E3S Web of Conferences 336, 00044 (2022) https://doi.org/10.

1051/e3sconf/202233600044
ICEGC'2021

MPPT Based Artificial Neural Network Versus Pertub &

Observe For Photovoltaic Energy Conversion System
Karima Et-torabi1* and Abdelouahed Mesbahi1
1
LESE, Hassan II university, Casablanca, Morocco

Abstract. In this paper, we present a study of maximisation power tracking for photovoltaic energy
conversion chain. We give a comparison between two techniques of tracking maximum power; Perturb
&Observe (P&O) as first method versus a second method using artificial neural network (ANN). The two
methods are designed, modelled and simulated in MATLAB/Simulink environment. The simulation results
are discussed involving performance and constraints of these two algorithms.

1 Introduction
2 Modelling of the PV array and boost
In general, photovoltaic energy conversion converter
System (PECS) are used to produce electric power
for utilities along sunny day. To maintain The proposed system is shown in Fig.1 which
supplying load under variables irradiation the consists of a PV array, DC-DC boost converter,
current and voltage are adjusted to maximum load and MPPT controller [7].
power available for climatic conditions, MPPT
techniques are used for extracting the maximum
PV L
power from the PV arrays at different ARRAY iL
VPV ipv
environmental conditions such as temperature and ic
solar irradiance. Many MPPT algorithms have Cin ipv Cout

Load
MPPT
been used in the literature [1-5]. MPPT techniques Algorithm D
VL

are clustered into two major groups: the VPV

conventional and intelligent techniques. The

conventional MPPT algorithms include Hill-
Climbing (HC) [1], Perturb and observe (P&O)
[1,2], and incremental conductance (INC) [2].
These techniques show obvious shortcomings,
Fig.1. Block diagram of the studied system
such as low tracking efficiency during quickly
changing solar irradiation and fluctuations about
the point of maximum power. The intelligent 2.1 PV ARRAY
techniques, including fuzzy logic controller (FLC)
[3,4], artificial neural network (ANN) [5], particle The equivalent model used to represent parasitic the
swarm optimization (PSO) [1], genetic algorithm pv cell consists of a current source, one diode, and
(GA) [6]. These algorithms present many resistors which is illustrated in Fig.2 [7,8].The relation
advantages, such as low oscillations around the between the terminal current (Ipv) and voltage (Vpv) of a
point of maximum power and a speedy tracking PV array is expressed by following equation [8]:
response to changing conditions. The DC-DC
boost converter is designed to be placed between a = ℎ − −1 −
PV array and a load, it is used to transfer the
maximum power to load by changing the duty (1)
cycle of the DC-DC boost converter.

! "#
=
In this study, P&O and artificial neural network
With:
(ANN) techniques are simulated using with $
Simulink /Matlab. Thus, ANN is designed and
compared with P&O technique Where, Iph is the photovoltaic current, Is is the reverse
saturation current of the diode, Vt is the thermal voltage,

* Corresponding author: [email protected].

© The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(http://creativecommons.org/licenses/by/4.0/).
E3S Web of Conferences 336, 00044 (2022) https://doi.org/10.1051/e3sconf/202233600044
ICEGC'2021

n is the diode ideality factor, k is the boltzman constant

(1.38e-23J/K), q is the electron charge (1.602e-19C), T
is cell temperature in Kelvin, Nsc number of series cells
per module, Ns and Np are the number of series and
parallel modules, Rp and Rs are the shunt and series
resistors of the cell, respectively.

Fig.2. Equivalent circuit of the one diode model

The Canadian Solar CS6P-240P PV array is chosen
for our modeling and simulation. We used
MATLAB/Simulink, for the simulation. The I-V and P-
V characteristics of modeled PV array with variation of
temperature and irradiance shown in Fig.3 and Fig.4.

Fig.4. The I–V and P-V characteristics for PV array with

various temperature at 1000w/m²

2.2 Boost converter

The DC-DC converter acts as an interface between

the PV array and the load to boost voltage at PV
terminal. The bloc Boost is depicted in fig1 and it
consists mainly of inductor, transistor switch, diode and
filter capacitor [4].
The voltage conversion ratio of the boost converter can
be expressed as in (2) [9]:
&
%
=
&'(
(2)

Where, Vpv is input voltage (PV voltage), VL is output

voltage, and D is the duty cycle of the switch PWM
signal.
The inductor ) and capacitor *+, , are designed to limit
respectively current and voltage ripple. These main
elements are designed according to boost performances.
Table 1, summarizes the calculated values of inductor )
Fig.3. The I–V and P-V characteristics for PV array with

and capacitor *+, , as in (3), (4) [10]:

varying irradiation at 25°C

(
*+, = ∆0%
.. .
(3)
0%

2
E3S Web of Conferences 336, 00044 (2022) https://doi.org/10.1051/e3sconf/202233600044
ICEGC'2021

The P&O technique continuously increments

.(
)= or decrements the PV array terminal voltage with
.. .1
(4)
a small amount until reaching to the maximum
power point and compares the trend of the output
Where, δ is the percentage of the inductor current power with that of the perturbation. Consequently,
ripple, (∆VL/VL) is the percentage of output voltage The perturbation must continue in the same
ripple. direction if the output power increases and
reverses if it decreases. the algorithm continues in
Table 1. Parameters of the boost converter the same way.
Parameters Symbols Values
Input filter capacitor (µF) Cin 390 3.2 MPPT based Artificial Neural Network
Output filter capacitor (µF) Cout 470 In this work, the proposed (ANN) aims to predict the
Boost inductor (mH) L 10 maximum power point (MPP) of PV array.The four steps
for designing ANN can be summarized by:
Switchng frequency (KHz) fs 24
• Selection of ANN architecture
Resistive load (Ω) R 40
• Data collection
• Training the ANN
• Testing the ANN
3 MPPT algorithms We have used a Multi-Layer Perceptron (MLP)
network, contains three layers as shown in Fig.6.[12,13]:
To track the set point at optimal available
power, MPPT algorithms adjust voltage and • The first layer: where the inputs to the network are the
current to deliver its maximum power regarding PV current (Ipv) and voltage (Vpv) of PV system recorded
climatic circumstances. The following describes under different temperature and irradiance
two algorihtms Pertub and Observe . •The second layer: is the hidden layer, contains 10 hidden
neurons which receive data from the input layer and send
them to the output layer. Its activation functions are the
3.1 Perturb and Observe
sigmoid functions (Tansig)
The Perturb and observe (P&O) is the most • The third layer : is the output layer, contains one neuron
popular used MPPT, because of it simple duty cycle (D) whose activation function is linear
structure and easy implementation. The flowchart function (Purelin).
of P&O MPPT algorithm is shown in Fig. 5 [11].

Fig.6. The proposed neural network architecture

The training phase uses the data set obtained from the
simulation of the PV array in Matlab / Simulink using
the P&O method, the ANN inputs variables are the
voltage Vpv and current Ipv of PV array corresponding to
a given solar radiation and ambient temperature
temperature conditions, while the duty cycle D is chosen
as output. The nntool Toolbox can be used to design,
train, validate and test a neural network in Matlab, the
data set are divided into three parts 70% for training,
15% for validation, and 15% for testing. The neural
network is trained with Levenberg-Marquardt
backpropagation algorithm, which is a very fast and
accurate technique for minimizing the mean square error
(MSE) as shown in Fig.7. This algorithm provides better
results than others. As shown in Fig. 8, the regression
Fig.5. Flowchart of P&O MPPT algorithm

3
E3S Web of Conferences 336, 00044 (2022) https://doi.org/10.1051/e3sconf/202233600044
ICEGC'2021

analysis is performed to measure the correlation Table 2. Electrical parameters of the Canadian Solar CS6P-
between target and output. 240P

Parameters Symbols Values

Maximum power (w) Pmpp 240.097
Maximum power point current
Impp 8.03
(A)
Maximum power point voltage
Vmpp 29.9
(V)
Short circuit current (A) Isc 8.59
Open circuit voltage (V) Voc 37
Temperature coefficient of Isc
Ksc 0.0637
(A/K)
Temperature coefficient of Voc
Koc -0.3654
(V/K)
Series cells Nsc 60
Number of parallel modules Np 1
Number of series modules Ns 2

Fig.7. Training neural network

Fig.9. Slope variation of irradiance

As shown in Fig. 9, the irradiation level was suddenly

increased from 800 W/m2 to 1000 W/m2 then decreased
from 1000W/m² to 200W/m². The temperature is
constant at 25°C.
The simulation results are presented in Fig. 10 which
shows the generated power of the PV arrays with ANN
and P&O MPPT techniques. when decreasing or
increasing the irradiance level, the ANN rapidly tracks
the MPP with an efficiency of 97.64% compared to the
P&O.

Fig.8. Regression

4 RESULTS AND DISCUSSION

In this section, we present simulation results on the
MATLAB/ Simulink environment of the system
proposed (PV arrays, DC–DC converter, MPPT
controller and load). The electrical parameters of the PV
array are given in Table 2. Fig.10. output power from both method under fast-
changing solar irradiation

