Available online at www.sciencedirect.

com

ScienceDirect
Energy Procedia 111 (2017) 924 – 933

8th International Conference on Sustainability in Energy and Buildings, SEB-16, 11-13 September
2016, Turin, ITALY

A new MPPT-based ANN for photovoltaic system under partial

shading conditions
Loubna Bouselhama*, Mohammed Hajjia, Bekkay Hajji, Hicham Boualia
a
Renewable energy , embedded system and information processing laboratory, National Scool of applied sciences, mohamed first university,
60000, Oujda, Morocco

Abstract

In solar photovoltaic system, tracking the maximum power point (MPP) is challenging task due to varying climatic conditions.
Moreover, the tracking algorithm becomes more complicated under the condition of partial shading due to the presence of
multiple peaks in the power voltage characteristics. This paper introduces a novel method to track the global maximum power
point under partially shaded conditions. The method combines an artificial neural network controller with a scanning algorithm.
The PV system along with the proposed MPPT algorithm was simulated using Matlab/Simulink environment. The simulated
system was evaluated under uniform and non-uniform irradiation conditions. For comparison, an improved variable step P&O
with global scanning (PO&GS) and incremental conductance controller based on a fuzzy duty cycle change estimator (FLE)
with direct control were used and the results show that the proposed approach is effective in tracking the MPP and presents fast
response time.
© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of KES International.
Keywords: Photovoltaic system;Global maximum power point ;Partiel shading;Artificiel neural network

1. Introduction

For clean environment and a profitable economic gain of the sustainable solar energy source, recently a great
importance is given to the generation of electricity across the photovoltaic (PV) system. However, the main
weakness of the PV system is low efficiency of conversion of insolation into electricity. Furthermore, the power
generated by PV modules depends on environmental factors, i.e. solar radiation and atmospheric temperature [1].
These factors affect the both current–voltage (I-V) and power–voltage (P-V) characteristics of the PV system. Under
uniform irradiance, the P-V curve of PV array has one maximum power point (MPP). whereas, with non-uniform
irradiance, such as partial shadowing of some PV modules or even some PV cells, the P–V characteristics get

* Corresponding author. Tel.: +212 673-824131

E-mail address: l.bouselham@ump.ac.ma

1876-6102 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of KES International.
doi:10.1016/j.egypro.2017.03.255
 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933 925

more complex, displaying multiple peaks, only one of which is the global peak (GMPP); the rest are local peaks
(LMPPs).Thus, a control technique named Maximum Power Point Tracking (MPPT) is necessary to be applied for
optimally exploits the available power in all operation conditions.
To date, numerous MPPT controllers have been presented and implemented in the literature, these controllers
have some generic requirements such as low complexity, low cost, minimum output power fluctuation, and the
ability to track quickly when operating condition changes. The most widely used algorithms are perturb and observe
(P&O) [2,3] and incremental conductance (InC) [4]. These conventional methods achieve moderate performance
with an easy implementation and a low cost. For better transient and steady-state performance, artificial intelligence
based MPPT techniques have been suggested such as fuzzy logic [5] and artificial neural networks controller (ANN).
However, the ANN controller has proved good performance under rapidly varying irradiance, especially in terms of
efficiency and response time [6]. In addition, to address the partial shading effect, the ANN controller has been
improved by combining it with other MPPT methods. In [7], A MPPT system is proposed for partially shaded PV
array by using ANN and fuzzy logic with polar information controller. The ANN is trained to determine the global
MPP voltage under several partially shaded conditions. The global MPP voltage as a reference voltage is used in the
fuzzy logic with polar information controller to gain the required control signal for the power converter. The main
drawback of this method is the high cost and complexity, due to the combination of two smart methods. In [8], an
ANN based algorithm in conjunction with incremental conduction is proposed. The ANN is utilized to estimate
reference voltage of IncCond algorithm. The similar working procedure is proposed in [9] by combining the ANN
with P&O. The performance of these MPPTs is generally good except that they are very slow in tracking.
In order to improve the response time of ANN controller to track the GMPP and reduce the complexity of the
controller under PSC , this paper proposes a novel MPPT , it consists of simple algorithm that scans the P-V curve to
identify the GMPP, combined to ANN which gives the duty cycle corresponding to the GMPP. The remainder of the
paper is organized as follows: a description of the considered PV model and its characteristics are introduced in
Section 2; while section 3 describes the proposed MPPT method. Section 4 briefs the results and discussions.
Finally, conclusions are reported in Section 5.

