2023 International Conference on Recent Trends in Electronics and Communication (ICRTEC)

Smart Transformer – An Analysis of Recent

Technologies for Monitoring Transformer
Sanjay S Tippannavar Vijay Mishra Yashwanth S D
Department of ECE EvryVed Private Limited Department of ECE
Sri Jayachamarajendra College of J601, Prince Town, Shettihalli, Sri Jayachamarajendra College of
Engineering, Mysuru, India Jalahalli West, Bengaluru, India Engineering, Mysuru, India
[email protected] [email protected] [email protected]
2023 International Conference on Recent Trends in Electronics and Communication (ICRTEC) | 979-8-3503-9619-5/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICRTEC56977.2023.10111875

Rishitha R Gowda Sathvik H R Ajay M

Department of ECE Department of ECE Department of ECE
Sri Jayachamarajendra College of Sri Jayachamarajendra College of Sri Jayachamarajendra College of
Engineering, Mysuru, India Engineering, Mysuru, India Engineering, Mysuru, India
[email protected] [email protected] [email protected]

Abstract— A crucial part of the electrical power system is physical objects with electronics, software, sensors, actuators,
played by the transformer. Continuous monitoring has become and network connectivity that can identify. IoT-based sub-
an integral aspect of its operation since it deals with high voltage station monitoring and controlling offers a workable solution
and current, which may significantly damage the device's to these issues with a fully automated system that ensures
performance and physical state. The goal of this review article increased dependability while enhancing system performance
is to provide an overview of IoT (Internet of Things)-based through efficient equipment use. A fully automated IoT-based
electrical parameter monitoring and control technologies. The sub-station that allows only qualified workers to remotely
necessity for various monitoring systems that use various protect, monitor, and manage linked equipment at a very
electrical, mechanical, optical, chemical, and acoustic sensors to
cheap cost is highly desirable in power/energy management.
monitor the critical properties of the power transformer is due
to the significance of the continuous service of power
When creating a framework for a smart sub-station, reliability
transformers and their role in the reliability of the power and the reduction of workforce costs due to IoT technology
delivery. For the power/energy management department, are among the main concerns [1].
remote equipment monitoring and control, which is often One of the most vital pieces of electrical equipment used
carried out manually or with the help of a costly PLC and
in power transmission systems is the power transformer,
SCADA system, is a crucial problem. Real data from the
environment or objects is gathered using a variety of sensors.
which transforms voltage levels. Power transformer
People may establish a machine-to-machine link with the aid of maintenance is thus needed, but because they are typically
the IoT idea. Temperature, oil level, and vibration are dispersed geographically, it is not feasible to do periodic
significant indications used to assess the condition of the monitoring due to a lack of staff. Transformer failure, as
transformer. The stability, precision, and efficiency of the mentioned above, may happen, leading to an unexpected
transformer may all be improved while using less labour. With power shutdown of the transformer from the network. The
the advent of the internet and the computational age, it is power system network of a smart grid is of great importance
extremely desired to have an intelligent monitoring and in today's world because to the increasing demands for
trustworthy controlling system over the complete sub-station
reliable power delivery. Usually, it involves using several
equipment. This may be accomplished by using the Internet of
Things (IoT) technology. The Internet of Things (IoT) is a
strategies for monitoring, safeguarding, and controlling
network of physical objects that are integrated with electronics, processes. Transformer distribution is a critical element of
software, sensors, actuators, and network connection that have distribution networks. According to the Libyan perspective,
the capacity to recognize, collect, and share data. Each strategy such systems, which may be somewhat managed, can monitor
is then briefly explained along with the appropriate reading the power system network. Transformers last around 20 to 25
recommendations. A quick guide is therefore provided for years. The transformers are often installed just before their
professionals looking into the sensors used for power operational life expires. A system that monitors all operating
transformer monitoring. parameter changes and sends real-time data to the monitoring
station. As a result, important operational parameters are
Keywords—Transformer, Communication, Oil Monitoring,
Electricity, Theft monitoring, LoRa, Arduino, Machine Learning, tracked to offer functional data for the lifespan of
CNN, Efficiency, Embedded Systems. transformers, enabling their optimal usage and extending the
useful life of the asset. Additionally, it will be helpful to
I. INTRODUCTION identify issues prior to a failure, which will save money and
Remote supervision and administration of the sub-station lead to greater consistency. Due to the widespread use of
machinery is a critical concern for the power/energy internet and Wi-Fi devices for data transmission as well as
management department. Normally, this is accomplished other network applications, it will be a desirable alternative
manually or with the aid of an expensive PLC and SCADA [2].
system. In the era of the internet and computing, it is The most important component of the power system is the
extremely desired to have a smart monitoring and reliable power transformer. It functions according to the
controlling system over the complete sub-station equipment. electromagnetic induction theory. The formation of gases
This may be accomplished by applying Internet of Things inside the transformer oil is one of the main causes of faults
(IoT) technology. The Internet of Things (IoT) is a network of in transformers, while there are other contributing variables

979-8-3503-9619-5/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 as well. Although there are many monitoring strategies, DGA in many countries, which poses a significant obstacle to the
is the one that is most frequently employed (Dissolved Gas growth of the power sector. As of 10 October 2020,
Analysis). Gas chromatography is used in DGA to separate Bangladesh's total installed capacity for power generating was
the gases from the transformer oil tank and analyse them. 20813 MW, and both the country's per capita energy output
This provides the ppm amount of gases present. Gases such and consumption have increased dramatically over the last ten
as hydrogen (H2), methane (CH4), acetylene (C2H2), years. Approximately 97% of the entire population is already
ethylene (C2H4), ethane (C2H6), carbon dioxide (CO2), and online, and it is predicted that by end of 2021, the entire
carbon monoxide are produced in transformer oil (CO). population would be covered by the network [9].
Although DGA has proven to be a valuable strategy, it has a
significant difficulty with its reliance on knowledge. Energy flow is investigated using the energy theft
detection system. To discover unauthorised tapping on
Therefore, it is crucial to develop a more effective technique
distribution lines, power theft detection is utilised. However,
that is self-sufficient and does not require significant human
tapping is accomplished using wireless data transmission and
skill, such as an intelligent condition monitoring system. The reception techniques. This programme uses the same methods
DGA dataset was used by the author to classify defects using and costs the same, but it also includes extra features for
a machine learning technique. Support vector classification wireless meter reading. The distribution network will be
and multilayer neural networks are the algorithms employed. protected by this programme. The initial step in identifying
These learning algorithms are combined with the existing savings potential, both in commercial and covert applications,
DGA data to create models. These models will provide us is bounce testing. Due to the lack of information on quick
with the mapping function between characteristics (also house demolition, the alarming costs of important metrics, the
known as inputs and outputs), which is the hypothesis [5]. slow speed of inspection equipment, and other issues, air
testing and request business side management are currently
There are numerous explicit and implicit factors that can experiencing numerous difficulties. Energy conservation in
lead to a transformer's performance to deteriorate. Some of the modern world is crucial but challenging. Online mistakes
the most frequent events include partial discharges, insulation and power abuse will be provided by the proposed system.
deterioration, humidity, moisture, overheating, winding Energy conservation is a difficult and crucial task in the
resonance, loss of winding clamping, insulating oil solid modern society. Using a Wi-Fi connection, this function
contamination, lightning strikes, system faults, system includes keeping an eye on the machine [14].
overloads, switching operations, etc. The Infrared Emission
Testing (IET) technique, which is used to assess the Currently, theft is the major problem since other methods
temperature of the transformer's external surface, is the of metre manipulation and power theft are not taken into
account by the power metres used to monitor unit
traditional method of transformer health monitoring. The
consumption. Advanced Metering Infrastructure (AMI), as
following factors, like as current, voltage, or temperature,
opposed to Automated Meter Reading (AMR), permits two-
may contribute to transformer overheating. This technique way communication between the metre and the provider.
can be used to find hotspots sites, cooling system Wireless communications are feasible because the
obstructions, and electrical connection issues. Therefore, connections from the metre to the network are fixed wired
emerging technologies like IoT provide a powerful connections, such Power Line Carriers (PLCs). With the help
mechanism in cooperation with communication tools like of a permanent cable connection, a PLC, or Wi-Fi, data may
GSM and GPRS for the proper maintenance and monitoring be delivered from a power meter's position to the central
of transformers. Data is transferred in packets from the server or the other way around [15].
transmitting unit to the operating unit. GPRS supports
Internet protocol (IP) for communication. Several servers are Given that it is a significant problem in power system
utilised for the client-server request [6]. networks all across the globe, electrical power is unlawful and
ought to be completely outlawed. It is necessary to pinpoint
The distribution transformer, one of the most important the theft's location in order to apprehend the perpetrators and
components on the distribution side, serves as the connection put an end to energy theft. A current transformer, an LCD, an
between the producing station and the customers. If ESP32 module, a LoRa-1276, and an Arduino make up the
distribution transformers are not carefully inspected, circuit. High currents cannot be handled by metres, hence
transformer breakdowns may occur. Some examples of flaws current transformers are utilised to detect them. The current
include over-current, over-voltage, a rise in the core of the running through the load is measured by one CT, while the
transformer, and an increase or decrease in the oil level. We current flowing through the supply is measured by a second
have provided an external supply from a solar panel to the CT that is attached to the power supply terminals. to establish
microcontroller and the connection modules since the the electrical power output of the source of electricity [16].
transformer may be prone to voltage fluctuations. The
network connection module and micro-controller are All of the electrical protection systems in use today rely on
completely configurable and have been configured to do the electromechanical components. The percentage of total power
needed task alone. A micro-controller that gathers and generated by the grid lost to power theft is substantially higher
analyses the transformer vital parameters, which are collected in several other nations. There are numerous ways for users to
as part of the monitoring system, is in charge of it [7]. steal electricity from the grid. Completely avoiding the energy
meter is a widely used strategy. Despite being quite
The economy of the countries in South East Asia have dependable and affordable, these systems have several
grown quickly during the last ten years. Their ability to drawbacks. Relays for electromechanical protection are
generate more power is a key factor in their economic overly large and require frequent upkeep. It is not possible to
development. Unfortunately, transformer theft happens often have many functions. The transmission line's zone is secured

