN algorithm.
In the training stage, the KNN algorithm impartially gathers the dataset,
and after receiving newer data, it is classified into the same class alike to the
received data. It works well on noisy as well as large data [19, 20].

Figure 5.8 Positive linear relationship.

 5.3 Some Crucial Algorithmic Mathematical Models in Machine Learning 109

Figure 5.9 Negative linear relationship.

This workflow of this algorithm is as given below.

• Calculate the Euclidean distance as of the question instance toward the
labeled instance.
• The labeled instances are ordered by rising distance.
• Compute k of adjacent neighbors relying on root mean squared error.
• Compute a converse distance weighted average through the multivariate
neighbors of k-nearest.
The feature space with new arising data point is given in Figure 5.10 and
feature space with classified newer arising data point in Figure 5.11.
For this, Euclidian distance is calculated by the following formula:

Euclidian Distance = (x2 − x1 )2 + (y2 − y1 )2 . (5.10)

This equation provides the nearest neighbors for the newer data point in the
feature space as described for class C and class D in Figures 5.10 and 5.11.
The selection of value K in the model is completely trial and error based.
So, we have to go with some values for K, and if we get the intended results,
then we fix it further. Generally, K = 5 is favored in distinct issues. For lower
values of K, we find mostly outliers mean nosiness.
110 Introduction To Machine Learning Techniques

Figure 5.10 Feature space with new arising data point.

Figure 5.11 Feature space with classification of newer arising data point.

5.3.5 K-Means Clustering

In this repetitive algorithm, the unlabeled dataset is divided into k distinct
types of clusters in the format that every dataset fits in the individual group
which consists of the same possessions [20, 21].
It consists of the following tasks.
• Finds the finest value for K centroids by comprising iterative process.
 5.3 Some Crucial Algorithmic Mathematical Models in Machine Learning 111

• Allocates individual data points to neighboring k centroid. The data

points closer to the typical k centroid form cluster.
Figure 5.12 shows feature space with all data points and clusters of data points
are given in Figure 5.13.
Overall, the functioning of K-means lies in the most effective cluster
formed by it. The elbow method is used to find the value of K or the number of

Figure 5.12 Feature space with all data points.

Figure 5.13 Feature space with clusters of all data points.

112 Introduction To Machine Learning Techniques

clusters. It utilizes the perception of within cluster sum of squares (WCSS) to

describe the entire differences inside the cluster. The mathematical equation
to calculate the WCSS value for four clusters is given below:

WithinClusterSumofSquares(WCSS) =

P i in Cluster1 Distance(P i C1)2

+ P i in Cluster2 Distance(P i C2)2

+ P i in Cluster3 Distance(P i C3)2

+ P i in Cluster4 Distance(P i C4)2. (5.11)

Here, P i in Cluster1 Distance(P i C1)2 is the sum of the square of the
distances concerning individual data point and related centroid inside the
cluster and is similar for every cluster. The distance between data points can
be calculated either by Manhattan distance or by Euclidean distance. That is
all about K-means for solving real-world issues.

5.4 Pre-Eminent Python Libraries Intended for Machine

Learning
For real-world implementation, python is a very simple and robust program-
ming language to hack embedded platforms. There are multiple libraries
available in python for ML and some of them are given below.
• SciPy
• NumPy
• Scikit-learn
• Pillow
• Pandas
• Matplotlib
• PyTorch
• Keras
• TensorFlow
The OpenCV is a very popular computer vision and ML library for the
development of real-world applications. It also contains an inference module
for running deep neural networks with CUDA support. It is compatible
with several python libraries for ML applications on an embedded platform.
So, embedded vision has become a vast research area for real-world visual
 5.4 Pre-Eminent Python Libraries Intended for Machine Learning 113

applications [6]. Object detection and object tracking are now utilized in a
score of fields such as autonomous navigation, military applications, security
purposes, etc. [7].
Let us take some real-world examples with python programming.

5.4.1 Human Detection (OpenCV, HoG, SVM with

Multi-Threading)
Human detection is one of the crucial tasks related to security and has been
implemented with HoG and SVM [22].
Python code:
import cv2
import numpy as np
from imutils.object_detection import non_max_suppression
import imutils
from imutils.video import VideoStream
import datetime

image = cv2.imread("imagePath")
hog = cv2.HOGDescriptor()
hog.setSVMDetector(cv2.HOGDescriptor_getDefaultPeopleDetector())

while True:
imageDetect = imutils.resize(image, width=400)
timingdetails = datetime.datetime.now()
timingdetails = timingdetails.strftime(“%A %d %B %Y
%I:%M:%S%p”)
cv2.putText(imageDetect, timingdetails, (10, imageDetect.shape[0] - 10),
cv2.FONT_HERSHEY_SIMPLEX,
0.35, (255, 0, 255), 1)

(rectangle, weights) = hog.detectMultiScale(imageDetect,

winStride=(16,16),
padding=(16,16), scale=1.05)
rectangle = np.array([[x, y, x + w, y + h] for (x, y, w, h) in rectangle])
nms = non_max_suppression(rectangle, probs=None,
overlapThresh=0.65)

for (xa, ya, xb, yb) in nms :

114 Introduction To Machine Learning Techniques

Figure 5.14 Human detection output.

cv2.rectangle(imageDetect, (xa, ya), (xb, yb), (255, 255, 0), 4)

cv2.imshow(“human detection video frame”, imageDetect)

key = cv2.waitKey(1) & 0xFF

if key == ord(“g”):
break
cv2.destroyAllWindows()
Output Image:
Figure 5.14 shows the output of the human detection task using OpenCV,
HoG, and SVM with multi-threading for real-time consequences.

5.4.2 Instagram Filters – (OpenCV, Matplotlib, NumPy)

The interesting application of Instagram filtering is implemented below [23].
Python code:
import cv2
import numpy as np
import matplotlib.pyplot as plt
import imutils
matplotlib.rcParams[‘figure.figsize’] = (10.0, 10.0)
matplotlib.rcParams[‘image.cmap’] = ‘gray’
 5.5 Machine Learning Techniques in State of Affairs of Power Systems 115

threshold = 160
maximunValue = 255
framePath = “trumpGg.jpg”

frame = cv2.imread(framePath)
frame = imutils.resize(frame, 400)

src = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

dst = cv2.edgePreservingFilter(frame, flags=1, sigma_s=60, sigma_r=0.4)
framea = cv2.detailEnhance(dst, sigma_s=10, sigma_r=0.15)

kernelSize = 11
frame1 = cv2.GaussianBlur(src ,(3, 3),0,0)
laplacian = 10 * (cv2.Laplacian(frame1,cv2.CV_64F, ksize = kernelSize))
threshold, laplacian = cv2.threshold(laplacian , threshold, maximunValue,
cv2.THRESH_BINARY_INV)

gray, color = cv2.pencilSketch(frame, sigma_s=60, sigma_r=0.07,

shade_factor=0.05)
frameb = cv2.stylization(frame, sigma_s=60, sigma_r=0.07)

cv2.imshow(“Laplacian”, dst)
cv2.imshow(“framea”, framea)
cv2.imshow(“frameb”, color)
cv2.imshow(“framec”, gray)
cv2.imshow(“framed”, frameb)
cv2.waitKey(0)
cv2.destroyAllWindows()
Output Image:
Figure 5.15 shows the output images of Instagram filtering.

5.5 Machine Learning Techniques in State of Affairs of

Power Systems
The power system is a term representing a network consisting of electric com-
ponents installed to provide, move, and utilize electric power. An electrical
grid is a well-known model which supplies power to industries, homes, and
imperative places inside a large region [24]. The conventional grid utilizes
a finite single-route interaction means power is transferred from the power
116 Introduction To Machine Learning Techniques

Figure 5.15 Instagram filtering images.

station to the intended area. Whereas, the smart grid operates on double-
route interaction, and, in this, information and electricity are interchanged.
The energy Internet is also a booming concept which involves the trend of the
Internet of Things with smart grid power systems. The smart grid empowers
solar and wind types of renewable energies [25], [26]. Overall, ML techniques
help the modern hybrid power systems to predict and make decisions on
acquired data as shown in Figure 5.16. This makes the power system most
effective, credible, safe, and greener [29].
Let us have a look at some instances.
The composite inputs of time series are the major issue due to the
undetermined nature of supplies of renewable energy and it invites to apply
ML techniques such as convolutional neural network (CNN), multilayer per-
ception (MLP), long short-term memory (LSTM), recurrent neural network
(RNN), etc. These help to make predictions on short-term time-series in
resources of renewable energy and power system [28]. Specifically, solar
energy is the central resource of pure energy and is known for large electrical
 5.6 Conclusion 117

Figure 5.16 Machine learning in modern hybrid power system.

power stations related to the smart grid. The administration and steady func-
tioning of the power system are affected by intermittency and randomness
of solar energy. For this, the CNN and LSTM types of ML techniques
are utilized for forecasting the photovoltaic (PV) output power correctly.
These techniques have been applied for actual PV data in Belgium and
Limberg [27].
The hybrid power plant consists of a wind generator and solar panels. In
this, neural network type of advanced ML technique has been applied to direct
the electricity allocation generated in hybrid power plant via controller [30].
The ML techniques help for the smart controlling and output predictions in
power system applications [31]. At the length, the renewable modern hybrid
power system is now the need in distinct human working fields.

5.6 Conclusion
In conclusion, the ML technique is the finest revolution for advancement in a
score of human working fields. It is of three types such as supervised learning,
unsupervised learning, and RL technique. Supervised learning resolves two
kinds of problems such as classification and regression, whereas unsupervised
learning functions for clustering and association issues. RL is utilized for
118 Introduction To Machine Learning Techniques

decision-making and is classified into positive reinforcement and negative

reinforcement types. Overall, there are distinct algorithms involved under
the types of ML techniques for generating intended outturns as per the
occurred issues. The decision trees, k-means clustering, logistic regression,
linear regression, Q-value, SARSA, KNNs are the mostly utilized algorithms
for solving real-world issues.
There are several applications of ML in a realistic scenario such as power
systems, smart grids, energy internet, object detection, object classification,
speech recognition, Google translator and map, autonomous intelligent sys-
tems, etc. Advanced ML has become a robust technique for modern hybrid
power systems with renewable energy involving solar and wind energy,
hydropower energy, geothermal energy, etc. Smart controlling and decision-
making are the central tasks performed by ML techniques in power system
applications.

References
[1] A. Voulodimos, N. Doulamis, A. Doulamis, and E. Protopapadakis.
Deep learning for computer vision: A brief review. Computational
Intelligence and Neuroscience, 2018.
[2] E. Alpaydin. Introduction to Machine Learning, 2020, MIT Press,
Cambridge, MA, USA.
[3] T.D. Buskirk, A. Kirchner, A. Eck, and C.S. Signorino. An introduction
to machine learning methods for survey researchers. Survey Practice,
11(1), 1–10, 2018.
[4] M. Kubat. An Introduction to Machine Learning, 2017, Springer, Cham.
[5] J. Brownlee, Machine Learning Mastery with Python, 527, 100–120,
2016.
[6] O.S. Vaidya, R. Patil, G.M. Phade, and S.T. Gandhe. Embedded vision
based cost effective tele-operating smart robot. International Journal
of Innovative Technology and Exploring Engineering (IJITEE), 8(7),
1544–1550, 2019.
[7] R.R. Patil, O.S. Vaidya, G.M. Phade, and S.T. Gandhe. Qualified
scrutiny for real-time object tracking framework. International Journal
on Emerging Technologies, 11(3), 313–319, 2020.
[8] S.B. Kotsiantis, I. Zaharakis, and P. Pintelas. Supervised machine
learning: A review of classification techniques. Emerging Artificial
Intelligence Applications in Computer Engineering, 160(1), 3–24, 2007.
 References 119

[9] A. Singh, N. Thakur, and A. Sharma. A review of supervised machine

learning algorithms. In 2016 3rd International Conference on Comput-
ing for Sustainable Global Development (INDIA Com), 1310–1315,
2016, IEEE.
[10] P.C. Sen, M. Hajra, and M. Ghosh. Supervised classification algorithms
in machine learning: A survey and review. In emerging technology in
modelling and graphics, 99–111, 2020, Springer, Singapore.
[11] D. Maulud and A.M. Abdulazeez. A review on linear regression
comprehensive in machine learning. Journal of Applied Science and
Technology Trends, 1(4), 140–147, 2020.
[12] M.E. Celebi and K. Aydin. Unsupervised Learning Algorithms, 2016,
Springer International Publishing, Berlin.
[13] S. Kotsiantis and P. Pintelas. Recent advances in clustering: A brief
survey. WSEAS Transactions on Information Science and Applications,
1(1), 73–81, 2004.
[14] A.C. Nanayakkara, B.T.G.S. Kumara, and R.M.K.T. Rathnayaka. A
survey of finding trends in data mining techniques for social media
analysis. Sri Lanka Journal of Social Sciences and Humanities, 1(2),
2021.
[15] S. Padakandla. A survey of reinforcement learning algorithms for
dynamically varying environments. ACM Computing Surveys (CSUR),
54(6), 1–25, 2021.
[16] M.M. Afsar, T. Crump, and B. Far. Reinforcement learning based
recommender systems: A survey. arXiv preprint arXiv:2101.06286,
2021.
[17] C.Y.J. Peng, K.L. Lee, and G.M. Ingersoll. An introduction to logistic
regression analysis and reporting. The Journal of Educational Research,
96(1), 3–14, 2002.
[18] B. Charbuty and A. Abdulazeez. Classification based on decision
tree algorithm for machine learning. Journal of Applied Science and
Technology Trends, 2(01), 20–28, 2021.
[19] M. Steinbach and P.N. Tan. kNN: k-nearest neighbors. In The Top Ten
Algorithms in Data Mining, 165–176, 2009, Chapman and Hall/CRC,
Boca Raton, FL, USA.
[20] P. Aiwale and S. Ansari. Brain tumor detection using KNN & LLYOD
clustering. IJEAT, 10(12), 187, January 2020, ISSN 2249–8948.
[21] A. Coates and A.Y. Ng. Learning feature representations with k-means.
In Neural networks: Tricks of the trade, 561–580, 2012, Springer, Berlin,
Heidelberg.
120 Introduction To Machine Learning Techniques

[22] N. Dalal and B. Triggs. Histograms of oriented gradients for human

detection. In 2005 IEEE Computer Society Conference on Computer
Vision and Pattern Recognition (CVPR’05), Vol. 1, 886–893, 2005,
IEEE.
[23] G. Rajagopalan. Working with NumPy arrays. In A Python Data
Analyst’s Toolkit, 117–145, 2021, Apress, Berkeley, CA.
[24] Y.H. Song (Ed.). Modern Optimisation Techniques in Power Systems,
Vol. 20, 2013, Springer Science & Business Media, Berlin.
[25] L. Cheng and T. Yu. A new generation of AI: A review and perspective
on machine learning technologies applied to smart energy and electric
power systems. International Journal of Energy Research, 43(6), 1928–
1973, 2019.
[26] Y. Kabalci, E. Kabalci, S. Padmanaban, J.B. Holm-Nielsen, and F.
Blaabjerg. Internet of things applications as energy internet in smart
grids and smart environments. Electronics, 8(9), 972, 2019.
[27] G. Li, S. Xie, B. Wang, J. Xin, Y. Li, and S. Du. Photovoltaic power
forecasting with a hybrid deep learning approach. IEEE Access, 8,
175871–175880, 2020.
[28] M.M. Rahman, M. Shakeri, S.K. Tiong, F. Khatun, N. Amin, J.
Pasupuleti, and M.K. Hasan. Prospective methodologies in hybrid
renewable energy systems for energy prediction using artificial neural
networks. Sustainability, 13(4), 2393, 2021.
[29] C. Feng, M. Sun, M. Dabbaghjamanesh, Y. Liu, and J. Zhang. Advanced
machine learning applications to modern power systems. In New Tech-
nologies for Power System Operation and Analysis, 209–257, 2021,
Academic Press, Cambridge, MA.
[30] A. Gozhyj, P. Bidyuk, Y. Matsuki, V. Nechakhin, I. Kalinina, and
O. Shchesiuk. Hybrid power plant control system based on machine
learning methods. In Conference on Computer Science and Information
Technologies, 251–262, 2020, Springer, Cham.
[31] K.S. Perera, Z. Aung, and W.L. Woon. Machine learning techniques
for supporting renewable energy generation and integration: A survey.
In International Workshop on Data Analytics for Renewable Energy
Integration, 81–96, 2014, Springer, Cham.
 6
Machine Learning Techniques
for Renewable Energy Resources

K. Punitha∗ , S. Anbarasi, and T. Balasubramanian

P S R Engineering College, India

E-mail: [email protected]
∗ Corresponding Author

Abstract
At present, renewable energy resources like solar and wind attracted abundant
attention, and thanks to their green, clean, inexhaustible, and recycled nature
and them being free from carbon emission, renewable energy resources are
the foremost promising alternative to fossil fuels. Though renewable energy
resources are accessible freely, their higher upfront price, environmental
dependency, and lower efficiency act as a barrier to wider implementation.
The demerits of renewable energy resources are volatility, intermittence, and
uncertainty which affect the stability and reliability of large-scale renewable
integration into the power generation. Hence, researchers are exploring pos-
sibilities to boost accessibility and efficiency with the help of technology
such as machine learning. Deep learning, a promising kind of machine
learning technique, can be incorporated with renewable energy, especially
solar photovoltaic (PV) systems, in three major categories, such as fore-
casting, accessibility, and efficiency. Boosting the efficiency of a solar PV
system requires maximum power point tracking (MPPT), which maximizes
the extraction of available maximum power from PV modules. As the con-
ventional MPPT algorithms have no prior knowledge of the maximum power
point (MPP) at the beginning of the perturbation, these MPPTs demand
a long convergence time to achieve MPP. The need for prior knowledge
of MPP is necessary to start any conventional MPP algorithm, which the

121
122 Machine Learning Techniques for Renewable Energy Resources

deep learning based long short-term memory (LSTM) network provides in

this work. The goal of this book chapter is to implement deep learning
in the solar PV system forecasting maximum voltage to provide reference
value to its MPPT technique. The case study is also presented with a
conclusion.

Keywords: Machine learning, deep learning, LSTM network, prediction,

MPPT algorithm.

6.1 Introduction
Despite the abundance of green energy resources today, fossil fuels remain
the world’s primary resource of energy. Fuels that are considered fossil
are hydrocarbons or their derivatives, together with natural resources like
coal, petroleum oil, and natural gas. Fuels take a long time to form; so
the well-known oil resource reserves are depleting faster than they can be
replenished. Also they release greenhouse gases, which cause environmental
change such as global warming, putting the environment at risk. Green energy
has, therefore, gained a great lot of attention worldwide in recent years. A
renewable energy source can be recycled into new energy in nature, such
as solar power, wind energy, tidal energy, or geothermal energy. Renewable
energy has two major advantages over fossil fuels. A few things to remember
about renewable energy resources are that they are abundant, renewable,
and inexhaustible. The second benefit is that renewable energy is carbon-
free, green, and clean and, thus, benefits the environment. In particular,
renewable energy can effectively reduce the emission of carbon monoxide
(CO), sulfur dioxide (SO2), and dust, thereby reducing atmospheric pollution
and greenhouse gas emissions. Aside from that, the use of renewable energy
can diminish the need to exploit petroleum, and this will also help to protect
the environment. It can reduce solid waste discharge, which reduces soil
pollution. By using renewable energy, water resources are also protected by
reducing the waste gases and waste liquids emitted during the process. Thus,
renewable energy has become very popular in recent years [1].
In light of the rapid industrialization of our planet, it has become clear
that excessive consumption of petroleum will accelerate the loss of fossil
fuel reserves and harm the environment. Ultimately, this will lead to health
concerns and global climate change. The wind and sun are currently the
fastest-growing energy sources, along with petroleum and nuclear energy.
Solar, wind, hydropower, biomass, waves, tides, and geothermal heat are
 6.1 Introduction 123

reusable forms of energy to be recovered in nature. Renewable energy

poses several important challenges, notably the supply of energy, due to its
characteristics of sustainability and low environmental impact.
The sustainable energy refers to the inclusion of these sources into
existing and future energy supply systems [2]. Energy security and regional
energy shortages will be improved by developing renewable energy systems.
However, this generation of various energy sources is unpredictable and
chaotic as a result of the inherent instability of renewable energy and the
unpredictable nature of renewable energy. So, it is still a challenge to handle
with renewable energy statistics accurately. Energy monitoring with high
precision can increase energy efficiency.
Developing, managing, and making energy policy all rely on energy
forecasting. In the context of increased ways of supplying electricity from
renewable energy sources [3], it is imperative to develop technologies for
storing renewable energy. Several research papers have suggested that a num-
ber of machine learning models were used for renewable energy forecasts.
This data-driven approach enables predictions of renewable energy use. A
hybrid machine learning model was designed to boost renewable energy
prediction accuracy. Renewable energy has been predicted for different time
intervals, such as minutes, hours, days, and weeks, depending on the purpose
of the prediction. Typically, predictions for renewable energy are assessed
according to their accuracy and efficiency [4].
Solar irradiance reaches the earth at different rates based mostly on
moving clouds. It is necessary to incorporate cloud information directly
or indirectly into the formula to forecast the irradiance accurately. Physi-
cal models of clouds’ generation, propagation, and extinction are difficult
because clouds’ generation, propagation, and extinction are stochastic. Cloud
information can, therefore, usually be expressed using statistical methods
[5–7]. A special focus is on forecasting irradiance during very short periods
of time because clouds are persistent during that period.
A very short-term forecast is often used in large photovoltaic (PV)
installations, unlike those with longer horizons whose results are critical to
electricity grid operations. Maximum power point tracking (MPPT) algo-
rithms may benefit from anticipating the potential shading of a particular
section of a PV system [8]. A sub-minute forecast could also facilitate
improved control of ultra-capacitors [9, 10]. The current state-of-the-art
methods of forecasting very short-term irradiance were reviewed in Mane
[11]. The use of sky cameras [12–14] or sensors [15, 16] can be used to
analyze the data.
124 Machine Learning Techniques for Renewable Energy Resources

Several datasets related to solar engineering, such as satellite-derived

irradiance measurements [17], output measurements from hundreds of solar
power plants [18], and module-level data from solar power plants [19], meet
the HACE principle proposed by [20] which exemplifies big data. Raw
datasets have a lot of irrelevant data and noise embedded in them, which
makes processing them a challenge. Not only is it clean, green, and naturally
replenished in wide geography, but it also poses an unschedulable uncertainty,
which threatens the stability and reliability of energy systems, especially
with their large-scale integration. Considering the volatility, the intermittent
nature, and random nature of renewable energy, the price of electricity pro-
duction will undoubtedly increase due to increased system reserve capacity.
However, renewable energy utilizes a large percentage of power electronics,
therefore reducing the rotational inertia of the system and reducing the margin
of stability. Sustainable energy forecasting helps to reduce uncertainty, which
is vital to electrical system planning, management, and implementation [20].
The unpredictable, chaotic, and irregular nature of weather, accurate forecast-
ing of renewable energy sources remains a challenge. Numerous algorithms
exist for predicting renewable energy from 1 or 2 minutes up to one or
two days ahead. The three most common categories are physical methods,
statistical models, and artificial intelligence methods, along with their hybrid
methods [21].
Forecasting correct energy over time is crucial for the growth of PV
technology. Aspects such as how efficiently sunlight is converted into energy
and how this relationship varies over time are important. The MPPT solar
charge controller, which is also called a smart DC-DC converter, is essential
for any solar power system to extract maximum power from a PV module,
while simultaneously forcing it to work at voltage close to the MPP, thus
making it more efficient.
In this chapter, the MPPT technique was described using a deep archi-
tecture based on an long short-term memory (LSTM) network for solar PV
P max and V max forecasting. By using the forecasting results, the MPPT
technique can control and maintain reference voltage on a real-time basis.
During one switching period, the PV module voltage V pv and power P max
are checked, and a DC/DC converter is driven to operate at the voltage V max
corresponding to the actual MPP. MPPT units produce pulses that trigger the
MOSFET switch in the DC-DC boost converter. Therefore, the DC-DC con-
verter’s power flow is controlled by varying the switching period’s on/off duty
cycle. Data-driven problems have been a major focus of machine learning
techniques in recent decades. Machine learning techniques encompass a wide
 6.1 Introduction 125

range of topics, such as statistics, mathematics, artificial neural networks, data

mining, optimization, and artificial intelligence. Machine learning techniques
try to identify relationships between input and output data, regardless of
whether it is mathematically based. Once the machine learning models are
properly trained with training data, decision-makers can use the forecasting
input values to produce accurate forecasting results. In machine learning, data
preprocessing plays a crucial role and can lead to enhanced performance
[22]. Typically, machine learning uses three types of learning methods:
supervised learning, unsupervised learning, and reinforcement learning. In
the phase of training, supervised learning takes advantage of labeled data.
For training data that has not been labeled in advance, unsupervised learning
involves automatically classifying input data into clusters based on certain
criteria. As such, clustering criteria determine how many clusters are created.
Reinforcement learning’s purpose is to obtain feedback from the external
environment so as to maximize expected benefits. Numerous theoretical
mechanisms and applications are proposed in accordance with three basic
principles [23]. Due to rapid technological advancements in hardware and
software, deep learning, a subfield of machine learning, has seen tremendous
growth in the past few years. A wide range of applications have used deep
learning to achieve satisfying results as it enables the realization of non-
linear attributes and invariant data configurations at a high level [24]. A
single machine learning model is also able to forecast renewable energy [25].
However, utilizing a single machine learning model cannot achieve better
results owing to the difference in datasets, time steps, prediction ranges,
settings, and performance indicators. Consequently, some studies focused on
developing hybrid machine learning models or overall prediction methods
for renewable energy predictions to improve their performance. The field of
machine learning has been increasingly focusing on support vector machines
and deep learning methods [26].
Solar radiation forecasting problems depend on previous time steps.
These problems have been successfully addressed by deep learning [27, 28].
The recurrent neural network (RNN), i.e., LSTMs and gated recurrent units
(GRUs), features that allow them to learn long-term dependencies are,
therefore, more suitable for these problems [29, 30]. Deep learning based
forecasting is mainly used for short-term forecasting within 1 or 2 hours
to a few days [31, 32]. Solar irradiance forecasts for a period of 120 min-
utes are made in [33]. LSTM was used to forecast solar radiation for the
upcoming year [34]. These methods did not account for maximum PV system
power influence because they were trained with a supervised method without
126 Machine Learning Techniques for Renewable Energy Resources

optimal feature extraction. Compared to conventional power production tech-

nologies, PV systems have become increasingly economically competitive
over the past few years. Thus, they can be deployed as standalone units and in
grid-connected applications, including in conjunction with electrical energy
storage devices [35].
This book chapter is organized as the following sections: Related Lit-
erature, Background and Motivation, Overview of Machine Learning, Deep
Learning Architecture, LSTM Network, Concepts of Solar PV and its MPPT
Technique, Discussion of Simulation Results, and Conclusion.

6.2 Overview of Machine Learning

In machine learning, deep learning is a budding technique of soft computing.
In the 1950s, artificial intelligence was implemented on hardware systems
for the very first time. Within the 1960s, the concept of machine learning
emerged with several systematic theorems. In addition, its recently evolved
branch, deep learning, was first mentioned in the 2000s, and shortly after-
ward, its rapid application to numerous fields as a result of its new prediction
concept derived from big data. The difference between deep learning with
machine learning and artificial intelligence is shown in Figure 6.1.

Figure 6.1 Difference between deep learning with machine learning and artificial
intelligence.
 6.2 Overview of Machine Learning 127

Figure 6.2 Difference between model (a) development of traditional programming and (b)
development of machine learning programming.

Before getting into machine learning, we should understand the difference

between humans and machines. If we ask “who is your father?” to a baby and
a machine, the baby can recognize, but it cannot answer. But the machine
cannot recognize or answer. This is because humans can learn from parents,
friends, relatives, teachers, colleagues, society, books, and experience. So
we should make the machine learn through learning algorithms to answer
the above question. For that, we should feed problem content, data, and
algorithms to the machine. In traditional programming, we generally provide
programs and data as input to the computer/machine to get the desired2
output. But in machine learning programming, we should provide data and
output to the computer to get the program as output. This can be depicted in
Figure 6.2.
In general, machine learning technology combines three methods of
learning: supervised learning, unsupervised learning, and reinforcement
learning. Labeled information is beneficial to supervised learning during
the training phase. To perform an unsupervised learning task, input data
is spontaneously assigned to certain clusters based on certain criteria for
training data that has not been labeled yet. In reinforcement learning, the
external environment uses feedback to maximize expected advantages. The
development of machine learning is depicted in Figure 6.3.
The first and one among the foremost crucial things to seek out in machine
learning is to collect sufficient data and find what are the inputs and also the
expected outputs. It is vital to keep in mind that machine learning can solely
be utilized to learn patterns that exist in the training data; so it can identify
what we have got before. So split the collected data into training as well as
testing data. Once using machine learning, we tend to assume that the future
will behave just like the past, and this is not invariably true. For this purpose,
we want to divide our information into two parts: a training set that is used to
learn the model; a test set to calculate the generalization performance of the
128 Machine Learning Techniques for Renewable Energy Resources

Figure 6.3 Development of machine learning model.

model. Then model assessment metrics are vital to enumerate the model. The
model assessment aims to evaluate the generalization accuracy of a model
on future information. Deep learning, a promising kind of machine learning
able to predict non-linear features, has been reported in the literature. Its
architectures are reported in the next section.

6.3 Deep Learning Architecture

Deep learning algorithm utilizes machine learning to achieve artificial intel-
ligence. It has also shown hopeful outcomes in the unmanned vehicle,
computer vision, audio processing, and data mining. Deep learning algo-
rithms imitate human-level intelligence in machines to solve any problems
for a given region. Deep learning architecture includes feed-forward neural
network, RNN, and RNN extensions, i.e., LSTM and GRU.

A. Feed-Forward Deep Neural Network

A feed-forward deep neural network (FFNN) is the uncomplicated type of
artificial neural network. Figure 6.4 depicts a simple example of an FFNN
where i represents input, and h and o represent hidden and output, respec-
tively. These neurons are interconnected. There are three sorts of layers
referred to as the input, hidden, and output layers. In the hidden layer,
the input is fed forward until the output is determined through the use of
activation functions in each node and initialized weights and biases. At last,
weights are changed utilizing the back-propagation algorithm and the loss
functions in order to get the best results. In FFNN, there is no feedback
 6.3 Deep Learning Architecture 129

Figure 6.4 Simple FFNN.

connection; they do not have an instrument to recollect past outputs, unlike

the RNN. Due to this, they are unsuitable for forecasting time series.

B. Recurrent Neural Network

RNN is an associate extension of a conventional FFNN where the output from
the last step is taken as input to the current step. In other words, each node
test will be based on the time step output from the previous test. An RNN is
depicted in Figure 6.5.
The hidden state ht is defined as follows:

a(t) = b + W h(t − 1) + U x(t); )); h(t) = tanh(a(t)) (6.1)

o(t) = c + V h(t); y(t) = sof t max(o(t)) (6.2)

where
x(t) is taken as the input to the network at time step t;
h(t) represents a hidden state at time t;
130 Machine Learning Techniques for Renewable Energy Resources

Figure 6.5 Simple RNN.

o(t) illustrates the output of the network;

y(t) is true target;
U is input to hidden connections parameterized by a weight matrix;
W is hidden-to-hidden recurrent connections parameterized by a weight
matrix;
V is hidden-to-output connections parameterized by a weight matrix.
After the hidden state is obtained, in a fully connected layer, RNN output o(t)
is directly calculated from the current hidden state.
Unfortunately, it is a troublesome task to train an RNN as well as it
cannot process very long sequences when tanh or relu is employed as an
activation function. It is additionally determined that its gradient vanishes and
tends to explode throughout training RNN to capture long-term dependencies.
To handle the aforementioned problems, LSTM unit was introduced, and,
afterward, GRUs were introduced recently.

