Solar Energy 157 (2017) 171–186

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

Global MPPT based on flower pollination and differential evolution

algorithms to mitigate partial shading in building integrated PV system
Ahmed A. Zaki Diab b, Hegazy Rezk a,b,⇑
a
College of Engineering at Wadi Addawaser, Prince Sattam Bin Abdulaziz University, Saudi Arabia
b
Electrical Engineering Department, Faculty of Engineering, Minia University, Egypt

a r t i c l e i n f o a b s t r a c t

Article history: The implications of partial shading condition (PSC) on building integrated photovoltaic power system
Received 16 August 2016 (BIPVPS) and its mitigation is introduced. Tracking global maximum power point (MPP) for BIPVPS during
Received in revised form 18 June 2017 PSC based on the traditional maximum power point tracking (MPPT) algorithms like hill climbing, and
Accepted 7 August 2017
perturb & observe is incompetent. Therefore, a global MPPT based on meta-heuristic optimization tech-
niques is an important point of research to increase the performance of partially shaded BIPVPS. This
work presents a powerful technique called Flower Pollination Algorithm (FPA) to mitigate PSC in
Keywords:
BIPVPS. The feasibility and effectiveness of FPA technique for extracting global MPP are validated with
Terms
global MPPT
Building integrated PV system
various shadow patterns. For illustrating the capabilities of FPA method, it is compared with other opti-
Partial shading mization techniques such as differential evolution and particle swarm optimization. The results confirm
Flower pollination algorithm that FPA ensures exact catching global MPP under different PSC patterns. Moreover, FPA has best perfor-
Differential evolution mance when compared with the other two optimization techniques.
Particle swarm optimization Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Nowadays, looking for different but environmentally friendly
resources of energy becomes a must. This is because of the dra-
matic drawbacks of the conventional fossil fuel energy. Gases, such
as CO and CO2, which emitted from the combustion of fossil fuel is
negatively affecting the living organs. These gases also deteriorate
the Ozone layer, which in turn changes the climate and increases
the temperature of the Earth. Additionally and most importantly,
its bad effect on the human health especially the pulmonary sys-
tem. Because of the above, one of the national research projects
in worldwide is the field of Renewable Energy. Solar Energy is def-
initely one of this type. Photovoltaic power system (PVPS) has
many merits, such as: pollution-free, little maintenance and
noise-free (Rezk and Dousoky, 2016; Rezk, 2016).
For supporting grid-connected PVPS in the world the Feed-in-
tariff system (FITS) is created. FITS aimed to encourage investment
in renewable energy sector. It is designed for providing price cer-
tainty to renewable electricity generators (Murphy and
McDonnell, 2017). Fig. 1 illustrated the electricity costs and FIT
in Germany as an example. From the figure, it can be see that the
feed-in-tariff for PV rooftop up to 10 kW is decreased by 8.9%
⇑ Corresponding author at: College of Engineering at Wadi Addawaser, Prince
Sattam Bin Abdulaziz University, Saudi Arabia.
E-mail address: Hegazy.hussien@mu.edu.eg (H. Rezk).
during the period form 2000 into 2015. In addition FIT for PV free-
standing is decreased by 11.2% for the same period.
Recently, there is an increasing interest in using building inte-
grated photovoltaic system (BIPVPS) as a new source of electrical
energy. BIPVPS provides energy to buildings. Furthermore, PV pan-
els have been used as structural elements for the buildings.
Accordingly, it decreases concentrations of additional weight
above ceiling. On the other hand, under implementation feed-in-
tariff program, BIPVPS produces enough revenue for paying back
cost of PV system (Salam et al., 2015). The Energy Payback Time
(EPBT) for a grid-connected BIPVPS system, including PV modules,
inverter, and cables, varies based on the type of system and loca-
tion it is installed in. For monocrystalline panels it ranges from
under one year to just above 2 years. Due to more efficient produc-
tion methods, the EPBT has been cut to less than half of its 1990
figures as illustrated in Fig. 2 (Bhandari et al., 2015).
The problem of how can increase the efficiency of PVPS, has
more interesting from the researchers. Moreover, developing
materials which can offer a high conversation efficiency at low cost
under research. There is a wide range of PV solar cell technologies.
Such technologies include; crystalline silicon, thin-film and multi-
junction solar cells. The crystalline silicon like mono-crystalline,
poly-crystalline, multi-crystalline and ribbon multi-crystalline.
The thin-film technologies like amorphous silicon, cadmium tel-
luride (CdTe) and copper-indium-gallium-diselenide (CIGS). Every

http://dx.doi.org/10.1016/j.solener.2017.08.024
0038-092X/Ó 2017 Elsevier Ltd. All rights reserved.
172 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186

Fig. 1. Electricity costs and Feed-In Tariffs in Germany.