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 by the differential relay. The design of this project is unique As a result, failure in terms of the crucial aspect of the
in that it uses a faster Atmega328 microcontroller to drive a transformers or power transformers becomes completely
relay. Based on fundamental building blocks, this protection unacceptable. Significant damage could be done to the
was created. The price and size of the relay will be too high if environment as well as the asset itself as a result of this failure.
it is developed in the same facility as the electromechanical The problem cannot be solved by time-based replacement. As
component. All of the components are integrated within the a result, data collection and monitoring are done using modern
controller. The device's controller functions as its brain, and technologies. Modern power systems use automatic systems
additional hardware is employed for signal conditions and like micro-controllers, microprocessors, Arduino, etc. As part
comparison. This relay has two current settings and a of transformer commissioning, as a diagnostic tool for
maximum current handling capacity of 15A [23]. condition evaluation, and as an acceptance test right after
transformer repair or refurbishment, on-site tests on high
People's everyday activities depend heavily on electrical voltage power transformers are becoming more and more
power. The IoT technique is used in most real-time popular [30].
applications due to the growth of technology. To gather
accurate data from the environment or things, many sensors The papers were segregated based on three ways;
are employed. People can establish a machine-to-machine
connection with the aid of the IoT concept. Temperature, oil • Based on years from which the papers were found.
level, and vibration are key indicators used to assess the • Based on methodologies.
transformer's condition. This technology boosts the stability, • Based on the journals / conferences from which the
accuracy, and efficiency of the transformer while reducing the papers were chosen.
need for labour. The controller receives sensor data and
checks the indication limit values. The message is transmitted
to the involved parties if the indicator's value exceeds the
threshold values. All sensor readings are transmitted over IoT
to the Android phone [27].

Ensuring the reliability and health of the distribution

power substation has long been a problem. Because of this,
maintenance personnel would occasionally shut off
transformers and circuit breakers to check on their
functionality. This approach still has catastrophic failures, not
to mention the needless maintenance. In response to a growing
demand for more affordable and effective diagnostic tools,
online monitoring and artificial intelligence analysis
techniques have been applied to the electrical distribution Fig. 1. Graph depicts the papers segregated based on methodologies.
power substation. Mobile networks and GSM devices,
including GSM modems, are extensively used, and their lower
By screening the articles, we were able to remove the
costs make them an enticing choice for other major network
outdated approaches and methodologies and identify the finest
applications as well as voice media The online tracking system
is made up of an embedded system, a GSM modem, mobile recent research that would improve the survey. Consequently,
users, cellular networks, and sensors placed at the future students will be positively impacted.
transformer's position [28]. II. METHODOLOGY
A system that demonstrates the connection between the This section has combined related approaches into a
operator and the system by integrating a transformer, micro- single header that contains six parts with thorough
controller, logic level converter, and GSM (Global System for information in order to effectively describe the best strategies
Mobile Communication Modem) can be used for monitoring based on comparable concepts and suggested methods.
the transformer in real-time. This GSM modem transmits
A. Transformer Monitoring using IOT protocols and
messages to the system and monitors the transformer's
condition. Low oil levels and extremely hot winding technologies:
temperatures cause voltage fluctuations, which ultimately lead There are many methods through which a transformer can
to insulation failure and system failure. Some power providers be monitored, one among them being using IOT. For that
typically use Supervisory Control and Data Acquisition matter, monitoring and fault detection using IOT devices is the
(SCADA) to properly monitor Power Transformers online. most optimum method when it comes to monitoring many
However, this system's amplification is very expensive [29]. transformers from a single control centre. IOT devices like
GSM/GPS modules, controllers like Arduino Mega,
Transformers transfer energy between two or more circuits Raspberry Pi and communication modules like ESP32 and
using the electromagnetic induction process without changing ESP8266 can be used [1-2][7]. These devices are further
the frequency. Although we are aware that a transformer is a integrated with a suite of sensors, which monitor various
static device, all transformers need to be regularly maintained parameters of the transformer: Temperature of the core, oil,
and repaired in order to work according to manufacturer and and Winding defects and the electrical parameters like
industry requirements. As time passes and transformers are primary and secondary winding currents, Load power etc. To
installed in more residential and commercial settings, the regulate certain parameters we have actuators which receive
likelihood of repairs and maintenance increases. Any type of commands from the control centre(could be a mobile phone,
transformer failure is going to result in serious consequences. Computer or Terminals) through the IOT devices[27][28][30].

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 calculated, and the units are then uploaded to the server. The
system relay operates and the power supply is turned off in the
event of a theft. If the customer has not paid the consumption
fees, the utility may even remotely turn on or off the system
and cut off the home's supply. This may even eliminate the
need for human labour to physically extinguish the fuse. Here,
the system is connected to an IoT server via the Blynk TCL
protocol [8][15].

Fig. 2. Transformer Monitoring Using IOT technologies.

During the transformer operation various factors affect its

performance and its crucial to regulate and monitor these
parameters to avoid a drop in the efficiency. The sensors are
constantly monitoring these parameters and feeding the data
to the control centre through the communication
devices(ESP8266,ESP32,Sim800/900 and protocols(MQTT,
Wi-Fi, ZIGBEE, Bluetooth) [1-2][6-7][15-16][27-30].The
controllers like Raspberry Pi, Arduino Mega and other
microcontrollers control these sensors and the other peripheral
devices like power sources, auxiliary communication devices
and ports(USB, RS-232, UART etc). Fig. 3. Electricity Theft detection in transformers and the proposed model
The data sent through the communication devices travel C. Transformer Condition Monitoring and Fault Detection
through a secure network established for that purpose and data Using Machine Learning and Deep Learning
is usually stored, compared and modified using a database.
Although there are many monitoring strategies, DGA is
These databases are stored in secure locations, usually in the
the one that is most employed (Dissolved Gas Analysis). Gas
control centres. The technician/operator at the control centre
chromatography is used in DGA to first remove the gases from
monitors the data coming from many transformers and
the transformer oil tank and then analyse them. This provides
responds according to the situation. For Ex: one of the
the ppm amount of gases present. Gases such as hydrogen
transformers has an issue with the oil temperature and certain
(H2), methane (CH4), acetylene (C2H2), ethylene (C2H4),
gases have accumulated which and it needs to be regulated.
ethane (C2H6), carbon di-oxide (CO2), and carbon monoxide
The operator sends a command to that transformer monitoring
are produced in transformer oil (CO). The DGA dataset can
system to open a valve to let the gases out of the oil
be used to classify defects using machine learning algorithms.
chamber[16].
Multilayer neural networks and support vector classification
B. Electricity Theft detection and monitoring in power grids are the algorithms employed. These learning techniques are
Microcontroller, LCD, GSM Module, Current used to create models from the available DGA data. The
Transformers [14][23], IR sensors to detect proximity to the mapping function between characteristics, or inputs and
transformer, a Hall effect sensor to detect magnetic field and outputs, will be provided by these models. Carbon monoxide,
proximity, and weight sensors to detect changes in the carbon dioxide, and (CO) [5].
transformer's weight [9]. To measure the current power per After receiving the vibration data from the IoT-based
load, one current transformer is attached to the side of the load, monitoring system, EML (Ensemble Machine Learning) is
while the other current transformer is linked to the supply used to extract characteristics and detect transformer
terminal to measure the source's current supply. The output is problems. The EML is constructed in the research utilising
taken into the microcontroller whenever there is a line-to- RVMs, DBNs, and SDAs with different activation functions.
ground fault in the protected zone and the current bleeds on DBNs, SDAs, and RVMs are each used in turn to extract
the short-circuited channel, causing the readings of the two characteristics from vibration signals. Then, based on the
C.T.s to vary. using HC-05 Bluetooth module, the characteristics, the features are used to detect faults.
microcontroller also activates the relay and buzzer with an
SMS alert [14][23]. The analogue current value was read by A typical DL approach is DBN, which consists of several
the AC712. Using the ACS712 data sheet, the relevant value RBMs. A visible layer v and a hidden layer h make up RBM's
is converted, and the associated voltage's Root Mean Square structure. Units make up both the visible layers and data, and
(RMS) is calculated. Current and power calculations follow the hidden layer h learns the features [11]. Conditional
the voltage calculation. When the threshold is reached, the distributions of each visible layer and hidden layer are as
units are computed, a previously determined tariff is follows;
multiplied in accordance with the unit, and the consumption
charges are calculated, consumption charges are then

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 𝑝(ℎ𝑗 = 1|𝑣) = 𝑓 (∑ 𝑤𝑖𝑗 𝑣𝑖 + 𝑎𝑗 ) (1)
𝑖

𝑝(𝑣𝑖 = 1|ℎ) = 𝑓 (∑ 𝑤𝑖𝑗 𝑣𝑖 + 𝑏𝑖 ) (2)

𝑗

Ten different activation functions, including the Arc- Tan,

TanH, Sinusoid, Softsign, rectified linear unit (ReLU), leaked
rectified linear unit (Leaky ReLU), SoftPlus, bent identity,
Gaussian, and Sigmoid functions, are used to generate 20
distinct qualitative DL methods for DBNs and SDAs,
respectively.

