Adaptive maximum power point tracking using neural networks for a photovoltaic systems according grid

Power Stations, Grids and Systems
ISSN 2074-272X. Electrical Engineering & Electromechanics, 2021, no. 5 57
© H. Sahraoui, H. Mellah, S. Drid, L. Chrifi-Alaoui
UDC 621.3 https://doi.org/10.20998/2074-272X.2021.5.08
H. Sahraoui, H. Mellah, S. Drid, L. Chrifi-Alaoui
ADAPTIVE MAXIMUM POWER POINT TRACKING USING NEURAL NETWORKS
FOR A PHOTOVOLTAIC SYSTEMS ACCORDING GRID
Introduction. This article deals with the optimization of the energy conversion of a grid-connected photovoltaic system. The novelty
is to develop an intelligent maximum power point tracking technique using artificial neural network algorithms. Purpose. Intelligent
maximum power point tracking technique is developed in order to improve the photovoltaic system performances under the
variations of the temperature and irradiation. Methods. This work is to calculate and follow the maximum power point for a
photovoltaic system operating according to the artificial intelligence mechanism is and the latter is used an adaptive modified
perturbation and observation maximum power point tracking algorithm based on function sign to generate an specify duty cycle
applied to DC-DC converter, where we use the feed forward artificial neural network type trained by Levenberg-Marquardt
backpropagation. Results. The photovoltaic system that we chose to simulate and apply this intelligent technique on it is a stand-
alone photovoltaic system. According to the results obtained from simulation of the photovoltaic system using adaptive modified
perturbation and observation – artificial neural network the efficiency and the quality of the production of energy from photovoltaic
is increased. Practical value. The proposed algorithm is validated by a dSPACE DS1104 for different operating conditions. All
practice results confirm the effectiveness of our proposed algorithm. References 37, table 1, figures 27.
Key words: artificial neural network, grid-connected, adaptive modified perturbation and observation, artificial neural
network-maximum power point tracking.
Вступ. У статті йдеться про оптимізацію перетворення енергії фотоелектричної системи, підключеної до мережі.
Новизна полягає у розробці методики інтелектуального відстеження точок максимальної потужності з використанням
алгоритмів штучної нейронної мережі. Мета. Методика інтелектуального відстеження точок максимальної потужності
розроблена з метою поліпшення характеристик фотоелектричної системи в умовах зміни температури та опромінення.
Методи. Робота полягає в обчисленні та відстеженні точки максимальної потужності для фотоелектричної системи,
що працює відповідно до механізму штучного інтелекту, і в останній використовується адаптивний модифікований
алгоритм збурення та відстеження точок максимальної потужності на основі знаку функції для створення заданого
робочого циклу стосовно DC-DC перетворювача, де ми використовуємо штучну нейронну мережу типу «прямої подачі»,
навчену зворотному розповсюдженню Левенберга-Марквардта. Результати. Фотоелектрична система, яку ми обрали для
моделювання та застосування цієї інтелектуальної методики, є автономною фотоелектричною системою. Відповідно до
результатів, отриманих при моделюванні фотоелектричної системи з використанням адаптивних модифікованих збурень
та спостереження – штучної нейронної мережі, ефективність та якість виробництва енергії з фотоелектричної енергії
підвищується. Практична цінність. Запропонований алгоритм перевірено dSPACE DS1104 для різних умов роботи. Усі
практичні результати підтверджують ефективність запропонованого нами алгоритму. Бібл. 37, табл. 1, рис. 27.
Ключові слова: штучна нейронна мережа, підключена до мережі, адаптивне модифіковане збурення та спостереження,
штучна нейронна мережа-відстеження точки максимальної потужності.
Introduction. Nowadays, the electric power
generation mainly uses fossil and fissile (nuclear) fuels.
The widespread use of fossil fuels, such as gasoline, coal
or natural gas, allows for low production prices. On the
other hand, their use results in a large release of
greenhouse gases and polluting gases. Electricity
production from fossil fuels has a great responsibility for
global CO2 emissions, hence pollution, according to the
last International Energy Agency report [1]. Nuclear
power, which does not directly release carbon dioxide, the
risks of accident linked to their exploitation are very low
but the consequences of an accident would be disastrous.
Although the risks of accident linked to their exploitation
are very low, but the consequences of an accident would
be disastrous and we must not forget the Fukushima
Daiichi nuclear disaster in Japan. Furthermore, the
treatment of waste from this mode of production is very
expensive; the radioactivity of the treated products
remains high for many years [2], that’s what prompted to
the propose a nuclear plant waste management policies
and strategies [2], and some researcher suggests to build a
regional and global nuclear security system [3]. Finally,
uranium reserves are like those of limited oil [4].
Although the world is in surplus in electricity
production today, the future is therefore not promising on
fossil fuel resources whose reserves are constantly
decreasing and whose prices fluctuate enormously
depending on the economic situation [5]. The future
preparations in the fields of energy production to satisfy
the humanity needs should be foreseen today, in order to
be able to gradually face the inevitable energy changes.
Each innovation and each breakthrough in research
will only have repercussions in about ten years at best, the
time to carry out the necessary tests and to consider
putting into production without risk for the user as much
for his own health than for its electrical installations, to
avoid the problems of pollution in the production of
electricity, alternative solutions can be photovoltaic (PV),
wind, or even hydroelectric sources [4, 5].
The use of PV solar energy seems to be a necessity
for the future. Indeed, solar radiation constitutes the most
abundant energy resource on earth. The amount of energy
released by the Sun, for one hour could be enough to
cover global energy needs for a year, for that we should
better exploit this energy and optimize its collection by
PV collectors [6].
The basic element of a PV system is the solar panel
which is made up of photosensitive cells connected to
each other. Each cell converts the rays from the Sun into
continuous type electricity. PV panels have a specific
highly non-linear electrical characteristic which appears
clearly in the current-voltage and power-voltage curves
[7]. Its electrical characteristics have a particular point
Electronic copy available at: https://ssrn.com/abstract=3945562