C. Long Short-Term Memory

LSTM captures long-term dependencies to solve the gradient problems.
Three gates are present in an LSTM unit. They are input gate, output gate,
and forget gate with internal memory and is shown in Figure 6.6. An LSTM
unit is ready to decide whether or not to retain or forget the current memory
via the introduced gates. LSTMs use their input gate to decide which current
data to pass. The forget gate determines what information has to be passed.
The output gate specifies the state information to be passed.
 6.3 Deep Learning Architecture 131

Figure 6.6 LSTM unit.

D. Gated Recurrent Units

LSTM is similar to GRU as it memorizes important information as well as
transports it over long distances, capturing long-term relationships. In addi-
tion to being less complex, it is also more efficient in terms of computation.
Certain data shows that it performs faster and better than LSTM. A GRU
also features two gates, namely a reset gate and an update gate. Figure 6.7
illustrates how the above gates modulate the flow of data within the unit
without having distinct memory cells.
GRU reset gate is defined as follows:

z = σ(Wz .xt + Uz .h(t − 1) + bz ; ; r = σ(Wr .xt + Ur .h(t − 1) + br (6.3)

Figure 6.7 GRU unit.

132 Machine Learning Techniques for Renewable Energy Resources

‘
h = tanh(Wh .xt +r ∗Uh .h(t−1)+bz ; ; h = z ∗h(t−1)+(1−z)∗h (6.4)
where
r is the relevance gate;
h is the current cell state;
h(t - 1) is the previous cell state;
W, U are weights;
z is the update gate.

6.4 LSTM Network Based Prediction

In situations where the order of information is crucial, normal neural net-
works do poorly. To overcome this constraint, RNNs were developed. An
RNN cell does not just consider its current input but also the output of
RNN cells preceding it for its current output. RNNs are smart at managing
sequential data; however, they run into problems when the context is much
away. LSTM cell is a special type of RNN that is designed to overcome the
problems associated with RNN in learning long-range dependencies. This
is a new kind of neuron designed to overcome the limitations associated
with RNN in learning long-term dependencies. As shown in Figure 6.8,
LSTM networks are formed by repeating modules of different structures. As
shown in Figure 6.8, LSTM networks are formed by repeating modules of
different structures. In addition, LSTM consists of internal state data which
is transmitted from one cell to another and also modified by operation gates.
Additionally, they contain a sigmoid neural network layer and a point-wise

Figure 6.8 LSTM network model.

 6.4 LSTM Network Based Prediction 133

multiplication step. The sigmoid type layer delivers output as a number which
is between 0 and 1. That output number describes what quantity of every
part ought to be let through. The output value of 0 suggests that “let nothing
through,” whereas a value of 1 suggests that “let everything through.” An
LSTM has three of those gates to save and control the cell states. They are
forget, input, and output gates. Every LSTM module comprises four layers
or structures. The primary layer of LSTM is to come to a decision on what
data is suitable to throw out from the past states (ht−1 and xt ) and is set by
sigmoid or forget gate layer based on output Ct−1 . If Ct−1 is 1, it denotes
“completely keep” and else “completely get rid of” which is shown in the
following equation:

ft = σ(Wf .[ht−1 , xt ] + bf ). (6.5)

The next step is to resolve what new data is going to store in the state. This
has another two layers whose function is shown in the following equations:

it = σ(Wi .[ht−1 , xt ] + bi ) (6.6)

Ct = tanh(WC .[ht−1 , xt ] + bC ). (6.7)
The first layer is called the sigmoid layer or the input gate layer selects
which value is going to be updated. Next tanh layer produces a vector of the
latest values which is able to be added to the state shown in the following
equation:
∼
Ct = f ∗Ct−1 + it ∗ Ct . (6.8)
t
Creating and updating the state is done with the third layer, which com-
bines the two layers above. The fourth or final state is to update the previous
state Ct−1 into the latest current state Ct . At last, multiplication is performed
on the previous state by ft , after forgetting then to add. Equation (6.7) is that
the new value decided to update every state value. In the future, the output
will be based on a filtered version of the cell state. There can be a sigmoid
layer that determines what parts of the state are going to be the output. To
place tanh in between –1 and 1, multiply the output of sigmoid gate by the
state, and because of that, the only output that is determined is given in the
following equations:
Ot = σ(WO [ht−1 , xt ] + bo ) (6.9)
ht = Ot ∗ tanh(Ct ) (6.10)
134 Machine Learning Techniques for Renewable Energy Resources

where
ft is the forget gate;
it is the update gate;
Ct is the current cell state;
ht−1 is the previous output;
xt is the input;
W is the weight;
Ot is the output gate.

6.5 Concepts of Solar PV and its MPPT Techniques

The solar PV systems transform solar daylight directly into electricity. How-
ever, economical utilization of PV systems is achieved through MPPT, as this
guarantees maximum possible power transport to the grid. At a single point
on its current-voltage (I–V) or power-voltage (P–V) characteristics curve of
a PV cell, called MPP, the PV system functions with good efficiency and
delivers high output power. There are various types of MPPT algorithms
projected in the literature. The conventional MPPT techniques for example
perturb and observe (P&O) is a simple, easily implemented, and efficient
method. The PV module provides an output that is directly related to the
irradiation and is inversely related to its temperature. The concepts behind the
P&O MPPT method are given below. If dP/dt > 0, the perturbation should
be in one direction and if dP/dt < 0, the perturbation should be reversed. The
process should be repeated periodically until dP/dt = 0 reaches the maximum
power point (MPP). This perturbation MPPT demands an extended response
time to attain the MPP and more oscillation around MPP, as they do not
have a previous value of the MPP at the beginning of the perturbation. This
P&O algorithm is commenced from some random low value of voltage or
some percentage of its open-circuit voltage and progresses via successive
perturbations to attain its MPP in [10]. To achieve this approach, the range of
increment or decrement value is crucial. Large ranges result in the algorithms
obtaining MPP quickly but in oscillations around it. In contrast, a small range
results in little oscillation around the MPP; however, the rate of convergence
is also reduced. To overcome this problem, variable step ranges are used. To
accomplish this, complicated control technology is needed. Less oscillation
and quick tracking can be carried out in P&O with previous knowledge
of reference voltage value. For the P&O MPPT, a deep learning based
LSTM network predicted V max is projected to offer a reference voltage V ref
online.
 6.6 Simulation Results and Discussion 135

6.6 Simulation Results and Discussion

Figure 6.9 depicts the overall circuit design. It consists of a cascaded (cas-
caded buck and boost) converter with a 300-Wp PV array similar to real PV
system specification (given in Table 6.1), P&O MPPT controller, blocking
diode, and resistive loads. The simulation was done for a 1000-Ω resistive
load, the inductor values were chosen as 564 μH and 0.78 H, and output
capacitor value is taken to be 22 μF and 220 μF. P&O MPPT controller gets
the starting value of perturbation derived from deep learning based LSTM
network prediction.

6.6.1 Modeling and Performance Analysis

A. Solar PV System
PV cells are modeled by evaluating the I–V and P–V curves to simulate the
real PV cell of varying atmospheric values. The most popular concept is to
employ the single-diode model of the electrical equivalent circuit. This model
comprises three parameters like open-circuit voltage, short circuit current,

Figure 6.9 Overall system developed in MATLAB software.

Table 6.1 PV array specification.

S. No. Parameters Rating
1. Maximum power (Pmax) 300 Wp
2. Open-circuit voltage (Vocn) 36.1 V
3. Short circuit current (Iscn) 8.3 A
4. Maximum power point voltage (Vpm) 44.6 V
5. Maximum power point current (Imp) 8.87 A
136 Machine Learning Techniques for Renewable Energy Resources

diode identity factor, as well as the modified one that includes a series resistor.
Figure 6.10 shows a PV cell with a single diode.

Figure 6.10 (a) PV cell single-diode model. (b) PV array configuration.

 6.6 Simulation Results and Discussion 137

In a PV array, PV cells are connected in series and parallel. Arrays with

parallel connections usually have an increased current, while modules with
series connections have an increased voltage. Solar cells can typically be
modeled as a current source connected to an inverted diode parallel to it.
It has its own series and parallel resistance as shown in Figure 6.10.

i = Iph− Is(T ) (exp{q(Vin + i.Rs )/KT A} − 1 (6.11)

where i is the PV panel output current, Iph is photocurrent, Is (T ) is the

reverse saturation current, q (= 1.6 × 10−19 ) is an electron charge, Vin is
the terminal voltage of the PV panel, RSH is the PV panel series resistance,
A is the ideal factor of the PN junction of the PV diode, which varies in the
range of [1, 2], and k (= 1.38 × 10−23 J/K) is the Boltzmann constant. The
photocurrent is then found using the following equation:

Iph = [Isc + Ki (T − Tref )].G (6.12)

where Isc is the short circuit current provided by the PV panel at a reference
temperature and an irradiance of 1 kW/m2 , Ki (= 3 mA/◦ C) is the temperature
coefficient, G is the solar irradiance in kW/m2 , and T and Tref are measured
temperature and reference temperature, respectively. The output current is
then given in the following equation:

Ist (T ) = Is (Tref ) exp{Ks (T − Tref )} (6.13)

where is (Tref ) is the reverse saturation current (Tref = 295 K) and Ks (≈

0.072/◦ C) is the temperature coefficient of the PV panel.
Figure 6.11(a) shows the I–V characteristics of a typical solar cell. Multi-
plying the voltage and current characteristics gives us the P–V characteristic
shown in Figure 6.11(b). At MPP, the panel power output is at its maximum.
Table 6.2 lists the standard PV array specifications based on NAVISOL
datasheets.
The PV cell and array are modeled using a single-diode equivalent circuit
and its performance characteristics are analyzed under varying atmospheric
as shown in Figures 6.12(a) and (b).

B. Cascaded Converter Modeling

A design of cascaded buck and boost converters has been proposed in this
work and whose circuit diagram is depicted in Figure 6.13.
The design equations consist of buck and boost converters combined
with effects of inductor coupling on key converter performance parameters
138 Machine Learning Techniques for Renewable Energy Resources

Figure 6.11 (a) I–V characteristics. (b) P–V characteristics.

Table 6.2 PV array specification.

S. No. Parameters Rating
1. Maximum power (Pmax ) 300 Wp
2. Open-circuit voltage (V ocn ) 44.6 V
3. Short circuit current (I scn ) 8.87 A
4. Voltage at maximum power point (V pm ) 36.1 V
5. Current at maximum power point (I mp ) 8.3 A

Figure 6.12 (a) V–I curve. (b) P–V curve under varying atmosphere condition.

like inductor ripple current, input ripple current, and minimum load. The
following sections summarize the design of the DC-DC buck and boost
converters formula and the values are tabulated in Table 6.3.
 6.6 Simulation Results and Discussion 139

Figure 6.13 Circuit diagram of cascaded DC–DC converter.

Table 6.3 Cascaded buck and boost converter parameters.

S. No. Parameters Theoretical values
1. Input voltage V in (avg) 9.5 V
2. The efficiency of the converter (η) 99%
3. Buck Inductor (L1 ) 0.78 H
4. Boost_Inductor (L2 ) 564 μH
5. Buck_Capacitor (C1 ) 220 μF
6. Boost_Capacitor (C2 ) 22 μF
7. Switching frequency (FS ) 25 kHz
8. Buck_Duty cycle (D1) 42.10%
9. Boost_Duty cycle (D2) 20.83%
10. Load resistance (RL ) 1000Ω
11. (ΔV out)(ΔV out) Output voltage ripple 1.63 V
12. Inductor ripple current (ΔIL ) 1.350 A
13. Output voltage (V out ) 12 V (boost)
14. Output voltage (V out ) V (buck)

1. Input voltage ranges:

Input voltage V in (min) = 8.0 V
Input voltage V in (max) = 11.0 V
Input voltage V in (avg) = 9.5 V
2. Output voltage and current ranges:
Output voltage Vo (boost) = 12.0 V
Output voltage Vo (buck) = 4.0 V
Output current Io = 1.0 A
140 Machine Learning Techniques for Renewable Energy Resources

3. Calculation of duty cycle (boost):

V in(avg)
Duty cycle D1 = = 20.83% D1 = 1 −
V out
4. Calculation of duty cycle (buck):
V out
Duty cycle D2 = = 42.10% D2 =
V in(avg)
5. Calculation of inductor (boost):
D1(V out − V in(avg))
Inductance L1 = = 564 μH L1 =
F s ∗ ΔIL1
6. Calculation of inductor (buck):
D2(V out − V in(avg))
Inductance L2 = = 0.78 H L2 =
F s ∗ ΔIL2
7. Calculation of inductor ripples current (boost):
Inductor ripple current ΔIL1 = 20%-40% of Io = 0.03 A
8. Calculation of inductor ripples current (buck):
Inductor ripple current ΔIL2 = 20%–40% of Io
= 0.118 mA
9. Output capacitor selection
 (boost): 
I0 .D1
Output capacitance CO = = 22μF
F S.ΔV out
where  
I0 ΔIL
Output voltage ripple ΔV out = ESR + = 0.378 V
1 − D1 2
10. Output capacitor selection
 (buck): 
ΔIL2
Output capacitance CO = = 220 μF
8 ∗ F S.ΔV out
where  
I0 ΔIL
Output voltage ripple ΔV out = ESR + = 0.00268 mV.
1 − D2 2

C. Proposed MPPT Controller

To obtain the maximum power from a solar array, MPPT algorithms are
required. The MPPT of solar panels varies with the amount of irradiation
and temperature; so MPPT algorithms are necessary. Figure 6.14 illustrates a
modified P&O algorithm flowchart.
A comparison of the proposed and conventional algorithms for power-
conditioning of PV systems is shown in Figure 6.15. Both algorithms respond
differently to the same variation in irradiation. The proposed algorithm took
0.2 seconds lesser to reach the steady-state than the conventional P&O
algorithm.
 6.6 Simulation Results and Discussion 141

Figure 6.14 Flowchart of modified P&O MPPT algorithm.

6.6.2 Prediction or Forecasting Methodology

Time series analysis denotes the analysis of variation of data over a period
of time. One of the applications of time series analysis is solar forecasting.
In this section, time series analysis is performed using an LSTM network to
predict maximum voltage (V mpp ) of 300 Wp real-time solar PV system based
on its MPP of the past one year. The data used in this work is collected from
29th September 2018 to 29th September 2019. For the purpose of training,
MPP data from 29th September 2018 to 29th July 2019 is taken (75%).
For the sake of prediction, MPP data from 1st August 2019 to 29th September
142 Machine Learning Techniques for Renewable Energy Resources

Figure 6.15 (a) Influence of irradiation variation. (b) Influence of temperature variation. (c)
Response of the proposed MPPT algorithm in PV system.

2019 is utilized (25%). To estimate the performance of the proposed algo-

rithm compared with the real MPP for the same month data as well. The
database contains ten columns: date, time, voltage, ampere, kWh, PF, kW,
Pmax , V max , and hours. The predicting value is V max ; therefore, there is no
need to give any interest to the rest of the column data.
Figure 6.16 shows the recorded value of V max collected from the data
logger of real solar 300-Wp PV system for the past one-year day-wise 365
counts. It is found to be extremely non-linear and it is very difficult to capture
the trend using this information. Therefore, that LSTM is utilized in this work.
LSTM is a type of RNN able to recall the previous information, and while
predicting the upcoming values, it captures this previous information into
account.
The steps involved in LSTM network development are the same as any
other machine learning problem. The aim of this chapter work is to predict
V max and to provide a reference voltage for the P&O MPPT method. LSTM
network training, LSTM network testing, and sequence data loading are the
steps required for time series forecasting utilizing deep learning. The first step
is to import the dataset and normalize/scale our data between 0 and 1 using
MinMaxScaler. The succeeding step is to divide the data into training as well
as testing data. Among 364 data, first 300 data are divided as training data
and the remaining 64 as testing data. The fourth step is to create an LSTM
model. Figure 6.17 shows LSTM network based test and predicted (scaled)
value of V max . In the output waveform, the actual V max (scaled) value for
the July month represents the blue line and August 2019, while the red line
presents the predicted V max value. It is clearly shown that the predicted value
follows the tested value.
 6.6 Simulation Results and Discussion 143

Figure 6.16 The recorded value of Vmax for 365 days.

The comparison has also been made for the recorded V max from the data
logger and LSTM algorithm predicted value and is tabulated in Table 6.4.
It is also observed that the error percentage in both cases is very less in the
proposed LSTM algorithm. The training progress of V max is also shown in
Figure 6.18.

6.6.3 Utilizing Predicted Value in MPPT Technique

In the P&O-based algorithm, MPP may be attained by comparing the power
difference dP. LSTM network training, LSTM network testing, and sequence
data loading are the steps required for time series forecasting utilizing deep
learning. This voltage reference (V ref ) can either be maximized or minimized
with a small value that is forced to operate the PV array. The most appropriate
way to use the conventional P&O MPPT method is to have a previous
understanding of the V ref value, which will diminish the tracking period of
the MPPT method. For that, we can make use of deep learning based LSTM
144 Machine Learning Techniques for Renewable Energy Resources

Figure 6.17 The predicted value of Vmax for another 365 days.

Figure 6.18 Training progress of V max .

 6.7 Conclusion and Future Directions 145

Table 6.4 Comparison of recorded and predicted Vmax .

Day Date Recorded value Predicted value Error percentage (%)
351st 29.07.2019 28.8033 24.03 −8.26974
352nd 30.07.2019 24.0425 26.21 3.494287
353rd 31.07.2019 28.078 27.13 −8.49543
354th 01.08.2019 27.0213 29.53 −8.4984
355th 02.08.2010 22.8388 24.96 −10.311
356th 03.08.2019 21.2563 23.7 −1.98981
357th 04.08.2019 25.0024 25.51 −8.26974

Figure 6.19 Converter output voltage with and without the utilization of LSTM Vref value
in P&O MPPT.

network forecasting parameters as the reference values. Figure 6.14 depicts

the flowchart of P&O MPPT which utilized day-wise LSTM predicted V max
as V ref .
Figure 6.19 inferred that the P&O MPPT with reference value oscillates
less and reaches its maximum value 0.15 seconds earlier than the other case.

6.7 Conclusion and Future Directions

This chapter presents the modified P&O MPPT algorithm whose initial
perturbation or reference value is provided by a deep learning based LSTM
network to increase its tracking speed and response. Day-wise maximum
voltage of 365 days is collected from 33-kW solar PV systems to predict
maximum voltage which is further utilized for reference voltage or initial per-
turbation value for P&O MPPT. To determine how well the LSTM network
predicts, the statistical error is calculated. It is found that LSTM is superior in
146 Machine Learning Techniques for Renewable Energy Resources

prediction methodology. Ultimately, it can be inferred that the proposed P&O

MPPT provides a good response, less tracking time, and very little oscillation
around the MPP value. This section also briefly presents some interesting
research directions, which are worth investigating further. GRU method can
be implemented instead of LSTM unit in the same problem to accomplish
more precise prediction. In near future, we can try to implement it in Arduino
UNO hardware as an MPPT controller to evaluate the proposed methodology.

References
[1] Kotsiantis, S.B. Supervised machine learning: A review of classification
techniques. Informatica 2007, 31, 249–268.
[2] Qiu, J., Wu, Q., Ding, G., Xu, Y., Feng, S. A survey of machine learning
for big data processing. EURASIP J. Adv. Signal Process. 2016, 2016,
67.
[3] Gu, J., Wang, Z., Kuen, J., Ma, L., Shahroudy, A., Shuai, B., Liu, T.,
Wang, X., Wang, L., Wang, G., et al. Recent advances in convolutional
neural networks. Pattern Recognit. 2017, 1, 1–24.
[4] Amasyali, K., El-Gohary, N.M. A review of data-driven building energy
consumption prediction studies. Renew. Sustain. Energy Rev. 2018, 81,
1192–1205.
[5] Wang, H.Z., Lei, Z.X., Zhang, X. A review of deep learning for
renewable energy forecasting. Energy Convers. Manage. 2019, 198,
111799.
[6] Olabi, A.G. Renewable and energy storage system. Energy 2017, 136,
1–6.
[7] Zendehboudi, A., Baseer, M.A., Saidur, R. Application of support vector
machine models for forecasting solar and wind energy resources: A
review. J. Clean. Prod. 2018, 199, 272-285.
[8] Bosch, J., Kleissl, J. Cloud motion vectors from a network of ground
sensors in a solar power plant. Sol. Energy 2013, 95, 13–20.
[9] Dong, Z., Yang, D., Reindl, T., Walsh, W.M. Satellite image analysis and
a hybrid ESSS/ANN model to forecast solar irradiance in the tropics.
Energy Convers. Manage. 2014, 79, 66–73.
[10] Hohm, D.P., Ropp, M., 2000. Comparative study of maximum power
point tracking algorithms using an experimental, programmable, max-
imum power point tracking test bed, in: Conference Record of
the Twenty-Eighth IEEE Photovoltaic Specialists Conference, 2000,
pp.1699–1702.
 References 147

[11] Inman, R.H., Pedro, H.T., Coimbra, C.F. Solar forecasting methods for
renewable energy integration. Prog. Energy Combust. Sci. 2013, 39,
535–576.
[12] Mahamadou, A.T., Mamadou, B.C., Brayima, D., Cristian, N. Ultra-
capacitors and batteries integration for power fluctuations mitigation
in wind-PV-diesel hybrid system. Int. J. Renew. Energy Res. 2011, 1,
86–95.
[13] Yang, D., Sharma, V., Ye, Z., Lim, L.I., Zhao, L., Aryaputera, A.W.
Forecasting of global horizontal irradiance by exponential smoothing,
using decompositions. Energy 2015, ISSN 0360-5442.
[14] Teleke, S., Baran, M., Bhattacharya, S., Huang, A. Rule-based control
of battery energy storage for dispatching intermittent renewable sources.
IEEE Trans. Sustain. Energy 2010, 1,117–124.
[15] Quesada-Ruiz, S., Chu, Y., Tovar-Pescador, J., Pedro, H., Coimbra,
C. Cloud-tracking methodology for intra-hour {DNI}forecasting. Sol.
Energy 2014, 102, 267–275.
[16] Yang, H., Kurtz, B., Nguyen, D., Urquhart, B., Chow, C.W., Ghonima,
M., Kleissl, J. Solar irradiance forecasting using a ground based sky
imager developed at UC San Diego. Sol. Energy 2014, 103, 502–524.
[17] Guerriero, P., Napoli, F.D., Vallone, G., d’Alessandro, V., Daliento, S.
Monitoring and diagnostics of PV plants by a wireless self-powered
sensor for individual panels. IEEE J. Photovolt. 2016, 6, 286–294.
[18] Nikitidou, E., Kazantzidis, A., Tzoumanikas, P., Salamalikis, V., Bais,
A. Retrieval of surface solar irradiance, based on satellite-derived cloud
information, in Greece. Energy 2015, 90(Part 1), 776–783.
[19] Yang, D., Quan, H., Disfani, V.R., Liu, L., Reconciling solar forecasts:
Geographical hierarchy. Solar Energy 2017, 146, 276–286.
[20] Wu, X., Zhu, X., Wu, G.Q., Ding, W. Data mining with big data. IEEE
Trans. Knowl. Data Eng. 2014, 26, 97–107.
[21] Wang H., Ruan J., Wang G., et al. Deep learning based interval state
estimation of AC smart grids against sparse cyber attacks. IEEE Trans.
Ind. Inf. 2018, 14(11), 4766–4778.
[22] Benjamin, P., Marc, M., Philippe, P., Batista, D. J. Historical trends in
global energy policy and renewable power system issues in Sub-Saharan
Africa: The case of solar PV. Energy Policy 2019, 127, 113–24.
[23] Yongning, Z., Lin, Y., Zhi, L., Xuri, S., Yansheng, L., Su, J. A novel
bidirectional mechanism based on time series model for wind power
forecasting. Applied Energy 2016, 177, 793–803.
148 Machine Learning Techniques for Renewable Energy Resources

[24] Laura, F.-P., Fermín, M., Martín, G.-R., Teresa, L. Assessing energy
forecasting inaccuracy by simultaneously considering temporal and
absolute errors. Energy Convers Manage 2017, 142, 533–546.
[25] Short, W., Packey, D.J., Holt, T.A. Manual for the Economic Evaluation
of Energy Efficiency and Renewable Energy Technologies. Golden,
CO, USA: U.S. Department of Energy Managed by Midwest Research
Institute for the U.S. Department of Energy, 1995.
[26] Gasparin, A., Lukovic, S., Alippi, C. Deep learning for time series
forecasting: The electric load case. arXiv, 2019.
[27] Jafar, A., Lee, M. Performance Improvements of Deep Residual Con-
volutional Network with Hyperparameter Opimizations. Seoul, Korea:
The Korea Institute of Information Scientists and Engineers, 2019,
pp. 13–15.
[28] Sezer, O.B., Gudelek, M.U., Ozbayoglu, A.M. Financial time series
forecasting with deep learning: A systematic literature review.
2005–2009, arXiv, 2019.
[29] Bengio, Y., Simard, P., Frasconi, P. Learning long-term dependencies
with gradient descent is difficult. IEEE Trans. Neural Netw. 1994, 5,
157–166.
[30] Chung, J., Gulcehre, C., Cho, K., Bengio, Y. Empirical evaluation of
gated recurrent neural networks consequence modeling. arXiv, 2014.
[31] Aslam, M., Lee, J., Kim, H., Lee, S., Hong, S. Deep learning models for
long-term solar radiation forecasting considering microgrid installation:
A comparative study. Energies 2019, 13, 147.
[32] Che, Y., Chen, L., Zheng, J., Yuan, L., Xiao, F. A novel hybrid model
of WRF and clearness index-based Kalman filter for day-ahead solar
radiation forecasting. Appl. Sci. 2019, 9, 3967.
[33] Zhang, X., Wei, Z. A hybrid model based on principal component
analysis, wavelet transform, and extreme learning machine optimized by
bat algorithm for daily solar radiation forecasting. Sustainability 2019,
11, 4138.
[34] Lai, C.S., McCulloch, M.D. Levelized cost of electricity for solar
photovoltaic and electrical energy storage. Appl. Energy 2017, 190,
191–203.
[35] Huang, C., Wang, L., Lai, L. Data-driven short-term solar irradiance
forecasting based on information of neighboring sites. IEEE Trans. Ind.
Electron. 2019, 66, 9918–9927.
 7
Application of Optimization Technique
in Modern Hybrid Power Systems

D. Lakshmi1 , R. Zahira2 , C. N. Ravi3 , P. Sivaraman4 , G. Ezhilarasi5 ,

and C. Sharmeela6
1 Academy of Maritime Education and Training (AMET), India
2 BSA Crescent Institute of Science and Technology, India
3 Vidya Jyothi Institute of Technology, India
4 Vestas Technology R&D Chennai Pvt Ltd, India
5 Sri Sairam Institute of Technology, India
6 Anna University, India

E-mail: [email protected]; [email protected];

[email protected]; [email protected]

Abstract
The load frequency control (LFC) is one of the operational problems in the
power system. The supply of reliable and quality power to its users is the
main aim of any electric power utility. The reliability of the power supply
requires that the demand and losses should be equal to power generation
in the power system. This balance is measured with the help of frequency
in the output line. Hence, controlling the generation based on the change
in load is carried out by the LFC. The frequency in the LFC areas and the
tie line connecting the LFC areas should be the same to ensure reliabil-
ity. In this research work, deregulated market structure two-area LFC and
a tie line connecting them are considered. Intelligent algorithms provide
better performance compared to the conventional optimization technique.
For the LFC problem in the two-area power system, this paper considers
the intelligent algorithms, namely flower pollination algorithm (FPA) and
differential evolution (DE). The hybrid DE-FPA algorithm is being developed
to manage the generating plant’s generation based on load variations at the

149
150 Application of Optimization Technique in Modern Hybrid Power Systems

critical damped time. Hybrid intelligent algorithms use the advantages of

the DE and FPA and give better results as compared to conventional control
techniques.
The need of the day is to use renewable energy sources, and, in partic-
ular, in this work, wind energy generation is considered. In wind energy
generators, doubly fed induction generators (DFIGs) are commonly used.
Controlling the electric power output of DFIG is complex and requires better
control techniques. The main control variable of the DFIG is inertia, and it is
controlled by the controller. Control of the frequency of generated power is
controlled by the kinetic energy in the wind turbine blade. LFC consisting
of DFIG is considered and controllers are designed to take care of the
response of conventional generators and DFIG for the change in load demand.
The results of the developed algorithm are compared with the conventional
controller and discussed in the research work.

7.1 Introduction
The main purpose of power system dynamics is to maintain a balance
between production and load. The proper quality of the power supply index
requires that the frequency and voltage be maintained within acceptable
limits. To improve system efficiency and meet demand, two or more regions
are interconnected by a tie line. The load variation within an area affects
the remaining areas, which will reflect the frequency [load frequency control
(LFC) - actual power] and voltage (AVR - reactive power changes).
In this work, we are considering the real power output of the generating
unit (LFC). LFC plays an important role in power system operation. The
objective of LFC is to control the real power of the generation to meet the
dynamic change in load demand. Thus, LFC regulates the MW power output
of generators within a control area in response to the changes in system
frequency and tie-line power, called area control error (ACE) [1].
In the literature, there are various types of control methods or log-
ics/algorithms proposed by various researchers for LFC regulation. In the
early stages integral, proportional–integral (PI), and proportional-integral
derivative (PID) have been used. Among them, PI is mostly used because
of its simplicity and is tuned using the Ziegler-Nichols method [2]. There
are numerous artificial intelligence based controllers used for LFC regulation
to reduce the settling time, undershoot, and overshoot. Some of the artifi-
cial intelligence based controllers are two-stage neural network fuzzy logic,
 7.2 Modern Power System 151

neuro-fuzzy controllers, etc., and they are more suitable for modern complex
power systems with nonlinear characteristics [3, 4].
In the modern era with new technologies in the power industry, the
electrical power system is transferred to more emphasis on restructuring and
deregulation. The deregulated power system splits the conventional single
power system (from generating station to distribution) into three different
sections/verticals, namely generation companies in short form GENCOs,
transmission companies in short TRANSCOs, and distribution companies in
short DISCOs. These three different verticals are managed by independent
service or system operator (ISO). Various researchers have analyzed the LFC
problem in a deregulated environment [5, 6]. The DISCO participation matrix
(DPM) based LFC regulation for area control error participation factor (APF)
in the deregulated electrical power systems was discussed in [7].
In general, electrical power system consists of energy generation from
conventional as well as non-conventional energy resources, i.e., hydro, coal,
nuclear, solar, wind, biomass diesel, etc.; nowadays, non-conventional energy
resources are gradually replacing conventional sources because they are free
from environmental pollution and emissions, zero or lesser running cost, etc.
Wind energy is the domestic source of energy and is available in abundance.
In a deregulated power system, LFC becomes complicated when wind units
are integrated with the system.
Inertial control of the wind units provides coordination with conventional
systems [8, 9]. A wind unit supports the system by inertial control, pitch
control, and speed control [10]. Participation of doubly fed induction gener-
ator (DFIG) is analyzed through modified inertial control, which reflects the
frequency deviation by making use of kinetic energy of the turbine blades to
improve the frequency [11, 12].
Research on three areas deregulated power systems with optima PID
using imperialist competitive algorithms was discussed in [13]. Dynamic par-
ticipation of DFIG for hydrothermal deregulated power systems with fuzzy
controllers was discussed in [14]. Participation of DFIG in an asynchronous
power system using feedback control is discussed in [15] and the use of
intelligent algorithms such as firefly for deregulated power systems in [16].