Fig. 5. Fault diagnosis process.

D. Transformer Oil condition/oil level detection monitoring

and monitoring using Transformer Oil
Standard testing or regular tests should be used to confirm
the essential characteristics of the transformer oil. The
electrical properties of insulating oil, including the dielectric
constant (), dielectric loss, tan delta (dissipation factor),
resistance, and resistivity of vegetable insulating oils, were
measured using an automatic ADTR-2K Plus apparatus in
accordance with standard IEC 60247. The reliability
Fig. 4. Transformer Monitoring Using IOT and Ensemble Machine technique is used to simulate the material strength and time
Learning. to failure of an electrical or mechanical system's component
A nonredundant activation function is therefore installed parts [22]. Two single-phase transformers of the same design
in each DBN and SDA, and an RVM is used as the classifier were made by design engineers. The only difference between
in the output layer. To extract features and locate transformer these transformers was the insulating fluid used in them. Both
defects in the work, 20 DLs, numbered DL1 through DL20, transformers' low voltage windings were short-circuited, and
are merged to create the EML. In relation to many ensemble the converter's high voltage windings received a short-circuit
learning techniques, the majority voting strategy is a voltage [24]. The well-known ageing indicators interfacial
straightforward and efficient method [11]. tension and total acid number provide a possible signal for oil
The thermal properties of transformers can be precisely reclamation. The fleet's acidity is also observed to study the
and efficiently reflected in infrared pictures. As a starting degradation of the oil/paper insulation system [26].
point, the infrared picture data from the transformers is
gathered. As additional differentiating characteristics, the
temperature, texture, and form attributes of the infrared
picture are extracted. The two categories into which the
example case library is separated are the labelled dataset and
the unlabelled dataset. For the labelled dataset, the sample's
imbalance ratio is first calculated. The imbalance ratio is
calculated by dividing the total number of samples by the total
number of samples in the minority subgroup. If the imbalance
ratio exceeds the predefined threshold, it is presumed that the
labelled dataset has a sample imbalance issue. It is necessary
to utilise GAN to generate samples for the minority subgroup. Fig. 6. The test configuration of BHT AC generator.
Unpredictable noise vector z is sent into the generator G.
The output is produced using synthesised sample data G, and Fibre-optic probes (Neoptix - 400 m sensitive area) were
the noise is generally Gaussian or uniformly distributed (z). used to more accurately and with little transmission loss
The discriminator D receives input from actual data. It aims to measure the temperature in windings (HV and LV locations)
perform binary classification on the supplied data x and G. (x). and oil. The degree of polymerization (DP) is an important
If an actual sample from the input source can be confirmed, it marker that offers information on the mechanical strength
produces one. If not, it returns zero. Back-propagation would and, as a result, the lifespan of the paper [24]. However,
then be used to direct the discriminator's output result in order physical modifications, particularly viscosity (), have an
to optimise the parameters and so increase G's capability. The effect on the heat resistance of the liquid insulating substance.
discriminator simultaneously enhances its categorization a dynamic thermal model based on the relationship between
performance [18].

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 temperature and viscosity proposed by D. Susa may provide ITSM2000, are employed. Various analyses, such as the
more precise results than the traditional approach [22]. autocorrelation and partial autocorrelation functions plot, the
Box-Pierce test, the Ljung-Box test, the McLeod-Li test, the
turning points test, the Wallis and Moore phase-frequency
E. Localized Transformer monitoring (Real-time)
test, and the Jarque-Bera test, and the order of minimum
Data loggers are used to collect a variety of data, display corrected Akaike's information criteria, are used to detect
it, and record it for further analysis. The proposed approach transformer winding faults [20]. The 110kV three-phase
is put to the test on actual hardware utilising a lab-available double-winding transformer's external leads, inside the
CORTEX M4 processor. If any of the parameters under windings, and surrounding screens are all injected with high-
consideration are exceeded in respect to the established limit, frequency small signals to simulate the transformer's external
the calculated Ff alerts the monitoring and preventative discharge, the discharge inside and outside of the windings,
system. With the aid of this information, precautions are in order to investigate the differences between the internal
made in order to maintain the transformer operational for a and external partial discharges of the transformer received
long time. In addition to monitoring, an Adaptive Power signals at the detection end and develop an efficient
Differential Protection (APDP) strategy is recommended positioning method [21]. Dissolved gas analysis is a frequent
since it will protect the transformer against internal issues. method for identifying internal faults in transformers (DGA).
Online monitoring and adaptive power differential protection This approach uses the Doernenburg ratio and the Rogers
methods are both put into use simultaneously in the hardware ratio. Uncertainty in the gas analysis for identifying winding
processor to increase transformer reliability in the smart grid. problems is the main drawback of DGA [25].

Fig. 7. Generalised schematic diagram for transformer monitoring and

protection.
Fig. 8. The multi-terminal detection and positioning flow chart of
transformer.
Based on the computation of the average power of all
three phases on both sides of the transformer, the proposed III. RESULTS AND DISCUSSIONS
APDP scheme was developed. The input power should equal
A. Transformer monitoring based on IOT protocols and
the total of the output power plus any transformer losses. The
technologies
asset is deemed to have an internal defect if the difference in
power measured at the two ends is greater than the established After communication devices are successfully linked to
threshold, and the relay will then trip. If there is a problem the server, the web server is immediately given with the oil
tank vacancy level, oil deterioration level, and CTs current.
before the transformer is powered up, the output from the two
When their expected values were not met, a warning message
sides of the PT will not be equal, or Vp ≠Vs, and this could was sent to the web server along with the necessary remedial
lead to the fault continuing during the inrush situation [17]. action. Table 1 shows the different degrees of oil deterioration
F. Transformer Fault detection based on transformer and oil tank vacancies together with the relevant alarm
Windings message. Fig. 11 and 12 illustrate the front end of the website,
which will display all the data from the sub-station equipment.
To study the temperature state of the transformer in the
early phases of insulation breakdown, an electromagnetic-
thermo-fluid finite element technique (FEM) is used. The
transformer is modelled with two FEM software tools, one
for electromagnetic simulation and one for thermo-fluid
simulation, which interact and share data to give the needed
multi-physic modelling framework [10]. For improved
understanding and analysis of frequency response analysis
(FRA) results, time series analysis is integrated with Fig. 9. Webpage for sub-station monitoring and control.
frequency response analysis (FRA), and healthy baseline
residual data are modelled using time series analysis for
transformer winding defect diagnosis. To evaluate measured
data, various statistical applications, such as R 4.0.2 and

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 Fig. 10. Webpage during the runtime for oil monitoring.

TABLE I. ALARM MESSAGE SETTING Fig. 14. Dual load monitoring (Theft).
Oil deterioration The graph depicted below shows the amount of
Oil vacancy Message to server
level electricity used by the original user and the fraud one.
<75% <12cm Normal
>75% <12cm Oil reconditioning is necessary.
<75% >12cm Add more coil.
Oil adding and reconditioning is
>75% >12cm necessary.

The received data are still maintained in the cloud, making

it possible for anybody with access to use a login to view the
data and control the system whenever they want, from any
location. The command button used to control the
transformer's tap position and the servo motor's (CB state) Fig. 15. Power consumption by True and Fraudulent user.
from a website is shown in Figs. 13 and 14. The results of the
experiment support the assertion that when applied to a real Additionally, any increase in load results in the power
circuit, the innovative IoT-based sub-station regulating and supply to the load being shut off. The electric board is
monitoring system worked well. notified, as well as the people who own the home, via GSM,
and a buzzer is also installed to warn theft to have happened.
A button feed with the two states "Open" and "Close" can
It then generates a bill depending on the amount of power
be used to regulate the circuit breaker. The differential
protection is shown in Fig. 14. used. A Think Speak Software is used to retain the typical
load used, identify any increases in load caused by theft, and
identify the crime itself. The load in this project will be
continually monitored by the Arduino Uno depending on
parameters like voltage, current, and power. The Arduino
Uno keeps track of the load power once again.
Fig. 11. CB controller button on the webpage.
C. Transformer Condition Monitoring and fault Detection
Using Machine Learning and Deep Learning
Transformers' infrared imagery were gathered from a
Chinese electric firm. There are 359 unlabelled samples and
182 labelled samples, respectively. The latter number includes
141 hot issues and 41 faulty defects. Therefore, the proposed
Fig. 12. Differential protection on the webpage. GAN must be used to create the minority labelled samples.
The infrared image's temperature, texture, and form features
are first extracted as the characteristic features. Then, data is
generated for the labelled samples of the defective faults using
the GAN model. These defining features of the labelled and
unlabelled samples are then provided into the semi-graph
classification model's model inputs for training. A trained
Fig. 13. Slider for changing the tap position on the webpage. classification model is then used to test the suggested strategy
The transformer tap position can be controlled from a in the real world. The transformer's tested infrared picture is
slider button of the web page which is shown in Fig. 15. shown in Fig. 18.