58 ISSN 2074-272X. Electrical Engineering & Electromechanics, 2021, no. 5
called Maximum Power Point (MPP). This point is the
optimal operating point for which the panel operates at its
maximum power, MPP is highly dependent on climatic
conditions and load, which makes the position of the MPP
variable over time and therefore difficult to locate [8].
A Maximum Power Point Tracking (MPPT) control
is associated with an intermediate adaptation stage,
allowing the PV to operate at the MPP so as to
continuously produce the maximum power of PV,
whatever the weather conditions (temperature and
irradiation), and whatever the charge. The converter
control places the system at MPP this point defined by
current Impp and voltage Vmpp. There are several MPPT
techniques that aim to extract maximum power from the
solar cells outputs [9-12] the interested reader is referred
to [11] for more details. Classic techniques such as the
Incremental Conductance technique and Perturbation and
Observation (P&O) technique, these two methods are the
most used and easy to implement methods but have
drawbacks [10-13].
New techniques based on artificial intelligence, such
as Fuzzy Logic Control [14, 15] Squirrel Search
Algorithm [16], Particle Swarm Optimization [17], Levy
Flight Optimization [12], Artificial Neural Networks
(ANN) [18, 19], and other propose a hybrid techniques
[20-23].
The methods based on ANN allow solving non-
linear problems and more complicated by a very fast way
since they are represented by non-linear mathematical
functions [24]. A different ANN-MPPT algorithm for
maximization of power PV production have been studied
in many research papers [18, 19, 25]. Messalti et al. in
[19] proposes with experimental validation two versions
of ANN-MPPT controllers either with fixed or variable
step. The aim of their works was to propose an optimal
MPPT controller based on neural network for used it in
the PV system. Different operating climatic conditions are
investigated in the ANN training step in order to improve,
tracking accuracy, response time and reduce a chattering.
Kumar et al. in [26] propose two Neural Networks
(NN) in the purpose of PV grid-connected with multi-
objective and distributed system; one is for assuring
MPPT and the other for the generation of reference
currents; the NN used for MPPT is based on hill climbing
learning algorithm, and use a NN version of a Power
Normalized Kernel Least Mean Fourth algorithm control
(PNKLMF-NN) to generate a reference currents.
Tavakoliel al. in [27] propose an intelligent method
for MPPT control in PV systems, this study establishes a
two-level adaptive control framework to increase its
efficiency by facilitating system control and efficiently
handling uncertainties and perturbations in PV systems
and the environment; where the ripple correlation control
is the first level of control and the second level is based
on an adaptive controller rule for the Model Reference
Adaptive Control system and is derived through the use of
a self-constructed Lyapunov neural network. However,
this approach did not been applied in the purpose of grid
connected PV system.
In [28] the authors have made a new technique – a
Adaptive Modified Perturbation and Observation
(AMPO), which reduces the MPP search steps this last
based on the function which widely used in sliding mode
control sign function, by this technique of reducing the
calculation time and the chattering; compared to classic
P&O technique.
Many authors study the issues of PV system grid
connection [26], [29-31]. Slama et al. in [29] offer a
clever algorithm for determining the best hours to switch
between battery and PVs, on the other hand, Belbachir et
al. in [30] seeks the optimal integration, both for
distributed PVs and for the batteries, other deal with the
management of electricity consumption [31].
In this paper, we propose an adaptive P&O
algorithm technique based on neural network with the PV
system to increase the power of PV and operate at MPP,
whatever the climatic variation such as the radiation and
the temperature, according a grid demand.
The main contribution of this work is to present an
Adaptive Modified Perturbation and Observation –
Artificial Neural Networks (AMPO-ANN) based on the
sign function which simplifies and reduces the step and
time of calculating MPPT point which allows minimizing
both the calculation time, the structure of the classic P&O
algorithm and the AMPO-ANN designing. Furthermore,
in the purpose of real-time application the major problem
in a neural network framework is the large size of the
created program who add a difficult to implant, especially
for the realization of a complex system with small time
constant.
We are mainly interested in the development of a
control system based on ANN which allows the
continuation of the MPP by simulation and experiments,
which allows increasing the performance of the neural
MPPT chart compared to the other maximization
methods. Figure 1 below presents a proposed system of
control.
β
T
ipv
Vpv
iout
u
Vout
Converter
DC-DC
Vin
PV
ANN
.
Inverter
DC/AC
Three –phase
Grid
AC Load
LC Filter
Fig. 1. Schematic diagram of the PV system under study
with AMPO-ANN strategy
The goal of the paper is to develop a technique of
maximization power point tracking search based on the
function sign which simplifies and reduces the step size
and the computation time of the maximum power point
tracking point which minimize both calculation time.
Subject of investigations. This paper is valid power
maximization technique-based neurons networks by a test
bench with a dSPACE DS1104.
Description and modeling of proposed PV system.
Equation (1) describes the PV cell model, this model
(Fig. 2) can be definite by the application of standard data
given by the manufacturer. The equivalent circuit for PV
cell is presented as follow [32].
The typical equation for a single-diode of PV panel
is as follows:

sh
pv
s
pv
aV
I
R
V
s
ph
pv
R
I
R
V
e
I
I
I T
pv
s
pv
















1 , (1)
where Ipv is the current generated by PV panel; Iph is the
generated photo-current; Is is the current of saturation;
Vpv is the voltage of PV panel; Rs is the array’s equivalent
series resistance; a is the constant of the ideal diode VD;
VT is the thermal voltage of PV (VT = KT/q, where K is the
Boltzmann’s constant; T is the temperature of PV; q is the
charge of an electron); Rsh is the array’s equivalent
parallel resistance.
Iph
Rs
Rsh
VPV
VD
Is
Ish
I0
Fig. 2. The PV cell equivalent circuit
Figures 3 and 4 illustrate the PV panel’s
characteristics as they change; both temperatures between
25 to 75 C and irradiation between 200 to 1000 W/m2
respectively.
Left MPP
Right MPP
0 5 10 15 20 25
Vpv, V
0
10
20
30
40
50
60
70
Ppv, W
Fig. 3. Characteristics P = f(V) of the PV panel under variation
of temperature
Ppv, W
0
10
20
30
40
50
60
70
80
90
0 5 10 15 20 25
Vpv, V
Fig. 4. Characteristics P = f(V) of the PV panel under irradiation
variation
Buck converter modeling. The load is connected to
the DC bus via a DC-DC buck power converter [33]
(Fig. 5), which allows it to be controlled.
Vin VD
Q L iL
IC
C
IR
R Vout
Fig. 5. Buck converter
To model the converter, state space average
equations are employed, as shown in the equation (2) [34]







,
;
2
3
1
2
2
2
1
1
1
x
k
x
k
x
x
k
uV
k
x in


(2)
where u is duty cycle and k1 = 1/L; k2 = 1/C; k3 = 1/RC.
The steady state is given by
[x1 x2] = [iL Vout]. (3)
 The DC/AC inverter model. Figure 6 presents the
structure of three-phase voltage source inverter (VSI).
Fig. 6. Structure of a three-phase VSI
The switching function is Cii = A, B, C as
bellow[35]:
 if Ci = 1, then Ki is OFF and Ki is ON;
 if Ci = 0, then Ki is ON and Ki is OFF.
The outputs voltage of the inverter UAB, UBC, UCA
can be write as:











.
;
;
Ao
Co
CA
Co
Bo
BC
Bo
Ao
AB
U
U
U
U
U
U
U
U
U
(4)
Since the phase voltages are star-connected to load
sum to zero, equation (4) can be written:
 
 
 















.
3
1
;
3
1
;
3
1
BC
CA
Cn
AB
BC
Bn
CA
AB
An
U
U
U
U
U
U
U
U
U
(5)
For the phase-to-neutral voltages of a star-connected
load obtain this model:











.
;
;
Co
no
Cn
Bo
no
Bn
Ao
no
An
U
U
U
U
U
U
U
U
U
(6)
and we conclude that:
 
Co
Bo
Ao
no U
U
U
U 


3
1
. (7)

For ideal switching can be obtained:
Uio = CiUc – Uc /2, (8)
with
 
 
 











.
5
,
0
;
5
,
0
;
5
,
0
C
C
Co
C
B
Bo
C
A
Ao
U
C
U
U
C
U
U
C
U
(9)
Substitution of (6) into (7) obtain [30]:




















.
3
2
3
1
3
1
;
3
1
3
2
3
1
;
3
1
3
1
3
2
Co
Bo
Ao
Cn
Co
Bo
Ao
Bn
Co
Bo
Ao
An
U
U
U
U
U
U
U
U
U
U
U
U
(10)
Setting (9) with (10), obtain:
.
2
1
1
1
2
1
1
1
2
3
1






































C
B
A
C
Cn
Bn
An
C
C
C
U
U
U
U
(11)
The Conventional P&O Algorithm (CPOA). In
the P&O process the voltage is increased or decreased
with a defined step size in the direction of reaching the
MPP. The method is carried out again and again till the
MPP is attained. In steady condition the operational point
oscillates about the MPP, the oscillation is highly
dependent on step size, so that when using a small step
size, it can reduce volatility but can reduce system
dynamics as well. On the other hand, while using a large
step size it can improve system dynamics, but it can
increase volatility around MPP as well [36]. Figure 7
illustrates the flowchart of the CPOA algorithm.
AMPO-ANN algorithm. Many MPPT approaches
have recently been created and developed. In terms of
accuracy, for real time implementation the P&O MPPT
method is more practical than other MPPTs because it is
easier to implement [38]. The P&O MPPT technique is
primarily based on the perturbation of the PV output
voltage V(t) and related output power P(t), which is
compared to the prior perturbation P(t +1). Keep the next
voltage shift in the same direction as the previous one if
the power increases.
Measurement VPV (K+1) IPV (K+1)
Initialisation values VPV (K) IPV (K), PPV(K)
Measurement PGPV (K+1)
PPV(K+1) > PPV(K)
PPV(K)-PPV(K+1)>0
VPV(K+1)> VPV(K)
u=u-Δu
u=u-Δu u=u+Δu
u=u+Δu
VPV(K)=VPV(K+1)
IPV(K)=IPV(K+1)
Yes
Yes
Yes
No
No
No Yes
Return
VPV(K+1)> VPV(K)
Conventional steps
Fig.7. Flowchart of the CPOA algorithm
Artificial intelligence is used in many areas of
research, and ANN is a bright and promising part of these
technologies, where process control and monitoring,
recognition of patterns, power electronics, finance and
economics, and medical diagnosis are only a few of the
applications where ANNs have proven their worth [37].
In this paper, we will use two neural networks at the
same time; the first network whose role is to estimate the
output current which corresponds to the maximum power,
and the second is used to estimate the voltage which
corresponds to the maximum power too[28].
However, if the steps of the algorithm are tracking
speed has been increased, as has the accuracy. and
rapidity are increased (dPpv/dVpv>0), but with high
increasing in the oscillation, resulting in comparatively
low performance and vice versa, In this paper, an AMPO
algorithm method is dedicated to find a simple
implantation in comparison with classical CPOA
algorithm, and the AMPO can be written as follows:
u() = Uc( – 1) + sign(P), (12)
where  is fixed step and Uc is the voltage control;
P = P() – P( – 1),
if P > 0 then increase Uc, else P < 0 decrease Uc.
A power of the panel (Ppv) sensor is connected to the
P&O algorithm unit in order to detect the power in state 
and compare it with next value (+1). At a certain point,
when the difference between Ppv() and Ppv( +1) is ΔPpv
then the algorithm will recognize that there is a powerful
change and the algorithm should start from the beginning
(). The value of u() (u is voltage control of P&O) is set
to depend on the value of  of the ΔPpv criteria and it is
different from irradiation values of PV, the ΔPpv/ΔVpv
change value around at point MPP, the duty cycle follows
this change, view the duty cycle varying between values
positive, zeros, negative.
In this article, we replace this variation of power
u( +1) show in equation (12) by function sign(Ppv) play
the role of conventional step of algorithm P&O, with
rapidly responses, equation (13) can be written in the
following form:
 