7.2 Modern Power System

The electric power industry has operated as a vertically integrated utility
(VIU) for many years. VIU is self-sufficient in terms of power generation,
transmission, and distribution. VIU owns all levels of the power supply circle,
152 Application of Optimization Technique in Modern Hybrid Power Systems

including generation, transmission, and distribution and so has the sole right
to set the price of electric energy. As a result, VIU refers to the electric power
industry that is monopolized by a single utility.
Furthermore, while such interconnection improves reliability, it comes
with some drawbacks, including inefficient production, large losses, decaying
infrastructure, and poor management. As a result, deregulation in the electric
power industry is required, as is the creation of a separate autonomous
company for generation, transmission, and distribution.

7.2.1 Deregulated Power System

Deregulation and market-based competition are taking place in the electric
power industry. Retail power sales and the separation of power generation
and related services are among the responsibilities of this transformation
(http://nptel.ac.in/courses).
Deregulation of the power industry was motivated by several factors.
1. Electricity rates are lower as a result of competition.
2. Customers have a wide range of options, resulting in greater plans,
reliability, and quality.
3. Service enhancement, cost reduction, and profit maximization.
4. Innovation in the operation of GENCOs and DISCOs to improve service,
resulting in cost savings and profit maximization.
Deregulated power system components are shown in Figure 7.1.

7.2.2 Components of Deregulation

Several new entities have emerged in the power industry as a result of
deregulation.

Figure 7.1 Components of deregulation.

 7.2 Modern Power System 153

• GENCOs (Generating Companies):

GENCOs are a group of one or more generating units that are sold as a whole
through a single ownership company structure. As a result, GENCOs produce
electricity and have the option to sell it to companies with which they have
concluded sales agreements.
• TRANSCOs (Transmission Companies):
The delivery of power from GENCOs to DISCOs or retailers is the respon-
sibility of TRANSCO. The TRANSCO maintains and operates the transmis-
sion system in certain geographical regions to assure the overall reliability of
the electric power system.
• DISCOs (Distribution Company):
The electricity is distributed to customers through a DISCO. The DISCO
purchases wholesale electric power from GENCOs directly or through spot
markets and distributes it to end customers. The DISCO is responsible for
maintaining the availability and reliability.
• RESCOs (Retail Energy Service Company):
A RESCO is a company that sells electricity. RESCO buys power from
GENCOs and sells it directly to customers.
• Customer:
In deregulation, consumer has a variety of options, including bidding for
power on the spot market or contracting directly from a DISCO or a GENCO.
• ISO (Independent Service Operator):
An ISO is an independent organization that is not involved in market trad-
ing. It is responsible for assuring the electricity system’s dependability and
security. The ISO encourages ancillary services such as frequency regu-
lation, emergency reserve supply, and reactive power from other system
organizations.
• Ancillary services:
Ancillary services can be defined as “the services associated with those
activities on the deregulated power system that are necessary to support the
transmission of power by maintaining reliable operation and ensuring the
required degree of quality and safety.” North American Electric Reliability
Corporation (NERC) has suggested 12 ancillary services. Their performances
are as follows:
• to manage generation and load balance (control of frequency);
154 Application of Optimization Technique in Modern Hybrid Power Systems

• to ensure reactive power support and the voltage;

• to retain transmission reserves and generation;
• to be ready in the event of an emergency (i.e., control of stability and
system restart).
LFC is one of the ancillary services provided to maintain real power balance.
The LFC also maintains the net scheduled area interchange and keeps the sys-
tem frequency deviation at zero. Thus, the performance of LFC is important
and a few of the qualities do.

1. On a market-price basis, LFC can be traded as an ancillary service.

Because of the flexibility of participation in LFC, both generation and
loads will be more dynamically regulated.
2. The nature of tie-line transactions is very dynamic. To guarantee that
tie-line transactions remain stable, this must be taken into account when
designing LFC controllers.
3. LFC should be able to track the generators’ participation, which is also
dynamic because DISCOs have the opportunity to contract with different
GENCOs.

7.2.3 Types of Transactions

There are three types of structures based on the transaction among DISCOs
and GENCOs. They are (1) pool-co or charged transaction (contract between
DISCO and GENCO of the same area), (2) bilateral transaction (DISCOs
have a contract with a GENCO of another area), and (3) charged-cum-
bilateral transaction [18]. In this chapter, we will consider only the bilateral
transaction in detail.

7.2.3.1 Bilateral transactions

A DISCO can deal with any GENCO in the same area or another area. As
a result, the idea of physical control area has been superseded with virtual
control area (VCA). The GENCO and DISCO affiliates are encircled by
VCA’s contract, which has adjustable borders. Each DISCO is in charge
of managing the tie-line power exchange with its neighbors by securing
transmission and generation as needed in a bilateral transaction (Bevrani et
al. 2004). The behavior of the power system before and after deregulation is
shown in Figure 7.2.
 7.2 Modern Power System 155

Figure 7.2 Structure of power system before and after deregulation.

7.2.3.2 DPM and APF

The DPM is used to depict the contracts that are formed between a DISCO
and a GENCO [4]. Equation (7.1) shows the DPM for the power system,
where “ijth” entry of the matrix (called generation participation factor) repre-
sents the fraction of the total contract by the DISCO “j” with a GENCO “i,”
and the sum of all the entries in a particular column is unity. Equation (7.1)
gives the DPM matrix
⎡ ⎤
gpf 11 gpf 12 gpf 13 gpf 14
⎢ ⎥
4 
⎢ gpf 21 gpf 22 gpf 23 gpf 24 ⎥ 4
DPM = ⎢ ⎢ gpf
⎥ and
⎥ gpf ij = 1 (7.1)
⎣ 31 gpf 32 gpf 33 gpf 34 ⎦ j=1 i=1
gpf 41 gpf 42 gpf 43 gpf 44

where gpf ij = Demand of DISCO “j” from GENCO “i”

Total Demand of DISCO “j” .
The difference between an area’s actual interchange and its scheduled
interchange is known as ACE since there are several GENCOs in each area
keeping the frequency at its specified value (ACE). Thus, control operations
require ACE, which must be divided among them in accordance with their
156 Application of Optimization Technique in Modern Hybrid Power Systems

participation in the LFC. ACE participation factors (Apf) indicate the ele-
ments that provide ACE to the participating GENCOs, and the sum of the
Apf in a given area must equal unity as shown in following equation:

m
Apf j = 1. (7.2)
j=1

The generation of a GENCO in MW expressed in terms of gpfs is shown

in the following equation:

m
ΔPG = gpf ij ΔPLoj . (7.3)
j=1

Equation (7.4) shows the contract of DISCO with GENCO during the
steady-state condition.

n
ΔPLi,Loc = ΔPLi . (7.4)
i=1

The scheduled steady-state tie-line power flow from area i to area j is

given by the following equation:
 n m m n
ΔPtie1,2schedule = gpfij ΔPLoj − gpfij ΔPLoj .
i=1 j=i i=j j=1
(7.5)
In bilateral transactions, the tie-line power will not settle at zero, but rather
at a value determined by a bilateral contract between GENCOs in one area
and DISCOs in another.

7.2.4 Renewable Energy Sources

While harnessing nature’s power is typically considered new technology, it
has long been used for heating, transportation, lighting, and other purposes.
The wind has propelled ships across the oceans and mills that process
grain.

7.2.4.1 Doubly fed induction generator

Electrical power generation from renewable sources, such as wind, is drawing
increased interest as a result of environmental concerns and a scarcity of
traditional energy sources.
 7.2 Modern Power System 157

In general, wind turbines do not work with traditional energy sources.

However, with advancements in controller technology, kinetic energy stored
in wind turbines may be retrieved using variable speed generators. Wind
turbines based on DFIG can run at varying speeds and synchronize with the
frequency regulation of conventional energy sources. A dynamic model of
the DFIG wind unit is shown in Figure 7.2.
In the emulation control of DFIG, a control signal is provided to modify
the power setpoints ΔPf *, which is a function of deviation and rate of change
of frequency. To get the most power, the controller will aim to keep the turbine
at its ideal speed. ΔPω *. The controller provides the set power depending on
the measured speed and measured electrical power. From eqn (7.6), ΔPN C
has two components they are additional reference points based on frequency
change as shown in eqn (7.7) and the PI controller calculates the optimum
turbine speed as a function of wind speed:

ΔPNC = ΔPf∗ + ΔPω∗ (7.6)

1
ΔPf∗ = ΔX2 . (7.7)
Rw
Contribution of DFIG toward system inertia is given by the following
equation:
2H dΔf
= ΔPG + ΔPNC − ΔPtie1,2 − ΔPD − DΔf. (7.8)
f dt
From Figure 7.3, it is seen that additional reference power setting based
on the change in frequency using a washout filter with a time constant Tw and

Figure 7.3 DFIG-based wind unit model.

158 Application of Optimization Technique in Modern Hybrid Power Systems

Rw the droop constant and ΔX 2 measured change in frequency to which the

wind system is connected.
DFIG wind unit reacts to frequency deviations during transients by using
the stored kinetic energy and does not provide permanent system frequency
deviation. DFIG inertia contributes to the system, and the real power injected
by the wind unit is ΔPN C . This injected power is compared with ΔPN Cref
to obtain maximum power output.

7.2.4.2 DFIG in deregulated power system

Overall transfer modeling of deregulated power systems is shown in
Figure 7.4. It shows the transfer function model of the system, in which R1 ,
R2 , R3 , and R4 are the governor regulation parameters of thermal units for
area 1 and area 2 in Hz/MW, respectively. Each area consists of a two-speed
governing system and two reheat turbines. The transfer function model of the
speed governor is given by the following equation:
1
GGj (s) = (7.9)
1 + sTGj
where TGj is the time constant of the jth governor. The transfer function
model of the steam turbine is given by the following equation:
1
GT (s) = (7.10)
1 + sTT j
where TT j is the time constant of the jth reheat turbine, KRj is the gain
constant of reheater, and TRj is the time constant of the reheater. The transfer
function of power system is as shown in the following equation:
KP i
GP i (s) = (7.11)
1 + sTP i
where KP i is the gain of the ith area power system and TP i is the time constant
of the ith area power system. The values of KP and TP are given by the
following equations:
1
KP i = (7.12)
Di
2H
TP i = (7.13)
f Di
where H is the per unit inertia constant, f is system frequency, and Di
is expressed as percent change in load by percent change in frequency.
 7.2 Modern Power System 159

Figure 7.4 Transfer function model of two areas’ deregulated power system.
160 Application of Optimization Technique in Modern Hybrid Power Systems

The scheduled steady-state tie-line power flow from area 1 to area 2 is the
difference between the demand of DISCOs in area 2 from GENCOs in the
area 1 and the demand of DISCOs in area 1 from the GENCOs in area 2,
which is given in the following equation:
2 4 4 2
ΔPtie1,2schedule = gpfij ΔPLjo − gpfij ΔPLoj .
i=1 j=3 i=3 j=1
(7.14)
The tie-line power flow from area 1 to area 2 is the product of the tie-line
coefficient and the difference between the change in frequency in area 1 and
the change in frequency in area 2, as given by the following equation:
2ΠT12
ΔPtie1,2actual = [Δf1 − Δf2 ] . (7.15)
s
Equation (7.16) gives the error in the tie-line power flow from area 1 to
area 2, which is the difference between the actual and scheduled value of the
tie-line power:

ΔPtie1,2error = ΔPtie1,2actual − ΔPtie1,2schedule . (7.16)

Equation (7.17) gives the error in tie-line power flow from area 2 to area 1:

ΔPtie,2,1error = a12 ΔPtie1,2error (7.17)

since both the areas are assumed to be identical a12 = -1.

Equations (7.18) and (7.19) give the ACE of area 1 and area 2

ACE1 = B1 Δf1 + ΔPtie12error (7.18)

ACE2 = B2 Δf2 + a12 ΔPtie12error . (7.19)

As there is more than one GENCO in each area, ACE signals have to
be given in proportion to their participation in LFC. The coefficient that
distributes the ACE to all GENCOs is called “ACE participation factor”
(apf). The summation of Apf should be unity for each area. Hence, ACE
participation factors for area 1 are apf11 and apf12 , and apf21 and apf22 for
area 2. The generated power (or) contracted power supplied by the GENCOs
in (MW) is given as shown in the following equation:


4 
4
ΔPgei = gpf ij ΔPLoj − apf ij ΔPU Ci . (7.20)
i=1 j=1
 7.3 Optimization Techniques and Proposed Technique 161

7.3 Optimization Techniques and Proposed Technique

The different techniques are proposed for tuning the PI controller.

7.3.1 Controllers
The tuning of controllers is more important for any system design; if tuning
is not done properly, the characteristics get affected and the system becomes
unstable. For the system under consideration, a Ziegler-Nichols-tuned con-
ventional PI controller is used.

7.3.2 PI Controller
When the change in error is high, a proportional controller will be beneficial
since it increases transient performance. When the error is small, the integral
control mode is effective, and the steady-state is improved. Although it has
the advantage of reducing overshoot, the derivative control mode increases
noise and makes the system less stable due to its high sensitivity [17]. As
the load changes, the derivative mode leads the system to become unstable.
Figure 7.5 shows the structure of a PI controller, with KP and KI denoting
proportional and integral gain values, respectively [18].
In this chapter, because of simplicity, flexibility, and easy design PI
controller has been chosen. Eqn (7.21) shows the mathematical model of the
PI controller t
UPI = KP ACEi + KI ACEi dt (7.21)
0
where U P I is the controlled output of the PI controller, KP is the proportional
gain, and KI is integral gain, and ACE of the concerned area is ACE.

Figure 7.5 Transfer function model of PI controller.

162 Application of Optimization Technique in Modern Hybrid Power Systems

7.3.3 Artificial Optimization Algorithm for Tuning PI

While designing the optimized PI controller, by ensuring system stabil-
ity, the objective function is defined based on required specifications and
parameter values [19]. In this chapter, the integral square error (ISE) is
utilized as the objective function for optimizing the PI controller parameters.
By employing this index, errors are decreased while the response time is
increased.
The performance index for minimization of the error is shown in the
following equation:
tsim
objfn = ISE = (|Δf1 | + |Δf2 | + |ΔPtie1,2 |)2 dt (7.22)
0
where Δf 1 and Δf 2 are the changes in frequency of area 1 and area 2,
respectively, and ΔPtie1,2 is the change in tie-line power flow.

7.3.3.1 Differential evolution

Differential evolution (DE) is capable of solving real-world problems that are
nonlinear, noncontinuous, and nondifferential. DE places a greater empha-
sis on mutation than recombination or crossover when compared to other
population-based meta-heuristic techniques [20]. It mutates vectors by select-
ing a pair of vectors from the same population at random. The mutation
directs the vector toward the global optimum. The difference between two
randomly picked vectors has a distribution determined by these vectors’
distribution. As a result, DE can work more reliably and as a generic global
optimizer [21]. DE is based on the idea of a population of vectors, with each
vector representing a set of decision variables. Choice variables are chosen
based on their impact on the problem to be solved. These choice variables
must be encoded, and a set of starting values must be picked from the
solution space. Mutation and recombination are used to create new vectors.
The selection technique determines the optimal vectors based on the selection
criterion [22–24].
A. Encoding
Encoding is defined as the process of converting a set of decision variables
into a vector and an objective function into a fitness function. The capacity of
DE to cope with floating-point and mixed-integer data simplifies the process
of encoding decision variables into vectors. The number of option variables
determines the size of the vector, and each vector represents a single solution
from the problem’s solution space.
 7.3 Optimization Techniques and Proposed Technique 163

B. Mutation
The goal of mutation is to increase search variety in the parameter space while
also directing current vectors with a reasonable level of parameter variation
in a way that produces better results at the right moment. It maintains the
search’s robustness while also exploring new parts of the search area. There
are four types of mutation [8]. DE/rand/1/bin – Yi = Xr1 + F*(Xr2 – Xr3 )

DE/rand/2/bin – Yi = Xr1 + F*(Xr2 – Xr3 ) + F*(Xr4 – Xr5 )

DE/best/1/bin – Yi = X best + F*(Xr1 – Xr2 )
DE/best/2/bin – Yi = X best + F*(Xr1 – Xr2 ) + F*(Xr3 – Xr4 )
r1 = r2 = r3 = r4 = r5 are randomly selected.
C. Recombination
By creating offspring individuals from current individuals or vector parame-
ters, recombination or crossover tries to reinforce previous successes. If the
freshly formed individual is to be recombined, the cross-over constant is
employed. Binomial and exponential cross-over are the two forms of cross-
over. In the binomial approach, a random number is created to form the
trailing vector. The adjusted vector variable is used if the value is smaller
than the cross-over constant; otherwise, the target vector variable is used.
D. Selection
The fitness of the trail and target vectors is compared, and the vector with the
lowest objective value for the next generation is chosen. This assures that the
population size does not fluctuate over time.

7.3.3.2 Flower pollination algorithm

Flower pollination algorithm (FPA) is a population-based algorithm inspired
by nature. The fundamental goal of this FPA is to ensure that plant species
reproduce optimally by surviving the fittest of flowering plants [25, 26].
There are millions of flowering plants in our universe, with flowering species
accounting for 80% of the total. A flower’s ultimate goal is to proliferate
by pollinating other flowers. Pollen transfers from one flower to another on
the same plant (self-pollination-abiotic) or another plant (cross-pollination-
biotic) [27]. This transformation occurs by pollinators such as wind, birds,
insects, bats, and other animals. FPA performs better when compared with
others in terms of accuracy and convergence speed.
164 Application of Optimization Technique in Modern Hybrid Power Systems

The following four rules were employed to explain the concept of flower
pollination.
1. Cross and biotic pollination are considered global pollination, and
pollinator movement is compared to levy flight movement.
2. Abiotic and self-pollination take place as local pollination.
3. Pollinators such as birds and insects acquire flower constancy, which is
proportional to the resemblance of the two flowers involved and equal to
the reproduction likelihood.
4. The chance of switching from local to global pollination or vice versa
can be regulated, p = 0.7.
In global pollination, flower pollen is carried by pollinators like birds, wind,
and insects. This global pollination, i.e., rules 1 and 3 can be written as in the
following equation:
 
∗
xk+1
i = x k
i + γL (λ) g − x k
i (7.23)

where xi k is the flower i at iteration k, and g* is the current best solution

among the solutions for the current iteration. Here, γ is the scaling factor
used to control the step size and its value is 0.3 L (λ). Pollination strength is
measured by the step size parameter in specific levy-flights movements. Levy
distribution is complicated by the fact that pollinators fly over a great distance
with varying distances. It is used as shown in the following equation:
λΓ (λ) sin (πλ/2) 1
L≈ , (S >> 0) (7.24)
π S 1+λ
where Γ(λ) is the standard gamma function and levy distribution will be valid
for longer steps S > 0.
Rules 2 and 3 are for local pollination and they are shown in the following
equation:  
xk+1
i = x k
i + ε x k
j + x k
m (7.25)
where ε is local random whose values lie between 0 and 1.

7.3.3.3 Hybrid algorithm

A superior controller is required to control the frequency and tie-line power
deviations; hence, to improve the system dynamic performance, a novel
optimization technique is needed [28].
 7.4 Simulation Results and Discussion 165

A. Requirement for Hybrid Algorithm

A hybrid algorithm combines two or more approaches to solve the problem,
picking one (depending on the data) or switching between them throughout
the procedure.
B. Hybrid DE-FPA
DE focuses on vector mutation by using a pair of vectors from the same
population that is randomly selected. The search for the global best solution
becomes more diverse as the number of mutations increases.
FPA is a simple and efficient pollination procedure with no mutation. DE
mutation is merged with FPA to get a hybrid DE-FPA method to solve the
LFC in the deregulated environment.

7.3.3.4 Design of a hybrid DE-FPA algorithm for LFC

DE-FPA is implemented in deregulated as follows.
Step 1: Initialize the flower population (variables of LFC).
Step 2: Find each flower’s fitness function and get the global flower
that has minimum ISE from the population.
Step 3: For each flower, initiate a random number.
Step 4: If the random number is below switch probability, go to step 5;
else, apply cross-pollination (global) and go to step 6.
Step 5: Implement self-pollination (local).
Step 6: Execute mutation for enhancement of diversity.
Step 7: Repeat steps 2-6 till the stopping criterion is satisfied.
Step 8: Get the fittest solution.

7.4 Simulation Results and Discussion

Case Study – Bilateral Contract Between Areas:

Here, a DISCO may transact power from a GENCO of its area or with
other areas. Equation (7.26) shows DPM which is considered participation
of GENCOs in LFC is shown by the following ACE participation factor Apfs
in two areas, i.e., Apf11 = 0.75, Apf12 = 1 - 0.75, Apf21 = 0.5, Apf22 = 1 - 0.5.
DISCO’s demands (pu MW) are ΔPLo1 = ΔPLo2 = ΔPLo3 =
ΔPLo4 = 0.1.
166 Application of Optimization Technique in Modern Hybrid Power Systems

ΔPU C1 = 0 and ΔPU C2 = 0. In this case, uncontracted load is zero.

Hence, there is no contract violation. ΔPtie1,2schedule = -0.02 (pu MW), using
Equation (7.26) scheduled tie-line error under steady-state.
⎡ ⎤
0.5 0.25 0.5 0.3
⎢ ⎥
⎢ 0.2 0.25 0.2 0 ⎥
DPM = ⎢ ⎢ 0 0.25 0.2 0.7 ⎥ .
⎥ (7.26)
⎣ ⎦
0.3 0.25 0.1 0
GENCO’s power generation (pu MW) by using Equation (7.3) is shown
below:
ΔPge1 = 0.1545, ΔPge2 = 0.0665, ΔPge3 = 0.135, and ΔPge4 = 0.0655.
Figures 7.6 and 7.7 show the frequency deviation (Hz) in areas 1 and 2.
Table 7.1 gives a comparison of PI and hybrid DE-FPA tuned PI controllers
in terms of undershooting, overshoot, and settling time. From the dynamic
responses and comparison table, conventional PI controllers have more oscil-
lations and overshoot. FPA-PI is superior and faster than PI controllers.
Adding DFIG in both areas stabilizes the frequency response of the system.
From Figure 7.6, it is seen that the settling time for PI controller is
47 seconds, whereas for DE-FPA tuned PI controller, it is 11 seconds. Sim-
ilarly, from Figure 7.7, it is observed that the settling time of the DE-FPA
tuned PI controller is 9 seconds, whereas for the conventional PI controller,
it is 35 seconds. As a result, the hybrid DE-FPA tuned PI controller is better
when compared with the PI controller.

Figure 7.6 Frequency deviations in area 1.

 7.5 Conclusion 167

Figure 7.7 Frequency deviations in area 2.

Table 7.1 Frequency deviation of each area.

Controllers Settling
used Area Peak shoot (Hz) Undershoot (Hz) time (seconds)
Frequency deviation (Hz)
PI 1 0.38 −0.52 47
2 0.98 −1.7 43
DE-FPA 1 0.28 −0.41 11
2 0.7 −1.45 9

From Table 7.1, the performance of the PI and DE-FPA tuned PI con-
trollers in area 1 declares that with peak overshoot and peak undershoot as
the performance index; the DE-FPA tuned PI controller has 27% and 26%
improvement over the conventional PI controller. The settling time of the
DE-FPA tuned PI controller is around 82% less than the PI controller.
Similarly, for area 2 with peak shoot and undershoot as the performance
index, the DE-FPA tuned PI controller showed an improvement of 27% and
23% over conventional PI controllers. The settling time of the DE-FPA tuned
PI controller is found to be 77% better when compared with the PI controller.

7.5 Conclusion
The deregulated power system has bilateral transactions for the effective
utilization of available resources. The RESs are intermittent for the continuity
of the power supply. Wind energy systems for electric power generation are
considered in this work. The major issue to solve in the bilateral transaction
168 Application of Optimization Technique in Modern Hybrid Power Systems

violation during practical aspects is that it depends on the demand variation

and available resources. The solution to identify and solve this is the mea-
surement of frequency deviation in the areas where the violation is taking
place. For the deregulated power system two-area LFC with wind energy
generator having a DFIG simulation model is developed. The frequency
deviation of areas 1 and 2 is measured and controlled by using a PI controller.
Conventional techniques tuning to find the coefficient of PI controller are
measured. Hybrid intelligent algorithm DE-FPA is developed and used to
find the coefficient of PI controller for the better performance of the LFC.
The performance measures are overshoots, settling time, and oscillations
before settling. Developed hybrid intelligent algorithms give better perfor-
mance compared with the conventional PI controller. The frequency deviation
waveforms of area 1, area 2, and tie lines are discussed for the conventional
and hybrid intelligent algorithm. The hybrid intelligent algorithm DE-FPA
gives better performance.

Appendix

Table 7.2 System parameters.

Parameters, symbols (units) Values
Governor time constants Gj (s) 0.08
Rating, Pri (MW) 2000
Di system damping coefficient (pu MW/Hz) 8.333× 10−3
Power system time constant, TP i (s) 20
Turbine time constants, TT j (s) 0.3
Gain constant of power system, KP i (Hz/pu MW) 120
Bias factor, Bi (pu MW/Hz) 0.425
Speed regulation, Rj (Hz/pu MW) 2.4
Inertia constant, Hi (s) 5
Tie-line power constant, T 12 0.0826
DFIG gain constant (Kwi1 , Kwi2 ) 0.5
DFIG gain constant (Kwp1 , Kwp2 ) 1
System frequency, f (Hz) 60
Wind inertia constant (He1 , He2 ) (pu MW s) 3.5
DFIG time constant (Ta1 , Ta2 ) s 0.2
DFIG time constant (Tr 1 , Tr 2 ) s 15
DFIG time constant ((Tw 1 , Tw 2 ) s 6
 References 169

References
[1] lgerd, O. I. “Electric energy systems theory: An introduction.” 1982.
[2] Kumar, J., Ng, K-H., and Sheble, G. “AGC simulator for price-based
operation-part II: Case study results.” IEEE Transactions on Power
Systems 12.2 (1997): 533–538.
[3] Christie, R. D., and Bose, A. “Load frequency control issues in power
system operations after deregulation.” IEEE Transactions on Power
Systems 11.3 (1996): 1191–1200.
[4] Donde, V., Pai, M. A., and Hiskens, I. A. “Simulation and optimization
in an AGC system after deregulation.” IEEE Transactions on Power
Systems 16.3 (2001): 481–489.
[5] Aditya, S. K., and Das, D. “Battery energy storage for load frequency
control of an interconnected power system.” Electric Power Systems
Research 58.3 (2001): 179–185.
[6] Selvaraju, R. K., and Somaskandan, G. “Impact of energy storage units
on load frequency control of deregulated power systems.” Energy 97
(2016): 214–228.
[7] Sasaki, T., Kadoya, T., and Enomoto, K. “Study on load frequency con-
trol using redox flow batteries.” IEEE Transactions on Power Systems
19.1 (2004): 660–667.
[8] Chidambaram, I. A., and Paramasivam, B. “Optimized load-frequency
simulation in restructured power system with redox flow batteries and
interline power flow controller.” International Journal of Electrical
Power & Energy Systems 50 (2013): 9–24.
[9] Shayeghi, H. A. S. H., Shayanfar, H. A., and Jalili, A. “Load frequency
control strategies: A state-of-the-art survey for the researcher.” Energy
Conversion and Management 50.2 (2009): 344–353.
[10] Tyagi, B., and Srivastava, S. C. “A LQG based load frequency con-
troller in a competitive electricity environment.” International Journal
of Emerging Electric Power Systems 2.2 (2005).
[11] Zamani, A. A., et al. “Optimal fuzzy load frequency controller with
simultaneous auto-tuned membership functions and fuzzy control rules.”
Turkish Journal of Electrical Engineering & Computer Sciences 22.1
(2014): 66–86.
[12] Zamani, A. A., et al. “Optimal fuzzy load frequency controller
with simultaneous auto-tuned membership functions and fuzzy con-
trol rules.” Turkish Journal of Electrical Engineering & Computer
Sciences 22.1 (2014): 66–86.
170 Application of Optimization Technique in Modern Hybrid Power Systems

[13] Demiroren, A., and H. L. Zeynelgil. “GA application to optimization

of AGC in three-area power system after deregulation.” International
Journal of Electrical Power & Energy Systems 29.3 (2007): 230–240.
[14] Bhatt, P., Roy, R., and Ghoshal, S. P. “Optimized multi area AGC
simulation in restructured power systems.” International Journal of
Electrical Power & Energy Systems 32.4 (2010): 311–322.
[15] Fathima, A. P., and Abdullah Khan, M. “Design of a new market
structure and robust controller for the frequency regulation service in the
deregulated power system.” Electric Power Components and Systems
36.8 (2008): 864–883.
[16] Zahira, R., et al. “Modeling and simulation analysis of shunt active filter
for harmonic mitigation in islanded microgrid.” Advances in Smart Grid
Technology. Springer, Singapore, 2021. 189–206.
[17] Lakshmi, D., and Zahira, R., “Load frequency control in deregulated
power system.” International Journal of Research in Arts and Science 5
(2019): 124–133. Special Issue in Holistic Research Perspectives.
[18] Lakshmi, D., et al. “Flower pollination algorithm in DPS integrated
DFIG for controlling load frequency.” 2020 International Conference
on Power, Energy, Control and Transmission Systems (ICPECTS). IEEE,
2020.
[19] Yang, X.-S. Engineering Optimization: An Introduction with Meta-
heuristic Applications. John Wiley & Sons, 2010.
[20] Jin, L., et al. “Robust delay-dependent load frequency control of wind
power system based on a novel reconstructed model.” IEEE Transac-
tions on Cybernetics (2021).
[21] Panda, A., et al. “A PSO based PIDF controller for multiarea multisource
system incorporating dish stirling solar system.” 2019 International
Conference on Intelligent Computing and Control Systems (ICCS).
IEEE, 2019.
[22] Tang, J., Liu, G., and Pan, Q. “A review on representative swarm
intelligence algorithms for solving optimization problems: Applications
and trends." IEEE/CAA Journal of Automatica Sinica 8.10 (2021):
1627–1643.
[23] Zahira, R., and Lakshmi, D. “Control techniques for improving quality
of power-A review.” International Journal of Research in Arts and Sci-
ence 5 (2019): 107-123. Special Issue in Holistic Research Perspectives
[Volume 4].
 References 171

[24] Pahadasingh, S. “TLBO based CC-PID-TID controller for load fre-

quency control of multi area power system.” 2021 1st Odisha Inter-
national Conference on Electrical Power Engineering, Communication
and Computing Technology (ODICON). IEEE, 2021.
[25] Chen, X., et al. “A survey of swarm intelligence techniques in VLSI
routing problems.” IEEE Access 8 (2020): 26266–26292.
[26] Zahira, R., Lakshmi, D., and Ravi, C. N. “Power quality issues in
microgrid and its solutions.” Microgrid Technologies (2021): 255–286.
[27] Rehman, U., et al. “Load frequency management for a two-area sys-
tem (thermal-PV & Hydel-PV) by swarm optimization based intelli-
gent algorithms.” 2021 International Conference on Emerging Power
Technologies (ICEPT). IEEE, 2021.
 8
Application of Machine Learning
Techniques in Modern Hybrid Power
Systems – A Case Study

B. Koti Reddy1 , Krishna Sandeep Ayyagari2 ,

Raveendra Reddy Medam3 , and Mohemmed Alhaider4
1 Department of Atomic Energy, India
2 Department of Electrical & Computer Engineering, The University
of Texas at San Antonio, USA
3 Department of EEE, Maturi Venkata Subba Rao Engg (MVSR) College,

India
4 College of Engineering at Wadi Addawaser, Prince Sattam bin Abdulaziz

University, Saudi Arabia

E-mail: [email protected]; [email protected];
[email protected]; [email protected]

Abstract
Due to the rapid adoption of intelligent power electronic devices and digital
technologies, traditional vertically designed power systems are being phased
out and replaced by modern hybrid power systems. Since its inception, the
power system has undergone numerous changes that have increased system
efficiency, increased the share of renewable energy, and made it easier to
control. However, such a rapid revolution in electrical power systems during
the current Industrial Revolution has increased its complexity. The primary
concerns are cybersecurity, forecasting supply and demand, optimal power
allocation, power quality maintenance, and a skilled workforce shortage.
Digital tools aid in load management and the optimization of various power
resources. Modern hybrid power systems, artificial intelligence techniques
such as machine learning and optimization algorithms, are emerging in

173
174 Application of Machine Learning Techniques in Modern Hybrid Power

the power sector for better control. Nonetheless, little research has been
conducted on machine learning applications in industries with integrated
power resources. Machine learning techniques will be used in the industries to
forecast supply and demand, make the best use of energy resources, etc. This
chapter aims to discuss the use of machine learning techniques in modern
hybrid power systems. A well-known large industry with multiple energy
resources has been considered for this purpose. All components of the power
system network are modeled, and simulations are run to determine the best
way to use them under various generation and load scenarios, weather condi-
tions, and financial conditions. For the case study considered, the simulated
results are validated using field data and ETAP software, and the results are
encouraging.
Keywords: Energy resources, machine learning (ML), artificial intelligence
(AI), optimization, ETAP, modern hybrid power systems (MHPS).
Notation: Upper-case (lower-case) boldface will denote matrices (column
vectors); (·)τ for transposition; (·)∗ for complex-conjugate; and (·)−1√ for
inverse. Re denotes the real part of the complex number, and j := −1
the imaginary unit. For a given N ×1 vector x, diag (x) returns N ×N
matrix with the elements of x in its diagonal. Finally, IN denotes N ×N
identity matrix; and 0N , 1N the N-dimensional vectors with all zeroes and
ones, respectively, and 0N ×M is N ×M matrix with all zeroes.