Nomenclature

PV photovoltaic
MPPT maximum power point tracking
GMPP global maximum power point
LMPP local maximum power point
PSC partial shading conditions
ANN artificial neural network
STC standard conditions
SP shading pattern
THR threshold
G irradiance level
VGMPP corresponding voltage to GMPP (V)
Vpv PV output voltage (V)
Ppv PV output power (W)
PM maximum power (W)
D duty cycle
Inc incremental conductance
PO&GS perturb and observe and global scanning
FLE fuzzy logic estimator
926 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933

2. PV modeling and characteristics

2.1. PV module model

There are many models for characterizing solar cells. One of the most known models is the single diode shown in
Fig.1.a. The relation between the module current output and its voltage is given by following equation (1):

Where is the photo-generated current in STC, is the diode reverse saturation current, is the series
resistance, is shunt resistance, is diode ideality factor, is module output current, is module output voltage,
is Boltzmann constant, is module temperature, is electron charge and is series number connected cells in
module.
This equation (1) shows clearly the dependence of the model on the solar irradiance and temperature conditions.

2.2. Effect of PSC

To supply with high current and reach a given power, PV strings are connected in parallel. Each string is formed
of a specific PV module number connected in series in order to obtain the required voltage. Fig. 1.b shows PV array
system structure used in this work while table 1 reports the electrical features of the PV module.
Under a normal condition, when the entire strings PV receive a uniform irradiance, the typical P-V curve shows
a single MPP, as illustrated in curve (a) of Fig.2. However, when partial shading occurs in one of the module
composing the PV array, the shaded modules will not be able to produce as much current as the unshaded modules,
causing the so called hot-spot heating [10]. This drawback is overcame by using an external bypass diode that
conducting every time when the solar cell is reversed biased, thus allowing the current of unshaded cells to flow
externally to the shaded cell. Therefore, a staircase current waveform is created on the I–V curve, while the
corresponding P–V curve is characterized by multiple maxima points, as depicted in curve (b) of Fig. 2

3. The proposed MPPT method

A method is developed to track the GMPP for standalone photovoltaic system under any weather conditions. It
combines an ANN controller with scanning procedure. The block diagram of a standalone PV system with the
proposed MPPT is depicted in Fig. 3. The scanning procedure aims to identify the GMPP and its corresponding
voltage , after that, the ANN is activated to give the appropriate duty cycle (D).

(a) (b)

Fig.1. (a) Equivalent circuit of PV cell, (b) PV array system structure

3.1. Scanning procedure

The scanning procedure carried out by comparing the actual power with the previous value
 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933 927

and storing the maximum power where , ). Afterwards, to obtain ,

each maximum power reached must be compared to the previous maximum ,
where , ). The flowchart of scanning procedure is reported in Fig.4.
The scanning procedure is called at every change in irradiance. The change in irradiance is detected when the
output power variation ΔP is greater than a suitable threshold (THR) [11] that is set by users according to the
configuration of PV arrays. In our case, simulation results show the change of power is greater than 2.25 W when
the irradiance changes, so the restart condition is set to ΔP > 2.25. Fig. 5 shows the process of the scanning
procedure.