B. Electricity Theft detection and monitoring in power grids

The inputs are current, voltage, power, and transmission
loss. These components will help increase the project's
accuracy. The difference from the prior measurement, which
often happens, may be compared and efficiently utilised to
identify any theft from the voltage provided to a dwelling
using the formula V=IR. As shown in Fig. 16, this method
produces the best results for identifying power theft as soon
as it happens.
Fig. 16. Infrared image from the test results.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 Additionally, the algorithm uses timely transmission of the
information to the field employees. The transformer was
damaged at the location indicated by the black box in the Fig
.18, based on an examination by the technicians. According to
the findings in Table II, efficient feature extraction can
considerably increase the model's capacity to classify
transformer defects. The transformer has an overheating issue,
according to the analysis of the algorithm described in this
work.

TABLE II. EFFECTS OF DIFFERENT FEATURE EXTRACTION METHODS (a)

Fault type Method ACC (%) REC (%) PRE (%)

Method 1 52.2 53.1 54.7
Method 2 62.4 64.8 63.6
Equipment
Method 3 68.5 70.2 71.1
defect
Method 4 74.5 72.6 76.8
Proposed 82.2 84.7 83.1
Method 1 57.5 60.1 58.8
Method 2 65.5 68.2 67.7 (b)
Overheating Method 3 72.3 77.2 74.1
Method 4 78.8 75.6 80.1
Proposed 86.2 84.8 83.5

This is because, despite the input vector's complexity, the

model's small sample size prevents it from being used to its
full potential for learning. As a result, categorization accuracy
suffers. Simple feature extraction can be used to somewhat
enhance the model's effect. This is so that the characteristics
of the input information can be accurately represented without
using a single type of biased feature quantity. (c)

D. Transformer Oil condition/oil level detection monitoring Fig. 17. The (a) dielectric dissipation factor, (b) relative permittivity and (c)
and monitoring using Transformer Oil resistivity of MO, PO, and RBO under different temperatures.
The relative permittivity, resistivity, and dielectric E. Localized Transformer monitoring(Real-time)
dissipation factor of MO, PO, and RBO are shown in Fig. 19
(a)–(c) and measurement data. The temperatures used to test The fitness function (Ff) as it is presented in Equation is
assessed for the deterioration of various parameter
each sample ranged from 30 to 90 degrees Celsius. As can be combinations. Fig. 20 displays the plot of the combination of
shown in Fig. 19(a)-, vegetable oils (PO and RBO) have a various parameter modifications v/s calculated fitness
higher dissipation factor (DDF) than mineral oil for all function (a). The change in temperature, efficiency, and
temperature ranges (c). At 90 C, RBO has a dissipation factor losses are also calculated and shown in Figs. 20(b), (c), and
that is 0.02168, seven times larger than mineral oil's (d), respectively, as functions of load variation. In order to
(0.00322). However, PO has a dissipation factor of 0.03846, account for variations in magnetic imbalance, winding
which is 12 times higher than that of mineral oil. According to temperature, harmonic content of current, and overloading
IEC 60247, all oil samples satisfied the requirement for circumstances. The relevance of fitness function is used to
transformer liquid insulation for DDF values, which is less score the transformer's overall health in this instance (Ff).
than 0.005 for MO and less than 0.05 for vegetable oil. Due to Here, we have utilised Ff's ranges to represent the economic
the somewhat more polar nature of their molecules compared situation: 1 Ff > 0.9 = Healthier, 0.9 Ff > 0.7 = Moderate, 0.7
to mineral oils, vegetable oils have a higher dissipation factor Ff > 0.5 = Poorer, and 0.5 Ff = Worse. When estimated Ff
value. In comparison to other vegetable oils, RBO has a lower advances into the region of bad condition, the suggested
DDF value than PO, which may be due to the molecular method raises an alert and, in the worst case, trips the
makeup of RBO having 20% less fatty saturated acids than transformer's circuit breakers [17].
PO's (50%) does. Higher fatty acid levels will result in more
polar oil.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 Utilizing two rheostats, various resistances are implemented.
It allows for the parallel connection of rheostats and a section
of the winding. The simulations and experimental results are
utilised to show how the recommended approach in this study
is a potent predictive strategy and may pinpoint the
insulation's susceptibility in the very first stages of a failure.
The temperature variations near to the winding are big enough
for the available temperature sensors to pick up on them. As a
consequence, the performance of the suggested approach
lacks the drawbacks of the current methods in the first phases
of fault [10].
Fig. 18. (a) Parameter variation v/s fitness function, (b)
Loading v/s efficiency, (c) Loading v/s temperature and (d)
Loading v/s losses.

F. Transformer Fault detection based on transformer

Windings
A 20 kVA transformer with access to the terminals of different
windings is taken into account in order to perform
experimental testing, demonstrate the accuracy of the
simulation findings, and demonstrate the decrease in
insulation resistance. Since the current in the winding is the
main factor causing heat and temperature fluctuations, if
current variations in tests and simulations follow the same
pattern, the same will be true for temperature. A network
analyzer used for data collection is the Nipower Unilyzer 902. Fig. 19. Comparison of simulated and experimental result

TABLE III. COMPARISON OF METHODOLOGIES BASED ON VARIOUS PARAMETERS.