8.1 Introduction
The electrical power system (EPS) has undergone various modifications since
its inception. This is due to rapid technological improvements, computer
percolation, and intelligent electronic devices (IED) in EPS. Furthermore,
while the rising use of renewable energy (RE) such as wind and solar may
have given conventional EPS a facelift, it has made the system more complex
due to its intermittency and reliance on weather conditions [1]. Figure 8.1
depicts the main distinction between standard EPS and modern hybrid power
system (MHPS).
In addition, the EPS has modified its previous vertical shape of power
flow from top (generation) to bottom (consumption) (T2B) with bottom-to-
top (B2T) approach of inventory flow from consumer to generator to a two-
way approach, as illustrated in Figure 8.2.
Furthermore, the evolution of de-regulated electricity markets (DEM),
distributed generation (DG), microgrids (MGs), smart grids (SGs),
 8.1 Introduction 175

(a) Traditional EPS.

(b) MHPS.
Figure 8.1 Electrical power system. (a) EPS. (b) MHPS.
176 Application of Machine Learning Techniques in Modern Hybrid Power

Figure 8.2 Power and inventory flow (traditional at left and MHPS at right).

smart appliances (SAs), and energy storage systems (ESS) has changed the
name of EPS to MHPS. Moreover, there are many expectations on EPS
globally in terms of the five Es, which are energy, efficiency, environment,
equipment, and economics. In addition to meeting these expectations, the
MHPS provides many benefits to humanity, including the utilization of
abundantly available renewables, de-centralized EPS, reduced operation and
maintenance costs, reduced losses, environmental protection, and so on.
However, such a rapid revolution in EPS during the current Industrial Revo-
lution (IR 4.0) has increased its complexity. The technical issues involved in
MHPS, as well as their mitigation strategies, are discussed in the following
sections of this chapter.

8.2 Technical Issues in Modern Hybrid Power Systems

The accelerated advancement of EPS has resulted in rapid industrial
development and the extension of power supply to every corner of the
globe, with an increased share of renewables. However, due to the three
Ds (de-carbonization, de-centralization, and digitalization), this type of
paradigm shift has resulted in the following technical issues.
 8.2 Technical Issues in Modern Hybrid Power Systems 177

8.2.1 Power Quality

The power quality (PQ) issues related to the MHPS are as follows:
1. voltage variations and unbalance among phases in case of poly-phase
systems;
2. low voltage and high voltage ride through [2];
3. frequency fluctuations;
4. harmonics due to the presence of power electronic devices (PEDs) [3];
5. voltage transients;
6. voltage flickering, swell, and sags;
7. reactive power management especially in case of inverter-operated
power systems having no-reactive power support [4].

8.2.2 Demand–Supply Management

Here, the main issues are as follows:
1. load forecasting;
2. weather forecasting;
3. load shedding;
4. balance between demand and supply;
5. optimal allocation, and operation of various energy resources and
storage systems [5].

8.2.3 Synchronization and Islanding

Here, the major issues to be dealt with are as follows:
1. connection and re-connection of various resources with proper synchro-
nization;
2. anti-islanding facility and islanding operations;
3. maintaining stability of generators.

8.2.4 Protective Devices, Safety, and Environment

The important safety-related issues are as follows:
1. operational strategies;
2. maintenance of equipment;
3. correct operation of protective devices;
4. proper coordination among various protective devices;
5. environmental issues with respect to use of semiconductors, and energy
storage batteries;
178 Application of Machine Learning Techniques in Modern Hybrid Power

6. estimation of proper settings of protective devices [6];

7. perfect grounding system;
8. reverse power flow conditions;
9. island and grid-interactive operational procedures;
10. maintenance of reliability due to lack of proper sizing and mis-
positioning of DGs.

8.2.5 Human Factor

The main issues are as follows:
1. availability of skilled manpower;
2. training and retaining of skilled manpower;
3. re-training and skill development of human resources.

8.3 Application of ML and Optimization Techniques

in MHPS
Most of the technical issues mentioned above can be mitigated with the
help of machine learning (ML). Supervised learning (classification and
regression) and unsupervised learning (clustering) are the two main types
of ML [7]. Regression, support vector machines (SVMs), artificial neural
network (ANN), and other common algorithms all use supervised learning.
Unsupervised learning employs K-means clustering, predictive maintenance,
component analysis, and other techniques. The necessary steps for the suc-
cessful development of an ML model are described in the box given below.
 8.4 A Prediction Case Study of ML in MHPS 179

Table 8.1 ML application in MHPS.

Sl. MHPS Application examples of ML
no. component
1 Generation • Renewable energy integration [8]
• Optimal allocation of power [9]
• Forecasting of energy from weather-dependent resources
such as solar and wind [10]
2 Transmission • Optimal allocation of transmission lines and its assets
• Implementation of FACTS (flexible AC transmission
system) [11]
• Use of UHV and EHV DC systems
• Monitoring of lines at remote or hill station or higher
elevations
3 Distribution • Integration of distribution generation
• Optimal operation of microgrids and smart grids
• Adaptive protection system of DG, MG, and SGs [12]
• Integration of smart appliances with system
• Optimal allocation of energy resources
• Effective load shedding as and when required [13]
• Occupancy sensors and automation of HVAC functions
4 Others • Augmentation and optimum utilization of other assets
• Safety and security of human and assets
• Demand-supply forecasting and its management
• Predictive maintenance of equipment [14]
• Energy storage system management [15]
• Optimal cost of operation and maintenance
• State evaluation and fault detection [16]

The paradigm shift in EPS needs new technologies such as ML and artificial
intelligence (AI), whereas ML is the way to achieve AI. ML is used in various
constituents of MHPS as shown in Table 8.1.

8.4 A Prediction Case Study of ML in MHPS

Digital tools aid in the better management and optimization of various energy
resources, particularly RERs. AI techniques like ML are emerging in the
power sector to improve RE prediction, which is mostly intermittent. To rely
on RERs for optimization and better control, industries must be able to predict
an hour or a day ahead of time power with varying weather conditions. For
this purpose, a case study of a well-known large-scale continuous process
chemical industry that has both a cogeneration power plant (CPP) and a solar
photovoltaic power plant (SPP) as DG is used to forecast solar irradiance
180 Application of Machine Learning Techniques in Modern Hybrid Power

using the ML technique. Figure 8.3 depicts a single line diagram (SLD) for
the case study industry. The following are the details for the MHPS.
This section describes the power distribution network, CPP setup, and the
SPP details for the case study considered.
A coal-fired steam boiler is installed to meet the process plant’s need
for high-pressure steam at higher temperatures. Excess or waste steam is
used to power a turbine-generator (TG) set. The process plant’s total power

Figure 8.3 Single line diagram of the industry.

 8.4 A Prediction Case Study of ML in MHPS 181

requirement is 18 MW, with a process load demand of 15 MW required

during normal industry operation. The remaining 10 MW is always required
to support plant base loads like water systems and ventilation. When the CPP
is turned off, the utility grid will supply the base loads of 10 MW using
a demand management controller, which is used to control the unwanted
demand spikes caused by the large induction motors’ sudden turn-on and
turn-off. As a continuous process chemical industry, it runs almost continu-
ously if the industry is operational. Furthermore, during the night, some of
the loads that run during the day are turned off. During the night, however,
the lighting load will balance the net load. As a result, the load profile over
the course of a full day is constant. A 2.3-MVAr shunt capacitor is connected
to a 6.6-kV bus to cater plants’ reactive power load.
CPP with Operational Issues:
In the mentioned industry, a CPP with a steam boiler and a TG is installed. A
radiant boiler with integral superheaters, a forced flow section, and a tubular
air heater with two forced draft and two induced draft fans comprise the steam
generator. The turbines are extraction condensing and have TG. The TG is a
two-pole, three-phase, air-cooled motor that is directly connected to the steam
turbine. Their maximum output is rated at 6.6 kV and is rated at 18 MW at
0.85 power factor. In addition, as shown in Figure 8.3, the CPP is linked to
a utility grid for initial operation. It is worth noting that the minimum power
output of each turbine, from an operational standpoint, would be at least 60%
of the 18 MW (i.e., 10.8 MW). The TG will provide 18 MW of active power,
which is the industry’s exact requirement. Later, the industry installed a 6-
MW peak capacity SPP to meet statutory requirements and to promote RE.
Furthermore, during the SPP generation time (approximately 5-6 hours per
day), the CPP operation is kept running at a lower output of around (13 MW),
with SPP supplying the remainder to satisfy power balance. As a result of
the reduction in CPP output, the overall efficiency of the steam turbine falls
from 34% to 32%, and even the extraction of steam into the process system
becomes more complicated.
Solar Photovoltaic Power Plant with Operational Issues:
The total SPP was divided into two divisions of 3 MW each. A total of 23,530
PV modules with a peak power rating of 255 W, open circuit, and nominal
voltages of 37.6 and 30.5 V, respectively, short circuit and nominal currents
of 8.95 and 8.42 A, respectively, panel efficiency of 15.5%, and annual power
182 Application of Machine Learning Techniques in Modern Hybrid Power

de-rating of 4% [17] were installed with six 1-MW inverters. These inverters
are grid-interactive and do not have reactive power capability. The power was
evacuated at the 6.6-kV level. The levelized cost of energy was estimated to
be US $ 0.1 (in July 2021).
The following operational interlocks are required for the joint manage-
ment of the CPP and SPP. If the CPP’s TG is tripped, the industry continues
to operate with assistance from the utility grid and SPP when it is available.
If the process load is suddenly reduced due to any abnormality in process
operations, the CPP will be islanded, and the SPP will be kept off to avoid
unauthorized power export to the grid and to prevent sudden voltage rise at
plant buses. Circuit breakers have an additional mechanical interlock for this
purpose (CB). When the CPP is not producing power, the SPP trips within
0.08 seconds (i.e., the trip time of CBs at CPP and SPP together). This
is well within the 0.16 seconds specified by [18]. Even though such local
PV integration reduces power losses by 3%-9% [19], this scenario indicates
a certain curtailment of RE; in this case, PV, for the efficient and secure
operation of power systems that are integrated with grids [20]. The design
efficiency of a CPP is 34% up to 16 MW, and it drops to 32% when the
load on the CPP is 14 MW. This is primarily due to the operation of SPP.
With the operational issues of CPP and SPP, there is an urgent need to find
an optimal solution for such industries. It is also necessary to investigate
irradiance forecasting techniques for a large-scale PV plant. This is the theme
of the ensuing subsection.

8.4.1 Forecasting Irradiance of SPP

The schematic diagram of 6-MW SPP is shown in Figure 8.4 The optimiza-
tion of EPSs with DG is typically done with prior information on load profile
and weather forecasting data to ensure satisfactory operation. The prediction
of solar power generation for a day or a specific time ahead based on such
weather forecasts will undoubtedly aid power generation stations in planning
and dispatch. The prediction should be automatic and more accurate by
considering seasonally varying weather conditions, which is possible with
the help of ML techniques, as used by many researchers [21, 22]. However,
every model has limitations, and there are some deviations. Figure 8.5 depicts
a dataset of every hourly reading on March 1, 2021, at the case study
industry’s site location (latitude of 17.90 N and longitude of 80.80 E) which
amounts to 300 datapoints. The ambient temperature and relative humidity
are input variables in this data, and the output variable is global horizontal
 8.4 A Prediction Case Study of ML in MHPS 183

Figure 8.4 Schematic diagram of 6-MW SPP.

Figure 8.5 Site weather conditions for solar irradiance.

irradiance (GHI). As pointed out earlier, the performance of predictor plays

an important role in the operation of the CPP. The useful measures necessary
to evaluate the predictions’ performance are listed in the ensuing subsection.
184 Application of Machine Learning Techniques in Modern Hybrid Power

8.4.2 Metrics for Understanding the Performance of Predictions

.

using ML Methods
Some of the metrics used in regression analysis are briefly explained
here [23].
1. Mean absolute error (MAE) is the absolute difference between the true
values and the predicted values. Since it is absolute, any negative sign in
the result is to be ignored. It takes the average of error from each sample,
as in the following equation:
MAE= True values – Predicted values

= |y − 
y|/n (8.1)

where y = true value

y = predicted value and
n = number of observations.
It is generally used for continuous variable data and the lower value of
MAE gives a better regression model. It treats the errors equally whether
it is large or small.
2. Mean squared error (MSE) is estimated by considering the averages of
the squares of the difference between the true values and the predicted
values of the dataset as in the following equation:

MSE = 1/n (y− ŷ)2 . (8.2)

Here, by squaring the difference between the true values and the pre-
dicted values, the higher error value can be penalized. It is the most
widely used metrics when the dataset contains too high or too low
values, but it is not useful when the data contain huge noise. It is the
variance of the error value.
3. Root mean squared error (RMSE) is the square root of MSE, which
gives better accuracy of the regression model, as in the following
equation: 

(y− ŷ)2
RMSE = . (8.3)
n
It is more useful when large errors are present in the dataset. It is also
the standard deviation of errors.
 8.4 A Prediction Case Study of ML in MHPS 185

4. R-squared: It is the coefficient of determination and indicates the

qualitative information of a dataset, as in the following equation:
SSEM − SSER
R2 = . (8.4)
SSEM
Here, SSEM is the sum of squared errors by mean line and SSER is the
sum of the squared errors by regression line. Here, the amount of error is
reduced since the regression model is used instead of the mean of mean
line. Its value lies between 0 and 1 where the value 1 or close to 1 gives
a better performance regression model, i.e., a perfect model and a value
equal to zero indicates that the model does not fit for the given data.
5. P-value: It indicates the statistical significance of coefficient relationship
of a model. A lower value of less than 0.05 indicates a meaningful
addition to the dataset. It tests the null hypothesis of each term of
coefficients. A lower p-value means a significant change in prediction
value for a change in response variable. It gives a direction of which
term to be retained in a regression model.

8.4.3 Model-Based and Model-Free Regression Techniques

In general regression analysis (supervised learning), the statistical relation-
ship between a dependent variable and one or more independent variable are
estimated. Given k training points are (x1 , y1 ) , . . . , (xk , yk ).
The model-based regression analysis solves the following optimization
problem, as in the following equation:
{k}
1   T 2
(P1) min β x i − yi . (8.5a)
{β} n
{i=1}

Upon solving the optimal β, the test points are then used to obtain the
estimated values (ŷ1 , . . . , ŷk ) to unforeseen inputs (x1 , . . . , xn ). Note that
the problem (P1) is differentiable with respect to βand can be solved via
off-the-shelf convex optimization techniques. In this study, the predictions
from regression from problem (P1) are termed as model-based because the
coefficients of the input variables are exactly known.
The attention is then turned to the prediction techniques by ANNs which
are generally termed as model-free.
186 Application of Machine Learning Techniques in Modern Hybrid Power

ANNs are widely used in a wide range of engineering applications,

including forecasting, modeling complex patterns, image recognition, and
regression. Solar radiation prediction models are developed using ANNs in
this work. A neuron is the basic unit of ANN, and it generates output using
a transfer function. Each input will be multiplied by a weight, which will act
as a link between the input and the neuron as well as between the various
layers of neurons. The neuron uses a transfer function to obtain the result in
the final stage of ANN. The advantage of ANN techniques is that they do not
require knowledge of mathematical calculations between parameters (model-
free), but they do require less computational effort and provide a compact
solution for multivariable problems.
The ANNs find a nonlinear mapping between the training and test inputs
given as Ω : ϕk (xk ) → yk . For simplicity, let vectors dk , θk and bk
represent the trainable parameters of the ANN. In the task of supervised
learning, the ANN is trained using back-propagation algorithm based on
gradient descent which minimizes the training loss given as in the following
equation: 
(P2) min ||yk − gk (dk , θk )||2 . (8.5b)
θk
k

Upon training, the ANN finds the optimal parameters (θ∗ , b∗ ) for each
layer. The trained ANN with optimal parameters is then used to estimate
the values for the unanticipated inputs [24]. When compared to model-
based techniques, ANNs (model-free) can model any degree of nonlinearity.
However, in practice, it is vital to evaluate the performance of both model-
based and model-free prediction strategies. Because the investigated case
study is a continuous-running process plant, the optimal judgments made
while accounting for predicting mistakes in the GHI play an important part in
decision making. As a result, the prediction accuracies of both model-based
and model-free models are analyzed in the next section. Figure 8.6 describes
the complete methodology of the optimal decision making in the MHPS using
model-based and model-free approaches.

8.4.4 Prediction Block

This block preprocesses the training dataset, which consists of one month
of data with appropriate inputs required to predict GHI, and then trains the
corresponding model-based and model-free regression models. The trained
models are then used to estimate/predict the GHI on test day. Furthermore,
it is assumed that the models are trained offline, which means that when the
 8.4 A Prediction Case Study of ML in MHPS 187

Figure 8.6 Complete methodology for optimal decision making.

site conditions change, the training must be repeated, and the corresponding
models are updated using the cloud software installed in the process plant.
Upon getting the prediction of GHI from model-based and model-free regres-
sion models, the best option is to be selected based on the following criteria
for each method:
1. smaller value of RMSE, MSE, and MAE;
2. R-squared value closer to 1.0;
3. simplicity and faster response.

8.4.5 Forecasting of Solar Irradiance with a Model-Based

Regression Approach
In the current study, one-month data collected in the plant location used for
the model-based regression approach and simulations carried out on various
regressions models and the results are shown in Table 8.2.
With the above criteria, the linear regression-linear with R-squared of
0.76, RMSE of 131.31 and MSE of 17243 with a fastest response time of 0.97
seconds is selected as the best fit prediction model. The regression results are
shown in Table 8.3.
In this selected model-based regression model, the parameters with “p”
value of more than 0.05, i.e., wind speed and sky cover, are ignored since
188 Application of Machine Learning Techniques in Modern Hybrid Power

Table 8.2 Comparison of parameters of various regression models.

Model Model RMSE R-squared MSE MAE Training Remarks
Parameter Time (s)
Linear Linear 131.31 0.76 17,243 95.37 0.97 Our
Regression best
Interactions 127.72 0.77 16,312 96.82 5.15 -
Robust 139.24 0.73 19,387 91.24 4.84 -
Stepwise 128.34 0.77 16,470 92.10 12.01 -
Tree Fine 156.71 0.66 20,559 99.92 2.38 -
Medium 144.10 0.71 20,766 93.59 2.88 -
Coarse 162.16 0.63 26,296 117.35 2.70 -
Support Linear 134.63 0.75 18,124 91.19 3.35 -
vector
machine
(SVM)
Quadratic 132.40 0.76 17,526 96.80 3.57 -
Cubic 132.57 0.75 17,576 88.97 5.23 -
Fine 143.38 0.71 20,559 100.86 8.07 -
Gaussian
Medium 175.20 0.78 15,675 84.18 7.80 -
Gaussian
Coarse 132.50 0.75 17,756 87.31 5.17 -
Gaussian
Gaussian Rationale 116.56 0.81 13,587 81.86 14.04 -
process quadratic
regres-
sion
(GPR)
Squared 115.90 0.81 13,433 82.10 12.61 System
exponen- best
tial
Matern 5/2 116.34 0.81 13,356 81.31 13.05 -
Exponential 117.84 0.81 13,886 81.35 14.16 -
Ensembles Boosted 131.01 0.76 17,163 90.20 10.34 -
of trees trees
Bagged 129.65 0.73 16,808 84.40 12.93 -
trees

they do not have much significance on irradiance values. Training the field
data without wind speed and sky cover is done and the results are shown in
Table 8.4. Here, a model with the R-squared of 0.76, RMSE of 132.06, and
MSE of 17439 with a fastest response time of 0.82 seconds is selected as the
best prediction model. The final prediction plot is as in Figure 8.7.
 8.4 A Prediction Case Study of ML in MHPS 189

Table 8.3 Initial regression results for irradiance prediction.

Statistics Estimate SE t-Stat. p-Value
(Intercept) −6102 505.57 −12.07 1.53 × 10−26
Amb. temp (◦ C) 50.87 6.04 7.95 8.39 × 10−10
Rel. humidity (%) 9.90 2.19 4.51 7.44 × 10−5
Dew point (◦ C) 106.86 8.94 11.96 8.36 × 10−25
Wind speed (m/s) −33.56 17.49 -1.91 0.58
Sky cover (%) −0.47 1.05 −0.46 0.64
Precipitation (%) 151.24 10.98 13.77 1.82 × 10−28

Table 8.4 Final regression results for irradiance prediction.

Statistics Estimate SE t-Stat. p-Value
(Intercept) −6244 510.92 −12.22 4.59 × 10−28
◦
Amb. temp (( C)) 51.94 6.37 8.15 1.04 × 10−14
Rel. humidity (%) 10.89 2.27 4.79 2.68 × 10−6
◦
Dew point ( C) 106.72 9.06 11.79 1.69 × 10−26
Precipitation (%) 151.01 11.16 13.52 8.97 × 10−33

Figure 8.7 Final prediction plot.

With the above results, the relation function for solar irradiance is
formulated as in the following equation:

Ĝ{m−based) = 51.94 x1 + 10.89x2 + 106.72 x3 + 151.01 x4 + (−6244)

(8.6a)
190 Application of Machine Learning Techniques in Modern Hybrid Power

where x1 , x2 , x3 , and x4 are, respectively, the ambient temperature, relative

humidity, dew point, and precipitation. The intercept is given by the last term
in the equation, which is −6244.

8.4.6 Forecasting of Solar Irradiance with a Model-Free

Regression Approach (ANNs)
In the absence of solar radiation measuring instruments, the design of the
ANN model is critical for estimating global solar radiation. This ANN model
will be extremely useful in maximizing the utilization of a large amount of
free, environmentally friendly solar energy for a variety of applications. The
feed-forward networks are the most common type of multi-layer perceptron
(MLP). A typical MLP has three distinct layers, which are referred to as the
input, hidden, and output layers; in that order, x1 , x2 , x3 , and x4 are the input
signals. The input layer neurons, which only serve as buffers, distribute to the
hidden layer neurons. Figure 8.8 [25] depicts each neuron located within the
hidden layer after weighing the input signals. It adds them together with the
assigned connection strengths from the layer in the input and output. The
output can also be calculated using the sum function [26]. A set of input–
output data can be used to train and estimate the value of an ANN model.
In a multilayer feed-forward network, the most used algorithms for training
are Levenberg–Marquardt (LM) back-propagation, Bayesian regularization
(BR), and scaled conjugate gradient (SCG). The ANN model used in this
study was built in MATLAB version 2019 and consisted of three layers of
feed-forward network with tangent sigmoid activation function in hidden
layers and linear activation function in output layer. Four parameters are used
as inputs: ambient temperature, relative humidity, dew point, and precipita-
tion, with only one parameter, monthly average global solar radiation (GHI),
predicted as an output.

Figure 8.8 Architecture of artificial neural network [26].

 8.4 A Prediction Case Study of ML in MHPS 191

8.4.7 Normalization, Training, and Testing for Model-Free

Regression
The training set is used in a model-free regression (ANN) to learn by
optimally adjusting the weights on the neural network as described in the
preceding section. In other words, the ANN model should predict new data
(unforeseen data) with as little error as possible. The testing set is only
used to assess the efficacy of an ANN model that has been built. In the
current study, one-month data collected in the plant location was normal-
ized to a range between -1 and +1 before the training process [28]. After
normalization, the entire dataset was divided into three categories for training
(70%), testing (15%), and validation (15%) of an ANN model. Choosing the
number of neurons in a hidden layer and the number of hidden layers for
any ANN model is difficult. For most complex applications, one hidden layer
is sufficient.
The training of an ANN model begins with a random initial weight. It is
critical to identify the optimal number of neurons within the hidden layer to
achieve the lowest possible error in ANN model output. Based on MSE and
R2 results, an acceptable range of neurons in the hidden layer is chosen. The
optimal number of hidden neurons was determined when the MSE was the
lowest and the linear correlation coefficient (R) was the highest.
To optimally find the number of hidden neurons in the hidden layer,
the ANN model’s training was done by increasing the number of neurons
one by one until it converged into a minimum mean square error. Among
the three algorithms, the R-value for the LM algorithm was in the range
of 0.9000. When the value of R exceeds 0.9, it indicates that most pre-
dicted values match the measured values. As a result, the LM algorithm
was considered in the current work to determine the optimal number of
neurons in the hidden layer. According to the preceding discussion, nine
neurons in the hidden layer were discovered to be the best at predicting GHI.
Figure 8.9 depicts the regression analysis and Figure 8.10 depicts the training
performance.
The GHI predicted using model-free regression approach can be given as
in the following equation:

Ĝ{m−free} = (xn , g k (θ∗ , b∗ )) (8.6b)

It is worth mentioning that the model-free prediction model outperformed

the model-based prediction model in the case study considered with a slightly
higher R2 on the training dataset. This is because the model-free approach
192 Application of Machine Learning Techniques in Modern Hybrid Power

Figure 8.9 Regression analysis of the model-free approach.

models the nonlinear interactions between the inputs necessary to predict

GHI. The key problem here is that when these models are utilized to estimate
the actual test day GHI forecasts, the GHI must be estimated with as little
error as possible. Because a big estimation inaccuracy can have a negative
impact on the CPPs’ performance. Therefore, the optimization-assisted ML
approach is necessary to gauge the accuracies of the predictions. These
predicted forecasts are now fed into the optimization block to determine the
optimal CPP day-ahead decisions.
 8.5 Optimization Block in MHPS 193

Figure 8.10 Training performance.

8.5 Optimization Block in MHPS

In this block, the GHI on the test day is predicted using trained model-based
and model-free regression models, and then the processing plants’ setpoints
are optimized to find the optimal setpoints.

8.5.1 Optimization-Assisted ML of MHPS

In this section, the forecasts which are estimated from the model-based and
model-free regression approaches are given as input to the optimization block
to estimate the SPP PV production. To solve the optimization problem, a
detailed mathematical model of the process plants’ DN is required.
In the case study considered, the distribution network of the industry
is modeled as a single-feeder radial distribution grid consisting of N + 1
buses and lines connecting these buses is modeled by a tree graph. Let
N0 := {0, . . ., N } denote set of all nodes and L := {1, . . ., N } denote
set of distribution lines. The substation node (root node) is indexed as node
0 that connects to the external transmission network. All the non-substation
194 Application of Machine Learning Techniques in Modern Hybrid Power

nodes corresponding to N := {1, . . ., N } represent the non-substation nodes.

For each node n ∈ N0 , let vn,t denote its squared voltage magnitude at time
t and sn,t = pcn,t + jqn,t
c denote the complex power injected to the node n.