4
E3S Web of Conferences 336, 00044 (2022) https://doi.org/10.1051/e3sconf/202233600044
ICEGC'2021

International Conference of Computer Science and

Renewable Energies (ICCSRE), (2019, July).
https://doi.org/10.1109/ICCSRE.2019.8807457
4. A. Abo-Sennah, M. A. El-Dabah, A. El-Biomey
Mansour, "Maximum power point tracking
techniques for photovoltaic systems: a comparative
study", International Journal of Electrical and
Fig.11. Output power Overshoot with both methods in Computer Engineering (IJECE), vol. 11, p. 2088-
the case of sudden increase and decrease of irradiation 8708 (2021)
http://doi.org/10.11591/ijece.v11i1.pp57-73
The overshoot in case of sudden increase or decrease 5. F. Tsai, C. S Tseng, K. T Hung., S. H Lin, "A Novel
in irradiation is more important with P&O MPPT DSP-Based MPPT Control Design for Photovoltaic
controller compared to overshoot using the proposed Systems Using Neural Network Compensator †"
artificial neural network MPPT controller, as shown in Energies, vol.14, p. 11 (2021)
Fig.11. https://doi.org/10.3390/en14113260
In this test, ANN technique can quickly track the
MPP under fast-changing solar irradiation, while P&O 6. K. H. Chao, M. N. Rizal, "A Hybrid MPPT
fails to track MPP when irradiation changes fast.in Controller Based on the Genetic Algorithm and Ant
addition, ANN has small oscillation about the maximum Colony Optimization for Photovoltaic Systems
power point, while P&O technique has high oscillation under Partially Shaded Conditions" Energies,
around a maximum power point MPP. vol.14, p.10 (2021).
https://doi.org/10.3390/en14102902
7. S. Messalti, A. Harrag, A Loukriz, "A new variable
5 Conclusion step size neural networks MPPT controller:
Review, simulation and hardware implementation”,
This paper presents MPPT based ANN to obtain
Renewable and Sustainable Energy Reviews,
maximum power for PECS at any atmospheric
vol.68, p.221–233 (2017)
conditions and comparing the results with conventional
P&O method. to find which algorithm gives better https://doi.org/10.1016/j.rser.2016.09.131
performance. Inputs ANN are electrical parameters 8. K. Et-torabi, I. Nassar-eddine, A. Obbadi, Y.
obtained from PV array under changing atmospheric Errami, R. Rmaily, S. Sahnoun, A. El fajri, M.
conditions and output is duty ratio for boost converter. Agunaou, "Parameters estimation of the single and
DC-DC converter is used to track MPP of PV array. double diode photovoltaic models using a Gauss–
These techniques are simulated using with Simulink Seidel algorithm and analytical method: A
/Matlab. The simulation results it is concluded that ANN comparative study", Energy Conversion and
based MPPT gives better performance in terms of Management, vol.148, p. 1041-1054 (2017)
dynamic and fast response, less steady state oscillations https://doi.org/10.1016/j.enconman.2017.06.064
under fast-changing solar irradiation. On the contrary,
9. A. Hassoune, M. Khafallah, A. Mesbahi, D.
P&O technique fails to track MPP when irradiation
Breuil, "Control strategy for a boost converter
changes fast.
operating as a battery charger of photovoltaic-grid
system for electric vehicles charging station", in
References Proceedings of the International Conference on
Automation, Control Engineering and Computer
1. S. Motahhir, A. El Hammoumi, A. El Ghzizal, "The Science (ACECS17), Tangier,2017
most used MPPT algorithms: Review and the http://dx.doi.org/10.13140/RG.2.2.29850.47047
suitable low-cost embedded board for each
algorithm" Journal of cleaner production, vol. 246, 10. M. M. Shebani, T. Iqbal, J. E. Quaicoe, "Comparing
p. 118983 (2020) bisection numerical algorithm with fractional short
circuit current and open circuit voltage methods for
https://doi.org/10.1016/j.jclepro.2019.118983 MPPT photovoltaic systems", in Proceedings of the
2. P. Pandiyan, S. Saravanan, N. Prabhakaran, R. International Conference on Electrical Power and
Tiwari, T. Chinnadurai, N. R. Babu, E.Hossain, Energy Conference (EPEC), (2016, October)
"Implementation of Different MPPT Techniques in https://doi.org/10.1109/EPEC.2016.7771689
Solar PV Tree under Partial Shading Condition",
Sustainability, vol.713. p.13 (2021). 11. M.Mudassar, M. R. Fazal, M. U. Asghar, M. Bilal,
https://doi.org/10.3390/su13137208 R. Asghar, "Implementation of improved Perturb &
Observe MPPT technique with confined search
3. I.Nassar-Eddine, A. Obbadi, K. Et-Torabi, H. space for standalone photovoltaic system", Journal
Mokhliss, A. Elamiri, R. Rmaily, Y. Errami, A. El of King Saud University-Engineering Sciences,
Fajri, S. Sahnoun, M. Agunaou,"A new Fuzzy vol.32, p. 7 (2020)
Logic architecture to control the DC-bus voltage in
grid connected photovoltaic system", in https://doi.org/10.1016/j.jksues.2018.04.006
Proceedings of the International Conference on 12. V. Viswambaran, A. Bati, E. Zhou, "Review of AI
based maximum power point tracking techniques &

5
E3S Web of Conferences 336, 00044 (2022) https://doi.org/10.1051/e3sconf/202233600044
ICEGC'2021

performance evaluation of artificial neural network

based MPPT controller for photovoltaic systems",
International Journal of Advanced Science and
Technology·, vol. 29, p. 8159-8171(2020)
13. S. Gowid, A. Massoud, "A robust experimental-
based artificial neural network approach for
photovoltaic maximum power point identification
considering electrical, thermal and meteorological
impact", Alexandria Engineering Journal, vol. 59,
p.3699-3707 (2020)
https://doi.org/10.1016/j.aej.2020.06.024