;
0
1
;
0
0
;
0
1
sign
pv
pv
pv
pv
P
if
P
if
P
if
P (13)
The adding the variation of power (ΔPpv) and
voltage (ΔVpv) can be whiten equation (13) as follow:
 




















.
0
1
;
MPP
at
0
0
;
0
1
sign
pv
pv
pv
pv
pv
pv
pv
pv
V
P
if
V
P
if
V
P
if
V
P (14)
Equation (14) can be written as follow:
 = sign{(Ppv() – Ppv(+1))(Vpv() – Vpv(+1))}. (15)
To simplify the writing of equation (15) can be
written in the following form:
 = sign(ΔPpvΔVpv). (16)
State of the voltage control  of AMPO can be
summarized in Table 1.

Table 1
Variation of MPP in algorithm
sign(ΔPpv()) sign(ΔPpv(+1))  u duty cycle State of MPOA
–1 –1 –2 +1 Left MPP
–1 +1 0 0 at MPP
+1 –1 0 0 at MPP
+1 +1 +2 +1 Right MPP
The Table 1 presented the variation of MPP point in
algorithm by 4 cases:
 Case 1. If state of changing algorithm is
sign(ΔPpv()) = –1, then sign(ΔPpv(+1)) = –1, and  = –2,
the MPP moving to left; there for u = +1, to increase
power of PV (Ppv).
sign(ΔPpv()) = +1, then sign(ΔPpv(+1)) = –1, and  = 0,
at point MPP; there for u = 0, no changing in the power of
PV (Ppv).
sign(ΔPpv()) = –1, then sign(ΔPpv(+1)) = +1, and  = 0,
at point MPP; there for u = 0, no changing in the power of
PV (Ppv).
sign(ΔPpv()) = +1, then sign(ΔPpv(+1)) = +1, and  = +2,
the MPP moving to right; there for u = +1, to decrease
power of PV (Ppv).
After this case the variation of δ and u can be
thought in AMPO-ANN for desired voltage regulation
(for regulate the desire voltage), as shown in Fig. 8.
The value of δ presented the variation of power of
panel ΔPpv we can add to value of duty cycle u for adjust
at point MPP can be written as follow:
u() = u() + u( + 1). (16)
After equations (15) – (17) can be designing the
flow chart of the MPOA algorithm modified shown in
Fig. 8 and presented a new step has determined by
previous equation.
Measurement
VPV (+1), IPV(+1), PPV (+1)
Initialisation values
VPV(), IPV(), PPV()
u()= u() + δu(+1)
VPV() = VPV(+1)
IPV() = IPV(+1) Return
δ=sign(ΔVPV (+1)ΔPPV(+1))
ANN
Fig.8. Flowchart of the adaptive ANN-AMPO
Simulation of proposed system. The simulation of
the Intelligent Maximum Power Point Tracking (IMPPT)
based on AMPO-ANN makes it possible to verify that
neural networks approach, after learning is effectively
capable of predicting the desired output for the values of
the data at the input which are not used during learning.
We should always compare the true exit from the
trajectory of neural networks with the trajectory of the
model of PV cells.
The simulations results given in Fig. 9 and represent
the electrical characteristics of the stand-alone PV system
controlled by AMPO-ANN under standard climatic
conditions (1000 W/m² and 25 °C). The powers obtained
from the proposed technique stabilize in a steady state
around the optimal values delivered by PV (Pmpp = 111 W,
Vmpp = 26 V and Impp = 4.4 A); AMPO controller allows
us for parts per million (PPM) to be attained in 0.06 s,
whilst the ANN algorithm allows for PPM to be obtained
in 0.02 s only. In addition Fig. 9 shows that in steady state
the maximum power supplied by the PV system
controlled by the AMPO-ANN is more stable and closer
to the PPM compared to AMPO control; the AMPO
control give a power oscillates around the MPP which
resulting in power losses.
From Fig. 9,a we observe that Ppv takes 0.02 s in
transient state to stabilize at a steady – state value which
is MPP in the neighborhood of 111 W.
Figure 9,b summarizes a comparison between the
MPPT of PV output power controlled by AMPO and
AMPO-ANN. We see that the power generalized based
on AMPO algorithm has more pikes and is more oscillate
compared to the improved technique based on ANN.
0 0.2 0.4 0.6 0.8 1
0
50
100
AMPO
AMPO-ANN
t, s
Ppv, W
a
0.43 0.44 0.45 0.46
109.5
110
110.5
111
111.5
AMPO
AMPO-ANN
t, s
Ppv, W
b
Fig. 9. a – power of PV (AMPO-ANN);
b – zoom power of PV (AMPO-ANN)
From Fig. 10 we observe the during the period from
0 s to 0.05 s the voltage decreases with significant
oscillations, then it stabilizes at the maximum value 26 V.
Figure 11 shows the load current curve based on
AMPO-ANN techniques, we note that its value in steady
state stabilizes around 4.4 A.