Furthermore, for each line n ∈ L, let zn = rn + jxn denote its impedance

and Sn,t = Pn,t + jQn,t denote the complex power flowing from the
sending bus πn . Also let Cn denote the set of children nodes corresponding to
node n.
It should be noted that, even though the considered industry operates
with poly-phase lines and balanced loads, the three-phase lines are con-
verted to single-phase using their positive sequence impedance adopting the
method in [27].
Modeling of Solar Power Plant:
Consider the SPP plant connected to node n with active power ppv n,t . Let
dpv
n , A pv pv
n , η n , and G pv
n,t , respectively, denote the panel derating factor, panel
area, panel efficiency, and solar irradiance of the SPP. Due to the intermit-
pv
tency of solar irradiance, it should be noted that the solar irradiance Gt
is estimated using model-based and model-free methods as described in the
preceding sections rather than assuming a perfect forecast. Additionally, in
this work, it is assumed that the reactive power capability of the SPP is absent,
implying that SPP can only support active power. Since the high penetration
of ppv
n,t complicates the operation of captive power plant, the excess PV power
pv
can be curtailed which is denoted by αn,t . The following equations, i.e.,
Equation (8.7a)–(8.7d), describe the operation of SPP:

ppv pv pv pv pv
n,t = dn An ηn Ĝn,t (8.7a)

onppv pv
n,min ≤ pn,t ≤ onpn,max
pv
(8.7b)
 pv  pv
ppv
n,t = 1 − αn,t pn,t (8.7c)
pv
0 ≤ αn,t ≤ 0.3∀n ∈ N . (8.7d)
In this work, the cost for operating SPP plant is given by ΓPV
n = k pv ppv
n,t ,
where k pv = $0.1/kWh.
Modeling of Captive Power Plant:
The equations governing the operation of CPP are given as follows:

n,min ≤ pn,t ≤ pn,max

pCPP CPP CPP
(8.8a)
 8.5 Optimization Block in MHPS 195

CPP
qn,min ≤ qn,t
CPP
≤ qn,max
CPP
(8.8b)
where pCPP CPP
n,t and qn,t , respectively, denote the active and reactive powers of
the CPP. The cost for operating CPP is given by ΓCPP n = knCPP pCPP
n,t with
knCPP = $0.86/kWh. Since the efficiency of the CPP varies with the pCPP ,
t
this time-varying efficiency is not modeled in this work. However, we gauge
the efficiency of the CPP by the optimal pCPP t , i.e., when the optimal value
of the CPP is close to its maximum value, then the CPP operates at the peak
efficiency.
Furthermore, let pBL L BL L
n,t , pn,t , qn,t , and q n,t denote the active and reactive
power baseloads and loads of the processing plant.
The focus then shifts to branch-flow modeling of the processing plant’s
DN. Because power flow equations are nonlinear, the LinDistFlow approxi-
mation of power flow equations [28] are employed to solve the power flow
equations provided as follows:

Pn,t = Pj,t −pcn,t ∀n ∈ N0 (8.9a)
j∈Cj

Qn,t = Qj,t −qn,t
c
−qnsh vn,t ∀n ∈ N0 (8.9b)
j∈Cj

vn,t = vπn,t − 2Re [zn∗ Sn,t ] n ∈ N (8.9c)

where qnsh is the reactive power injected by the capacitor  bank connected
τ
to node n. Furthermore, define vectors pt = p1,t , . . ., pcNc,t , qt =
c
 τ
c , . . ., q c
q1,t denote the net active and reactive power consumption. Also,
Nc,t
let vt = [v1,t . . ., vNt ]τ collect the squared voltage magnitude for all nodes
n ∈ N . By adopting [29], the squared voltage magnitude relates to the net
active and reactive power injections as given in the following equation:

vt = Rp(t) + Xq(t) + 1N T v0 (8.9d)

where R = 2F diag (r) F τ , X = 2F diag (x) F τ . Equation (8.9d) relates

power injections pt and qt to the squared voltage magnitude. If the entries
of r and x are positive, then the matrices R and X are symmetric and
positive definite. Also, at each time period t, the nodal voltages should
satisfy the limits dictated by ANSI C.84.1 given as in the following
equation:
vmin ≤ vt ≤ vmax . (8.9e)
196 Application of Machine Learning Techniques in Modern Hybrid Power

Since the CPP plant operates in the grid-connected mode, the bounds
on active and reactive powers supplied by the utility grid are given as
Equation (8.9f)–(8.9g)
min ≤ pt
pUG ≤ pUG
UG
max (8.9f)
UG
qmin ≤ qtUG ≤ qmax
UG
. (8.9g)
The cost of the active power imported from the utility grid is given as ΓUG
t =
k UG pUG
t , where k
UG = $0.11/kWh.

The optimization problem for process plant can be stated as follows:

 PV 
(P 3) min Γ + ΓCPP + ΓUG
{pt ,qt , vt , αt }

s.t. (8.7a) −(8.9g)

The problem (P3) is a linear programming problem because the objective
and constraints are linear and can be easily solved with off-the-shelf linear
programming solvers with guaranteed global optima. It is worth mentioning
that, while the linear approximation of the AC power flow is used to solve
the problem (P3), the optimal setpoints are validated using the nonlinear
solution ETAP, i.e., when the optimal setpoints are fed into the ETAP solver,
the voltage magnitude accuracy of the LinDistFlow and ETAP equations is
compared, with the LinDistFlow equations overestimating the actual non-
linear voltages. If the voltage magnitudes computed from the ETAP exceed
Equation (8.9e), the lower or the upper bound of Equation (8.9e) is tightened
further to improve approximation. Algorithm 1, as shown in the box below,
gives the steps to implement the proposed.

Algorithm 1: Algorithm to implement the proposed approach

Input: Fetch the load quantities and operational constraints
from the processing plant
{pL , q L , pBL , q BL , vmin , vmax , pUG
min , qmin , pmax , qmax , pmin , qmin , R, X }
UG UG UG CPP CPP

Outputs: αPV , pUG , qUG , pCP P , qCP P , pPV

Step 1: Train the model-based and model-free regression models
using previous inputs and get optimal β ∗ and (θ∗ , b∗ )
Step 2: Predict the test day GHI using the trained models
Step 3: Perform optimization (P3)  
min{pt ,qt , vt , αt } ΓPV +ΓCP P +ΓUG
s.t. (8.7a) − (8.9g)
Step 4: Fetch the optimal outputs
Step 5: Validate with ETAP and compare models
 8.5 Optimization Block in MHPS 197

Table 8.5 Industry technical parameters.

Parameter Nominal Minimum Maximum
pCPP (MW) 16 16 18
q CPP (MVAR) 12 0 13.5
pUG (MW) 0.5 −5 7
q UG (MVAR) 2 0.3 6
pPV (MW) - 0 6
pL (MW) 15 0 15
q L (MVAR) 12 0 12
pBL (MW) 10 10 10
q BL (MVAR) 2.25 2.25 2.25
r1 (Ω) 0.081 - -
r2 (Ω) 0.0027 - -
x1 (Ω) 0.1165 - -
x2 (Ω) 0.0188 - -
v1 (pu)2 1 - -
v2 (pu)2 1 (0.91)2 (1.05)2
v3 (pu)2 1 (0.91)2 (1.05)2
αPV - 0 0.3
q sh (MVAR) 2.3
dpv , Apv , η pv (0.85, 35295, 0.155)

8.5.2 Experimental Setup

The details of real-time operating conditions with constraints, decision vari-
ables, and objective functions of the case study industry to solve problem (P3)
are given in Table 8.5. Furthermore, for simulations in this work, a laptop
with 11th Gen Intel, core i5-1135G7 at 2.40 GHz, 16.0-GB RAM, and a 64-
bit operating system, x64-based processor, is used with MATLAB 2018a.
The load profile of the considered industry is constant because the load on
the process plant is constant. The load, PV, and net power injection of the
industry are shown in Figure 8.11. It can be seen that the net power injections
from the utility grid resembles the duck curve [30]; this is due to excessive
PV penetration in the afternoon and may result in the overvoltage issues and
deteriorate the efficiency of the CPP. The optimization problem (P3) is solved
with time granularity of 1 hour.

8.5.3 Validation Block

Using the actual nonlinear power flow solver ETAP, this block validates the
optimized setpoints computed in the previous stage. This block aims to exam-
ine the effects of optimized setpoints calculated from estimated GHI values,
198 Application of Machine Learning Techniques in Modern Hybrid Power

Figure 8.11 Load, PV, and net injections from the UG.

specifically the nodal voltages of the process plant, which are computed using
ETAP to see if the voltage magnitudes obey their respective lower and upper
bounds. The nodal voltage magnitudes should be well within the range for the
process plant to operate satisfactorily; otherwise, technical issues may arise,
resulting in abnormal operation. Specifically, the plants’ turbo generators and
SPP may trip if the voltage falls below the specified limits and results in the
complete production loss.

8.5.3.1 Thorough comparisons in voltage-magnitudes

for the actual test day for model-based and model-free
approaches
Here, the optimization problem is repeated to compute the optimal setpoints
for the test day considering the predictions from model-based and model-free
approaches and the corresponding voltage magnitudes are computed from the
ETAP.
The optimal inputs computed from the GHI prediction from model-based
and model-free regression approaches, as well as the true field GHI value
(benchmark), were used to solve the ETAP power flow. The power flow in the
ETAP for the model-based regression approach is depicted in Figure 8.12 for
a single snapshot. The voltage magnitude accuracy between the LinDistFlow
 8.5 Optimization Block in MHPS 199

Figure 8.12 SLD of industry (left) and power flow result with ETAP (right).

and the ETAP are below 0.02% pu for all optimizations, i.e., model-based,
model-free, and true-prediction.
Next, nodal voltage magnitudes computed from the model-based, model-
free, and true-prediction optimization are depicted in Figure 8.13, from which
the following conclusions can be drawn.

Figure 8.13 Nodal voltage magnitude profile.

200 Application of Machine Learning Techniques in Modern Hybrid Power

Table 8.6 Optimal value comparison.

Sl. no. Γ∗model−based Γ∗model−free Γ∗true

1 0.0602 0.0602 0.0601

Key Observations:
The voltage magnitudes of the model-based and model-free regression
approaches are nearly close when compared to the voltage magnitudes of
the true-GHI, as shown in Figure 8.13. Furthermore, when compared to
the model-based approach, the model-free regression approach performed
slightly worse in the first 3 hours. However, for the remainder of the time,
the model-free regression approach closely tracks the true-prediction. It is
worth noting that both the model-free and model-based approaches follow
their respective true values during the final hours, i.e., hours 8-10. Although,
both model-based and model-free regression models have training R2 values
in the range of 0.76-0.81, the process plants’ nodal voltage magnitude is well
within the bounds. As a result, it can be concluded that for the specific site
conditions, if the process plant loads are fixed, the solar forecasting error can
range between ±15%. According to the simulations, the majority of the SPP
active power generation is exported back to UG, allowing the CPP to operate
at full capacity, resulting in improved CPP efficiency. The focus then shifts to
the optimal value comparisons.
Optimal Value Comparisons:
In Table 8.6, Γ∗ represents the optimal value. According to Table 8.6, even
though both model-based and model-free regression models achieved the
same optimal value, this optimal value is less than the true-predicted optimal
value. However, the difference is negligible. As a result, even though the
respective R2 values of ML models are not close to 1.0, the ML methods
can still achieve the lower optimal value. Therefore, the process plants can
still be benefited even if the conventional forecasting methods are replaced
by the optimization-assisted ML methods.

8.6 Conclusion
The ML techniques are used to evaluate the optimal operation of a grid-
connected CPP coupled with an SPP. The case study of a well-known industry
of a comparable sort is chosen. Forecasting algorithms for solar irradiance
 References 201

are proposed, with promising results. In addition, optimization-assisted ML

techniques are used to solve the real-time problem. The simulation findings of
the proposed approach are satisfactory when validated with ETAP, a standard
power system analysis software used in the industry. This work was primarily
concerned with the optimal operation of PV integrated co-generation plants
with integration concerns. Other integration issues such as reactive power
and protective system are not considered and can be addressed in the future
work. Future work will also consider load uncertainty and poses the stochastic
optimization problem to ensure smooth operation of the process plant. Other
pertinent research directions include the use of decentralized control and
blockchain technology schemas for peer-to-peer power trading between the
industries with SPP and battery energy storage.

References
[1] M. L. Merlin Sajini, S. Suja, and S. Merlin Gilbert, “Impact analysis of
time-varying voltage-dependent load models on hybrid DG planning in
a radial distribution system using analytical approach,” IET Renewable
Power Generation, vol. 15, pp. 153–172, 2021.
[2] R. H. T. Moger, “Comprehensive review on low voltage ride through
capability of wind turbine generators,” International Transactions on
Electrical Energy Systems, vol. 30, 2020, Art. no. e12524.
[3] D. Chakravorty, J. Meyer, P. Schegner, S. Yanchenko, and M. Schocke,
“Impact of modern electronic equipment on the assessment of network
harmonic impedance,” IEEE Transactions on Smart Grid, vol. 8, no. 1,
pp. 382–390, Jan. 2017.
[4] B. K. Reddy and A. Kumar Singh, “Reactive power management and
protection coordination of distribution network with high solar photo-
voltaic penetration,” in Proceedings of 12th International Renewable
Engineering Conference (IREC), 2021, pp. 1-6.
[5] H. Wang, J. Wang, Z. Piao, X. Meng, C. Sun, G. Yuan, and S. Zhu,
“The optimal allocation and operation of an energy storage system with
high penetration grid-connected photovoltaic systems,” Sustainability,
vol. 12, no. 15, 2020.
[6] M. Usama, H. Mokhlis, M. Moghavvemi, N. N. Mansor, M. A.
Alotaibi, M. A. Azam, and B. A. Akram, “A comprehensive review
on protection strategies to mitigate the impact of renewable energy
202 Application of Machine Learning Techniques in Modern Hybrid Power

sources on interconnected distribution networks,” IEEE Access, vol. 9,

pp. 35740–35765, 2021.
[7] G. Shobha and S. Rangaswamy, “Machine learning” in Handbook of
Statistics, N. G. Venkat and C. R. Rao, Eds. Elsevier, vol. 38, 2018,
pp. 197–228, Ch. 8.
[8] A. Mosavi, M. Salimi, S. F. Ardabili, T. Rabczuk, S. Shamshirband,
and A. R. Varkonyi-Koczy, “State of the art of machine learning models
in energy systems: A systematic review,” Energies, vol. 12, 2019, Art.
no. 1301.
[9] A. Ioannis, R. Valentin, C. Benoit, K. Desen, N. Sonam, K. Aristides, F
David, E. G. Sergio Elizondo, and W. Steve, “Artificial intelligence and
machine learning approaches to energy demand-side response: A sys-
tematic review” Renewable and Sustainable Energy Reviews, vol. 130,
Sep. 2020.
[10] J. P. Lai, Y. M. Chang, C. H. Chen, and P. F. Pai, “A survey of machine
learning models in renewable energy predictions,” Applied Sciences,
vol. 10, no. 17, 2020.
[11] N. Kalfallah, F. Benzergua, I. Cherki, and A. Chaker, “USE of genetic
algorithm and particle swarm optimisation methods for the optimal con-
trol of the reactive power in Western Algerian power system,” Energy
Procedia, vol. 74, 2015.
[12] H. Khalid and A. Shobole, “Existing developments in adaptive smart
grid protection: A review,” Electric Power Systems Research, vol. 191,
2021.
[13] Y. Xiao, R. Zhao, and W. Wen, “Deep learning for predicting the
operation of under-frequency load shedding systems,” in Proceedings
of 2019 IEEE Innovative Smart Grid Technologies - Asia (ISGT Asia),
2019, pp. 4142-4147.
[14] Z. M. Çınar, A. N. Abdussalam, Q. Zeeshan, O. Korhan, M. Asmael,
and B. Safaei, “Machine learning in predictive maintenance towards
sustainable smart manufacturing in Industry 4.0,” Sustainability, vol. 12,
no. 19, 2020.
[15] G. Tianhan and L. Wei, “Machine learning toward advanced energy
storage devices and systems,” iScience, vol. 24, no. 1, 2021, Art.
no. 101936.
[16] V. T. Nikita, G. K. Victor, N. S. Denis, and V. Z. Alexey, “Machine learn-
ing techniques for power system security assessment,” IFAC-Papers
Online, vol. 49, no. 27, pp. 445–450, 2016.
 References 203

[17] REC-Solar. [Online]. Available: http://www.solardesigntool.com/comp

onents/module-panel-solar/REC-Solar/2307/REC255PEBLK/specific
ation-data-sheet-html (accessed on 10 April 2020).
[18] IEEE Standard for Interconnection and Interoperability of Distributed
Energy Resources with Associated Electric Power Systems Interfaces,
IEEE 1547-2018.
[19] S. P. Burger, J. D. Jenkins, S. C. Huntington, and I. J. Perez-Arriaga,
“Why distributed,” IEEE Power & Energy Magazine, vol. 17, no. 2,
pp. 18–24, Mar./Apr. 2019.
[20] K. Ogimoto and H. Wani, “Making renewables work,” IEEE Power &
Energy Magazine, vol. 18, no. 6, pp. 47–63, Nov./Dec. 2020.
[21] J.-P. Lai, Y.-M. Chang, C.-H. Chen, and P.-F. Pai, “A survey of machine
learning models in renewable energy predictions,” Applied Sciences,
vol. 10, no. 17, Aug. 2020, Art. no. 5975.
[22] A. B. Guher, S. Tasdemir, and B. Yaniktep, “Effective estimation
of hourly global solar radiation using machine learning algorithms,”
International Journal of Photoenergy, vol. 2020, 2020.
[23] B. Koti Reddy and A. K. Singh “Optimal operation of a photovoltaic
integrated captive cogeneration plant with a utility grid using opti-
mization and machine learning prediction methods,” Energies, vol. 14,
no. 16, 2021, Art. no. 4935.
[24] B. Facundo, G. Juan, and F. Víctor, “Prediction of wind pressure coeffi-
cients on building surfaces using artificial neural networks,” Energy and
Buildings, vol. 158, pp. 1429–1441, Jan. 2018.
[25] A. R. Pazikadin, D. Rifai, K. Ali, M. Z. Malik, A. N. Abdalla, and
M. A. Faraj, “Solar irradiance measurement instrumentation and power
solar generation forecasting based on artificial neural networks (ANN):
A review of five years research trend,” Science of the Total Environment,
vol. 715, 2020, Art. no. 136848.
[26] S. Leholo, P. Owolawi, and K. Akindeji, “Solar energy potential fore-
casting and optimization using artificial neural network: South Africa
case study,” in Proceedings of 2019 Amity International Conference on
Artificial Intelligence (AICAI), 2019, pp. 533-536.
[27] W. Kersting, Distribution System Modeling and Analysis, 4th ed. 2017.
doi: 10.1201/9781420009255
[28] M. E. Baran and F. F. Wu, “Network reconfiguration in distribution
systems for loss reduction and load balancing,” IEEE Transactions on
Power Delivery, vol. 4, no. 2, pp. 1401-1407, Apr. 1989.
204 Application of Machine Learning Techniques in Modern Hybrid Power

[29] H. Fontenot, K. S. Ayyagari, B. Dong, N. Gatsis, and A. Taha,

“Buildings-to-distribution-network integration for coordinated voltage
regulation and building energy management via distributed resource
flexibility,” Sustainable Cities and Society, vol. 69, 2021, Art.
no. 102832.
[30] Available: https://www.caiso.com/documents/flexibleresourceshelprene
wables_fastfacts.pdf (accessed on 10 July 2021).
 9
Establishing a Realistic Shunt Capacitor
Bank with a Power System using
PSO/ACCS

Ali Mohamed Eltamaly1,2,3 , Osama El Sayed Morsy4 ,

Amer Nasr Abd Elghaffar4,5,* , Yehia Sayed Mohamed5 ,
and Abou-Hashema Ahmed5
1 Electrical Engineering Department, Mansoura University, Egypt
2 Sustainable Energy Technologies Center, King Saud University,
Saudi Arabia
3 K.A.CARE Energy Research and Innovation Center, Saudi Arabia
4 Alfanar Engineering Service, Alfanar Company, Saudi Arabia
5 Electrical Engineering Department. Minia University, Egypt

E-mail: [email protected]
∗ Corresponding Author

Abstract
It is important to search for the optimum method to compensate the system
reactive power in power systems. The capacitor bank is considered as one of
the efficient methods to compensate for the reactive power with a low running
cost. Optimal capacitor placement and sizing are key issues to improve the
impact of the capacitor bank for enhancing the quality and reliability of the
distribution system. Particle swarm optimization (PSO) algorithm has been
used in this chapter to select the optimal busbar to add the capacitor and to
design the optimal size of this capacitor bank to the distribution network.
This chapter discusses the importance of using the shunt capacitor bank for
reactive power compensation in terms of improving reliability, loadability,
and reduction of power losses. Moreover, this research discusses the ability
of automatic capacitor control scheme (ACCS) to add the optimal value

205
206 Establishing a Realistic Shunt Capacitor Bank with a Power System

online of the shunt capacitors. The proposed method in this chapter has been
validated with the IEEE 15-buses power system at 11.0 kV as an example to
select the optimal placement and sizing of capacitor banks by using a PSO
algorithm automated by the ACCS module for enhancing the quality of the
power system.

Keywords: Reactive compensation, capacitor bank, power quality, particle

swarm optimization (PSO), automatic capacitor control scheme (ACCS),
distribution network.

9.1 Introduction
Power systems around the world are faced with great challenges to keep up
with the increasing demand for electric power. Therefore, it is important to
ensure the stability, reliability, and efficiency of the power system [1]. The
reactive power compensation can improve the power system performance
and ensures a high level of stability, reliability, loadability, and efficiency.
The main challenge of reactive power compensation is to design the security
system with a stable reliability power system [2]. Undervoltage and low/poor
power factor in the distribution system will directly affect the system effi-
ciency and increase the power losses. Using the advanced control techniques
with the electrical power system can treat the shortage distribution efficiency
by compensating the reactive power using shunt capacitor banks, flexible
AC transmission system (FACTS) devices with the distribution network [3].
FACTS devices are more accurate and operate fast because it depends on the
thyristor for switching, but the FACTS devices are costlier to use with the
small-scale distribution network [4]. On the other hand, using the distributed
generation (DG) near the load is highly preferred because it can save the
generated power from conventional sources and decrease the power losses in
long transmission system. But the DG is depending on the renewable energy
sources that are directly related to natural environmental causes [5]. Using
the capacitor banks parallel with the system under control by the automatic
capacitor control scheme (ACCS) techniques can enhance the system voltage
by compensating the reactive power. Also, the capacitor bank is very easy
to install by using any spare feeder to connect the capacitor bank with
the system, which is more economic compared to the active filters or the
electric synchronous motors [6]. On further consideration, the capacitor bank
installation should be according to the system status and the value of bus
reactive power load requirements. Several methods can be used to select
 9.1 Introduction 207

Figure 9.1 SLD shunt capacitor bank for reactive power compensation.

the capacitor allocation, for example, artificial intelligence (AI) [7], genetic
algorithm (GA) [8], etc. In this chapter, the particle swarm optimization
(PSO) algorithm has been used to optimize the location and size of the
capacitor bank in the power system to increase the reliability, stability, and
efficiency. PSO algorism has been chosen as an optimization tool because it
is very simple, effective, and accurate to catch global optimal solutions [9].
Figure 9.1 shows the single line diagram (SLD) of the shunt capacitor bank
for the reactive power compensation. The objective of this chapter can be
justified in two main points: the first one is studying the system offline which
is simulated by the known stable IEEE-15 bus system with expected load
extension which can be solved by the PSO algorithm for a more stable and
reliable system, and reduce cost. The results of using the PSO in this chapter
can be compared with another research’s method to describe the techno-
economic asset of using the PSO algorithm. The second contribution in this
chapter is the online working of the shunt capacitor bank with the power
system, where all researchers study the variant method to find the capacitor
bank value and allocation, but it is important to consider the load variant
during 24 hours duty time. The load variant is solved in this chapter by using
the ACCS new advanced control techniques.
208 Establishing a Realistic Shunt Capacitor Bank with a Power System

9.2 Problem Statement

Low power factor loads have bad effects on the power quality and the perfor-
mance of the power systems. Furthermore, the low power factor loads directly
increase the power losses, the overheating of feeders/circuits, and minimize
the rated transmission power [10]. The capacitor bank draws a leading current
and partly the lagging reactive component of the load current which can
improve the power factor of the loads [11]. By adding the shunt capacitor
bank at the load bus, the power factor will be improved which directly reduces
the reactive power consumption cost, increases the rating transmission energy
(loadability), and decreases the power losses [12]. Additionally, the power
factor correction will improve the overall performance of the power system
and will increase the useful life of the loads [13]. The shunt capacitor allo-
cation is considered one of the most important challenges in improving the
power system quality. The optimum capacitor allocation will help to improve
the voltage profiles and the transmission losses reduction. To determine the
best nodes for capacitor placement in the distribution network, it is required to
have basic voltage value and power loss reduction data [14]. The application
solution methods are depending on the standard capacitor sizes and associated
costs. Furthermore, in the objective function determination, the capacitor
placement must affect only the flow of reactive power by fixing the real power
flow, and the rated voltages are not affected at every point on the feeder; the
only balanced three-phase loads are considered. Figure 9.2 shows the variant
of electrical loads during 24 hours for three months for Thailand country as
an example of the average loads in the world [15]. This example proves that
it is not practical to add the permanent fixed capacitor value with the power

Figure 9.2 Average 24 hours load value during three months of Thailand.
 9.2 Problem Statement 209

system; so the use of the ACCS is to work with the busbar online to add the
required optimal value.

9.2.1 Power Flow Equations

Distribution load flow problems are the main challenges in the power system
design. Generally, the ratio R/Xratio in the radial distribution networks is
very high; so the power flow problems can be affected by the capacitor
placement [16]. Figure 9.3 shows the SLD for the sample distribution network
to be used for the power flow calculation. This can be achieved by a set of
recursive equations derived from the equivalent current injection at the nth
busbar as shown in the following equation [17]:
 ∗  
Sn Pn + jQn ∗
In = = (9.1)
Vn Vn
where In is the current at nth busbar, Vn is the voltage at nth busbar, Pn
is the active power at nth busbar, and Qn is the reactive power at nth
busbar.
Branch current “B” in the line section between busbars “n” and “n + 1” is
calculated by the Kirchhoff’s current law as

Bn, n+1 = In+1 + I n+2 .

So
B 1 = I 2 + I 3 + I 4 + I 5 + I 6 , B2 = I 3 + I 4 + I 5 + I 6
B3 = I4 + I 5 + I6 , B4 =I 5 + I6 , B5 =I6 .

Figure 9.3 SLD n-busbars sample distribution system.

210 Establishing a Realistic Shunt Capacitor Bank with a Power System

The branch current can be written in generalized matrix form as shown in

the following equation:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
B1 1 1 1 1 1 I2
⎢ B 2 ⎥ ⎢ 0 1 1 1 1⎥ ⎢ I 3 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ B 3 ⎥ = ⎢ 0 0 1 1 1⎥ ∗ ⎢ I 4 ⎥ . (9.2)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ B 4 ⎦ ⎣ 0 0 0 1 1⎦ ⎣ I 5 ⎦
B5 0 0 0 0 1 I6
The matrix can be expressed by considering the bus injection current in
the branches as BIBC to be as shown in the following equation:
[B] = [BIBC] ∗ [I] . (9.3)
The relationship between the branch currents and the busbar voltage can
be expressed as shown in the following equation:
Vn+1 = Vn − B n+1 (rn, n+1 + jxn, n+1 ). (9.4)
So, the power loss in the line section between busbars “n” and “n + 1” can
be derived as follows [17]:
 
2
Pn,n+1 + Q2n,n+1
PLoss n, n + 1 = rn,n+1 (9.5)
|V n+1 |2
where P n, n+1 is the active power between branch “n” and branch “n + 1,”
Q n, n+1 is the reactive power between branch “n” and branch “n + 1,” and
rn,n+1 is the resistance between branch “n” and branch “n + 1.”
Finally, the total power losses in the system can be defined by the formula
shown in the following equation:

nb
PT Loss = PLoss n, n + 1. (9.6)
n+1

9.2.2 Mathematical Representation

The mathematical equation for minimizing the sum of the power loss and
the capacitor costs should be considered the operational and power balance
constraints, as shown in the following equation [18]:

n 
K
Minimize Kei Ti PLi + KCj Cj (9.7)
i=1 j=1
 9.2 Problem Statement 211

where Kei is the constant cost for the energy load level i, Ti is the duration of
load level i, PLi is the power loss at ith load level with corresponding time
duration Ti , and Cj is the injection kVAR at the jth node.
The minimization of the objective function shown in Equation (9.7)
is subjected to power flow balance expressed as shown in the following
equation:
F(X,y) = 0. (9.8)
where X is the unknown power flow variables and y is the known or specified
(independent) variables.
Finally, the limits on voltage magnitude are expressed as follows:

V min ≤ Vm ≤ V max (9.9)

where V min is the minimum acceptable voltage, Vm is the voltage at

mth node, and V max is the maximum system voltage (1.0 pu practical
value).

9.2.3 Sensitivity Calculations

Using the shunt capacitors bank in the correct locations will have different
efficiency. The power losses at different bus are subject to many operating
factors [19]. The higher sensitive buses are considered the main reference
locations to install the capacitor bank [20]. Equation (9.10) shows the varia-
tion of the active power losses due to the reactive power compensation at the
busbar distribution system

∂PL  K
= 2 (∝ij Qj + βij Pj ) (9.10)
∂Qi
j=1

where PL is the total power losses in the system, Qj is the reactive power at
busbar j, and K is the number of locations
rij Cos (θi − θj ) rij Sin (θi − θj )
∝ij = , βij =
Vi Vj Vi Vj
where θi is the voltage angle at node i, θj is the voltage angle at node j, and
rij is the real part of impedance between nodes i and j.
The sensitivity value can be calculated by Equation (9.4) at all nodes. So,
the higher sensitive location can be used as the coordinate busbar.
212 Establishing a Realistic Shunt Capacitor Bank with a Power System

Figure 9.4 Current–voltage characteristic of a capacitor.

9.3 Capacitor Bank Operation Strategies

Using the capacitor bank with the distribution networks can control the
reactive power compensation, voltage regulation, power factor correction,
and power/energy losses reduction. Planning to select the optimal capacitor
allocation and determine the capacitor bank size with the radial distribution
system is very important to reach the optimum reactive compensation [21].
The stable voltage system is depending on the optimum selection of the
capacitor allocation and size. The main contributions of shunt capacitors with
the power system are the flexibility of installation, low cost and operation,
a great reduction in power losses, and improvement in voltage levels of
weakened busbars. The shunt capacitor bank output characteristic (I–V) is
a linear rate value between voltage and current as shown in Figure 9.4. The
capacitor bank operation with the power system characteristic is defined by
Equation (9.11). In the case of transmission systems, shunt capacitors are
used to compensate for inductive (ωLI2 ) losses and to ensure satisfactory volt-
age levels during heavy load conditions [22]. Capacitor banks are switched
either manually or automatically by voltage relays

V C = −jXC (jIC ) = XC IC (9.11)

VC
where IC = X C
= ωCVC and QC = ωCVC2 .
Using the shunt capacitor banks with the distribution power system will
provide voltage support to a busbar system in case of failure of a tie line or
sudden dip in the voltage [23]. Figure 9.5 shows a short transmission system
mentioned for the line impedance; the voltage equation in this sample system
follows the formulas shown in Equations (9.12)–(9.17):
  
P + jQ = Vr e−jθ Vs − Vr ejθ (gsr + jbsr )
 9.3 Capacitor Bank Operation Strategies 213

Figure 9.5 A short transmission power system for the power flow indication.

  
= Vs Vr cos θ − Vr2 gsr + Vs Vr bsr sin θ +
  
+ j Vs Vr cos θ − Vr2 bsr − Vs Vr gsr sin θ (9.12)

At neglecting the resistance

P = Vs Vr bsr sin θ (9.13)

 
Q = Vs Vr cos θ − Vr2 bsr . (9.14)
Also, at the receiving end, load changes by a factor P + Q; so

P = (Vs bsr sin θ ) V + (Vs Vr bsr sin θ ) θ (9.15)

and

Q = (Vs cos θ − 2Vr ) bsr V − (Vs Vr bsr sin θ ) θ (9.16)

where Vr is the scalar change in voltage Vr and θ is the change in angular

displacement. If θ is eliminated from Equation (9.12) and the transmission
system resistance is neglected, the dynamic voltage equation of the simulation
system is obtained as shown in the following equation:
   2 
Vr4 + Vr2 2QXsr − VS2 + Xsr 2
P + Q2 = 0. (9.17)

As the displacement angle θ is usually very small, from Equation (9.14),

we can obtain Equation (9.18) as follows:
Q Vs − 2Vr
= . (9.18)
V Xsr
The receiving end short circuit current can be calculated by the following
equation:
Vs
Ir = . (9.19)
Xsr
214 Establishing a Realistic Shunt Capacitor Bank with a Power System

This assumes that the system resistance is very smaller than the system
reactance. So, at no load, Vs = Vs ; therefore,
Vs Vs
Ir = − =− . (9.20)
Xsr Xsr
Thus,  
 ∂Q 
 
 ∂V  = Short circuit current. (9.21)

Finally, the voltage changes ratio can calculate related to the short circuit
level system Ssc , as shown in the following equation:
Vs Vs Q
≈ ≈ . (9.22)
V V Ssc
So, Equation (9.21) means the voltage regulation is equal to the ratio of
the reactive power change to a short circuit. The obvious results are indicating
the receiving end voltage falls with the decrease in system short circuit
capacity or increase in system reactance.