Fig. 2. EPBT of multi-crystalline PV rooftop systems installed in Southern Europe.

technology have different material and processing requirements,
leading to distinct emission profiles. Fig. 3 shows the development
of solar cell efficiencies during the period from 1993 into 2015
(Green et al., 2016).
Under uniform solar irradiance, the voltage-power curve BIPVPS
contains only one maximum power point (MPP) and such point can
be easily tracked by any conventional maximum power point
tracking (MPPT) method without any problems (Ahmed and
Salam, 2014). Many efficient traditional MPPT techniques was con-
ducted to improve the conversion efficiency of BIPVPS without tak-
ing the problem of partial shading condition (PSC) into
consideration (Ma and et al., 2013). These algorithms include; hill
climbing, incremental conductance (INC), incremental resistance
(INR), fuzzy logic control, perturb & observe (P&O), and neural net-
work (NN) (Rezk and Eltamaly, 2015; Rezk and Hasaneen, 2015;
Rezk, 2016; Salah et al., 2008). The problem occurs under PSC as
voltage-power characteristic contains multiple local MPPs and
only one global MPP. Therefore, the tracking of the MPP is difficult
and the conventional techniques fail in tracking global MPP since
they cannot differentiate between global and local peaks. Accord-
ingly, overall efficiency of the BIPVPS decreases. To solve this prob-
lem MPPT based on optimization algorithms is required.
The idea of using meta-heuristic optimization techniques for
global search is proposed in Karlis et al. (2007), Liu and et al.
(2012), Tey et al. (2014), Daraban et al. (2013), Phimmasone
et al. (2010), Jiang et al. (2013), Chen et al. (2014), Sarvi et al.
(2015), Mirhassaniet et al. (2015), Ishaque and Salam (2013), Liu
et al. (2012), Salam and Saad (2013), Yang (2012), Abdelaziz
et al. (2016), Rezk et al. (July 2017), Rezk and Fathy (2016), Fathy
and Rezk (2016, 2017), Mao et al. (2017), Chao et al. (2015),
Sundareswaran et al. (2015), Kumar (2017), Kumar et al. (2017),
Koad and Zobaa (2017), Rahmani and Yusof (2014),
Seyedmahmoudian et al. (2016), Dileep and Singh (2017), Sridhar
et al. (2016), Kofinas et al. (2015), Ahmed and Salam (2015). Such
algorithms include; genetic (Daraban et al., 2013) and particle
swarm optimization (Liu et al., 2012). In addition, ant colony opti-
mization and cuckoo search (Ahmed and Salam, 2014). They are
developed to deal with global MPPT problem. Based on literature
study, Flower Pollination Algorithm (FPA) didn’t utilize in the field
of tracking global MPP under abnormal conditions. FPA has been
considered simple in implementation and faster to extract optimal
solution in many engineering problems. This encourages us for
proposing FPA for dealing with tracking global MPP. It is noticed
from the previous studies that some works taken either only single
 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186
Fig. 3. Development of solar cell efficiencies.
irradiance pattern of PSC or few patterns into consideration for
checking the capability of global MPPT. This motivated the authors
to introduce a comprehensive comparison for FPA against Differen-
tial Evolution (DE) and Particle Swarm Optimization (PSO) based
global MPPTs under several different irradiance patterns via exten-
sive statistical analysis. The idea of varying PSC patterns is to
change the location of global MPP from left to right or middle for
measuring the response of FPA technique with various situations
and guarantee the reliability of FPA in tracking the global MPP.
The tracking performances of studied MPPT techniques are evalu-
ated for each pattern using MATLAB simulation.
This paper is organized as follows: Section 2 presents the previ-
ous works in the tracking of global MPP for PVPS under PSC.
Arrangement of BIPVPS and its characteristics are introduced in
Section 3. An introduction to PSO, DE and FPA are presented in Sec-
tions 4–6, respectively. Section 7 discuss the obtained results.
Finally, the conclusions are addressed in Section 8.
2. Literature review
Many researches tried to tackle the tracking of the global MPP
under PSC via using different optimization techniques (Karlis
et al., 2007; Liu et al., 2012; Tey et al., 2014; Daraban et al.,
2013; Phimmasone et al., 2010; Jiang et al., 2013; Chen et al.,
2014; Sarvi et al., 2015; Mirhassaniet et al., 2015; Ishaque and
Salam, 2013; Liu et al., 2012; Salam and Saad, 2013; Yang, 2012;
Abdelaziz et al., 2016; Rezk et al., 2017; Rezk and Fathy, 2016;
Fathy and Rezk, 2016, 2017; Mao et al., 2017; Chao et al., 2015;
Sundareswaran et al., 2015; Kumar et al., 2017, 2017; Koad and
Zobaa, 2017; Rahmani and Yusof, 2014; Seyedmahmoudian et al.,
2016; Dileep and Singh, 2017; Sridhar et al., 2016; Kofinas et al.,
2015; Ahmed and Salam, 2015). A summary of their contributions
is shown in Table 1. Mingxuan et al. proposed a MPPT method with
reduced steady-state oscillation based on a two-stage PSO tech-
nique. A hybrid shuffled frog leaping technique (SFLA)-PSO algo-
rithm (PSO-SFLA), to make sure fast and accurate searching of
the global extremum. Moreover, an adaptive speed factor pre-
sented to improve the convergence speed (Mao et al., 2017). Chao
et al. introduced a modified PSO for tracking MPP with shaded PV
systems. A control method has been proposed to adapt the PSO
parameters for enhancing convergence speed. The hardware
implementation using PIC microcontroller is introduced to
173
evaluate the proposed approach (Chao et al., 2015). Kinattingal pro-
posed a hybrid algorithm based on ant colony optimization and
local search capability of P&O method to improve the tracking per-
formance. Computed and measured results are presented for eval-
uating the proposed technique (Sundareswaran et al., 2015). The
integration between Jaya and Differential Evolution (DE) is done
by Nishant et al. The proposed technique is tested by simulation
work. The results showed that it tracked global MPP for different
studied cases with a suitable accuracy (Kumar et al., 2017). Nishant
introduced a hybrid approach from Whale optimization (WO) with
Differential Evolution (DE); for fast tracking of the global MPP
without oscillation (Kumar et al., 2017). Ramdan et al. proposed
a novel tracker based on the integration between Lagrange Interpo-
lation (LI) and PSO for eliminating the traditional methods initial-
ization problem. This can be done by migrating the particles close
to the best position directly using LI. Such technique minimizes
the PSO number of iterations for reaching the MPP. The results
are compared with P&O, INC, and the conventional PSO. The com-
parison indicated that the proposed technique enhances the search
stability and the tracking speed capability under any abrupt
change (Koad and Zobaa, 2017). Rahmani and Yusof proposed a
novel optimization technique named Radial Movement Optimiza-
tion (RMO) based on swarm population. This technique employs
spherical boundaries in the search-space. It needs only to adjust
few numbers of the parameter which makes the search process
robust and fast (Rahmani and Yusof, 2014). Seyedmahmoudian
et al. proposed a tracking system for global MPP for PV system
under PSC based on RMO technique. The speed of convergence,
efficiency, stability and computational complexity were considered
in the comparison with other techniques (Seyedmahmoudian et al.,
2016). A comprehensive study for soft computing methods like NN,
FLC, Chaos theory, probability, non-linear predictor, evolutionary
algorithms (GA and DE) and swarm intelligence algorithms (ACO
and PSO) is introduced for application for the MPPT by Dileep
et al. .The merits and weaknesses of each reported algorithm are
discussed (Dileep and Singh, 2017). TLBO is introduced as efficient
technique for global MPPT in Rezk and Fathy (2016). In Fathy and
Rezk (2016), mine blast algorithm (MBA) and TLBO are employed
to capture the global MPP for partially shaded PV system. The algo-
rithms have been tested under various cases of PSC such as, uni-
form and non-uniform patterns with various positions of global
MPP. In Sridhar et al. (2016), a shuffled frog leap method (SFLA)
was conducted to identify the global MPP. The results of SFLA
174 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186