0 0.2 0.4 0.6 0.8 1
0
10
20
30
t, s
Vpv, V
Fig. 10. Voltage of PV (AMPO-ANN)
0 0.2 0.4 0.6 0.8 1
0
1
2
3
4
5
t, s
Ipv, A
Fig. 11. Current of PV (AMPO-ANN)
In order to verify the robustness and the reliability of
the proposed method, we will test the performance of
AMPO-ANN by performing separately under climate
condition variation, we make variations on solar
irradiation and we assume that the temperature is a
constant equal to 25 °C, where we suppose that the
irradiation drops from 500 to 1000 W/m2
, at 0.5 s.
According Fig. 12 we note that the maximum power
delivered by the PV varies proportionally with irradiation.
When the irradiation is 500 W the Ppv stabilizes around 38 W.
But when the sun goes from 500 to 1000 W/m² the Ppv
rises to 111 W.
In addition, the simulation result presented by Fig. 9,
shows that the AMPO-ANN represent better
performances compared to AMPO; since they converge
quickly towards the new Pmpp with reduced the chattering.
Figures 13, 14 show current and voltage of PV based on
AMPO-ANN techniques respectively.
0 0.2 0.4 0.6 0.8 1
0
20
40
60
80
100
120
t, s
Ppv, W
Fig. 12. Power of PV (AMPO-ANN)
0 0.2 0.4 0.6 0.8 1
0
1
2
3
4
5
t, s
Ipv, A
0 0.2 0.4 0.6 0.8 1
0
10
20
30
t, s
Vpv, V
We will test the performances of the AMPO-ANN
algorithm previously developed with the purpose of the
grid connection and the climatic conditions are fixed in
standard conditions, then connect the PV system to the
electrical networks. Figure 15 shows the simulation
results. We observe better results for Imes (current
measured by the network), Ich (load current) also the
three-phase currents Ia, Ib, Ic (Fig. 16).
0 0.2 0.4 0.6 0.8 1
-2
-1
0
1
2
I
a
I
b
I
c
t, s
I, A
a
0.2 0.22 0.24 0.26 0.28 0.3
-2
-1
0
1
2
t, s
I, A Ia Ib Ic
b
Fig. 15. a – three-phase currents (AMPO-ANN);
b – zoom three-phase currents (AMPO-ANN).
0 0.2 0.4 0.6 0.8 1
-2
-1
0
1
2
Ich
Imes
t, s
I, A Ich Imes
a
0.2 0.21 0.22 0.23 0.24
-2
-1
0
1
2
Ich
Imes
t, s
I, A Ich Imes
b
Fig. 16. a – current Ich and Imes (AMPO-ANN);
b – zoom current Ich and Imes (AMPO-ANN)

The results confirm the correct functioning of the
two controllers AMPO and AMPO-ANN, but also show a
better functioning of the AMPO-ANN. The latter has
proven to have better performance, fast response time and
very low, steady state error, and it is robust to variations
in atmospheric conditions.
Experimental results. The proposed AMPO-ANN
controller has been put to the test in order to improve its
performance. Instead of a solar panel, an experimental
setup of a system made of a PV emulator coupled to a
DC-DC converter is shown in Fig. 17. LA-25NP and
LV-25P are sensors of the current Ipv and voltage Vpv. The
proposed control is implemented on the dSPACE
DS1104.

DS1104 dSPACE
Master : Power PC 604e
Slave : DSPTMS320F240
Interface
Load
PV Array Emulator
Voltage & current sensors
(LEM) AC-DC
&
DC-DC
Converter
Voltage & current sensors
(LEM)
Software Model of
PV-Array
Irradiation Temperature
Emulator control
PWM
DAC
Fig. 17. Structure of the laboratory setup
In the simulation part we assume that all
components are perfect (simplifying assumptions, losses
and switching phenomena are ignored), so the DC-DC &
AC-DC converters has an almost perfect operation.
On the other hand, the tests which we carried out in
the laboratory take into account the saturation of the used
components and the switching phenomena, these tests
consist to validating the proposed technique which
applied to a DC-DC converter then connected to an
AC-DC converter (inverter).
From Fig. 18-20 we observe that Ppv takes 0 s in
transient state to stabilize at a steady – state value which
is MPP in the neighborhood of 100 W, current and
voltage also taken point MPP at values 3.4 A and 29 V.
t, s
Ppv, W
50
100
0
10
0 20 30
t, s
Ipv, A
5
0
10
0 20 30
t, s
Vpv, V
10
0
10
0 20 30
20
30
We will test the performances of the AMPO-ANN
algorithm previously developed with the purpose of the
grid connection and the climatic conditions are fixed in
standard conditions, then connect the PV system to the
electrical networks. Figures 21, 22 show the experimental
results. We observe better results for I (current measured
by the network), VAC (load voltage), also single phase of
current and voltage.
t, s
VAC, V
0
10
0
20
40
–40
–20
20 30
Fig. 21. Voltage of AC bus (AMPO-ANN)
t, s
I, A
10
0
0.1
0
–0.1
–0.2
20 30
Fig. 22. Current of AC bus (AMPO-ANN)
The Figures 23-27 show the results of realizing the
output power of the PV, its operating voltage and current,
and the duty cycle (at the frequency of 3000 Hz) for the
AMPO-ANN and the conventional disturbance and
observation (P&O) AMPO-ANN using a converter and
inverter environmental conditions. It is clearly seen how
the AMPO-ANN algorithm reduces the response time of
the PV system. Obviously, the system with AMPO has a
great loss of energy in the transient state, that when the
increase in power is the result of the increase in
illumination in sinusoidal form between 500 W/m2
and
1000 W/m2
, the reversal of the direction of illumination
produced by the AMPO-ANN algorithm causes the
increase of the power at the MPP point to 101 W and at
the same time the output voltage of the inverter, the
output voltage VAC is 28 V and current at 0.15 A are
illuminate also the harmonics in grid connected, the MPP
starts close to the operating point, but the P&O algorithm
detects that and moves the operating point in the right