Fi
g.2.
I-V
and
P-V
chara
cteris
tics
under
unifo
rm
(curv
e a)
and
partia
l
shadi
ng
(curve b) conditions

Fig.3. The bloc diagram of the PV system with the proposed MPPT

Table 1. Electrical characteristics of PV module under STC

Designation Values
Maximum power (Pmax) 115 W
Voltage at Pmax (Vmax) 17.2 V
Current at Pmax (Imax) 6.69 A
Open circuit voltage (Voc) 21.8V
Short circuit current (Isc) 7.24 A
928 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933

Fig.4.
The
flowc
hart
of the
scann
ing
proce
dure

Fig.5.
The
scann
ing
proce
dure
to
obtai
n the
GMP
P

3.2. A
rtifi
cial
neut
ral network design

An artificial neural network is a mathematical model or computational model that is inspired by the structure
and/or functional aspects of biological neural networks. It is accepted as a technology offering an alternative way to
solve complex problems.
In this work, the developed neural network is a feed forward network including three layers as shown in Fig.6.a The
input layer consists of two neurons: the resulting GMMP and VGMPP of scanning procedure. The second layer is the
hidden layer with three neurons whose activation function is the sigmoid and output layer with one neuron to give
the duty cycle with a linear activation function. The choice of number of neurons in the hidden layer is made
according to the following relationship [12]:
And (2)

Where is the hidden layers, is the output layers and is the input layers.
The weight adjustment of the neural network is called training. The procedure of training a neural network
((
includes modification of the weights of the network to enhance network performance. Throughout the training, the
connection weights are modified until the best fit is attained for the input–output patterns based on the minimum
errors. In our case, the performance is evaluated in terms of mean square error (MSE). The synaptic weights are
adjusted until the MSE is minimized. Fi.g.6.b shows the supervised learning design. However, the training
procedure needs a set of samples of appropriate network behaviour inputs and target outputs. In this study, The
ANN is trained for a specific type of PV module; the dataset used (GMMP, VGMPP, D) covers a large range of
 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933 929

environmental conditions. It covers the range between 100 W/m2 and 1100 W/m2 for irradiance, and between 10 °C
and 50 °C for cell temperature. 70% of samples were devoted for training, 15% were devoted for cross validation,
and 15% were devoted for testing.

3.3. DC-DC boost converter

To reach the desired maximal voltage level, a DC–DC boost converter is used as the power stage interface
between the PV system and the resistive load. The pulse width modulation (PWM) technique achieved by the MPPT
duty cycle is applied to electronic switch of converter. The output voltage is expressed by the following
equation:
uaation:
u

Where is the duty cycle of the switching period, is the average output voltage of DC–DC converter and
is the PV array output voltage.
Table 2 contains the DC–DC specifications.

4. Simulation results and discussion

The PV system is simulated using Matlab/ Simulink environment. The simulations are carried out under different
weather conditions: constant and uniform irradiation, rapidly varying irradiation and partially shading or non
uniform irradiation condition. The performances are evaluated in terms of response time and the efficiency tracking.
The efficiency tracking [13] evaluated by using the following equation (5):

(a) (b)

Fig.6.
(a)
Neur
al
netw
ork
struct
ure,
(b)
Super
vised
learning design

Table 2. DC-DC boost converter characteristics

Designation Values
Capacitance (Cout) 200 μF
Capacitance (Cin) 10 μF
Inductance (L) 10 mH
Frequency (f) 20khz

(5)
930 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933

Where is the start-up time of the system and is the close-down time of the system, is the array output
power, and is the theoretical maximum array power.
On the other hand, to show the effectiveness of the developed MPPT algorithm, a performance comparison
between the proposed method and two others MPPTs is presented. The MPPT methods subject to comparison are:
- The proposed method in [11] which consist of an improved variable step P&O and Global scanning
(PO&GS) method to achieve the maximum power point tracking.
- InC controller based on a fuzzy duty cycle change estimator with direct control proposed in [14] to achieve
the maximum power point tracking when large changes occur in the irradiance. A fuzzy logic estimator
(FLE) is used to estimate the new duty cycle used to track the PV array maximum power point.