Table 1
Comparison of different global MPPT approaches from previously works.

Authors Year MPPT technique Grid Control DC-DC Controller Contribution

connection parameter converter
Mao et al. 2017 Hybrid shuffled frog leaping No Duty Boost DSP and Increasing speed of MPPT via
(2017) algorithm (SFLA)-PSO algorithm cycle converter simulation improvement of PSO using hybridization
(PSO-SFLA) results with SFLA
Ahmed and 2014 Cuckoo Search & PSO Off grid Voltage Buck- Matlab Studying CS performance under gradual,
Salam boostconverter Simulation step and rapid changes in irradiance and
(2014) temperature
Sundareswaran 2015 hybrid of ant colony (ACO) and No Duty Boost DSP and Increasing the accuracy of P&O using ACO
et al. (2015) P&O; (ACO- P&O) cycle converter simulation
results
Kumar et al. 2017 Jaya and differential evolution No Duty Boost DSP and A proposed hybrid algorithm based on
(2017) cycle converter simulation Jaya and DE algorithms
results
Kumar et al. 2017 Whale optimization (WO) with No Duty Boost DSP and A proposed hybrid algorithm based on
(2017) differential evolution cycle converter simulation WO and DE algorithms
results
Sridhar et al. 2017 shuffled frog leap algorithm (SFLA) No Duty Boost simulation Measuring the performance of SFLA
(2016) cycle converter results method
Rezk and Fathy 2016 Teaching–learning-based No Duty Boost simulation Evaluating the performance of TLBO
(2016) optimization cycle converter results algorithm
Fathy and Rezk 2016 Mine blast optimization No Duty Boost simulation Evaluating the performance of MBO
(2016) cycle converter results algorithm