9.4 Particle Swarm Optimization

PSO is a metaheuristic evolutionary technique that can be used to reach an
effective solution to many engineering problems. The PSO can be extended to
absorb the continuous variables in the system [24]. Also, the PSO application
can act with the system by considering the evaluation part as a volume-less
particle in the dimension search space [25]. PSO method consists of velocity
change of each particle at every time step to reach the better individual and
global best locations. At every step, the PSO accelerates at random separate
term, toward the best global and individual location [26]. Using the PSO
algorithm with the population to select the best global (gbest) and the best
particle [27]. The general engineering optimization problem can be defined
as minimizing the f (x).
This general problem can be simulated by Equation (9.22) by considering
the particle in d-dimensional space represented is defined by ith

xii = {xi1 , xi2 , xi3 , . . . , xid } . (9.23)

The best previous position of ith particle is represented as follows:

pbestii = {pbesti1 , pbesti2 , . . . , pbestid } (9.24)

 9.4 Particle Swarm Optimization 215

where the change of the position of the particle “I” is represented as follows:
 
Si = Si1 , Si2 , Si3 . . . , Sik . (9.25)

The modified velocity and position of each particle can be calculated

using the current velocity and the distance from particle best (pbest ) to
global best (gbest ) as follows:
   
Sik+1 = w∗Sik + C1 ∗rand ( )∗ pbestki − xki +C2 ∗rand ( )∗ gbestki − xki
(9.26)
S min ≤ Si ≤ S max (9.27)
xk+1
i = xki + Sik+1 (9.28)
where Sik is the velocity of individual constant, pbest is the best value
attained by individual i, and gbest is the global best value i at iteration k.
C1 , C2 is the acceleration of the group, and W is the inertia weight factor.
rand ( ) is a uniform random number between 0 and 1, and Sid k is the current

position of individual i at iteration k.

In the above procedures, the S max parameters can be used to determine
the specified resolution and fitness, which searched for areas between the
current location and the target location for a very high value of S max , and also
the particles will fly past the good solutions [28]. As the S max is a very low
value, it may not explore sufficiently beyond local solutions. Considering the
S max set from the 10% to 20% range of the dynamic variable, the searching
operation becomes stable and reliable.
C1 and C2 are the constants of the acceleration terms which pull each
particle toward pbest and gbest. Before the particles are tugged back, the
result of the low value will allow searching out far of the target zone and the
higher results in the movement toward the target regions. So, the settings of
the C1 and C2 are often 2.0 according to the basic standard [24].
In Equation (9.25), the inertia weight w is the balance between global
and local exploration and the exploitation on average results, where w often
decreases from 0.9 to 0.4, linearly decreasing as shown in the following
equation:  
wmax − wmin
w = wmax − ∗ iter (9.29)
itermax
where wmax is the maximum value of inertia weight (0.9 in this study), wmin
is the minimum value of inertia weight (0.4 in this study), itermax is the
maximum iteration number (generations), and iter is the current iteration
number.
216 Establishing a Realistic Shunt Capacitor Bank with a Power System

9.5 Limitation Treatment

The best particle is found regarding the fitness and converts the inequality
constraints to penalty function which added to the best particle. But, in this
method, the excellent particle can be rejected regarding the penalty factors.
Also, the empirical approach is directly affecting the penalty parameters for
solving the problem model. This affection can be avoided by using binary
fitness, where one binary fitness is for optimal particle and the other is for the
binding constraints.
The optimal particle fitness is the equalized value of Equation (9.25)
which indicates the cost of energy losses and the capacitors installed
[24–28].
By binding the fitness, it directly extends the violation level, which can
be calculated by the following equation:
⎧
⎨ zmin − z , z < zmin
Binding fitness (z) = z − zmax , z > zmax (9.30)
⎩
0 else

where z is the inequality value constraint, zmin is the lower limits inequality
constraints, and zmax is the upper limits inequality constraints. The binding
fitness constraints are considered the first best particle, but if the results are
not satisfied, the method will be regenerated. By this method, the particles can
be generated, which guarantees to complete the infeasible particles to violate
the binding constraints.

9.6 PSO Implementation for Offline Capacitor Study

Figure 9.6 shows the flowchart PSO algorithm for optimal capacitor allo-
cation. The optimum capacitor placement can be defined by the following
steps:
1. Defining the candidate allocation:
• Define the total active and reactive power demand at all busbars in
the distribution system.
• Define the power losses at all branches.
• Determine the candidate allocation.
• Find the sensitivity value.
• Select the higher sensitivity busbars as the optimum candidate
allocation.
 9.6 PSO Implementation for Offline Capacitor Study 217

Figure 9.6 Flowchart for optimum capacitor allocation by PSO.

218 Establishing a Realistic Shunt Capacitor Bank with a Power System

2. Applying data into the PSO algorithm.

• Input PSO control data.
• Generate random “n” particle numbers related to every busbar
number.
• Generate the particle velocity as V between the minimum and
maximum, where, Vmax = [(maximum capacitor bank rating in
MVAR - minimum capacitor bank rating in MVAR)/N number of
steps to move the particle from one position to another].
• Adjust the iter. count = 1.
• Running the load flows by placing a particle “I” at the candidate
busbar for the reactive power compensation.
• Determine the particle “i” fitness value (base power loss - pl) and
comparing the new value with the previous particle best value
(pbest).
• Determine the maximum global among the particle best values.
• Now, if the new global position is greater than the previous, use the
new global position.
• Using the new pbest and gbest, calculate the velocity
Equation (9.25).
• Set the new velocity value as the prober limit if the new velocity
value in limits (-vmax\, vmax).
• Repeat the gbest calculations to find the capacitor values at different
load conditions.
• Calculate the savings obtained with the resultant solution.

9.7 Simulation System for Optimal Capacitor Allocation

The PSO method can use the candidate of the capacitor allocation by deter-
mining the required kVAR compensation values of the capacitor bank at
varying loads. By assuming “n” is the different load levels and “k” is the
number of the points for capacitor allocated, PSO returns “nk” design vari-
ables. Meanwhile, the capacitor values cannot find all values in the market.
Here, the standard available values in the simulation system, which can be
placed in the optimum busbar are 200 and 1200 kVAR. This means, to get
400 kVAR, it is required to duplicate the 200-kVAR value with the optimum
busbar system.
 9.7 Simulation System for Optimal Capacitor Allocation 219

Figure 9.7 Schematic SLD of the IEEE-15-busbar system.

9.7.1 Modified System Data

The benefits of selecting the busbar system to install the shunt capacitor banks
are to reduce the power losses and cost. The simulation example used in this
part is the IEEE-15 busbar at 11-kV distribution power system. Figure 9.7
shows the SLD for the distribution power system, which presents the busbars
data in Table 9.1. The various constants used in the proposed algorithm are
Capmin = 200 kVAR, Capmax = 1200 kVAR, K = 0.7259, C1 = C2 = 2.05, and
w = 1.2. The total loads on the system are (1226.4 + j1251.2) kVA. From the
system data as given in Table 9.1, total real losses are 61.79 kW and reactive
power losses are 57.30 kVAR. Table 9.2 shows the MW load, MVAR loads,
and voltage value for the IEEE-15 busbar simulation system. The simulation
IEE-15 busbar loads can be illustrated as shown in Figure 9.8. The increase
of the demand power can be reached to 5.02% per year [5], which means the
loads in the simulation example will affect the voltage system. Finally, the
use of PSO for offline application algorithms can help to decrease the system
losses with the optimum capacitor value.
 220

Table 9.1 IEEE-15 busbar simulation system data.

Bus no. Bus 1 Bus 2 Bus 3 Bus 4 Bus 5 Bus 6 Bus 7 Bus 8 Bus 9 Bus 10 Bus 11 Bus 12 Bus 13 Bus 14 Bus 15
MVA 144.8 76.4 113.7 53.9 112.5 133.7 82.1 109.6 122.2 45.7 138.9 49.5 110.6 66.8 87.5
MVAR 18.9 5.2 21.3 10.8 23.6 14.1 9.3 32.7 20.5 16.1 33.1 7.6 19.5 8.5 13.2
Bus 1.0 0.971 0.956 0.949 0.949 0.958 0.956 0.956 0.967 0.966 0.949 0.945 0.944 0.948 0.94
voltage 009 283 669 952 918 232 008 954 971 897 952 829 517 608 844
Establishing a Realistic Shunt Capacitor Bank with a Power System
 9.7 Simulation System for Optimal Capacitor Allocation 221

Table 9.2 General data for IEEE-15 busbar system.

Data Sending end Receiving end X (Ohm) R (Ohm)
node node
Node-1 1 2 1.32349 1.35309
Node-2 2 3 1.14464 1.17024
Node-3 3 4 0.82271 0.841
Node-4 4 5 1.0276 1.52348
Node-5 2 9 1.3579 2.01317
Node-6 9 10 1.1377 1.68671
Node-7 2 6 1.7249 2.55727
Node-8 6 7 0.734 1.0882
Node-9 6 8 0.8441 1.25143
Node-10 3 11 1.2111 1.79553
Node-11 11 12 1.6515 2.44845
Node-12 12 13 1.3579 2.01317
Node-13 4 14 1.5047 2.23081
Node-14 4 15 0.8074 1.19702

Figure 9.8 MVA and MVAR loads in simulation IEEE-15 busbar distribution system.

9.7.2 Simulation Study

The optimal capacitor bank size at a single location busbar is shown in
Figure 9.9 and the corresponding power loss reduction is shown in
Figure 9.10, which discusses the highest saving power losses of 23.4 kW.
222 Establishing a Realistic Shunt Capacitor Bank with a Power System

Figure 9.9 Optimal singly capacitor bank value of the 15-busbar system.

Figure 9.10 Value of saving power losses in the 15-busbar system after using singly
capacitor bank.

Table 9.3 shows the result of voltage value after adding the capacitor bank.
Figure 9.11 shows the different voltage without adding the capacitor bank and
with adding the capacitor bank for IEEE-15 busbars without and with adding
the shunt capacitor bank. The power losses can be realized by adding a single
capacitor bank in busbar-3 which is 1013 kVAR. Also, at use two busbars
for compensation with using busbar-3 compensate system by 1013 kVAR,
the second capacitor of 321 kVAR at busbar-6 would provide more saving
power losses of 3.7 kW. Also, by adding the capacitor bank in busbar-6, it
must remove the capacitor by 142 kVAR from busbar-3 to get a further power
losses reduction by 0.5 kW; so busbar-3 was initially over-compensated.
 Table 9.3 Voltage value at IEEE-15 busbar simulation system after adding the capacitor bank.
Bus Bus 1 Bus 2 Bus 3 Bus 4 Bus 5 Bus 6 Bus 7 Bus 8 Bus 9 Bus 10 Bus 11 Bus 12 Bus 13 Bus 14 Bus 15
no.
Bus 1.0250 1.0224 1.0069 1.0009 0.99985 1.0086 1.0063 1.007263 1.0188 1.01768 0.9990 0.9955 0.9942 0.9985 0.99835
voltage
9.7 Simulation System for Optimal Capacitor Allocation
223
224 Establishing a Realistic Shunt Capacitor Bank with a Power System

Figure 9.11 Voltage value for IEEE-15 busbars without and with adding the shunt capacitor
bank.

But, when both busbar-6 and busbar-3 are compensated, the total power loss
reduction will be 32.7 kW by 1192 kVAR of shunt capacitor banks (871
kVAR at busbar-3 and 321 kVAR at busbar-6). To compare these values with
the fuzzy-reasoning approach method, which is illustrated in reference [26]
and [27], where this method requires adding 1193 kVAR by 805 kVAR at
busbar-3 and 388 kVAR at busbar-6. Finally, using the PSO method can save
0.1-kW power losses and it has decreased the capacitor value by 1.0 kVAR.

9.8 Automatic Capacitor Control Scheme

This part aims to discuss the new capacitor control techniques which can
contribute to enhancing the power system quality by using the ACCS design.
ACCS is the new suggestion system to control the reactive power using the
shunt capacitor bank with the distribution network. ACCS design is con-
sidered a supervising reactive power technique depending on the intelligent
electronic device (IED), which operates to save the stable system continu-
ously by comparing the reference setting values with the actual consumed
reactive power value.
 9.8 Automatic Capacitor Control Scheme 225

9.8.1 ACCS IED Scope

The ACCS/IED modular bay controller unit is considered a programmable
logic control (PLC) device, which can be used to control and monitor
the switch device. ACCS/IED is designed to operate with the switchgear
distribution units equipped with the electrical check-back signaling located
in medium-voltage or high-voltage substations [28, 29]. ACCS/IED relay
is used to supervise the busbar system in high-voltage substation by the
feedback indication to the numerical control relay, which is fed on the trans-
former or line circuit breaker status. Meanwhile, the ACCS relay contains the
integration binary inputs to be connected with the secondary coils to read the
voltage and current by the transformer and current transformer, respectively
[30]. Also, the ACCS/IED module can absorb more status for measuring
power in the power system, by using the measuring transducer which sends
the value (MVAR/voltage value) by DC mA DC. ACCS control module is
operating manual/automatic, at the releasing interlock for switching the shunt
capacitor bank feeder circuit breaker to link with the power system. The
ACCS design is considered a monitor continuous technology that improves
the action status to improve the power system and the maintenance cost.
ACCS/IED can be operated by the remote controller and the system value
can be monitored depending on the IEC 61850 communication system with
SCADA system [31].

9.8.2 ACCS Operation Logic Steps

Figure 9.12 shows the ACCS operation logic flowchart steps to control the
capacitor bank by adding with the system as required and isolate after the
system returns back to the normal operation. The main interlock to prevent
adding the shunt capacitor bank is the protection healthy condition to avoid
the switch on to the fault which can lead to damaging the capacitor bank.
Also, the satisfying condition allows adding two capacitor bank units with
the system as parallel in one busbar voltage or single unit capacitor, which is
used to control the closing time between the two capacitor groups [32, 33].
Using the timer interlock to prevent the close command after satisfying the
discharge of the electric charge due to operation. Also, the system reactive
power value is continuously compared with the reference set value which ini-
tiates to start the timer at the reactive power in the system equal to or greater
than the reference value. This timer is used to avoid adding the capacitor bank
with the system at the starting high loads, which can lead to overvoltage.
226 Establishing a Realistic Shunt Capacitor Bank with a Power System

Figure 9.12 Flowchart operation logic steps for ACCS module.

 9.8 Automatic Capacitor Control Scheme 227

9.8.3 ACCS Operation Sample

Generally, the ACCS/IED is the numerical digital control relay, which is
one of the various designers such as SIEMENS, ABB, Schneider, and so
on. Using the ACCS/IED to control the reactive power with the power
system will able to configure to operate with a multi-stage setting which
allows more accurate and flexible operation. Thus, it can control the reactive
power by classifying the busbar electric system into two or three parts. Also,
during a couple of two parts throw the busbar section to become one busbar
system it will digital selecting the busbar system status to select the required
optimum configuration. The following example discusses the actual control
steps to add the capacitor bank with one busbar, which is classified into
two sections with adding 7.1-MVAR capacitor bank with every part. In the
sample system, the distribution network is 11-kV voltage, and the selecting
busbar system is classified into two sections, which install one capacitor
bank unit with every part. The two-section busbar voltage is fed from a
separate power transformer 115/11 kV as shown in Figure 9.13, by using
the busbar section circuit breaker can couple the two sections to operate

Figure 9.13 Single line diagram for actual busbar voltage of 11 kV.
228 Establishing a Realistic Shunt Capacitor Bank with a Power System

Table 9.4 Setting values to operate ACCS/IED at different cases.

Quantities Setting values Time delay setting Values
MVAR 1 6 MVAR T1 – for close C1 and C2 at bus 5 minutes
tie CB open
MVAR 2 1 MVAR T2 – for open C1 and C2 at bus 5 minutes
tie CB open
MVAR 3 7 MVAR T3 – C1 close at bus tie CB close 7 minutes
MVAR 4 2 MVAR T4 – C1 open at bus tie CB close 7 minutes
MVAR 5 8 MVAR T5 – C2 close at bus tie CB close 10 minutes
MVAR 6 3 MVAR T6 – C2 open at bus tie CB close 10 minutes
V1 14.4 kV T7 open C1 and C2 at OV 5 second

the two power transformers in parallel. Table 9.4 shows the configuration
set values to operate the ACCS/IED; the operation logic is considered as
the continuous monitor to the actual reactive power in the two sections
individually.

Figure 9.14 Close and open individual logic ACCS/IED module command.
 9.8 Automatic Capacitor Control Scheme 229

Figure 9.15 ACCS/IED logic for capacitor bank-1 and capacitor bank-2 at the close the bus
tie circuit breaker.

For section-1, as the reactive power value reached the setting value
(MVAR-2), the ACCS/IED will start to initiate the internal timer-1 to close
the capacitor bank-1 (C1) feeder circuit breaker to improve the reactive power
with section-1. After that, if the reactive power decreased to the open setting
value (MVAR-2), ACCS/IED directly initiates the internal timer-2 to send the
open command to the capacitor circuit breaker. At overvoltage in the busbar
system which senses through the voltage transformer (Busbar VT-A), ACCS
will open the capacitor after timer-7, as shown in Figure 9.14 for the close
and open individual logic commands. By the same mirror logic, ACCS/IED
can be used to control the adding or isolating of the bus section-2 capacitor
230 Establishing a Realistic Shunt Capacitor Bank with a Power System

bank-2 (C-2), which also detects the undervoltage for section -2 through the
voltage transformer (Busbar VT-B). While closing the bus tie circuit breaker,
the ACCS/IED module will calculate the reactive power for section-1 and
section-2. Figure 9.15 shows the close and open logic commands to the two
capacitor bank feeder circuit breakers while closing the bus tie circuit breaker.
As the value reaches (MVAR-3), ACCS/IED will start to initiate the internal
timer-3; as it reaches (MVAR-4), it will initiate the timer-4 to add the shunt
capacitor bank unit-1 with section-1 after timer-3 target. At this time, if the
reactive power decreased to be less than the setting (MVAR-4) value, the
timer-4 will be released; else it will continue to close the capacitor bank-2
with the system after the timer-4 target. In this case (bus tie circuit breaker
closed and capacitor bank-1 and capacitor bank-2 in service), if the reactive
power decreases to the open setting value, the ACCS/IED will start to initiate
the internal timer-5 and internal timer-6 to open the circuit breaker feeder for
the capacitor bank-2 after timer-5 target. After that, if the reactive power is
still less than the setting value, it will open the circuit breaker feeder for the
capacitor bank-1 after timer-6 target; else it will be released to keep the shunt
capacitor bank-1 in service.

9.9 Conclusion
Due to the power system extension and the increase in the reactive loads,
the power system has become more sensitive and critical. This extension will
influence the distribution network in finding some non-preferred conditions
as undervoltage, sag, underfrequency, and so on. In recent years, the reactive
power compensation devices have been increased and improved to reach
optimum reactive compensation at a low cost. The main reason for using
the shunt capacitor bank is present in the power utilities, large industrial
to maintain voltage profile at acceptable levels, and to compensate reactive
power in distribution and industrial networks, wherever needed and quick
amortization of investment are some of the reasons why the using of shunt
capacitor banks, which represent an essential reactive power compensation
element with the distribution power system. This chapter presented the
optimal capacitor bank allocation in the distribution systems by using the
PSO method. The approached method in this chapter has been obtained with
IEEE-15 busbar voltage as a sample power system, which investigated to
provide a total reduction power loss of 32.7 kW by using 1192 kVAR of
capacitors by (871 kVAR at busbar-3 and 321 kVAR at busbar-6). Also, this
chapter discussed the ACCS to add the shunt capacitor bank with the system
 References 231

to improve voltage profile and reduce the active power loss by adding the
accurate required value. ACCS/IED control module simulated and discussed
in this chapter for the busbar voltage which classified to two sections, which
showed the operation steps at individual sections and at the close the bus
tie circuit breaker to be one part. Finally, this chapter aims to recommend the
power system designer and the researchers to use the PSO for offline planning
and ACCS to add the required capacitor value.

Conflict of Interest
The authors declare that this chapter has no conflict of interest.

Acknowledgement
Dr. Amer Nasr wishes to acknowledge Alfanar Company, especially Mr.
Amer Abdullah Alajmi (Vice President, Sales & Marketing, Alfanar Com-
pany, Saudi Arabia) and Mr. Osama Morsy (General Manager, Alfanar
Engineering Service, Alfanar Company, Saudi Arabia) for their supporting
throughout the completion of this research.

References
[1] A. M. Eltamaly and A. N. A. Elghaffar. (November, 2017). Modeling
of distance protection logic for out-of-step condition in power system.
Electrical Engineering, DOI: 10.1007/s00202-017-0667-3.
[2] A. M. Eltamaly, Y. Sayed, A. H. M. El-Sayed, and A. A. Elghaf-
far. (2020). Adaptive static synchronous compensation techniques with
the transmission system for optimum voltage control. Ain Shams
Engineering Journal, DOI: ttps://doi.org/10.1016/j.asej.2019.06.002
[3] A. M. Eltamaly and A. N. A. Elghaffar. (2017). Power flow control
for distribution generator in Egypt using facts devices. Acta Bulletin
of Engineering Journal, ISSN: 2067–3809.
[4] A. M. Eltamaly and A. N. A. Elghaffar. (August, 2018). Multi-control
module static VAR compensation techniques for enhancement of power
system quality. Annals of Faculty Engineering Journal. ISSN: 2601–
2332.
[5] A. M. Eltamaly and A. N. A. Elghaffar. (December, 2017). Techno-
economical study of using nuclear power plants for supporting electrical
grid in Arabian Gulf. 2(1), DOI: 10.1007/s40866-017-0031-8.
232 Establishing a Realistic Shunt Capacitor Bank with a Power System

[6] A. M. Eltamaly, Y. S. Mohamed, A. H. M. El-Sayed, A. N. A. Elghaffar,

and A. G. Abo-Khalil. D-STATCOM for distribution network compen-
sation linked with wind generation. In: A. M. Eltamaly, A. Y. Abde-
laziz, and A. G. Abo-Khalil (eds). (2021). Control and Operation of
Grid-Connected Wind Energy Systems. Green Energy and Technology.
Cham: Springer. DOI: https://doi.org/10.1007/978-3-030-64336-2_5.
[7] M. Aman, et al. (2014). Optimum shunt capacitor placement in
distribution system - A review and comparative study. Journal
of Renewable and Sustainable Energy Reviews, 30:429-439, DOI:
10.1016/j.rser.2013.10.002.
[8] M. Mahda, et al. (2017). Improve performance in electrical power distri-
bution system by optimal capacitor placement using genetic algorithm.
14th International Conference on Electrical Engineering/Electronics,
Computer, Telecommunications and Information Technology, DOI:
10.1109/ECTICon.2017.8096347.
[9] K. Prakash and M. Sydulu. (2007). Particle swarm optimization based
capacitor placement on radial distribution systems. Power Engineering
Society General Meeting, DOI: 10.1109/PES.2007.386149.
[10] A. Karaarslan, et al. (2011). A DSP based power factor correction con-
verter to reduce total harmonic distortion of input current for improve-
ment of power quality. Electrical Engineering Journal, 93(4):247–257,
DOI: https://doi.org/10.1007/s00202-011-0215-5.
[11] C. Ersavas, et al. (2016). Optimum allocation of FACTS devices under
load uncertainty based on penalty functions with genetic algorithm.
Electrical Engineering Journal, DOI: 10.1007/s00202-016-0390-5.
[12] A. M. Hemeida, et al. (2017). Moth-flame algorithm and loss sensitivity
factor for optimal allocation of shunt capacitor banks in radial distribu-
tion systems. 19th International Middle East Power Systems Conference
(MEPCON), DOI: 10.1109/MEPCON.2017.8301279.
[13] A. M. Eltamaly and A. N. A. Elghaffar. (2018). Enhancement of power
system quality using static synchronous compensation (STATCOM).
International Journal of Electrical and Computer Engineering, 8(30),
EISSN: 2305-0543.
[14] A. A. El-Fergany. (2013). Optimal capacitor allocations using evolu-
tionary algorithms. IET Generation, Transmission & Distribution, DOI:
10.1049/iet-gtd.2012.0661.
[15] P. P. Phyo and C. Jeenanunta. (January-March 2019). Electricity load
forecasting using a deep neural network. Engineering and Applied
Science Research, 46(1):10-17.
 References 233

[16] J. H. Teng. (2003). A direct approach for distribution system load flow
solutions. IEEE Transactions on Power Delivery, 18(3):882–887, DOI:
10.1109/TPWRD.2003.813818.
[17] N. Gnanase et al. (2016). Optimal placement of capacitors in radial dis-
tribution system using shark smell optimization algorithm. Ain Shams
Engineering Journal, 907-916, DOI: http://dx.doi.org/10.1016/j.asej.20
16.01.006.
[18] S. P. Singh and A. R. Rao. (2012). Optimal allocation of capacitors in
distribution systems using particle swarm optimization. Electrical Power
and Energy Systems Journal, 43:1267–1275, DOI: http://dx.doi.org/10.
1016/j.ijepes.2012.06.059.
[19] A. M. Eltamaly, M. Y. Sayed, A. H. M. El-Sayed, et al. (2020). Power
quality and reliability considerations of photovoltaic distributed gen-
eration. Technology and Economics of Smart Grids and Sustainable
Energy, 5:25, DOI: https://doi.org/10.1007/s40866-020-00096-2.
[20] M. R. Haghifam, et al. (2007). Genetic algorithm-based approach for
fixed and switchable capacitors placement in distribution systems with
uncertainty and time varying loads. IET Generation Transmission, &
Distribution, 1(2):244–252.
[21] A. R. Seifi. (2009). A new hybrid optimization method for optimum
distribution capacitor planning. Modern Applied Science Journal, 3(4).
[22] S. Corsi, Voltage Control and Protection in Electrical Power Systems,
Advances in Industrial Control, Chapter 2 Equipment for Voltage and
Reactive Power Control, DOI: 10.1007/978-1-4471-6636-8_2.
[23] J. C. Das. (2015). Power System Harmonics and Passive Filter Designs,
Book First Edition, chapter (11), John Wiley & Sons, Inc. IEEE, DOI:
10.1002/9781118887059.ch11.
[24] I. J. Hasan, et al., (2015). Optimal capacitor allocation in distribu-
tion system using particle swarm optimization. Applied Mechanics and
Materials, 699:770-775, DOI: 10.4028/www.scientific.net/AMM.699.7
70.
[25] M. A. Mohamed, et al. (2018). A PSO-based smart grid application for
optimum sizing of hybrid renewable energy systems. In Modeling and
Simulation of Smart Grid Integrated with Hybrid Renewable Energy
Systems, pp. 53–60, DOI: 10.1007/978-3-319-64795-1.
[26] M. H. Haque. (1999). Capacitor placement in radial distribution systems
for loss reduction. IEE Proceedings - Generation, Transmission, &
Distribution, 146(5), DOI: 10.1049/ip-gtd:19990495.
234 Establishing a Realistic Shunt Capacitor Bank with a Power System

[27] C.-T. Su and C.-C. Tsai. (1996). A new fuzzy-reasoning approach to

optimum capacitor allocation for primary distribution systems. Proceed-
ings of the IEEE International Conference on Industrial Technology.
[28] E. R. Yshi. (1999). Empirical study of particle swarm optimization.
Proceedings of the 1999 Congress on Evolutionary Computation. Pis-
cataway, 1945–1950.
[29] M. A. Mohamed, A. M. Eltamaly, and A. I. Alolah. (2017). Swarm
intelligence-based optimization of grid-dependent hybrid renewable
energy systems. Renewable and Sustainable Energy Reviews, 77: 515–
524.
[30] M. Ravindran and V. Kirubakaran. (2012). Electrical energy conserva-
tion in automatic power factor correction by embedded system. Energy
and Power, 2(4): 51–54, DOI: 10.5923/j.ep.20120204.02.
[31] P. S. R. Murty. (2017). Chapter 17 - Relaying and Protection. Electrical
Power System Book. 417–477, DOI: https://doi.org/10.1016/B978-0-0
8-101124-9.00017-6.
[32] Digital Capacitor Bank Control M-6280 Manual. (2008). Digital Capac-
itor Bank Control for Remote Capacitor Automation, Monitoring and
Protection. Beckwith Electric Co. 800-6280-SP-21MC4.
[33] Manual Micom C 264 / C 264C, Bay Computer C264/EN O/C80.
Schnider, operation guide.
[34] X.-m. Huang, Y.-j. Zhang, and H.-c. Huang. (2014). Automatic reactive
power control in distribution network based on feeder power factor
assessment. IEEE PES Asia-Pacific Power and Energy Engineering
Conference (APPEEC), ISSN: 2157-4839, DOI: 10.1109/APPEEC.2
014.7066071.
 10
Introduction to Blockchain Technologies

Xiaofeng Chen and Xiangjuan Jia

Hangzhou Qulian Technology Co., Ltd. China

E-mail: [email protected]; [email protected]

Abstract
The energy field is developing rapidly in the direction of a distributed,
low-carbon, digital, and intelligent energy Internet. With its characteristics
of decentralization, group collaboration, peer-to-peer (P2P) communication,
smart contracts, anti-counterfeiting and tamper-proofing, openness, and trans-
parency, blockchain technology fits the development direction of the energy
industry to improve quality, efficiency, innovation, and reform. At present,
the energy industry of many countries has explored related applications,
especially in the business fields of power trading, power metering, microgrid,
and renewable energy. As the energy industry pays more attention to energy
conservation and energy efficiency, blockchain applications are also attracting
attention for optimizing the energy industry.
The International Renewable Energy Agency (IRENA) also lists
blockchain technology as one of the key innovations in the field of renewable
energy. IRENA pointed out in a report that to accelerate the development of
low-cost renewable energy, the world needs at least 30 innovative technol-
ogy tools and enables them to benefit from the scale of renewable energy.
Blockchain technology is one of them.
Blockchain technology is a distributed network data management and
data computing technology that uses cryptography technology and distributed
consensus protocols to ensure network transmission and access security and
realizes multi-party data maintenance, cross-validation, uniformity across the
network, and resistance to tampering. As an important evolution of a new
generation of information and communication technology, the characteristics

235
236 Introduction to Blockchain Technologies

of non-tampering, transparent, and traceable data make the blockchain

technology an infrastructure to solve the mutual trust of industry chain
participants.

Keywords: Blockchain technology, energy, consensus mechanism, smart

contract, cryptographic algorithm, core technology, expansion technology,
supporting technology.