4.1. Constant and uniform irradiance

In this case, PV system was simulated with uniform irradiation on all the solar cells. A constant irradiation level
of G = 600 W/m2 and temperature level of T=25°C were maintained on the PV panels. Only one MPP occurs in this
case (390W). Fig.8 shows that PO&GS and FLE takes more time to converge to MPP estimated at 50ms and 35ms
respectively. However, the developed method track the MPP with fast time response estimated at 15ms, so almost
five times faster than the PO &GS and three times faster than the FLE. Moreover, sustained oscillations are present
in PV output power of FLE around MPP. Table 3 summarizes the performances of the compared methods under
STC.

Fig.8. The output power under constant and uniform irradiance

Table 3. The performances of the two compared methods at STC

Evaluated parameters Proposed method PO &GS FLE
Response time (ms) 15 50 70
Power production (W) 381 382.27 380.77
Efficiency (%) 98.77 99.1 98.7
Sensors Current, voltage Current, voltage Current,voltage

4.2. Rapidly varying irradiance

In order to simulate the rapidly varying irradiation condition, the irradiation level was suddenly decreased from
620 W/m2 to 400 W/m2 then increased to 750W/m². As shown in fig. 9, when increasing or decreasing the irradiance
level, the proposed method is rapid to track the MPP as compared to the PO&GS and FLE with an efficiency of
 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933 931

97.64%. Moreever, with decreasing of the irradiance , the FLE method takes more time to reach MPP.

4.3. Partial shading conditions

To evaluate the performances of the proposed MPPT controller during no-uniform irradiance levels,
four unshaded PV modules receive 900 W/m2 and the two remaining modules are partially shaded receiving 680
W/m2. From the output P–V characteristic curve of the PV generator depicted in fig.10.a, two power peaks are
observed corresponding respectively to the LMPP (394W) and the GMPP (494 W). As shown in fig.10.b the
proposed MPPT controller has the ability to distinguish between the global peak (GMPP) and local peak (LMPP)
through the scanning procedure. Furthermore, as shown fig.11, it can be noticed that the three methods are able to
track the GMPP but the proposed method still has a fast response time even under no-uniform irradiance estimated
at 15ms.

Fi
g.9.
The
PV
output
power
under rapidly varying irradiance

(a) (b)

Fig.10. (a) P–V characteristic curve of the PV under no – uniform irradiance, (b) PV power and GMMP determined by the scanning procedure
932 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933

Fig.11. The output power during the no-uniform insolation level

For an extensive verification, three Shading Patterns listed in table 4 are applied successively. Fig.12 and fig.13
show the P–V characteristic curve of the PV, PV power and GMMP determined by the scanning procedure and the
output power for the three shading patterns, respectively. It is worth noting that, the proposed MPPT controller has
the ability to identify the true peak (GMPP) among the multiple local peaks (LMPPs) through the scanning
procedure and track the GMPP with efficiency of 97.64%.

(a) (b)
Fig.12. (a) P–V characteristic curve of the three shading patterns , (b) PV power and GMMP determined by the scanning procedure for the three
shading patterns

Table 4. The shading patterns applied for the simulation under PSC and corresponding global peaks
Pattern number Shading pattern PGMMP (W)
[M1,M2,M3,M4,M5,M6]
SP1 [400,520,700,400,520,700] 290
SP2 [550,800,800,550,800,800] 396
SP3 [720,560,600,560,600,400] 340
 Loubna Bouselham et al. / Energy Procedia 111 (2017) 924 – 933 933

Fig.13. The output power during the three shading patterns

5. Conclusion

In this paper, an intelligent MPPT method has been proposed to track the GMPP under different weather
conditions for standalone PV systems. The proposed method integrates scanning procedure of P-V curve with ANN
controller. The simulation results demonstrate that the proposed method is able to distinguish between the GMPPP
and LMPPs and guarantees a rapid convergence to the GMPP with good efficiency estimated to 97.64 under
transient variation of shading patterns. A comparative study between the developed method and PO &GS and FLE
has been carried out. The results have shown that the three methods are able to track the GMPP but the proposed
method provides better results in term of response time. The response time given by the new MPPT method is
almost five times less than that given by the PO&GS and three times faster than the FLE.