method is compared with P&O and PSO algorithms. The PV system
is implemented in MATLAB/Simulink. Kofinas et al. proposed an
NN based MPPT. Moreover, adaptive neuron and learning tech-
nique of Big Bang-Big Cruch (BBBC) approach are applied for
enhancing the system performance and avoiding the local optima
(Kofinas et al., 2015). In Ahmed and Salam (2015); review of differ-
ent methods for MPPT is introduced. The techniques were P&O
algorithm, PSO, ant colony, NN, FLC, hill climbing, cuckoo search,
differential evolution, flashing firefly, and extreme seeking control
approach and irradiance measurement. In Ahmed and Salam
(2014), Jubaer and Salam proposed MPPT based on Cuckoo Search
(CS) optimization technique. The reported global MPPT based on
optimization techniques have been summarized in Table 1.
3. System description
3.1. Arrangement of BIPVPS
Fig. 4 illustrated a schematic diagram of BIPVPS for a dwelling
house. The BIPVPS installations are increasing every year. Archi-
tects and designers are developing new, creative methods for
employing these technologies and developers are integrating them
into green housing tracts, while manufacturers continue for creat-
ing new employing for meeting this market niche. In BIPVPS, the
solar panels integrated with building to be a part of the structure
(Salam et al., 2015). BIPVPS has multiple series-connected and
parallel-connected solar panels for creating a system with certain
rating. Therefore, the probability of occurrence the PSC has been
considered high (Ahmed and Salam, 2014). BIPVPS that usually
built in urban regions can exposed to PSC by taller adjacent build-
ings, utility grid towers, transmission cables, telephone poles and
trees.
3.2. BIPVPS under PSC
PV cell has a complicated relationship among irradiance, tem-
perature, and load demand. Accordingly, it exhibits nonlinear char-
acteristics among output power, current and voltage. Hence, MPPT
algorithms must be integrated with BIPVPS for enhancing its effi-
ciency (Liu et al., 2012; Tey et al., 2014). PSC is considered a condi-
tion when a part or the whole solar panel exposed a non-uniform
irradiance. Under PSC, the PV modules receiving similar irradiance
will continue operating at optimal efficiency. However, due to the
series configuration of solar cell in the panel, cells subject to sha-
dow, have to operate in the reverse bias voltage region for provid-
ing the current equal to that flowing in the unshaded cells.
Operating in such conditions has an inverse impact on the effi-
ciency of entire PV panel and may cause hotspots in solar cells
(Liu et al., 2012). In order to overcome that problem, bypass diodes
have been interconnected in parallel with each panel. Under

Table 2
Uniform irradiance and different partial shading patterns for 4S3P configuration
(W/m2).

Uniform PSC Pattern #2 PSC Pattern #3 PSC Pattern #4

Irradiance
G11 1000 1000 800 800
G12 1000 1000 400 1000
G13 1000 1000 500 300
G21 1000 1000 600 600
G22 1000 1000 700 700
G23 1000 1000 1000 800
G31 1000 400 300 300
G32 1000 400 1000 400
G33 1000 1000 700 700
G41 1000 400 200 200
G42 1000 400 700 200
G43 1000 1000 200 800
Fig. 4. Arrangement of BIPVPS.
 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186 175