Citation Methodology
Advantages Disadvantages Accuracy Error
Number (Dataset/Realtime)
Uses cutting edge communication tools to allow IOT devices to
Certain Sensors used are not novel and can provide with slightly
communicate and save data easily. Also, the monitoring system
1 Realtime less accurate readings. The oil level sensor uses an ultrasonic 75% 25%
includes activation, monitoring and warning systems working in
sensor which when covered with oil provides incorrect readings.
sync.
Most aspects of monitoring a transformer are addressed using A significant data delay exists with the node MCU when it sends
2 Realtime modern technology. The IOT devices used are appropriate for data through Wi-Fi. Also, better microcontrollers can be 80-85% 15-20%
the role. employed to increase throughput.
Better Data security and data transfer using Raspberry Pi.
Only Current sensors are employed for theft detection, which
3 Realtime Current sensors along with computational devices help in theft 70-75% 20%
cannot detect other theft methods employed.
detection. Also Cloud technology used provides data logging.
Reduces the significant power and economic losses caused by
The functionality of data storage by the meter for theft detection
customer power theft. Because the system is
4 Realtime is not feasible by the ordinary meter used here. Also email and 80% 20%
concealed, it is difficult for ordinary people to understand its
automated messages are used which are not feasible.
purpose.
Machine learning is being used with data from gas
It is feasible to depressurize gases that have gathered within the
chromatography, which calls for processing power for in-house
5 Dataset oil chamber by applying machine learning, which eliminates 85-90% 10-15%
detection. Additionally, performing gas
human error from the interpretation of gas chromatography data
chromatography within the transformer's main body is difficult.
Compared to conventional procedures, it offers better and more
The robustness and dependability of sensors and connected
6 Dataset precise findings. Less time is needed for fault detection, 80% 15%
networks are major problems.
protecting the system from catastrophic collapse.
Timesaving in finding faults is more when transformer is in
Initial cost of project is more includes components like solar
remote areas. The acquired information
7 Realtime panels, GSM module. Inefficient at night time solar doesn’t 85% 10-15%
can be used in the future and real-time data acquisition proves to
produce much energy required for power supply at all times.
be efficient
There's a significant reduction in power usage and a graphed
If theft is discovered, the whole power line is shut off, which
8 Realtime representation between the number of fraudulent and legitimate 80% 15-20%
inconveniences innocent/ other houses/buildings.
users helps in improving the system.
When considering the Hall effect, IR, weight sensor, and oil level
detectors, the GSM ideology is an additional benefit. When compared to other recent efforts, this study is unable to
9 Realtime With justification and, more significantly, considering how the demonstrate its correctness. Neither cutting-edge nor Wi-Fi 80% 15-20%
location of the sensors affects the ability to identify or Bluetooth-based technology is being employed.
abnormalities, so GPS module greatly helps in this regard
The pre-fault diagnosis and detection technique for winding and The suggested solution uses a powerful computing system
insulation failure utilising a temperature measure-based decision connected with the transformer to do thermofluid modelling of
10 Dataset 85-90% 10-15%
strategy is better than the post-fault diagnosis which many the temperature inside the bulk transformer for real-time
methods employ. problem identification which can incur more power.
This approach requires a more extensive data collection, which
This approach uses fewer physical calculations and predictions
takes time to create an ML model, making the creation of
and more machine learning methods to deliver more accurate
algorithms challenging.
11 Dataset information about the defects. 85-90% 10-15%
Additionally, developing the code for verifying and testing the
to reduce the frequency of operator checks, a sufficient dataset
obtained real-time data with training data is substantially more
of the transformer's most likely defects is being used.
difficult
Reveals information about the sensors that may be used to
monitor a number of different characteristics, including Doesn't include some modern sensors which can improve on
temperature, the make-up of the transformer's gases, partial performance of the monitoring system.
12 Dataset
discharges, voltage, current, moisture, and oil level. We can now Also, this paper surveys only sensors and not IOT devices which
analyse our work and utilise the sensors more efficiently thanks are compatible with certain sensors.
to this project.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 A few advantages of the FPI ultrasonic sensors include their high The construction process for the F-P sensor is more complicated
sensitivity to weak ultrasonic signals and their small size for and costly than other approaches, which is another drawback.
13 Dataset transformer installation. Insulated and easy to construct, Moreover, to increase the stability of the interferometric 85% 15%
interferometric ultrasonic sensors are produced by Michelson, ultrasonic sensors, the low frequency noise brought on by
Mach-Zehnder, and Sagnac. temperature change or random vibration should be eliminated.
The novel approach of identifying non-technical losses using While performing all of these tasks, the LCD display should be
Artificial intelligence and pattern separation methods used here, kept in mind. The fundamental flaw in this study is that no
14 Realtime 80-85% 15-20%
has shown how efficiently power theft can be identified by adequate communication channels have been used to convey or
observing customers and their power usage over a period of time. receive data.
As stated in the title and approach, this effort falls short of
demonstrating real-time analysis using digital metres. The
Here, the system is connected to an IoT server via the Blynk TCL
ESP8266 was not linked to any sensors and was just showing
protocol, which enhances the monitoring and communication
15 Realtime readings in all of the photographs. 75-80% 20-25%
experience. Since the utility used can remotely cut off power to
Further, when a current sensor malfunctions, the utility also loses
a fraudulent home, it eliminates physical visit to cut the fuse.
money and will not be able to check if the power line has been
tapped for theft if carried out.
The ESP8266 was not linked to any sensors and was just
When we can identify those who are abusing authority, the showing readings in all of the photographs. When a current
16 Realtime process of gathering data and manipulating it becomes quicker sensor malfunctions, the utility also loses money and will not be 75-78% 25%
and simpler. able to check if the power line has been tapped for theft if carried
out.
Estimate the transformer's health very accurately by utilising a
external problems may be unpredictable, the system is unable to
17 Realtime mathematical function to assess the transformer's condition and 75% 25%
precisely detect them.
protection. Improving the performance of grid tied transformer.
We can precisely measure a number of factors using infrared
The method of producing this dataset may be time-consuming,
pictures, which helps with defect identification, by compiling a
18 Dataset and it cannot always account for real-time modifications that 73% 27%
dataset from a number of photos, we are able to diagnose a
were not addressed during dataset collecting.
transformer issue with extreme accuracy.
The stability of nanoparticles is one of the key concerns when
Nanofluids may enhance the thermal characteristics of the base using them to enhance the fluids' thermal properties. there isn’t
19 Realtime 75% 25%
oil while also lowering the temperature of the oil. many experimental research on how oil-based nanofluids
transport heat.
Early abnormal event detection improves system performance, This system doesn't use any machine learning model to compute
20 Dataset 79% 21%
lowers costs, and delays the development of severe situations. so it requires more time to compute the equations
human assistance is required to make judgments, The
A linear time-invariant mechanism to detect discharges. lowers transformer's exterior discharge and internal non-winding
21 Simulation 70% 30%
the amount of time needed to compute while creating the mode. discharge contain distinct proportions of high frequency and low
frequency components hence this may be impossible to detect.
RBO has the greatest breakdown voltage for both test
This report made no mention of lowering the alternating current
configurations in the AC breakdown research. When it comes to
22 Realtime stress in the transformer. It also made no mention of alternatives 78% 22%
AC breakdown voltage, PO is the second-best oil sample, while
to oil.
MO has the lowest breakdown voltage.
This project works well in the laboratory and can withstand
Better communication approaches might be implemented and
23 Realtime currents of up to 30A. A bleeder bulb is connected to the line to 72% 18%
improved. Thence its correctness cannot be established.
test the relay.
Natural esters may replace mineral oil in distribution In this study, the heating impact brought on by windings has not
24 Realtime transformers since they degrade naturally. they allow us to been corrected. the life of the transformer was not extended by 76% 24
construct smaller transformers with the same rating. the use of natural esters.
Mega projects do not completely incorporate and use this.
25 Simulation The detection of inter winding faults which none were described Additionally, the majority of the parameters for the experiment 80% 20
were simulated.
This study sheds light on less pressing transformer-related
concerns such monitoring and ageing of insulation oil, as well as This research found no drawbacks to their approach when
26 Realtime 74% 26%
their thermal, electrical, chemical, and physical characteristics in applied to a transformer.
both.
This technique enhances the transformer's stability, accuracy,
Android phone's texting system is inefficient. Fewer metrics and
27 Realtime and efficiency while minimising the requirement for labour. It is 78% 22%
sensors were considered for assessing the transformer's health.
straightforward and fast to identify faults.
its resilience to electromagnetic interference, high sensitivity, Only current and temperature are considered; no additional
28 Realtime 82% 18%
superior insulation, and small optical sensor dimensions. characteristics are. To transmit messages, only GSM is utilised.
This technology is utilised to identify system problems before Node Red cannot be scaled since it is unsuitable for complex
29 Realtime they cause an unfavourable failure, extending the lifespan of the computations and many programming instructions, which 75% 25%
transformers relative to switches and relays. reduces performance in general.
The Arduino AT mega 328 lacks an integrated Wi-Fi
Construction and ease are both simple. The time and effort
30 Realtime microprocessor. More components are required, which would 80% 20%
needed for maintenance are reduced.
complicate system design.