direction. The AMPO-ANN algorithm gives a better
result than the classic algorithm AMPO.
t, s
Ppv, W
10
0
0
20 30
50
100
t, s
Ipv, A
10
0
0
20 30
2
4
t, s
Vpv, V
0
10
0
10
20
30
20 30
Fig. 25.Voltage of PV (AMPO-ANN)
t, s
VAC, V
0
10
0
20
40
20 30
–20
–40
Fig. 26. Voltage of AC bus (AMPO-ANN)
t, s
I, A
10
0
0.05
0
20 30
0.1
–0.05
–0.1
–0.15
Fig. 27. Current of AC bus (AMPO-ANN)
Conclusions.
We analyzed the electrical functioning of a
photovoltaic system, adapted by DC-DC converter,
regulated by an maximum power point tracking
command, to control maximum power point tracking of a
photovoltaic system based on neural networks were
presented and its architecture of neural networks was
used. The simulation and validation results show that this
system can adapt the maximum operating point for
variations in external disturbances.
We can say that artificial neural networks are
efficient and powerful modeling tools, their robustness
lies in the possibility of predicting the output of the
network even if the relationship with the input is not
linear.
The purpose of the modified algorithm adaptive
modified perturbation and observation – artificial neural
network is to reduce oscillation and achieve a high
response of the output power in response to changing
weather conditions and parameter variations. All of the
results show that the proposed technique control and our
improved maximum power point tracking approach are
effective.
Funding. This work was supported by the Franco-
Algerian cooperation program PHC-Maghreb.
Acknowledgement. The authors would like to thank
laboratory teams of research Propulsion Systems –
Electromagnetic Induction, LSPIE, University of Batna 2,
Batna, Algeria.
Conflict of interest. The authors declare that they
have no conflicts of interest.
REFERENCES
1. International Energy Agency. Global Energy & CO2 Status
Report 2019. Available at: https://www.iea.org/reports/global-
energy-co2-status-report-2019 (accessed 25 May 2021).
2. Wisnubroto D.S., Zamroni H., Sumarbagiono R., Nurliati G.
Challenges of implementing the policy and strategy for
management of radioactive waste and nuclear spent fuel in
Indonesia. Nuclear Engineering and Technology, 2021, vol. 53,
no. 2, pp. 549-561. doi: https://doi.org/10.1016/j.net.2020.07.005.
3. Zhou W., Ibano K., Qian X. Construction of an East Asia
Nuclear Security System. In: Zhou W., Qian X., Nakagami K.
(eds) East Asian Low-Carbon Community. Springer, Singapore,
2021, pp. 199-214. doi: https://doi.org/10.1007/978-981-33-
4339-9_11.
4. Rahman F.A., Aziz M.M.A., Saidur R., Bakar W.A.W.A.,
Hainin M.R., Putrajaya R., Hassan N.A. Pollution to solution:
Capture and sequestration of carbon dioxide (CO2) and its
utilization as a renewable energy source for a sustainable future.
Renewable and Sustainable Energy Reviews, 2017, vol. 71, pp.
112-126. doi: https://doi.org/10.1016/j.rser.2017.01.011.
5. Al-Maamary H.M.S., Kazem H.A., Chaichan M.T. The
impact of oil price fluctuations on common renewable energies
in GCC countries. Renewable and Sustainable Energy Reviews,
2017, vol. 75, pp. 989-1007. doi:
https://doi.org/10.1016/j.rser.2016.11.079.
6. Ahmadlouydarab M., Ebadolahzadeh M., Muhammad Ali H.
Effects of utilizing nanofluid as working fluid in a lab-scale
designed FPSC to improve thermal absorption and efficiency.
Physica A: Statistical Mechanics and its Applications, 2020, vol.
540, p. 123109 .doi: https://doi.org/10.1016/j.physa.2019.123109.
7. Abbassi A., Abbassi R., Heidari A.A., Oliva D., Chen H.,
Habib A., Jemli M., Wang M. Parameters identification of
photovoltaic cell models using enhanced exploratory salp
chains-based approach. Energy, 2020, vol. 198, p. 117333. doi:
https://doi.org/10.1016/j.energy.2020.117333.