10.1 Introduction and Classification

The International Renewable Energy Agency (IRENA) said that nearly 200
companies around the world are cooperating with blockchain technology
companies[1] . And blockchain-based solutions can support a wider range of
energy transactions and grid balancing solutions[2] . Blockchain-based smart
contracts can promote the modernization of the power grid and increase
the application of renewable energy, especially the intermittent power that
is difficult to absorb while reducing costs and speeding up the transaction
process[3] . Blockchain technology can realize peer-to-peer (P2P) electricity
transactions and can also manage renewable energy and carbon emission
reduction certificates and ensure that all transactions cannot be tampered
with, making renewable energy transactions and applications more reliable
and efficient[4] . So, what is blockchain technology?
Narrowly speaking, blockchain technology is a decentralized shared
ledger that combines data blocks in a chain into a specific data structure
in chronological order and cryptographically guarantees that it cannot be
tampered with or forged. Broadly speaking, blockchain technology is a new
decentralized infrastructure and distributed computing paradigm by using
encrypted chain block structure to verify and store data, using distributed
node consensus algorithm to generate and update data, and using automated
script code (smart contract) to program and manipulate data.
There are three types of blockchain technology: public chain, private
chain, and consortium chain.
The public chain has the characteristics of open source and anonymity,
is free to join and leave, like Bitcoin and Ethereum, and works based on the
consensus mechanisms as Proof of Work (PoW), Proof of Stake (PoS), etc.
However, anonymous access is currently not suitable for network real-name
supervision[5] .
A private chain is generally established inside an organization or enter-
prise, with stronger security and privacy protection capabilities, and can
 10.2 Blockchain Technology Characteristics 237

prevent internal and external malicious attacks. However, due to its closed
network, it is not suitable for applications that need to be deployed on the
Internet and provides services across domains.
Consortium chain is limited to the participation of consortium members.
The read and write permissions on the chain and the permission to participate
in accounting are set according to the rules of the consortium. The data is
limited to the consortium organization and its users to have permission to
access or update. Consortium chain has achieved partial decentralization and
strong controllability. Practical byzantine fault tolerance (PBFT) and Raft
are often used on consortium chains and private chains. These characteristics
make the consortium chain have manageable and controllable practical needs
and technical support[6] .

10.2 Blockchain Technology Characteristics

Blockchain is an innovative application model in the Internet era of com-
puter technologies such as distributed data storage, point-to-point transmis-
sion, consensus mechanism, encryption algorithm, etc. Its characteristics are
mainly manifested in the following aspects:

10.2.1 Multi-Centralization
From an architectural point of view, the blockchain is based on a P2P
network. The damage or loss of any node will not affect the operation of the
entire system. The system has excellent robustness. Therefore, the blockchain
is a multi-centralized architecture. In terms of storage, the blockchain is a
distributed storage technology. Data is distributed and stored in all nodes
and a consensus is reached. There is no single center that holds the storage
right. Therefore, the blockchain is also a multi-centralized storage[7] . In terms
of governance, the blockchain has no centralized organization or institution,
and the rights and obligations between any nodes are equal. The blockchain
prevents a few people from controlling the entire blockchain system through
a consensus mechanism; so the governance of blockchain is decentralized or
multi-centralized.

10.2.2 Tamper-Proof, Traceable, and Transparent

Blockchain technology records all transaction behaviors since the creation
of the system in the block, which can ensure that the data records on the
chain are not susceptible to human tampering, deletion, failure, etc.; so the
238 Introduction to Blockchain Technologies

information exchange activities can be queried and traced. This completely

transparent data management system provides reliable tracking shortcuts for
operations such as energy transaction processes, logistics tracking, purchase
records, use process, and auditing[8] .
The blockchain system is open and transparent. In addition to the encryp-
tion of the private information of the parties to the transaction, the data is
transparent to the entire network nodes[9] . Anyone or any participating node
can query the blockchain data records or develop related applications through
the public interface. This is the trustworthy foundation of the blockchain
system. Blockchain data records and operating rules can be reviewed by the
entire network nodes, with high transparency.

10.2.3 High Reliability

From a technical point of view, the essence of blockchain is a distributed
database and distributed computing system. Through the form of distributed
data storage, each participating node in the blockchain network can obtain a
copy of the complete database. Blockchain data is jointly maintained by all
nodes, and each node participating in the maintenance can make a copy of
a complete record. Therefore, the more the nodes participating in the system
and the stronger the computing power, the higher the data security in the
system[10] .
The database based on blockchain technology is expected to form several
global giant databases in the future. All human value exchange activi-
ties (including registration, account opening, payment, transaction, clearing,
etc.) can be completed in these databases. The business model is highly
expandable and inclusive.

10.3 Blockchain Technology Graph

Blockchain is an abbreviation of comprehensive technology. After years
of development, it has continuously integrated a variety of today’s tech-
nologies into one. It is mainly summarized into three categories: core
technology, expansion technology, and supporting technology[1] . The core
technology refers to the technology that a complete blockchain system must
have included, the expansion technology refers to the related technology
to further expand the blockchain service capabilities, and the supporting
technology refers to the related technology to improve the security of the
 10.3 Blockchain Technology Graph 239

Figure 10.1 Comprehensive graph of blockchain technology.

blockchain system and optimize the user experience. At this stage, blockchain
technology consisting of core technology, expansion technology, and sup-
porting technology has gradually formed. In the future, it will continue to
innovate and evolve in data circulation, network scale, technical operation
and maintenance, and platform security.
Blockchain technology, as shown in Figure 10.1, is a comprehensive
technology; its technical composition can be divided into three categories:
core technology, expansion technology, and supporting technology accord-
ing to the degree of importance. Core technology refers to the technology
that a complete blockchain system must include, including cryptographic
algorithm, P2P network, consensus mechanism, smart contract, and data stor-
age. Expansion technology refers to related technology that further expands
blockchain service capabilities, including scalability, interoperability, collab-
orative governance, security, and privacy. Supporting technology refers to
related technology such as improving the security of the blockchain system
and optimizing user experience, including system management, infrastruc-
ture, operation, and maintenance.

10.3.1 Core Technology Overview

The birth of Ethereum laid the foundation for the five core technologies of
the blockchain system, as shown in Figure 10.2. It includes cryptographic
algorithm, P2P network, consensus mechanism, smart contract, and data
storage.
240 Introduction to Blockchain Technologies

Figure 10.2 Core technology graph of blockchain technology.

(1) Cryptographic Algorithm

Generally speaking, cryptographic algorithms commonly used in blockchain
systems include hash algorithms and asymmetric encryption algorithm.
(a) Hash Algorithm
Hash algorithm refers to a method of extracting a fixed-length digital “finger-
print” from data of any length. At present, common hash algorithms include
MD5 (message digest), SHA-1 (secure hash algorithm), SHA-2, SHA-3, SM3
algorithms, etc.
The hash algorithm has an input and an output. The input is data of
any length. Within the algorithm, no matter what form the input data is,
it is processed as a simple bit sequence. The output of the hash algorithm
is a fixed-length hash value or hash value. Specifically, the hash algorithm
scrambles and mixes the data compresses it into a digest, makes the amount
of data smaller, and recreates a fingerprint called the hash value. In the
blockchain system, the hash algorithm is generally used in the process of
transaction verification and block construction, such as building a block. If
the data in the block is maliciously added or changed, the resulting hash
value will be completely different. In addition, each block header information
references the hash value of the previous block, which makes the blocks
interlocked and closely connected, thus forming a data chain that is difficult
to be tampered with[11] .
(b) Asymmetric Encryption Algorithm
An asymmetric encryption algorithm refers to a cryptographic algorithm that
uses different ciphertext for encryption and decryption. At present, common
asymmetric encryption includes RSA algorithm, Elgamal algorithm, SM2,
and elliptic curve digital signature algorithm (ECDSA), of which ECDSA
and SM2 are the most used algorithms in the blockchain.
 10.3 Blockchain Technology Graph 241

The asymmetric encryption algorithm is composed of a public key and

a private key. The public key and the private key are generated by pairing.
When the public key is used to encrypt data, only the corresponding private
key can be used to decrypt the data; the same applies when the private key
encrypts data - only the corresponding public key can be used to decrypt
it. Therefore, different keys are used for encryption and decryption, so
that its confidentiality will be relatively high, which also solves symmetric
encryption. The inconvenience that both parties need to share the key in the
algorithm eliminates the need for end-users to exchange keys.
In the blockchain network, each user has a unique pair of public and
private keys. The public key is like a bank card number that can be dis-
closed, and the private key is the non-public part, like the account password
corresponding to the bank card. Simply put, in the blockchain, whoever
controls the private key has all the data in the account corresponding to the
private key[12] . The typical application scenario of asymmetric encryption
in the blockchain system is the signature and verification of transactions.
The transaction initiator uses the private key to sign the original transaction
information and broadcasts the signed transaction and public key. Each
node receives after the transaction, the public key can be used to verify
whether the transaction is legal. In this process, the transaction initiator
does not need to expose its private key, thus achieving the purpose of
confidentiality.
(2) Peer-to-Peer Network
The network layer communication protocol of blockchain technology is a P2P
network protocol, also known as a P2P network. It is a channel for consensus
and information transfer between nodes. It is the network foundation of
blockchain technology. It is a non-central server and depends on each node.
A network system for exchanging information[13] . Different from a central
network service system with a central server, each node in a P2P network
can not only act as a requester of network services but also receive requests
from other nodes, respond, and provide resources, services, and content.
Blockchain technology adopts a P2P network architecture, such as transac-
tion broadcasting, combined with P2P network technology, to decentralize
transaction broadcasting to various nodes[14] .
Generally speaking, there are three types of P2P networks: structured
network, unstructured, and hybrid network.
Structured P2P network node presents a structure of a specific organiza-
tion, the node supports efficient search of files, and even the searched content
242 Introduction to Blockchain Technologies

is not widely used[15] . In most cases, this is due to the hash function, which
facilitates database searches. This is mostly used by consortium chain or
private chain which is deployed in private network.
The unstructured P2P network nodes do not have any specific orga-
nizational characteristics. Participating nodes randomly communicate with
other nodes. Compared with intense liquidity activities (for example, certain
nodes frequently join or leave the network), unstructured P2P networks are
considered to be robust. This is mostly used by the public chain which is
deployed in the Internet.
The hybrid P2P network combines the traditional client-server model
with a P2P architecture. For example, it can design a central server that
facilitates connections between peers. The interoperability between different
blockchains or on-chain and off-chain or blockchains and oracles may meet
hybrid networks.
Compared with the other two models, the hybrid model tends to improve
the overall performance. The hybrid network model combines the main
advantages of the two modes, while simultaneously achieving the two
characteristics of high efficiency and distribution.

(3) Consensus Mechanism

The consensus mechanism refers to the process of reaching a unified agree-
ment on the state of the network in a decentralized manner, also known as
the consensus algorithm, which is the key to ensuring the consistency of the
ledger data of each node on the blockchain platform. As a data structure
that stores data in chronological order, blockchain technology can support
different consensus algorithms.
During the operation of the blockchain system, the nodes in the dis-
tributed network need to use the main chain consensus to reach an agreement
on the final blockchain data in the blockchain tree[16] . The main chain con-
sensus refers to the distributed network nodes reaching an agreement on the
blockchain data the process of. Regarding whether the blockchain data meets
the final consistency, the main chain consensus can be divided into deter-
ministic consensus and non-deterministic consensus[17] . Non-deterministic
consensus is also called probabilistic consensus. Deterministic algorithms
include Paxos, Raft, and PBFT, and probabilistic consensus algorithms
include PoW and departmental PoS[18] .
The following is a description of the common consensus algorithms
mentioned above.
 10.3 Blockchain Technology Graph 243

(a) Probabilistic Consensus

Probabilistic consensus means that the blockchain data reach an agreement
with a certain probability, and the probability gradually increases over time.
There is no guarantee that blockchain data will not be changed in the future.
This has weak consistency. This type is usually applied in the public chain
area. This article will introduce the algorithms mentioned above as well as
their extended algorithms.
1. PoW:
PoW, the Proof of Work algorithm, which is used in Bitcoin, is an algo-
rithm to determine the contribution of each node according to the rights
in the computation process made. The simple understanding is proof to
confirm that you have done a certain amount of work. In the blockchain
network constructed based on the PoW mechanism, nodes compete for
the accounting right by calculating the numerical solution of the random
hash, and the ability to obtain the correct numerical solution to generate
a block is a specific manifestation of the node’s computing power. The
PoW mechanism has the advantage of being completely decentralized.
In the blockchain with the PoW mechanism as the consensus, nodes can
enter and exit freely. However, the computing behavior based on the
workload proof mechanism will cause a lot of waste of resources, and
the cycle required to reach a consensus is also long[19] . Therefore, PoW
is suitable for some resource-based and service-based projects.

2. PoS:
PoS, the Proof of Stake algorithm, is an improved consensus mechanism
designed for the shortcomings of the PoW. It is mainly through the
method of accounting for rights and interests to solve the problem of
inefficiency of the network, waste of resources, and consistency of each
node. Different from the PoW, which requires nodes to continuously
perform hash calculations to verify the validity of transactions, the PoS
adds node weights based on PoW and introduces tokens as the basis for
weighting and, based on the proportion and time of each node’s weight,
reduces the difficulty of PoW in proportion to speed up finding random
numbers. The advantage of the PoS is that it solves the problems of
waste of resources and low efficiency in PoW, but its decentralization is
weaker.
244 Introduction to Blockchain Technologies

3. DPoS:
DPoS, the deposit-based proof of stake algorithm, is an authorization
consensus mechanism similar to the board of directors. This mechanism
allows each token holder to vote on the nodes of the entire system, decide
which nodes can be trusted and act on their behalf for verification and
accounting, and generate a small amount of corresponding rewards.
The decentralization of the blockchain-based on the DPoS mecha-
nism relies on a certain number of representatives, rather than all users.
In such a blockchain, all nodes vote to elect a certain number of node
representatives, who will act on behalf of all nodes to confirm blocks and
maintain the orderly operation of the system. At the same time, all nodes
in the blockchain have the power to recall and appoint representatives
at any time. If necessary, all nodes can vote to disqualify the current
node representatives and re-elect new representatives to achieve real-
time democracy[20] . DPoS greatly improves the processing capacity of
the blockchain and reduces the maintenance cost of the blockchain, so
that the transaction speed is close to that of a centralized settlement
system.

(b) Deterministic Consensus

Deterministic consensus refers to the fact that once the blockchain data is
agreed upon, it cannot be changed and has strong consistency. This category
is usually the traditional distributed consensus algorithm and its improved
version.
1. Paxos:
Paxos algorithm is a consensus algorithm based on message passing and
is highly fault-tolerant. The problem solved by the Paxos algorithm is
how to reach an agreement on a certain value in a distributed system
where messages may be delayed, lost, and repeated, to ensure that
no matter any of the above exceptions occurs, the consistency of the
resolution will not be destroyed.
The Paxos algorithm can be applied to a variety of scenarios. For
example, in a distributed database system, if the initial state of each node
is the same, and each node performs the same sequence of operations,
then they can finally obtain a consistent state. To ensure that each node
executes the same sequence of commands, a “consensus algorithm”
needs to be executed on each instruction to ensure that the instructions
seen by each node are consistent[21] .
 10.3 Blockchain Technology Graph 245

2. Raft:
Paxos is indeed a very influential consensus algorithm. It can be said that
it has laid the foundation for distributed consensus algorithms. However,
due to its incomprehension and difficulty in implementation, it is very
difficult to implement a complete Paxos algorithm[22] . Therefore, there
are many Paxos variants, the most famous of which is the Raft consensus
algorithm.
Raft is a distributed consensus algorithm used to manage log repli-
cation consistency. Its functions are similar to Paxos, but Raft is easier
to understand and implement, and it is easier to apply to actual systems.
The difference is that Raft uses stronger assumptions to reduce the state
that needs to be considered. It decomposes the consistency problem into
three relatively independent sub-problems, namely leader election, log
replication, and security. This makes Raft easier to understand and easier
to apply to the establishment of actual systems. The Raft algorithm is
one of the more consensus algorithms adopted by the alliance chain.

3. PBFT:
PBFT, the practical Byzantine fault tolerance algorithm, is a distributed
system consensus algorithm that can tolerate Byzantine errors. The core
of PBFT is to form a consensus on the state of the network between
normal nodes[23] . It is a consensus mechanism that allows voting and
the minority obeys the majority. The algorithm can tolerate Byzantine
errors and can allow the participation of strong supervision nodes. The
algorithm has high performance and is suitable for the development
of enterprise-level platforms[24] . At present, the mainstream enterprise-
level blockchain solution Hyperledger Fabric and Hyperchain, which
were developed by Hangzhou Qulian Technology, both provide PBFT
implementation solutions. However, the original PBFT algorithm is not
perfect in terms of reliability and flexibility. Hyperchain has enhanced
reliability and flexibility and designed and implemented an improved
algorithm of PBFT, namely robust Byzantine fault tolerant (RBFT).
RBFT optimizes the execution process of the native PBFT algorithm,
adds dynamic data failure recovery and dynamic node addition and
deletion mechanisms, which greatly improves the stability, flexibility,
and availability of the system, and better meets the production envi-
ronment requirements of enterprise-level consortium chain. Realize the
single point automatic recovery, automatic recovery strategy of dynamic
246 Introduction to Blockchain Technologies

data of the ledger[24] . This strategy can ensure that the cluster dynam-
ically adds and deletes nodes and dynamic data failure recovery under
non-stop conditions greatly enhances the availability of the consensus
module and improves the overall transaction throughput and system
stability of the system while ensuring the strong consistency of the BFT
system performance, reaching ten thousand TPS and millimeter-level
delay[25] . RBFT’s automatic recovery mechanism makes the storage of
its node consistent with the latest storage state in the system as soon
as possible by actively requesting the block and the block information
that is being consensus. It solves the data recovery problem caused
by network jitter, sudden power failure, disk failure, etc., during the
operation of the blockchain network, which may cause the execution
speed of some nodes to lag behind most nodes or directly go down.
As shown in Table 10.1, the characteristics of common consensus
mechanisms are listed.

Table 10.1 Characteristics of common consensus mechanisms.

Algorithm Minimum Fault Fraud Advantage Disadvantage
node tolerance preven-
tion
PoW 1 The difficulty <50% Completely Violence causes a
of algorithm, decentralized, lot of waste of
computing Free to enter resources, and the
power and exit cycle for reaching
consensus is long.
PoS 3 The difficulty <50% Reducing the Waste of computing
of algorithm, number of resources
computing nodes
power, participating in
weight verification and
accounting can
speed up the
consensus cycle
DPoS 3 The difficulty <50% The number of Waste of computing
of algorithm, verification and resources
computing accounting
power, nodes is further
weight reduced, and
consensus
verification in
seconds can be
achieved
 10.3 Blockchain Technology Graph 247

Table 10.1 Continued

Algorithm Minimum Fault Fraud Advantage Disadvantage
node tolerance preven-
tion
Paxos 3 2F+1 F Efficient, Hard to understand
strict and implement,
mathematical crash fault-tolerant
proofs
Raft 3 2F+1 F Easier to Crash fault-tolerant
understand and
implement
PBFT 4 3F+1 F Consensus is The degree of
confirmation, decentralization is
transaction not as good as PoW,
confirmation and the reliability
time is short, and flexibility are
and efficiency not perfect
is high
RBFT 4 3F+1 F Supports The degree of
dynamic node decentralization is
management not as good as PoW
and failure
recovery
mechanisms
Note: F is the number of crash fault nodes or Byzantine fault nodes.

(4) Smart Contract

The smart contract refers to the custom program logic embedded in the
blockchain. It is responsible for implementing, compiling, and deploying
the business logic of the blockchain in the form of code, completing the
transaction trigger execution of the established rules, and minimizing manual
intervention.
According to Turing completeness, smart contracts are divided into two
types: Turing complete and non-Turing complete[26] . Turing’s complete smart
contract has strong adaptability and is suitable for businesses with more
complex logic, but there is a risk of falling into an endless loop. Non-Turing
complete smart contracts are simpler, more efficient, and safer, but they
cannot implement arrays or more complex data structures.
The development of blockchain technology has provided a good operating
foundation for smart contracts, and smart contracts have played an impor-
tant role in the blockchain. The smart contract of the blockchain was first
248 Introduction to Blockchain Technologies

established in Ethereum[27] . Ethereum is an open-source public blockchain

platform with smart contract functions. The Ethereum project draws on the
technology of the Bitcoin blockchain and expands its application range.
(5) Data Storage
Block is an important data structure in the blockchain ledger, which stores
core transaction information. The blocks of the blockchain are linked in a
certain order. This is a logical sequence, but it is not necessary to follow
this logical sequence when storing but to determine each other through a
hash pointer. The logical relationship between the hash pointers is a data
structure, to be precise, a pointer to the storage location of the data, and also
the hash value of the location data[28] . Compared with ordinary pointers, hash
pointers can not only tell you the storage location but also verify that the
data has not been tampered with. It is generally stored in the form of the
hash value of transaction information. In addition, some important additional
information, such as the generation time of the block, the hash value of this
block, and the sub-block, is used when nodes compete for the right to keep
accounts. Random numbers, etc., these blocks are arranged according to the
mutual relationship determined by the hash pointer to form a blockchain data
structure, which is the storage in the blockchain.

10.3.2 Expansion Technology Overview

With the continuous expansion of the breadth and depth of blockchain
applications, the industry’s requirements for blockchain are gradually rising.
Changes in industry needs have spawned a series of expansion technologies
to optimize the blockchain[29] ; as shown in Figure 10.3, it includes four
aspects: scalability, interoperability, collaborative governance, security, and
privacy.

Figure 10.3 Expansion technology graph of blockchain technology.

 10.3 Blockchain Technology Graph 249

(1) Scalability
Scalability has become a key technology to improve system perfor-
mance. Broader scalability includes two aspects: performance scalability
and function scalability. Performance scalability focuses on increasing trans-
action throughput through vertical expansion, and functional scalability
focuses on enhancing blockchain service capabilities through horizontal
expansion[30] .
In Bitcoin, the efficiency of the consensus mechanism is the bottleneck of
performance scalability[31] . The overall performance of the whole blockchain
system is also impacted by the bandwidth and speed of the P2P network,
the performance of sign, verification, a hash of cryptographic algorithms, the
single node performance, the IO rate of storage, the execute rate of virtual
machine, etc. Therefore, scalability has become a key technology to further
enhance the processing capabilities of the blockchain.
At present, common scalability technologies include sharding mech-
anism, namespace mechanism, lightning network, state channel, directed
acyclic graph (DAG) consensus, on-chain, off-chain, side-chain, child-
chain, etc.
(2) Interoperability
Interoperability refers to the ability to exchange information between a
blockchain system instance and other system instances or components and
to use the exchanged information. Among them, other system instances refer
to all external system instances except the blockchain system instance. It
includes three aspects: application layer interoperability, inter-chain interop-
erability, off-chain data interoperability.
It refers to the ability to exchange information between the upper application
system instance and the underlying blockchain system instance and to use
the exchanged information. Specifically, it contains two meanings: (1) data
circulation and value sharing between different applications through the
underlying chain; (2) docking interaction between the upper-level application
system instance and the underlying blockchain system instance.
It refers to the ability to exchange information between different blockchain
system instances and use the exchanged information. It can also be called
cross-chain. It is mainly manifested in the process of information interaction
between different blockchain system instances, including homogeneous chain
interoperability and heterogeneous chain interoperability.
250 Introduction to Blockchain Technologies

It refers to the ability to exchange information between the blockchain system

instance and the off-chain data system and to use the exchanged information.
It is mainly manifested in the process of secure interaction between the
blockchain system and the external data system.
(3) Security and Privacy
Blockchain is a decentralized ledger, and the security protection capabilities
of different nodes are uneven, leading to the risk of the system being attacked.
Smart contract developers have uneven capabilities, coupled with the lack of
convenient and effective smart contract automatic auditing schemes, resulting
in smart contract security accidents occurring frequently, and it has become
the hardest hit area for blockchain security. The blockchain is decentralized.
The characteristics of quasi-anonymity, coupled with the lack of effective
supervision methods, lead to the risk of abuse of the chain system. With the
continuous enrichment of data on the chain and the continuous expansion
of application scenarios, privacy issues in the data circulation process have
become increasingly prominent. For example, the smoothness of user iden-
tity information, asset information, transaction flow, and other information
requires certain technical means to improve the data circulation process
and privacy protection capabilities[32] . In addition to traditional data encryp-
tion, access control, and data processing, common methods also introduce
cryptographic algorithms such as zero-knowledge proofs and homomorphic
encryption, as well as TEE, MPC, and federated learning to enhance privacy
protection capabilities in multi-party collaboration scenarios. Realize the
“available but not visible” data.
(4) Collaborative Governance
Blockchain technology, as a decentralized ledger, emphasizes the model
of equal cooperation between all participants. Compared with traditional
centralized services, the cooperative model of equal status among multiple
parties increases the difficulty of collaborative governance. Blockchain gov-
ernance refers to the decision-making process of creating, modifying, and
updating system rules. It can be divided into two types: off-chain governance
and on-chain governance. The common form of off-chain governance is a
governance committee formed by core participants, and governance rules are
usually voted by the governance committee. On-chain governance means that
governance rules are usually encoded in a governance agreement, and each
participant makes online voting decisions[32] .
 10.3 Blockchain Technology Graph 251

10.3.3 Supporting Technology Overview

As a kind of software system, the actual application process of blockchain
requires supporting technology to improve system security, optimize user
experience, and accelerate the development of blockchain, including sys-
tem management, infrastructure, operation and maintenance, as shown in
Figure 10.4.
(1) System Management
The system management layer is responsible for the management of other
parts of the blockchain architecture, which mainly includes two types of
functions: authority management and node management. For permission
chains, authority management is a key part of blockchain technology, espe-
cially for data access. Node management is based on chain-level roles
and chain-level configuration, and chain-level administrators conduct node
management through proposal voting. Node management includes adding
nodes and deleting nodes. As a complex decentralized system that integrates
multiple technologies, blockchain is faced with system security and compli-
ance security issues during actual use. Like traditional centralized services,
chain system security currently mainly involves network attacks such as
DDoS attacks, Sybil attacks, and eclipse attacks. Common defense methods
include setting up proxy nodes, strengthening network identity authentication
systems, and network current limiting. To promote the stable and compliant
development of the blockchain system, it is necessary to strengthen the
implementation of the regulatory mechanism in terms of policies, regulations,
and technical tools.
(2) Infrastructure
The infrastructure of blockchain technology is divided into two categories:
general infrastructure and dedicated infrastructure. General infrastructure
refers to the software and hardware resources required during the use of

Figure 10.4 Supporting technology graph of blockchain technology.

252 Introduction to Blockchain Technologies

the chain system and traditional Internet services, and it is versatile, such
as communication networks, cloud platforms, etc. Dedicated infrastructure
refers to the software and hardware resources specifically required during
the use of the chain system, such as unified chain resource management sys-
tem, and digital identity management system. The general infrastructure has
been relatively mature after years of development. Dedicated infrastructure
involves the development of blockchain governance and related standards
and specifications. The business is in the early stage of the industry. With
the eruption of the interconnection needs of different chains, it will promote
the construction of blockchain dedicated infrastructure.
(3) Operation and Maintenance
As a complex system that integrates cryptographic algorithms, P2P networks,
consensus mechanisms, smart contracts, and other technologies, blockchain
is difficult to deploy, operate, and maintain. To lower the barriers to the
use of blockchain technology, some blockchain platform applications have
emerged, such as blockchain as a service (BaaS) and blockchain fusion server
(BFS),and they greatly reduce the difficulty and workload of blockchain
system operation and maintenance.
BaaS supports one-stop research and development (R&D) services such
as the one-click deployment of alliance chains, visual monitoring operation
and maintenance, and smart contract R&D, helping users to focus more on the
R&D and innovation of core business and achieve rapid business on-chain.
BFS provides blockchain one-click deployment networking services, pre-
senting the original cumbersome and time-consuming manual deployment
process of blockchain networks in an automated and visual way. At the same
time, BFS, which is developed by Hangzhou Qulian Technology, provides a
full range of operation and maintenance services, through node monitoring
visualization, to meet the needs of monitoring data visualization in various
scenarios.
For some common operations in operation and maintenance operations,
it is recommended that developers formulate a unified operation specification
according to their environment, which helps to minimize operation errors and
reduce operational risks. Before deployment, a reasonable resource assess-
ment can reduce the frequency of later expansion and enable the system to
effectively carry the existing business volume while coping with the subse-
quent business growth for a certain period. In actual deployment, according
to the security of the organization’s private key, it can be divided into two
 10.4 Conclusion 253

scenarios: single-institution deployment and multi-institution P2P deploy-

ment, and the deployment can be selected according to the actual scenario.
A complete monitoring system can provide early warning in time before the
event or provide detailed data after the event, which can be used to track down
the problem and effectively shorten the average repair time of anomalies.
The monitored indicators mainly include host status, node process, consensus
status, error log, and alarm information. Once the deployment is completed
and the operation is successful, it will enter the later stage of operation and
maintenance. The following operations must be performed in strict accor-
dance with the operation and maintenance specifications, including stress
testing, security control, organization management, group management, node
management, node upgrade, node migration, certificate management, disk
expansion, etc.