IV. CONCLUSIONS AND FUTURE SCOPE Low-power, very dependable, and long-distance
communication are advantages of this technique. An EML
The new, more intelligent transformer increases the was built using deep belief networks (DBNs), stacked
lifespan of the old one and enables power plant managers to denoising autoencoders (SDAs) with 10 different activation
make precise diagnoses using real-time information from the functions, 20 RVMs, and a brand-new combination technique.
linked sensors. The recommended GSM-based system offers The IoT-based monitoring system's communication test
a way to protect distribution transformers from burglars and revealed that the distance is more than 100 metres. Last but
vandals by keeping an eye on the area around the transformer not least, the transformer fault experiment demonstrated the
and sending an SMS to alert the user of any irregularities. To value of the suggested EML's development. The suggested
provide the authorities an accurate detection report, this EML is easily adaptable to other signal-based classification
system used three distinct types of sensors. The system also programmes. Future study will look at different DL
offered the authority transformer health monitoring services approaches, activation functions, and classifiers to improve
via SMS-sending transformer health metrics including the feature extraction and classification performance of the
transformer temperature and oil level. The authorities may EML. It is recommended that the design and operation of an
stop a transformer from being destroyed by utilising the report IoT smart energy metre be adopted in order to account for the
on the above-mentioned health characteristics of the reading on the metre for the quantity of power the transformer
transformer. delivered. Putting this into practise there requires less labour
A transformer problem detection approach based on and can do away with the need for human involvement for
Ensemble Machine Learning (EML) has been presented to metre reading computations and disconnection if payment is
monitor the IoT-based system, collect the transformer not made since it has the ability to remotely disconnect the
vibration signals, and communicate them to the distant server. system.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 Implementation costs may be decreased by using smaller [11] Zhang, C., He, Y., Du, B., Yuan, L., Li, B., & Jiang, S, “Transformer
ML models with less training data and group monitoring of fault diagnosis method using IoT based monitoring system and
ensemble machine learning”, Future generation computer systems,
transformers by a single microcontroller. Free Wi-Fi access 2020, pp. 533-545.
might eventually save operational costs. Depending on how [12] Hamed Samimi, M., & Dadashi Ilkhechi, H, “Survey of different
useful the suggested system is, other features could be added. sensors employed for the power transformer monitoring”, IET Science,
The future will need a focus on load balancing, energy use, Measurement & Technology, 2021, pp. 1-8.
automated billing, and safe data encryption. A better level of [13] Ma, G., Wang, Y., Qin, W., Zhou, H., Yan, C., Jiang, J., & Ju, Y,
technological preparedness is needed, and sensors must be “Optical sensors for power transformer monitoring: A review”, High
properly assessed. This operation may benefit from oil Voltage, 2021, pp. 367-386.
monitoring, pressure detection, and power theft. To achieve [14] Shewane, P., Dongre, S., Meshram, S., Lokhande, P., & Naik, A,
“Power Theft Identification. Power”, 2022.
reduced latency and greater value accuracy, a better and more
effective microcontroller or single-board computer may be [15] Sebastian, P. K., & Deepa, K, “Internet of Things based Smart Energy
Meter with Fault Detection Feature and Theft Detection”, In 2022
employed. More investigation into finding and shielding the International Conference on Electronics and Renewable Systems
transformer from external faults is necessary for this adaptive (ICEARS), 2022, pp. 494-500.
power differential protection system technology. Data [16] SV, A. K., Kanakaraja, P., Prasad, D. V. S., Reddy, T. G., &
security should be a top focus in the future with the addition Pardhasaradhi, N, “IoT Enabled LoRa based Power Theft Detection
of oil monitoring and air pressure monitoring in the current System”, In 2022 International Conference on Electronics and
transformer since it offers a hazard if not maintained correctly Renewable Systems (ICEARS), 2022, pp. 608-616.
in terms of safeguarding the data. As a backup with [17] Chothani, N. G., Raichura, M. B., Patel, D. D., & Mistry, K. D, “Real-
time monitoring & protection of power transformer to enhance smart
encryption, a Raspberry Pi or a more effective board with fail- grid reliability”, In 2018 IEEE electrical power and energy conference
safe capabilities might be used. The employment of (EPEC), 2018, pp. 1-6
appropriate communication techniques is required for data [18] Fang, J., Yang, F., Tong, R., Yu, Q., & Dai, X, “Fault diagnosis of
transmission and reception. More sensors and the addition of electric transformers based on infrared image processing and semi-
a Wi-Fi module may enhance the task. It is necessary to supervised learning”, Global Energy Interconnection, 2021, pp. 596-
develop a better method for processing infrared pictures that 607.
takes into account both the dataset and real-time [19] Taheri, A. A., Abdali, A., Taghilou, M., Alhelou, H. H., & Mazlumi,
K, “Investigation of mineral oil-based nanofluids effect on oil
modifications. To boost viability, better learning algorithms temperature reduction and loading capacity increment of distribution
may also be developed. transformers”, Energy Reports, 2021, pp. 4325-4334.
[20] Abbasi, A. R., Mahmoudi, M. R., & Arefi, M. M, “Transformer
REFERENCES winding faults detection based on time series analysis”, IEEE
[1] Hossain, M. S., Rahman, M., Sarker, M. T., Haque, M. E., & Jahid, A, Transactions on Instrumentation and Measurement, 2021, pp. 1-10.
“A smart IoT based system for monitoring and controlling the sub- [21] Li, Y., & Tang, Z, “Multi-terminal Detection and Localization of
station equipment” Internet of things, 2019. Partial Discharge in Transformer”, In Journal of Physics: Conference
[2] Hasan, W. K., Alraddad, A., Ashour, A., Ran, Y., Alkelsh, M. A., & Series (Vol. 2290, No. 1, p. 012013), 2021.
Ajele, R. A, “Design and Implementation smart transformer based on [22] Hamid, M. H. A., Ishak, M. T., Suhaimi, N. S., Adnan, J., Hashim, N.
IoT”, In 2019 International Conference on Computing, Electronics & I., Ariffin, M., ... & Abd Rahman, R, “Electrical properties of palm oil
Communications Engineering (iCCECE), 2019, pp. 16-21. and rice bran oil under AC stress for transformer application”,
[3] Darteh, Oscar & Liu, Xiaodong & Liu, Qi & Anaadumba, Raphael & Alexandria Engineering Journal, 2022, pp. 9095-9105.
Adjei, Charity & Sarker, Sajib, “Internet of Things Based Electricity [23] Mandake, Y. B., Bankar, D. S., Yadav, S., Vishal, R., & Singh, S,
Theft Detection Using Raspberry PI”, 2020. “Power Theft Detection Using Arduino”, 2022.
[4] Mucheli, N. K., Nanda, U., Nayak, D., Rout, P. K., Swain, S. K., Das, [24] Cilliyuz, Y., Bicen, Y., Aras, F., & Aydugan, G, “Measurements and
S. K., & Biswal, S. M, “Smart power theft detection system”, In 2019 performance evaluations of natural ester and mineral oil-immersed
Devices for Integrated Circuit (DevIC), 2019, pp. 302-305. identical transformers”, International Journal of Electrical Power &
[5] Saravanan, D., Hasan, A., Singh, A., Mansoor, H., & Shaw, R. N, Energy Systems, 2022.
“Fault prediction of transformer using machine learning and DGA”, [25] Ehsanifar, A., Allahbakhshi, M., Tajdinian, M., Dehghani, M.,
In 2020 IEEE International Conference on Computing, Power and Montazeri, Z., Malik, O. P., & Guerrero, J, “Transformer inter-turn
Communication Technologies (GUCON), 2020, pp. 1-5. winding fault detection based on no-load active power loss and
[6] Hazarika, K., Katiyar, G., & Islam, N, “IOT Based Transformer Health reactive power”, International Journal of Electrical Power & Energy
Monitoring System: A Survey”. In 2021 International Conference on Systems, 2021.
Advance Computing and Innovative Technologies in Engineering [26] Rao, U. M., Fofana, I., Bétié, A., Senoussaoui, M. L., Brahami, M., &
(ICACITE), 2021, pp. 1065-1067. Briosso, E. “Condition monitoring of in-service oil-filled transformers:
[7] Jeevanandham, A., Nitin, B., Maheshkumar, M., & Monish, B, “Solar Case studies and experience”, IEEE Electrical Insulation Magazine,
Powered Real Time Transformer Health Monitoring System Using 2019, pp. 33-42.
Internet of Things (IoT)”, In Journal of Physics: Conference Series, [27] Suja, K., & Yuvaraj, T, “Transformer Health Monitoring System Using
2022. Android Device”, In 2021 7th International Conference on Electrical
[8] Kumaran, K, “Power Theft Detection and Alert System using IOT”. Energy Systems (ICEES), 2021, pp. 460-462.
Turkish Journal of Computer and Mathematics Education [28] Ojo, T. P., Akinwumi, A. O., Ehiagwina, F. O., Ambali, J. M., &
(TURCOMAT), 2021, pp. 1135-1139. Olatinwo, I, “Design and Implementation of a GSM-based Monitoring
[9] Fahim, A., Chowdhury, M. A., Alam, M. F., Elahi, F., Shourov, E. C., System for a Distribution Transformer”, European Journal of
& Hasan, M, “Smart transformer theft protection and maintenance Engineering and Technology Research, 2022, pp. 22-28.
monitoring system”, In 2021 2nd International Conference on Robotics, [29] Baruah, D. J., Dey, S., Neog, P., & Saharia, K. T, “Design and Develop
Electrical and Signal Processing Techniques (ICREST), 2021, pp. 607- Transformer Monitoring System using NODE-MCU and Real Time
612. IOT Platform Connected using NODE-RED”, Int. J. Res. Appl. Sci.
[10] Soleimani, M., Faiz, J., Nasab, P. S., & Moallem, M, “Temperature Eng. Technol, 2022, pp. 562-568.
measuring-based decision-making prognostic approach in electric [30] Ghanmare, R., Ujawane, R., Shewane, P., & Meshram, P, “Iot Base
power transformers winding failures”, IEEE Transactions on Distribution Transformer Monitoring System Using Arduino”, IJRPR.
Instrumentation and Measurement, 2021, pp. 6995-7003.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:55:28 UTC from IEEE Xplore. Restrictions apply.
 Transformer 4.0 – A Smart Transformer for a
Smarter Living
Sanjay S Tippannavar Yashwanth S D
Department of ECE Department of ECE
Sri Jayachamarajendra College of Engineering, Mysuru, India Sri Jayachamarajendra College of Engineering, Mysuru, India
[email protected] [email protected]
2023 3rd International Conference on Innovative Sustainable Computational Technologies (CISCT) | 979-8-3503-0336-0/23/$31.00 ©2023 IEEE | DOI: 10.1109/CISCT57197.2023.10351417