8. Sivakumar L.P., Sivakumar S., Prabha A., Rajapandiyan A.
Implementation of particle swarm optimization for maximum
power absorption from photovoltaic system using energy
extraction circuit. 2019 IEEE International Conference on
Intelligent Techniques in Control, Optimization and Signal
Processing (INCOS), 2019, pp. 1-4. doi:
https://doi.org/10.1109/incos45849.2019.8951378.
9. Ramos-Hernanz J., Uriarte I., Lopez-Guede J.M.,
Fernandez-Gamiz U., Mesanza A., Zulueta E. Temperature
based maximum power point tracking for photovoltaic modules.
Scientific Reports, 2020, vol. 10, no. 1, p. 12476. doi:
https://doi.org/10.1038/s41598-020-69365-5.
10. De Brito M.A.G., Galotto L., Sampaio L.P., E Melo G.D.A.,
Canesin C.A. Evaluation of the main MPPT techniques for
photovoltaic applications. IEEE Transactions on Industrial
Electronics, 2013, vol. 60, no. 3, pp. 1156-1167. doi:
https://doi.org/10.1109/tie.2012.2198036.
11. Motahhir S., El Hammoumi A., El Ghzizal A. The most
used MPPT algorithms: Review and the suitable low-cost
embedded board for each algorithm. Journal of Cleaner
Production, 2020, vol. 246, p. 118983. doi:
https://doi.org/10.1016/j.jclepro.2019.118983.
12. Charin C., Ishak D., Mohd Zainuri M.A.A. A maximum
power point tracking based on levy flight optimization.
International Journal of Power Electronics and Drive Systems,
2020, vol. 11, no. 3, p. 1499. doi:
https://doi.org/10.11591/ijpeds.v11.i3.pp1499-1507.
13. Rezkallah M., Hamadi A., Chandra A., Singh B. Design and
implementation of active power control with improved P&O
Method for wind-PV-battery-based standalone generation system.
IEEE Transactions on Industrial Electronics, 2018, vol. 65, no. 7,
pp. 5590-5600. doi: https://doi.org/10.1109/tie.2017.2777404.
14. Nebti K., Lebied R. Fuzzy maximum power point tracking
compared to sliding mode technique for photovoltaic systems
based on DC-DC boost converter. Electrical Engineering &
Electromechanics, 2021, no. 1, pp. 67-73. doi:
https://doi.org/10.20998/2074-272x.2021.1.10.
15. Algazar M.M., Al-Monier H., El-Halim H.A., Salem
M.E.E.K. Maximum power point tracking using fuzzy logic
control. International Journal of Electrical Power & Energy
Systems, 2012, vol. 39, no. 1, pp. 21-28. doi:
https://doi.org/10.1016/j.ijepes.2011.12.006.
16. Fares D., Fathi M., Shams I., Mekhilef S. A novel global
MPPT technique based on squirrel search algorithm for PV
module under partial shading conditions. Energy Conversion
and Management, 2021, vol. 230, p. 113773. doi:
https://doi.org/10.1016/j.enconman.2020.113773.
17. Ishaque K., Salam Z., Amjad M., Mekhilef S. An improved
particle swarm optimization (PSO)–based MPPT for PV with
reduced steady-state oscillation. IEEE Transactions on Power
Electronics, 2012, vol. 27, no. 8, pp. 3627-3638. doi:
https://doi.org/10.1109/tpel.2012.2185713.
18. Bendib B., Krim F., Belmili H., Almi M.F., Bolouma S. An
intelligent MPPT approach based on neural-network voltage
estimator and fuzzy controller, applied to a stand-alone PV
system. 2014 IEEE 23rd International Symposium on Industrial
Electronics (ISIE), 2014, pp. 404-409. doi:
https://doi.org/10.1109/isie.2014.6864647.
19. Messalti S., Harrag A., Loukriz A. A new variable step size
neural networks MPPT controller: Review, simulation and
hardware implementation. Renewable and Sustainable Energy
Reviews, 2017, vol. 68, pp. 221-233. doi:
https://doi.org/10.1109/isie.2014.6864647.
20. Motamarri R., Nagu B. GMPPT by using PSO based on
Lévy flight for photovoltaic system under partial shading
conditions. IET Renewable Power Generation, 2020, vol. 14,
no. 7, pp. 1143-1155. doi: https://doi.org/10.1049/iet-
rpg.2019.0959.
21. Motamarri R., Bhookya N. JAYA algorithm based on Lévy
flight for global MPPT under partial shading in photovoltaic
system. IEEE Journal of Emerging and Selected Topics in
Power Electronics, 2021, vol. 9, no. 4, pp. 4979-4991. doi:
https://doi.org/10.1109/jestpe.2020.3036405.
22. Charin C., Ishak D., Mohd Zainuri M.A.A., Ismail B., Mohd
Jamil M.K. A hybrid of bio-inspired algorithm based on Levy
flight and particle swarm optimizations for photovoltaic system
under partial shading conditions. Solar Energy, 2021, vol. 217,
pp. 1-14. doi: https://doi.org/10.1016/j.solener.2021.01.049.
23. Abo-Elyousr F.K., Abdelshafy A.M., Abdelaziz A.Y.
MPPT-based particle swarm and cuckoo search algorithms for
PV systems. Green Energy and Technology, 2020, pp. 379-400.
doi: https://doi.org/10.1007/978-3-030-05578-3_14.
24. Mohd Adnan M.R.H., Sarkheyli A., Mohd Zain A., Haron
H. Fuzzy logic for modeling machining process: a review.
Artificial Intelligence Review, 2015, vol. 43, no. 3, pp. 345-379.
doi: https://doi.org/10.1007/s10462-012-9381-8.
25. Abo-Sennah M.A., El-Dabah M.A., Mansour A.E.-B.
Maximum power point tracking techniques for photovoltaic
systems: A comparative study. International Journal of
Electrical and Computer Engineering, 2021, vol. 11, no. 1, p.
57-73. doi: https://doi.org/10.11591/ijece.v11i1.pp57-73.
26. Kumar N., Singh B., Panigrahi B.K. PNKLMF-based neural
network control and learning-based HC MPPT technique for
multiobjective grid integrated solar PV based distributed
generating system. IEEE Transactions on Industrial Informatics,
2019, vol. 15, no. 6, pp. 3732-3742. doi:
https://doi.org/10.1109/tii.2019.2901516.
27. Tavakoli A., Forouzanfar M. A self-constructing Lyapunov
neural network controller to track global maximum power point
in PV systems. International Transactions on Electrical Energy
Systems, 2020, vol. 30, no. 6, pp. 1-15. doi:
https://doi.org/10.1002/2050-7038.12391.
28. Sahraoui H., Chrifi-Alaoui L., Drid S., Bussy P. Second
order sliding mode control of DC-DC converter used in the
photovoltaic system according an adaptive MPPT. International
Journal of Renewable Energy Research, 2016, vol. 6, no. 2, pp.
375-383. Available at:
https://www.ijrer.org/ijrer/index.php/ijrer/article/view/3369/pdf
(accessed 25 May 2021).
29. Slama F., Radjeai H., Mouassa S., Chouder A. New
algorithm for energy dispatch scheduling of grid-connected solar
photovoltaic system with battery storage system. Electrical
Engineering & Electromechanics, 2021, no. 1, pp. 27-34. doi:
https://doi.org/10.20998/2074-272x.2021.1.05.
30. Belbachir N., Zellagui M., Settoul S., El-Bayeh C.Z.,
Bekkouche B. Simultaneous optimal integration of photovoltaic
distributed generation and battery energy storage system in
active distribution network using chaotic grey wolf
optimization. Electrical Engineering & Electromechanics, 2021,
no. 3, pp. 52-61. doi: https://doi.org/10.20998/2074-
272X.2021.3.09.
31. Shavelkin A.A., Gerlici J., Shvedchykova І.О., Kravchenko
K., Kruhliak H.V. Management of power consumption in a
photovoltaic system with a storage battery connected to the
network with multi-zone electricity pricing to supply the local
facility own needs. Electrical Engineering & Electromechanics,
2021, no. 2, pp. 36-42. doi: https://doi.org/10.20998/2074-
272X.2021.2.06.
32. Jordehi A.R. Parameter estimation of solar photovoltaic
(PV) cells: A review. Renewable and Sustainable Energy
Reviews, 2016, vol. 61, pp. 354-371. doi:
https://doi.org/10.1016/j.rser.2016.03.049.
33. Wei Z., Zhang B., Jiang Y. Analysis and modeling of
fractional-order buck converter based on Riemann-Liouville
derivative. IEEE Access, 2019, vol. 7, pp. 162768-162777. doi:
https://doi.org/10.1109/access.2019.2952167.