10.4 Conclusion
Blockchain technology is born from Bitcoin, which is perfectly integrated
with PoW, P2P network, and cryptographic algorithms, and is developed and
grown from Ethereum which is a nicely used smart contract. Now that it is
stepping into stage 3, where is its development direction?
With the advantages of decentralization or multi-centralization, being
tamper-proof, traceability, transparency, and high reliability, blockchain tech-
nology will be evolving rapidly from three aspects: core technology, expan-
sion technology, and supporting technology. The innovation of core tech-
nologies which include cryptographic algorithms, P2P network consensus
mechanism, smart contract, and data storage will improve the performance
of blockchain systems. The innovation of expansion technologies which
include scalability, interoperability, collaborative governance, security, and
privacy will enhance the service capabilities and application breadth of the
blockchain system. The innovation of supporting technologies which include
system management, infrastructure, operation, and maintenance will improve
usability and ease of use of blockchain systems.
As a technology to reconstruct production relations and information
infrastructure, fusion with Internet of Things, artificial intelligence, Big
data, cloud, and 5G, blockchain technology will be strong enough to face
the challenges and risks on compliance, large-scale applications, the real
economy, and digital economy.
254 Introduction to Blockchain Technologies

References
[1] IRENA, Innovation Landscape for a Renewable-Powered Future:
Solutions to Integrate Variable Renewables. Preview for Pol-
icy Makers. Abu Dhabi: International Renewable Energy Agency,
2019.
[2] Hellani, H., Sliman, L., Samhat, A. E., and Exposito, E., On Blockchain
Integration with Supply Chain: Overview on Data Transparency. Logis-
tics, 2021, 5(3): 46
[3] IRENA, Innovation Landscape Brief: Blockchain. Abu Dhabi: Interna-
tional Renewable Energy Agency, 2019.
[4] Livingston, D., Sivaram, V., Freeman, M., and Fiege, M., Applying
Blockchain Technology to Electric Power Systems, 2018.
[5] Idrees, S. M., Nowostawski, M., Jameel, R., and Mourya, A. K. Security
Aspects of Blockchain Technology Intended for Industrial Applications.
Electronics, 2021, 10(8): 951.
[6] China Academy of Information and Communications Technology
(CAICT), Blockchain White Paper, 2020. Available: http://www.trus
tedblockchain.cn/#/result/result/resultDetail/3b1928a5c5404eaf80958
db5f39bef21/0
[7] China Academy of Information and Communications Technology
(CAICT), Blockchain Interoperability White Paper, 2020. Available:
http://www.trustedblockchain.cn/-/result/result/resultDetail/c90ee9
57aaf74f20960813999b7ea6bd/0
[8] Musleh, A. S, Yao, G. and Muyeen, S. M., Blockchain Application in
Smart Grid-Review and Frameworks. IEEE Access, 2019, 7: 86746-
86757.
[9] Blockchain Technology [EB/OL]. [Accessed: 2020-04-03]. Available:
https://www.tenet.eu/our-key-tasks/innovations/blockchain-technolo
gy.
[10] Yli-Huumo, J., Ko, D., Choi, S., et al., Where is Current Research on
Blockchain Technology-A Systematic Review. PLoS One, 2016, 11(10):
e0163477.
[11] Castro, M. and Liskov, B. Practical Byzantine Fault Tolerance and
Proactive Recovery. ACM Transactions on Computer Systems (TOCS),
2002, 20(4): 398-461.
[12] Castro, M. and Liskov, B., Practical Byzantine Fault Tolerance. In
Proceedings of the Third Symposium on Operating Systems Design and
Implementation, 1999, 173-186.
 References 255

[13] Available: https://en.wikipedia.org/wiki/Quorum _ (distributed - com-

puting)
[14] Schneider, F. B., Implementing Fault-Tolerant Services Using the State
Machine Approach: A Tutorial. ACM Computing Surveys, 1990, 22(4):
299-319.
[15] Musleh, A. S., Yao, G., and Muyeen, S. M., Blockchain Application
in Smart Grid-Review and Frameworks. IEEE Access, 2019, 7: 86746-
86757.
[16] Swan, M., Blockchain: Blueprint for a New Economy. USA: O’Reilly,
2015.
[17] Kushch, S. and Castrillo, F. P., A Review of the Applications of
the Blockchain Technology in Smart Devices and Distributed Renew-
able Energy Grids. ADCAIJ: Advances in Distributed Computing and
Artificial Intelligence Journal, 6(3): 75-84.
[18] Eranga, B., et al., A Blockchain Empowered and Privacy Preserving
Digital Contact Tracing Platform. Information Processing & Manage-
ment, 2021, 58(4): 102572-102572.
[19] A Brief Introduction to Several Consensus Mechanisms of Blockchain -
Katastros.
[20] ISO/TC 307 Blockchain and Distributed Ledger Technologies.
[EB/OL]. Available: https://www.iso.org/committee/6266604.html.
[21] Available: http://www.jos.org.cn/html/2021/2/6150.htm#outline_ancho
r_24.
[22] IEEE Xplore. [EB/OL]. Available: https://sagroups.ieee.org/cts-bsc/.
[23] IEEE P2418.5 - Standard for Blockchain in Energy. [EB/OL]. Available:
https://standards.ieee.org/project/2418_5.html.
[24] Available: https://www.hyperchain.cn/products/hyperchain.
[25] Available: http://kns.cnki.net/kcms/detail/32.1117.TV.20210412.1643.
006.html.
[26] Cheng, C.-y., et al., Vested Interests: Examining the Political Obstacles
to Power Sector Reform in Twenty Indian States. Energy Research &
Social Science, 2020, 70: 101766.
[27] Elliptic, The Global Standard for Blockchain Intelligence. [EB/OL].
[Accessed: 2017-06-10]. Available: http://www.elliptic.co/
[28] Blockchain Group, Blockchain Intelligence Group. [EB/OL].
[Accessed: 2017-06-10]. Available: http://Blockchaingroup.io/.
[29] Diestelmeier, L., Regulating for Blockchain Technology in the Electric-
ity Sector: Sharing Electricity – and Opening Pandora’s Box? [EB/OL].
[Accessed: 2017-05-09].
256 Introduction to Blockchain Technologies

[30] Dorri, A., Kanhere, S. S., and Jurdak, R., Blockchain in Internet of
Things: Challenges and Solutions. arXiv preprint arXiv:1608.05187,
2016.
[31] Cachin, C., Architecture of the Hyperledger Blockchain Fabric. In
Proceedings of the Workshop on Distributed Cryptocurrencies and
Consensus Ledger, 2016.
[32] Available: https://mp.weixin.qq.com/s/fjscjSNZf-kkqX97TBrNvQ.
 11
Blockchain Technologies for Renewable
Energy Resources with Case Study:
SHA–256, 384, and 512

Kaung Si Thu∗ , Shubham Tiwari, and Weerakorn Ongsakul

Department of Energy, Environment, and Climate, School of Environment,

Resources and Development, Asian Institute of Technology, Thailand
E-mail: [email protected]
∗ Corresponding Author

Abstract
Blockchain is a new technology that has captured the attention of energy
companies, supply companies, entrepreneurs, software companies, invest-
ment firms, government bodies, and academics. This chapter analyzes the
performance of the hardware used to simulate the peer-to-peer energy trading
between distributed generations with proof-of-work (PoW) consensus mech-
anism especially in the case of SHA-256, 384, and 512. The PoW mechanism
creates a blockchain environment to trade between parties without authoriza-
tion, and it establishes a systematic methodology for the trading of energy
with an advanced contract system. Performance is measured by the CPU and
GPU of the device during the complete period of simulation. Linear trendline
and percentage evaluation demonstrate different scenarios of three secure
hash algorithms (SHAs). The result shows that SHA-512 has the highest
performance efficiency in terms of hardware usage and transaction duration
in peer-to-peer energy trading systems. It means that SHA-512 takes more
memory size in creating a block than other SHAs; however, it has sustainable
development in the processing of the blockchain demonstration.

257
258 Blockchain Technologies for Renewable Energy Resources with Case Study

Keywords: Renewable energy, blockchain, secure hash algorithms (SHAs),

peer-to-peer, energy trading.

11.1 Introduction
Energy is an absolute necessity for developing a sustainable society. The
trading of energy is a vertical approach from generation to end-user; however,
technology advancements and market behavior have led to the inclusion of
customers into the electricity trading market, where prosumers can trade their
excessive amount of energy [14, 15]. This type of market can reduce carbon
emissions, generate profit for both prosumers and consumers, and benefit
the grid system without absolute centralized authorization. The decentralized
energy system is used to mitigate the information, communication, and tech-
nology in every aspect of the energy market to solve sustainable development
goals, energy efficiency, security, and optimization. Since diverse energy
sources exist, the market might be turbulent at any time [11–13].
The integration of blockchain is widely acknowledged as a paradigm shift
to a decentralized energy market with bi-directional power flow [16]. When
merged with smart contracts, it offers accessible, tamper-proof, and reliable
platforms that can allow technology solutions. This research allows us to
demonstrate the hardware performance in different usage of secure hash algo-
rithms (SHAs) at peer-to-peer energy trading between distributed generations
with a consensus mechanism. The proof-of-work (PoW) mechanism creates
a blockchain environment to trade between parties without authorization, and
it will establish a systematic methodology for the trading of energy with
an advanced contract system. The hypothesis is that transaction time in the
PoW mechanism with less computation power depends on the different use
of SHAs in the trading system. Additionally, the performance comparison is
evaluated on distinct SHAs used in the consensus mechanism-based trading
[17, 20–22].

11.2 Local Energy Trading and Consensus Algorithms

In the event of local energy trading, the community organizations will profit
as well as the utility grid infrastructure will be less stressed [7]. When it
comes to the energy market, the local grid can function as an ideal client
by supplying and selling electricity between grids [8]. While the local grid
systems have been generally approved after the feasibility pilot project, the
 11.2 Local Energy Trading and Consensus Algorithms 259

cost of installing renewable energy sources and storage technologies is still a

barrier that must be solved.
This research used the solar generation and consumption data from an
institute at Thailand – Asian Institute of Technology. Installed with solar
panels at Library and Energy department buildings, they produce electricity
during the daytime and consume, which acts as prosumer models. This
study compares the performance of blockchain simulation under various
SHAs (SHA-256, SHA-384, and SHA-512) based on the author’s previous
research [10]. Moreover, the research [9] demonstrated to promote the poten-
tial local energy trading within the institute if more participation is involved,
and [10] proved the laboratory simulation of blockchain-based energy trading
in prosumer bodies.
Demand-side integration is a critical component of dispersed generating
services, and it may be classified based on how demand fluctuations are gen-
erated by different factors. However, decentralized consortium blockchains
are spread across a variety of hardware owned by different entities [23–25]. A
ledger system or blockchain database, a chain of blocks network that reflects
transactions. Using a blockchain, people can send information directly to
one another, creating a link between them and respective data. Each client
joins the network using their blockchain application node. The identification
of a client is determined by a cryptographic key pair. A smart contract is
a protocol, necessarily for blockchain-based energy trading that optimizes
the execution of the corporate activity, and the right procedure is done by

Figure 11.1 Typical structure of blockchain [3].

260 Blockchain Technologies for Renewable Energy Resources with Case Study

the consensus mechanism. A cryptographic key is extracted, allowing for

accessing the user identification, which is an address that is unique to each
user’s server. A block typically has information of index, timestamp, nonce,
transaction data, a hash function of the previous block, and itself. The con-
sensus method ensures that a block record is replicated as distributed ledgers,
preventing tampering that may occur in many places at the same time. Each
consensus mechanism method uses a computational technique to calculate the
different hashes. Due to the high energy intensity of the consensus algorithms,
many sectors have changed to various mechanisms such as proof-of-stake
(PoS), zero-proof-knowledge (zk rollups), and proof-of-authority (PoA) in
the case of electricity trading [2].
Research [4] has shown that the PoW consensus mechanism which is the
most energy-intensive algorithm has the highest practice in 140 blockchain-
based energy sectors. Thus, nowadays, the transformation to low energy
consumption algorithms is one of the challenges and priorities in the energy
sector [18, 19]. Comparison between consensus mechanisms is performed
with a single-board computer in the study of Heck et al. [5]. It established the
performance of PoW and PoA in the local energy market on the Ethereum
platform which shows successful market-based solutions. PoA is also a more
cost-effective alternative.
A function that accepts an entry of any size and outputs a string message
of a specified size. Message integrity and digital signature systems can benefit
from the usage of encrypted hash functions since they have several extra
characteristics. National Institute of Standards and Technology (NIST) [6]
introduced SHA family which includes SHA-0, SHA-1, SHA-2, and SHA-3.
Non-linear functions are included to create complex and SHAs in which
they produce output lengths of 160, 224, 256, 384, and 512 bits as shown
in Figure 11.2. Within the SHA-2 family, the internal state size in SHA-256
and SHA-384 is 32 bits, while SHA-512 has 64 bits.

11.3 Simulation
11.3.1 Energy Trading Model and Case Study
Prior research by the author demonstrates the ability to trade using blockchain
by SHA-256-based PoW. A total of 45 transactions had occurred includ-
ing the genesis block, and the output graph of the trading can be seen in
Figure 11.2 which illustrates the demand of the library building because it had
demand, when the demand by energy building had not occurred. Figure 11.3
 11.3 Simulation 261

Figure 11.2 Demand power from Library building and net surplus solar power at Energy
building [10].

Figure 11.3 Visual model of energy trading.

depicts a visualization of the blockchain link between two prosumer mod-

els and advanced contract; equally, a smart contract is applied in the
simulation.
262 Blockchain Technologies for Renewable Energy Resources with Case Study

In the model of peer-to-peer energy trading, the demand and supply

energy data have been selected from one day and they are particularly
used in the simulation with a real-time trading period. The peers are solar
generation from the energy building which approximately had 8762 kWh and
64,550 kWh at the library building. There was sufficient surplus energy to
trade between these units even after own consumption in which 1 W is equal
to one token currency. MATLAB blockchain function [1] is used to execute
the market model and blockchain application.
The connection between two prosumer bodies assures the reliable estab-
lishment via local host which will be used to send information and transaction
data to each other. Each advanced contract will follow the protocol and energy
limit and then the token which is transferred into the wallet of two parties.
The surplus energy will be transferred directly to the demand side instantly
after the token is deposited into the seller’s wallet. The advanced contract
complies with the transaction termination protocol when the buyer reaches
insufficient balance and then sends information to both parties as regards the
closure condition.
The hardware used in the single computer to perform the simulation
is equipped with a 64-bit-based processor (Intel Core i7-10510U CPU @
1.80 GHz to 2.30 GHz), approximately 7.9 GB of Intel UHD GPU, 9.9 GB
of NVIDIA GeForce MX330 GPU, and 15.8 GB random access memory to
handle the cache. The setup is used to simulate the analysis and local host is
set up for peer-to-peer connection. Altering of SHAs inside the blockchain
functions can be completed as System.Security.Cryptography.SHA256/384/
512Managed.

11.3.2 Performance Result and Evaluation of the Models

at Different Hash Algorithms
Two MATLAB software were used separately to perform as blockchain
applications for both prosumer units. Each block has a single transaction
with a total of 44 transactions being executed. An example for the block
information which is data 1347 from different SHAs is applied to compare
the length of the output, as seen in Table 11.1, which includes distinct nonce
values, time stamps, and the hash function in hexadecimal characters.
After the successful establishment of the local connection between two
blockchain platforms, performances such as usages of GPU 1 (NVIDIA
GeForce MX330), GPU 2 (Intel UHD), CPU, the memory usage of
GPUs, and CPU power consumption are combined and monitored during
 11.3 Simulation 263

Table 11.1 Block information of SHA-256, 384, and 512.

SHA-256 SHA-384 SHA-512
Index 4
18-August-2021 18-August-2021 18-August-2021
Timestamp
16:00:51 18:01:54 23:45:51
Nonce 183,732 23,452 40,524
Data 1347
0x1B23B65F542EDF1B7
0xD4EEB25F60B833480 BD680591586C667F477
0x03DCF2C4DE2086B35 486EFBF2B417D3D641A BDD82449EC4B01A747
Previous 3455BF07784BD528A1A D60D52733305DD1648 5FB5FAAB
Hash in Hex 1F8A3C14CDB19BB8581 4211091 6414816520E76AC784A
A93C6 DCB81A20816210FEC44 9522C57EAE41E6244E6
EC5B1C5BD55051F5 F96B65075BE6DA0AB85
4E370EE11
0x4AE8FB686EB663746
0xBB1E3E4B2580A4DB1 977CE23FBC5D2C7A4D
0xF221CF3C46BD6B4EB 6C0AE62C2423D14E4EF BCFBD9750A172FECC9C
BE36AD33EA7FC7E9792 5FA924DBF84AB061335 CE0620A1AC
Hash in Hex 05E77898BC0351F9F71 01D8CD3B88 269D258CB46C85A290F
A05B4 394F9E7F59EFD0B094C 571EF3C5FFF2B1BEF97
CD4EE608EF7A 45081D30C5AE8EA232B
C178EF

transactions of two blockchain platforms every second. Heatmap correlation

is used to analyze their correspondence with the transaction as visualized in
Figure 11.4. The transactions in SHA-256 and 512 are positively affected by
the CPU than the GPU; nonetheless, SHA-384 is not highly dependent on
both but slightly by CPU.
The implementation of blockchain-based peer-to-peer energy trading
based on previous historical datasets confirms the successful transactions
using SHA-256, 384, and 512. The visualization is illustrated for every trans-
action without an idle period during the trading. As mentioned in Figure 11.5
the average transaction time for SHA 256, 384, and 512 are 71, 81, and
67 seconds, respectively. Hence, there are 3300–3500-second periods in
the comparison, which is a vast data point to evaluate. The comparison is
distinguished into three groups such as CPU usage, GPU usage, and GPU
memory consumption. CPU usage and its power consumption have directly
concerned each other and the CPU usages for three SHAs do not fluctuate
where they mostly remain between 10% and 20% as in Figure 11.6. During
the observation of two separate GPUs, GPU 2 has been utilized better than
GPU 1 as seen in Figure 11.7. GPU 1 merely alters a small amount of period,
and it does not occupy usage than GPU 2 which has a constant operation.
264 Blockchain Technologies for Renewable Energy Resources with Case Study

Figure 11.4 Heatmap visualization between transaction and hardware performance. (a)
SHA-256. (b) SHA-384. (c) SHA-512.

Figure 11.5 Duration of each block index in the transaction.

According to Figure 11.8, the trendline model is applied at the output to

minimize the standard deviations of the dataset and simplify the performance
variation between SHAs. It indicates that the CPU usage gradually decreases
with time, and the consumption by SHA-512 is roughly 13% which is less
than the others. CPU operation reaches 13.7% in SHA2-56 and 13.4% in
SHA-384.
 11.3 Simulation 265

Figure 11.6 (a) CPU usage and (b) power consumption in different SHA scenarios.

As it is mentioned earlier, operation by GPU 1 has not been impacted by

blockchain transactions, while GPU 2 has. GPU 1 usage is less than 0.3%
for variation of SHAs; hence, it can be neglected. The linear model of GPU
2 usage indicates the noticeable changes for each SHA, where SHA-256 has
operated below 0.5%, SHA-384 has the highest usage between 1.6% and 2%,
and the usage is below 1.5% in SHA-512.
266 Blockchain Technologies for Renewable Energy Resources with Case Study

Figure 11.7 GPU usage in different SHAs. (a) NVIDIA GeForce MX330. (b) Intel UHD.

11.4 Conclusion and Recommendations

This chapter focuses on the impact of changing SHA in consensus algorithms
on the hardware performance and the transaction in the blockchain-based
peer-to-peer energy trading; therefore, the implementation of variation in
SHA has been successfully demonstrated. Generally, the blockchain system
has integrated with the SHA-256 function, but the security of the block
depends on the variation of complex consensus algorithm approaches such
 11.4 Conclusion and Recommendations 267

Figure 11.8 Trendline models of hardware performance in SHAs. (a) GPU 1. (b) GPU 2. (c)
CPU.

as PoS, PoA, and so on. It has been pointed out that changing SHAs in
blockchain-based trading impacts the GPU, CPU, and RAM performance
significantly. Based on the result, the transaction system has been dominated
by the CPU and RAM, which never reaches the maximum in any hashing
algorithm scenario. Though they are two separate GPUs, GPU 2 performed
constantly, while GPU 1 simply oscillates as a spark at some certain period.
Other literature studies had measured the performance of blockchain trans-
actions and some by comparing different consensus mechanisms but never
in the same way as this research’s approach. This research indicates the
specific analysis of SHA roles in the trading model. As the nature of SHA,
512 bit is more secured than the others in the family of SHA-2 and has the
longest output ASCII code. When we compare each SHA model, it cannot be
indeed stated which one has the strongest and better performance. However,
in a conclusion, SHA-512 has accomplished the block transactions at the
lowest CPU and GPU usage at less duration than the SHA-256 and 384.
Furthermore, this approach is encouraged to perform in Ethereum’s platform
with improved long-term energy trading system whether other findings in the
aspect of security can be expected.
268 Blockchain Technologies for Renewable Energy Resources with Case Study

References
[1] Aarenstrup, R. (2021). MATLAB Blockchain Example, MATLAB Cen-
tral File Exchange. Retrieved: August 23, 2021. https://www.mathwork
s.com/matlabcentral/fileexchange/65419-matlab-blockchain-example
[2] Hartnett, S., Henly, C., Hesse, E., Hildebrandt, T., Jentzch, C., Krämer,
K., ... & Trbovich, A. (2018). The Energy Web Chain-Accelerating
the Energy Transition with an Open-Source, Decentralized Blockchain
Platform. Energy Web Foundation.
[3] Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system.
Decentralized Business Review, 21260.
[4] Andoni, M., Robu, V., Flynn, D., Abram, S., Geach, D., Jenkins, D.,
... & Peacock, A. (2019). Blockchain technology in the energy sector:
A systematic review of challenges and opportunities. Renewable and
Sustainable Energy Reviews, 100, 143–174.
[5] Heck, K., Mengelkamp, E., & Weinhardt, C. (2020). Blockchain-based
local energy markets: Decentralized trading on single-board computers.
Energy Systems, 12(3), 603–618. https://doi.org/10.1007/s12667-020-0
0399-4
[6] National Institute of Standards and Technology. (2015). Secure Hash
Standard (SHS). (U.S. Department of Commerce, Washington, DC),
Federal Information Processing Standards Publication (FIPS), 180–184.
https://doi.org/10.6028/NIST.FIPS.180-4
[7] Soshinskaya, M., Crijns-Graus, W. H., Guerrero, J. M., & Vasquez,
J. C. (2014). Microgrids: Experiences, barriers and success factors.
Renewable and Sustainable Energy Reviews, 40, 659–672.
[8] Tao, L., Schwaegerl, C., Narayanan, S., & Zhang, J. H. (2011, May).
From laboratory microgrid to real markets—Challenges and opportuni-
ties. In 8th International Conference on Power Electronics-ECCE Asia
(pp. 264–271). IEEE.
[9] Thu, K. S., & Ongsakul, W. (2020). Simulation of blockchain based
power trading with solar power prediction in prosumer consortium
model. In 2020 International Conference and Utility Exhibition on
Energy, Environment and Climate Change (ICUE). https://doi.org/10
.1109/icue49301.2020.9307086
[10] Thu, K. S., Aung, M. S., Ongsakul, W., & Manjiparambil, N. M. (2020).
The first national conference on engineering research. In Transition
of University to Prosumer Consortium Energy Model (pp. 168–172).
Yangon; Federation of Myanmar Engineering Societies.
 References 269

[11] Santofimia-Romero, M. J., del Toro-García, X., & López-López, J. C.

(2011). Artificial intelligence techniques for smart grid applications.
Green ICT: Trends and Challenges, 41-44.
[12] Dagdougui, H., Dessaint, L., Gagnon, G., & Al-Haddad, K. (2016).
Modeling and optimal operation of a university campus microgrid.
In 2016 IEEE Power and Energy Society General Meeting (PESGM)
(pp. 1–5).
[13] Oh, S., Kim, M., Park, Y., Roh, G., & Lee, C. (2017). Implementation
of blockchain based energy trading system. Asia Pacific Journal of
Innovation and Entrepreneurship, 11(3), 322–334.
[14] Plaza, C., Gil, J., de Chezelles, F., & Strang, K. A. (2018). Distributed
solar self consumption and blockchain solar energy exchanges on the
public grid within an energy community. In 2018 IEEE International
Conference on Environment and Electrical Engineering and 2018 IEEE
Industrial and Commercial Power Systems Europe (EEEIC/I&CPS
Europe) (pp. 1–4).
[15] Dib, O., Brousmiche, K.-L., Durand, A., Thea, E., & Hamida, E.
B. (2018). Consortium blockchains: Overview, applications and chal-
lenges. International Journal on Advances in Telecommunications,
11(1&2).
[16] Marwala, T., & Xing, B. (2018). Blockchain and artificial intelligence.
arXiv preprint arXiv:1802.04451.
[17] Penard, W., & van Werkhoven, T. (2008). On the secure hash algorithm
family. Cryptography in Context, 1–18.
[18] Maharjan, P. S. (2018). Performance Analysis of Blockchain Platforms
(Doctoral dissertation), University of Nevada, Las Vegas.
[19] Zheng, P., Zheng, Z., Luo, X., Chen, X., & Liu, X. (2018, May).
A detailed and real-time performance monitoring framework for
blockchain systems. In 2018 IEEE/ACM 40th International Confer-
ence on Software Engineering: Software Engineering in Practice Track
(ICSE-SEIP) (pp. 134–143). IEEE.
[20] Gligoroski, D., & Knapskog, S. J. (2007). Turbo SHA-2. IACR Cryptol.
ePrint Arch., 2007, 403.
[21] Hawkes, P., Paddon, M., & Rose, G. G. (2004). On corrective patterns
for the SHA-2 family. IACR Cryptol. ePrint Arch., 2004, 207.
[22] Saingre, D., Ledoux, T., & Menaud, J. M. (2020, November). BCT-
Mark: A framework for benchmarking blockchain technologies. In 2020
IEEE/ACS 17th International Conference on Computer Systems and
Applications (AICCSA) (pp. 1–8). IEEE.
270 Blockchain Technologies for Renewable Energy Resources with Case Study

[23] Li, Z., Kang, J., Yu, R., Ye, D., Deng, Q., & Zhang, Y. (2017).
Consortium blockchain for secure energy trading in industrial Inter-
net of Things. IEEE Transactions on Industrial Informatics, 14(8),
3690–3700.
[24] Hatziargyriou, N. (2014). Microgrids: Architectures and Control. Hobo-
ken, NJ: John Wiley & Sons.
[25] Basak, P., Chowdhury, S., nee Dey, S. H., & Chowdhury, S. (2012). A
literature review on integration of distributed energy resources in the
perspective of control, protection and stability of microgrid. Renewable
and Sustainable Energy Reviews, 16(8), 5545–5556.
 Index

A Distribution system 30, 56, 61, 74, 79

ACCS 205, 209, 224, 225, 227, 228 Distribution transformer 79, 81, 84
Artificial intelligence (AI) 21, 45, 93,
125, 126, 173 E
Automatic capacitor control scheme Energy 2, 29, 121, 156, 235, 259, 261
(ACCS) 205, 206 Energy resources 29, 30, 121, 151,
257
Energy trading 257, 258, 266, 270
B
ETAP 174, 196, 199, 201
Blockchain 21, 45, 235, 236, 238,
Expansion technology 238, 239, 248,
245, 259, 266
253
Blockchain technologies 235, 257
Blockchain technology 45, 47, 235,
237, 239, 248 F
Flower Pollination Algorithm 149,
163
C
Capacitor bank 64, 195, 205, 212,
222 G
Consensus mechanism 46, 236, 242, GSM 65, 67, 68, 85
246, 258
Core technology 238–240, 253
Cryptographic algorithm 239, 240, H
252, 253 Home automation 6, 25
Hybrid Power Systems 116, 118,
149, 173
D
Deep learning 93, 121, 126, 135
Distributed generation I
(DG) 79, 174, 206 Industry 4.0– 9
Distribution network 31, 180, 193, Internet of Things (Iot) 2, 30, 63, 79,
205, 227 80

271
272 Index

L Power systems 31, 115, 149, 173

Load Frequency Control 149, 150 Prediction 29, 126, 132, 141, 179,
Low voltage 61, 65, 66, 67 189
LSTM network 122, 126, 132, 145 PSO 205, 216–219

M R
Machine learning (ML) 8, 93, 94, 95, Reactive compensation 212, 230
112, 204 Renewable energy 29, 79, 121, 156,
Machine learning techniques 93, 96, 179, 235, 257
115, 121
Machine learning techniques
for renewable energy 121 S
Modern hybrid power systems Secure hash algorithms (shas) 258
(MHPS) 116, 149, 173, 176 Sensor 9, 12, 14, 15, 44, 83, 84
MPPT algorithm 141, 142 SHA 240, 257, 260, 263
Smart contract 47, 235, 247, 252, 261
Smart distribution system 61, 63, 79
O Smart grid 38, 56, 117, 179
Optimization 38, 149, 161, 178, 193, Solar PV 79, 121, 134
214 Supervised and Unsupervised
Optimization techniques 161, 178, Learning 100
185 Supporting technology 238, 251, 253

P U
Particle swarm optimization (PSO) Ubiquitous computing 5
205, 207
Peer-to-peer 45, 201, 235, 241, 257,
263 W
Power quality (PQ) 61, 177 Wind energy conversion system 76
 About the Editors

P. Sivaraman holds a B.E. in Electrical and Electronics Engineering and

an M.E. in Power Systems Engineering from Anna University, Chennai,
India in 2012 and 2014, respectively. He has more than seven years of
industrial experience in the field of power system analysis, renewable energy,
power quality and harmonic assessments, and microgrids, providing techno-
economical solutions to various power quality problems for industries all
over India. He is an expert in power system simulation software like ETAP,
PSCAD, DIGSILENT POWER FACTORY, PSSE, and MATLAB. He is a
working group member of various IEEE standards and task forces. He is a
senior member of the Institute of Electrical and Electronics Engineers (IEEE),
a member of the International Council on Large Electric Systems (CIGRE),
and an Associate Member of the Institution of Engineers (India). He received
Professional Engineers (PEng India) certification from the Institution of
Engineers (India).

C. Sharmeela holds a B.E. in Electrical and Electronics Engineering, M.E.

in Power Systems Engineering from Annamalai University, Chidambaram,
India, and a Ph.D in Electrical Engineering from Anna University, Chennai,
India in 1999, 2000, and 2009, respectively. At present, she holds the post of
Associate Professor in the Department of EEE, CEG campus, Anna Univer-
sity, Chennai, India. She has 20 years of teaching experience and has taught
various subjects to undergraduate and postgraduate students. She has done
a number of research projects and consultancy work in renewable energy,
power quality and design of PQ compensators for various industries. She is
a senior member of IEEE, Life member of CBIP, Fellow of the Institution of
Engineers (India), ISTE, and Life member of SSI, India.

Meera K. Joseph (Senior Member IEEE and Senior Member SAIEE, PMI-
ITPSA) currently works as an Independent Contractor at the Independent
Institute of Education, South Africa. She has been the doctoral supervisor
for DBA students at the Milpark Business School from January 2021. For a

273
274 About the Editors

short period, she served as the HoD, School of IT and Business at AIE, South
Africa up to the end of December 2021. She earlier worked as the Associate
Professor at the DFC, Department of Electrical and Electronic Engineering
Technology, School of Electrical Engineering, University of Johannesburg
(UJ) until 2018 and held various positions at University of Johannesburg,
South Africa for 18 years (permanent roles). She received the degrees of
D.Phil. Engineering Management (rural women and information and commu-
nication technology field), UJ in April 2014, a master’s degree in Computer
Applications in 1998 from Bangalore University and a B.Sc. in Chemistry
(Physics and Mathematics sub.) from the University of Kerala, India. She has
authored or contributed to 75 research works and has around 20 years lectur-
ing experience in the computer engineering/ICT field. At UJ she was lecturing
subjects with labs related to Java/UML, C, Javascript and HTML, Linux,
VB, VBA/ VB.Net, MS Access/ SQL, C++, MS Office Applications, MS
VISIO. Her multidisciplinary research interests are ICT4D (information and
communication technology for development), smart grids, cloud computing,
AI and machine learning for development, computer networks, femtocells,
ICT for renewable energy research, ICT for power engineering research, ICT
for empowerment, data analytics, Blockchain in cybersecurity and the use of
ICT in engineering education.

Sanjeevikumar Padmanaban (Member’12, Senior Member’15 IEEE)

received a bachelor’s degree from the University of Madras, India, in 2002,
a master’s degree (Hons.) from Pondicherry University, India, in 2006, and a
Ph.D. degree from University of Bologna, Italy, in 2012. He was an associate
with various institutions like VIT University India, National Institute of
Technology, India, Qatar University, Qatar, Dublin Institute of Technology,
Ireland, University of Johannesburg, South Africa. Currently he is working
as a Faculty Member with Aarhus University, Denmark. He is a fellow of the
Institution of Engineers, FIE, India, fellow of the Institution of Telecommuni-
cation and Electronics Engineers, FIETE, India and fellow of the Institution
of Engineering and Technology, IET, UK. He serves as an editor/associate
editor/editorial board member of refereed journals, in particular, the IEEE
Systems Journal, the IEEE Access Journal, the IET Power Electronics, Jour-
nal of Power Electronics, Korea, and the subject editor of the subject Editor
of IET Renewable Power Generation, the subject editor of IET Generation,
Transmission and Distribution, and the subject editor of FACTS journal,
Canada.