Abstract— Traditional transformers are not equipped with economic growth is closely related to the production and use
sophisticated monitoring and control features, which limits their of its energy. A nation's capacity to produce energy serves as
ability to operate and maintain efficiently. The suggested a critical indication of its economic development and is
system, which can be adapted to the current Distribution acknowledged as an important criterion by international
transformers in India, has been developed for 2% of the
estimated cost of the newer Smart Transformers on the market,
organisations. An online monitoring system is essential for
which cost a staggering 10 lakhs. This project focuses on turning ensuring their maintenance, monitoring them, and protecting
ordinary transformers into smart devices by integrating the them from outside disruptions. It allows for remote equipment
Raspberry Pi 4B, the Arduino Mega 2560, and Heltec ESP32, status tracking and real-time maintenance and control
which includes Wifi, LoRa, and Bluetooth in addition to an programmes.
OLED panel that provides additional functionality to the task.
The three boards are primarily separated into three distinct Transformers play an important part in power grid design
subsystems, with the Raspberry Pi serving as the primary today, and recent trends reveal that grid transformer efficiency
computer and the other two separate boards / subsystems
and longevity are dropping. To address these significant
serving as backup devices in case of unanticipated events.
several sensors, including the DHT 11, GPS, Accelerometer for issues, we need a system that complements the transformer
detecting tremors and accidents, the BMP280 for altitude and and is smart enough to use IOT technologies to boost both the
pressure sensing within the oil chamber, the HCSR04 for efficiency and longevity of grid transformers. The power
monitoring the oil level, and the Camera Module for transformer is a critical piece of hardware for the grid's
transmitting data over the internet. The database is managed by efficient and dependable energy transmission to users. Asset
an active DBMS that has been combined with an RFID scanner management and protection are the greatest strategies for
and a biometric fingerprint sensor for security reasons. Wi-Fi prolonging transformer lifetime and increasing grid reliability.
and LoRa protocols have been used to allow for distant However, diagnosing their flaws remains difficult. Due to the
communication while keeping the system operational round-
many benefits given by smart grid and technology, it is
the-clock. On the Raspberry Pi, a clever method of transmitting
Push notifications to all devices, including Smart phones, PCs, important to adjust the requirements of protective schemes in
tablets, and different Smart devices, has been implemented in order to incorporate a smart transformer in the distribution
order to save costs and eliminate GSM maintenance. For offline network and to keep up with the latest development in power
data logging, a USB device has been programmed to capture the systems. Today's procedures are linked to electrical
data on a Raspberry Pi that serves as a "Blackbox," and an SD characteristics, but such measurements don't give any
card using SPI communication interfaced with an Arduino information on the distribution transformer's internal health.
MEGA is very effective to understand the working statistics of This motivation paved the way for the Transformer 4.0 to be
the Smart Transformer. Additionally, by using a GPS module, used in the revolution of Industry 4.0.
it aids in the identification of power theft and transmits the
information to local authorities. Last but not least, this job
makes use of cutting-edge boards for highly effective, power- The main objectives of the proposed work are to;
efficient, accurate, and dependable working. Cloud services
from ThingSpeak and Adafruit have been enabled with • Keep track of and control important factors that affect
certificates that have been encrypted and decoded, as well as the the transformer's lifespan and effectiveness (Tank
use of the 8883 TCP port for safe Client communication on the pressure, Temperature, Oil viscosity).
MQTT protocol. With better maintenance procedures and less • Display a variety of errors and anomalies, enable data
downtime, this makes it possible to proactively identify probable
logging with plug-and-play support.
problems and abnormalities.
• Geotag the transformer using GPS - to find the
Keywords—Smart Transformer, Signal processing, Raspberry transformer easily.
Pi, microprocessors, communication, Cloud, IoT, MQTT, • Aggregating, analysing & managing transformer
Electricity, Lora, Wi-Fi, Bluetooth, OLED, DBMS, security. health data collected through sensor-based
technology.
I. INTRODUCTION
• Detect failures with early warnings and prevent
Power production and consumption have a significant critical events, thereby enabling faster response with
impact on a nation's economic growth. In fact, one of the predictive maintenance based on analysis and operate.
metrics used to gauge a nation's economic progress is its • Utilise State-of-Art computational devices to achieve
capacity to produce power. Due to energy theft, the majority higher efficiency, accuracy and reliable functioning.
of emerging nations have had unfavourable economic
repercussions in an effort to fulfil the need for power for The remaining sections are as follows: In Section II, the
industry and real estate. In the modern world, a nation's background research for the study was discussed. Section III

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:54:40 UTC from IEEE Xplore. Restrictions apply.
 provides a comprehensive and algorithmic explanation of the
proposed mythology. In Section IV, a comparative analysis is
used to analyse the results of the suggested method. Section
V, which also covers prospective lines of investigation, wraps
off the research effort.
II. RECENT WORKS
The technique provides a remote operator with sufficient
information to evaluate the kind and volume of oil that is
presently present in the transformer and oil circuit breaker. In
the case of any concerning conditions, the operator will get an
alert message outlining the required remedial action. Without
physically being at the sub-station, the operator may utilise
this technology to control circuit breakers, swap out relay
plugs, and modify transformer taps [1].
Fig. 1. Block diagram of Transformer 4.0 model
Long-distance monitoring, low power consumption, and
very reliable use are all advantages of the IoT-based
monitoring system, which is good for long-term signal
collecting and transmission. Long-distance monitoring, low
power consumption, and very reliable use are all advantages
of the IoT-based monitoring system, which is good for long-
term signal collecting and transmission. Since the methods
can automatically detect transformer problems using
computers and software, they are rapidly developing since
they do not need operators with substantial training and
experience [2].
Fig. 2. Summarized view of the proposed system depicted in different
A microprocessor, among other sensors, is coupled to the domains.
proposed system's temperature, current, vibration, oil level,
and humidity sensors. RTUs, or remote terminal units, are The flowchart of Transformer 4.0 is show in Fig. 3. The
used to transmit the input. The microcontroller transmits the blue line represents the genuine signal, the black line the flow,
signal to the website and LCD for display after input from and the red line the erroneous consequence. The moment the
several sensors. Following that, the website presents all of the system is switched on, it begins by looking for mistakes. If
physical inputs as numerical representations. A website at the any are discovered, it will notify the relevant workers and
monitoring node provides access to the whole system. An enter the information into the black box. As the processes
Arduino UNO and a Wi-Fi modem are further options for continue to operate, it checks for sensory information as well
monitoring the status [3]. as data that is available from it to see whether it is in the right
The system is proposed measures variables such as power, format or not. Signal processing and data collection identify
current, and voltage on the basis of energy is calculated and any irregularities, and as soon as they do, an alarm is sent to
payments are generated to the relevant mail. If the load rises, the appropriate authority through Push notifications and the
the electric board receives a notification concerning power cloud, instructing them to shut down the device, assign a
usage, which is then shown on the LCD. In the event that the different transformer to do the same job, and minimize power
load rises, the power supply to the load is also cut off. In loss in the currently operational environment. Upon recording
addition, if there is an increase in load, the electric board and any irregularities, a smart system determines if they are caused
the home's owners are alerted by GSM, and a buzzer is also by internal errors or external activities, depending on which
fitted to warn of theft. Depending on how much electricity is case additional action is taken. If any anomaly if found
consumed, it will then generate a charge [8]. externally, a camera records it, checks for structural problems,
This system of power differential protection (APDP) has and alerts the appropriate authorities. The same holds true for
been used. Monitoring is done while taking into consideration sensor testing, plug-in gadgets, and other failures.
a number of variables, including temperature, the effect of In the current context, security is crucial. Security of the
ageing, the deterioration of insulation, magnetic imbalance, devices is crucial in all facets of systems. Figure 4 shows the
etc. After real-time data capture, the processor is used to flow diagram outlining how the security enhancements
estimate the actual parameters [10]. implemented in Transformer 4.0 function. It utilises an
III. METHODOLOGY internal and exterior lock, an RFID and biometric fingerprint
sensor, and an active database management system (DBMS)
The block diagram of the proposed system is shown in Fig. built using Python SQL on a Raspberry Pi 4. When a person
1 with three different sub-systems having three boards
enters, low access level users use an RFID card to access the
primarily Raspberry Pi as the master device with Arduino
MEGA 2560 and Heltec ESP 32 for connecting to Cloud and exterior casing, whereas high access level users may use their
utilisation of on-board LoRa protocol. The summarized view biometric fingerprints to access both the inside and external
of Transformer 4.0 is depicted in Fig. 2. cases.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:54:40 UTC from IEEE Xplore. Restrictions apply.
 Fig. 5. RPi subsystem working in detail.

In order to detect various parameters, including

temperature, humidity, pressure, RFID entry, biometric entry,
oil level, and automatic temperature control, which entails
Fig. 3. Flow of the events occurring in Transformer 4.0. turning on the cooling fan at the appropriate time and logging
the sensory data for analysis, the primary sub-system used a
Raspberry Pi. The logic used in this process is explained in
Fig. 5.

Fig. 6. Circuit diagram of the Transformer 4.0 with all three sub-systems.

The detailed circuit diagram of the proposed Smart

Transformer is depicted in Fig. 6. It displays the usage of
various data communication protocols such as SPI, UART,
I2C, Wireless – Wi-Fi, LoRa and serial communication.
IV. RESULTS AND DISCUSSIONS
The outcomes of Transformer 4.0's hardware
implementation are covered in this section. Fig. 7 and Fig. 8
displays the principal sub-system's Hardware Integrated
Model. It has a number of sensors, including an accelerometer,
Fig. 4. Flow Diagram for secure access to Transformer 4.0 hardware.
an ultrasonic sensor for gauging oil level, a DHT-11 sensor for
humidity and temperature, an RFID sensor, and a fingerprint
sensor for unlocking the device's interior and exterior casing.
Additionally, a pen drive linked to a USB 3.0 connection for

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:54:40 UTC from IEEE Xplore. Restrictions apply.
 datalogging serves as a Blackbox for all the collected data,
storing it for offline use.

Fig. 7. Top-view of Raspberry Pi (primary) sub-system of Transformer 4.0.

Fig. 10. Security algorithm integrated with DBMS on primary sub-system.

The cloud interface of the primary sub-system is shown in

Fig. 11.

Fig. 8. Side-view of Primary sub-system (depicts RFID & Biometric system).