34. Sahraoui H., Drid S., Chrifi-Alaoui L., Hamzaoui M.
Voltage control of DC-DC buck converter using second order
sliding mode control. 2015 3rd International Conference on
Control, Engineering & Information Technology (CEIT), 2015,
pp. 1-5. doi: https://doi.org/10.1109/ceit.2015.7233082.
35. Alnejaili T., Drid S., Mehdi D., Chrifi-Alaoui L., Belarbi R.,
Hamdouni A.. Dynamic control and advanced load management
of a stand-alone hybrid renewable power system for remote
housing. Energy Conversion and Management, 2015, vol. 105, pp.
377-392. doi: https://doi.org/10.1016/j.enconman.2015.07.080.
36. Ahmed J., Salam Z. An improved perturb and observe
(P&O) maximum power point tracking (MPPT) algorithm for
higher efficiency. Applied Energy, 2015, vol. 150, pp. 97-108.
doi: https://doi.org/10.1016/j.apenergy.2015.04.006.
37. Bouchaoui L., Hemsas K.E., Mellah H., Benlahneche S.
Power transformer faults diagnosis using undestructive methods
(Roger and IEC) and artificial neural network for dissolved gas
analysis applied on the functional transformer in the Algerian
north-eastern: a comparative study. Electrical Engineering &
Electromechanics, 2021, no. 4, pp. 3-11. doi:
https://doi.org/10.20998/2074-272X.2021.4.01.
Received 17.07.2021
Accepted 05.09.2021
Published 26.10.2021
Hamza Sahraoui1,3
, Doctor of Engineering,
Hacene Mellah2
, Doctor of Engineering,
Said Drid3
, Professor, Dr.-Ing. of Engineering,
Larbi Chrifi-Alaoui4
, Dr.-Ing. of Engineering,
1
Electrical Engineering Department,
Hassiba Benbouali University of Chlef,
B.P 78C, Ouled Fares Chlef 02180, Chlef, Algeria,
e-mail: hamzasahraoui@gmail.com
2
University Akli Mouhand Oulhadj-Bouira,
Rue Drissi Yahia Bouira, 10000, Algeria,
e-mail: has.mel@gmail.com (Corresponding author)
3
Research Laboratory LSPIE,
University of Batna 2,
53, Route de Constantine, Fésdis, Batna 05078, Algeria,
e-mail: saiddrid@ieee.org
4
Laboratoire des Technologies Innovantes (LTI),
University of Picardie Jules Verne, IUT de l'Aisne,
13 Avenue François Mitterrand 02880 Cuffies-Soissons, France,
e-mail: larbi.alaoui@u-picardie.fr
How to cite this article:
Sahraoui H., Mellah H., Drid S., Chrifi-Alaoui L. Adaptive maximum power point tracking using neural networks for a photovoltaic
systems according grid. Electrical Engineering & Electromechanics, 2021, no. 5, pp. 57-66. doi: https://doi.org/10.20998/2074-
272X.2021.5.08.

Adaptive maximum power point tracking using neural networks for a photovoltaic systems according grid

More Related Content

What's hot (20)

Similar to Adaptive maximum power point tracking using neural networks for a photovoltaic systems according grid (20)

More from Mellah Hacene (20)

Recently uploaded (20)

Adaptive maximum power point tracking using neural networks for a photovoltaic systems according grid