References

[1] T. Esram and P. L. Chapman, “Comparison of Photovoltaic Array Maximum Power Point Tracking Techniques,” IEEE trans. Energy
Convers, vol. 22, no. 2, pp. 439–449, 2007.

[2] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, “Optimization of Perturb and Observe Maximum Power Point Tracking Method,”
IEEE Trans. Aerospace .Electro systems ,vol. 20, no. 4, pp. 963–973, 2005.
[3] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, “Predictive & Adaptive MPPT Perturb and Observe Method,” IEEE Trans.
Aerospace. .Electro, vol. 43, no. 3, 2007.
[4] A. Safari and S. Mekhilef, “Simulation and Hardware Implementation of Incremental Conductance MPPT With Direct Control Method
Using Cuk Converter,” IEEE Trans. Indus. Electron vol. 58, no. 4, pp. 1154–1161, 2011.
[5] R..Garraoui ; M .Ben Hamed.; L.Sbita, “ MPPT controller for a photovoltaic power system based on fuzzy logic” 10th International Multi-
Conference on Systems, Signals & Devices (SSD), pp:1-6,2013
[6] L. Bouselham, B. Hajji, and H. Hajji, “Comparative Study of Different MPPT Methods for Photovoltaic System,” 3rd International
Renewable and Sustainable Energy Conference (IRSEC), pp: 1-5,2015.
[7] Syafaruddin Karatepe E, Hiyama T. Artificial neural network-polar coordinated fuzzy controller based maximum power point tracking
control under partially shaded conditions. IET Renew Power Gener, vol.3, pp. 239–53,2009.
[8] Punitha K, Devaraj D, Sakthivel S. Artificial neural network based modified incremental conductance algorithm for maximum power point
tracking in photovoltaic system under partial shading conditions. Energy vol.62, pp: 330–40 ;2013
[9] Jiang L, Maskell DL. A simple hybrid MPPT technique for photovoltaic systems under rapidly changing partial shading conditions. 40th
photovoltaic specialist conference (PVSC). pp. 0782–7. ; 2014.
[10] M. E. Nezhad, B. Asaei, and S. Farhangi, “Modified Analytical Solution for Tracking Photovoltaic Module Maximum Power Point under
Partial Shading Condition,” 13th International Conference on Environment and Electrical Engineering (EEEIC), 2013.
[11] Q. Duan, J. Leng, P. Duan, B. Hu, and M. Mao, “An Improved Variable Step PO and Global Scanning MPPT Method for PV Systems
under Partial Shading Condition,” 7th International Conference on Intelligent Human-Machine Systems and Cybernetics 2015.
[12] S. Reza, S. Salwah, and B. Salim, “Neurocomputing Real-time frequency-based noise-robust Automatic Speech Recognition using Multi-
Nets Artificial Neural Networks : A multi-views multi-learners approach,” Neurocomputing, vol. 129, pp. 199–207, 2014.
[13] A. Bayod-ru and A. Cebollero-abia, “A novel MPPT method for PV systems with irradiance measurement,” solar energy vol. 109, pp. 95–
104, 2014.
[14] T. Radjai, L. Rahmani, S. Mekhilef, and J. Paul, “Implementation of a modified incremental conductance MPPT algorithm with direct
control based on a fuzzy duty cycle change estimator using dSPACE,” Sol. ENERGY, vol. 110, pp. 325–337, 2014.