The snippet of the data from sensors being collected is

visible in Fig. 9. It shows various parameters and triggered
values for the system to actuate. Fig. 10 shows the data being
uploaded to Cloud on ThingSpeak platform along with
numerical value and graphical chart in the public dashboard.

Fig. 11. Cloud interface for the primary sub-system of Transformer 4.0.

Fig. 12 displays the information from the USB device that

was used to enable the Blackbox capability. It is customary to
utilise a csv or txt file, however in practise, such files must
first be sorted before being used. Data is immediately saved in
an.xlsx file for ease of use and to avoid duplicate labour.

Fig. 9. Sensed data being displayed on the Raspberry Pi Shell.

Fig. 10 shows the DBMS being used for RFID and

fingerprint enrollment in the primary sub-system.

Fig. 12. Data stored in the USB device of the primary sub-system.

The hardware integration of the secondary sub-systems

utilizing the Arduino MEGA 2560 and HELTEC esp 32 is
depicted in Fig. 13.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:54:40 UTC from IEEE Xplore. Restrictions apply.
 Fig. 17. Output at the Heltec secondary subsystem with Wi-Fi and using
Fig. 13. Hardware Model of the secondary sub-system of Transformer 4.0. LoRa for data transmission.

The SD card module uses the SPI Communication

protocol to store the data collected from the Arduino sub-
system. The data is shown in Fig. 14 while being read and
visualized by the Arduino Mega or Serial Monitor at COM11.
The data that is kept on the SD card is shown in Fig. 15.
Secondary subsystems are primarily used to lighten the burden
on major systems and to speed up processing times.
Fig. 18. Certificate used with MQTT protocol for secure transmission using
port 8883 on Heltec secondary sub-system.

Port 8883 is used instead of the insecure 1883 port for

MQTT via TCP in order to safeguard the data sent over the
MQTT protocol to the cloud. The certificate created for
encryption along with port 8883 is shown in Fig. 18. Fig. 19
Fig. 14. Data acquisition on the secondary sub-system (Arduino Mega). shows the cloud UI for the Heltec sub-system.

Fig. 19. Cloud interface created as a backup on Heltec secondary sub-system.

Fig. 15. Stored data on the SD card interfaced to the Secondary subsystem.

If any anomalies are found, then an alert to the base-station

is sent via the secondary subsystem using GSM. Fig. 16
depicts the message received via GSM on a smart device.
Fig. 20. Received packet on the LoRa receiver.

Fig. 16. Alert sent to an authority on finding an anomaly.

As shown in Fig. 17, data is sent to the cloud via both the
Wi-Fi and LoRa protocols. In the absence of Wi-Fi, the Heltec
subsystem immediately switches to the LoRa protocol and
transmits the data in packets to the infrastructure.

Fig. 21. Proposed external casing using Acrylic and metal for the
Transformer 4.0 model.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:54:40 UTC from IEEE Xplore. Restrictions apply.
 Dr. U. B. Mahadevaswamy, the professor and head of the
Electronics and Communication Engineering department of
SJCE, JSS STU, Mysuru, has our sincere gratitude.
Throughout the project, Prof. Yashwanth S D, Assistant
Professor, Department of Electronics and Communication
Engineering, SJCE, JSS STU, Mysuru, has been our project
guide. We would like to take this opportunity to thank him for
his kind consent, constant encouragement, timely valuable
guidance, immense help, and cooperation. A special thanks to
our colleagues Sathvik H R, Rishitha R Gowda, and Ajay M
for their prompt efforts that enabled us to complete this
innovative work on schedule.
Fig. 22. Live-streaming feature on RPi as observed from the Transformer.

In order to showcase the components in a transparent

enclosure, the hardware for the suggested system is shown in REFERENCES
Fig. 21 built from metal and acrylic sheets. Using SolidWorks [1] Chen, Z., Chen, D., Zhang, X., Yuan, Z., & Cheng, X. (2021). Learning
CAD software, it clearly displays every piece of hardware that graph structures with transformer for multivariate time-series anomaly
was utilised and fixed. detection in IoT. IEEE Internet of Things Journal, 9(12), 9179-9189
[2] Ravindran, V., Ponraj, R., Krishnakumar, C., Ragunathan, S.,
V. CONCLUSIONS AND FUTURE SCOPE Ramkumar, V., & Swaminathan, K. (2021, November). IoT-Based
Smart Transformer Monitoring System with Raspberry Pi. In 2021
In order to avoid malfunctions, a number of codes have Innovations in Power and Advanced Computing Technologies (i-
been devised to alert the appropriate authorities if incoming PACT) (pp. 1-7). IEEE.
readings from the sensors installed on the transformer exceed [3] Sharma, R. R. (2021). Design of distribution transformer health
specified thresholds in accordance with the standard operating management system using IoT sensors. Journal of Soft Computing
Paradigm, 3(3), 192-204.
values. To a central monitoring station or any user connected
to that specific transformer, this real-time monitoring system [4] Pawar, R. R., & Deosarkar, S. B. (2017, July). Health condition
monitoring system for distribution transformer using Internet of Things
detects and sends data on all operating parameters. The (IoT). In 2017 international conference on computing methodologies
technology optimises the use of transformers, spots possible and communication (ICCMC) (pp. 117-122). IEEE.
problems, lowers costs, and guarantees that the same operate [5] Kamlaesan, B., Kumar, K. A., & David, S. A. (2017, August). Analysis
consistently by continually monitoring critical operating data. of transformer faults using IoT. In 2017 IEEE International Conference
After doing a thorough prior art search and carefully on Smart Technologies and Management for Computing,
examining the work that has previously been accomplished Communication, Controls, Energy and Materials (ICSTM) (pp. 239-
241). IEEE.
and published in various international publications and
[6] Srivastava, D., & Tripathi, M. M. (2018, October). Transformer health
conferences, we came to the conclusion that the monitoring system using internet of things. In 2018 2nd IEEE
aforementioned capabilities in the system are necessary. The International Conference on Power Electronics, Intelligent Control and
breakthrough is in the secured Blackbox feature's ability to Energy Systems (ICPEICES) (pp. 903-908). IEEE.
effectively monitor 10–12 parameters while combining [7] Sowbaranika, S. A., Madhumithra, U., Monisha, G., Yalhini, K. M.,
sensory input from three different MCUs to fulfil the task. Babu, W. R., & Vaiyapuriappan, R. (2022, March). Comprehensive
Additionally, Arduino is used as a fail-safe computing system Review on IoT Based Distribution Transformer Health Tracking
System. In 2022 8th International Conference on Advanced Computing
that activates in the event that the main computing system and Communication Systems (ICACCS) (Vol. 1, pp. 2017-2020).
malfunctions. Upgrading the present, out-of-date transformers IEEE.
with retrofitting of the Transformer 4.0 avoids the need to buy [8] YAMAN, O., & BİÇEN, Y. (2019). An Internet of Things (IoT) based
brand-new transformers for the same purpose. The main goals monitoring system for oil-immersed transformers. Balkan Journal of
of the proposed effort are to increase productivity and allow Electrical and Computer Engineering, 7(3), 226-234.
customers to live Smarter Lives. [9] Khandait, A. P., Kadaskar, S., & Thakare, G. (2017). Real time
monitoring of transformer using IOT. International Journal of
Electrical substations, industrial settings, and household Engineering Research & Technology (IJERT), 6(03).
distribution transformers with single- and three-phase lines all [10] Gajenthiran, G., Meyyappan, C., Vishnuprakash, J., Arjun, R., Viswak
make use of it. These techniques are used to transformers to Sena, T., Sethu, Y., & Sharan Prasanna, R. (2022). IoT-based smart
provide the user with information about the transformer and monitoring of online transformer. In Inventive Communication and
Computational Technologies: Proceedings of ICICCT 2021 (pp. 951-
any problems when he is unable to visit the spot. The user may 961). Springer Singapore.
monitor the sensor grid and controller and they can be fitted [11] Hamed Samimi, M., & Dadashi Ilkhechi, H. (2020). Survey of different
to an existing transformer system. sensors employed for the power transformer monitoring. IET Science,
Measurement & Technology, 14(1), 1-8.
ACKNOWLEDGMENT [12] Ma, G., Wang, Y., Qin, W., Zhou, H., Yan, C., Jiang, J., & Ju, Y.
Any task's delight and ecstasy upon its completion would (2021). Optical sensors for power transformer monitoring: A review.
be incomplete were it not acknowledged without High Voltage, 6(3), 367-386.
acknowledging the individuals who made it possible. Above [13] Sanjay S Tippannavar, Vijay Mishra, Yashwanth S D, Rishitha R
Gowda, Sathvik H R and Ajay M, "Smart Transformer - An Analysis
all, we should give our institution, Sri Jayachamarajendra of Recent Technologies for Monitoring Transformer," 2023
College of Engineering (SJCE), JSS Science and Technology International Conference on Recent Trends in Electronics and
University, Mysuru, the respect it deserves because it gave us Communication (ICRTEC), IEEE.
a fantastic platform and the chance to complete this project.

Authorized licensed use limited to: Sri Jayachamarajendra College of Engineering. Downloaded on January 16,2024 at 11:54:40 UTC from IEEE Xplore. Restrictions apply.