normal conditions, there is no influence of such diodes since they
are reversing biased. From another point of view, they are forward
biased and carry the current instead of solar panel under shadow-
ing effect (Tey et al., 2014).
Two different PV arrays are considered in this work. The first PV
configuration (4S3P configuration) comprises of 12 panels, config-
ured by three modules per string. The module has 72 series-
connected multi-crystalline silicon cells with maximum power of
102 W. The theoretical output power of the system is 1.2 kWp
(i.e. 12  51 W). For quantifying the impact of PSC, three arbitrarily
patterns are imposed, as illustrated in Table 2. G11 denotes the irra-
diance referred to the first series pair PV panels and the first paral-
lel branch whereas G12 represents irradiance of first series pair PV
panels and the second parallel branch and so on. Fig. 5 shows the
power against voltage, the power against current and the current
against voltage curves of BIPVPS, respectively, under uniform irra-
diance of 1 kW/m2 and under three different cases of PSC.
The second configuration (3S configuration) consists of three
modules in series with 153.38 maximum power under standard
condition and uniform irradiance. The simulations were imple-
mented in Matlab and run on Windows 8 operative system with
64-bit support with a 2.0 GHz Intel dual-core processor and 4 GB
RAM. Fig. 6 shows the Simulink model for designing the shadow
patterns. Table 3 and Fig. 7 illustrated the different cases of shadow
patterns.
Based on the above figures, it can see that; the current-voltage
graphs have multiple steps depending on the pattern of PSC.
Whereas, the power–voltage curves have multiple peaks which

Fig. 5. (a) Arrangement of 4S3P configuration (b) power – voltage characteristics and (c) current - voltage characteristics of BIPVPS under uniform irradiance and different
partial shading patterns for 4S3P configuration.

Fig. 6. Simulink model for drawing PV system characteristics under different cases different cases of shadow patterns of 3S configuration.
176 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186

Table 3
Location of global MPP under different cases of shadow patterns for 3S configuration.

Case G1 W/m2 G2 W/m2 G3 W/m2 Left MPP Center MPP Right MPP Global MPP, W
1 1000 1000 1000 – 152.38 – 152.38
2 800 1000 1000 99.18 – 132.2 132.2
3 400 1000 1000 99.18 – 69.55 99.18
4 1000 900 800 46.00 92.59 130.40 130.4
5 1000 800 300 45.98 84.00 60.00 84
6 1000 300 200 45.98 32.47 33.30 45.98
7 100 800 1000 45.98 84.00 17.00 84
8 100 600 1000 45.98 43.35 16.70 45.98
9 100 200 800 36.61 21.26 16.21 36.61

Fig. 7. (a) Arrangement of 3S configuration (b) power – voltage characteristics different cases of shadow patterns characteristics of BIPVPS under uniform irradiance and
different partial shading patterns for 3S configuration.

are equal to the number of different irradiance levels incident on
the PV modules. Therefore, traditional techniques fail in tracking
global MPP under PSC since they cannot differentiate between glo-
bal and local peaks. To solve this problem MPPT integrated by opti-
mization techniques is required. In sum, the number of MPP over
power–voltage characteristics depends on only the number of dif-
ferent irradiance levels incident on the PV system regardless the
cause of the shadow.
position and global best position. The local best solution, Pbest, is
the best solution recorded by the particle itself. The best solution
proposed through entire population is represented by the global
best solution, Gbest. In this work, the particle swarm position has
been represented with converter duty cycle. On the other side,
PV system output power is taken as the fitness function. The parti-
cle position di is updated by the following relation (Mirhassaniet
et al., 2015).

¼ di þ v kþ1
kþ1 k
di i ð1Þ
4. PSO based MPPT technique
v kþ1
i ¼ wv ki þ c1 r 1 ðPbest
di Þ þ c2 r 2 ðGbest
di Þ
k k
ð2Þ
PSO has been considered one of population-based evolutionary
algorithms which successfully employed for solving different engi- where;
neering optimization problems having multiple peaks. PSO pre-
sents some merits over other techniques employed for extracting vi denotes velocity component (step size) at iteration k+1;
global MPP under PSC (Sarvi et al., 2015). Such merits include; sim- w denotes inertial weight;
ple concept, easy implementation, robustness to control parame- c1, c2 denote acceleration coefficients;
ters. This technique was inspired by bird’s behavior. PSO employs r1, r2 denote random values;
a number of particles that constitute a swarm moving around in Pbest denotes best position of particle i;
the search space looking for the best solution. Each particle adjusts Gbest denotes best position in the entire population.
its travelling speed dynamically corresponding to the flying expe-
riences of itself and its colleagues. The particles present a candi- Considering (2), one may note that, classical PSO completely
date solution by exchange information obtained in their depends upon random values. Accordingly, the velocity equation
respective search process to find the best solution. The movement was modified by Ishaque and Salam (2013) in order to overcome
of the particles during searching process is based on simple rela- this problem. The modified velocity equation can be written as
tion (1). Each particle updates its position based on its current the following;
position, its current velocity, the distance among its current posi-
v kþ1
i ¼ wv ki þ Pbest þ Gbest
2di
k
ð3Þ
tion and local best position and the distance among its current
 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186 177

Fig. 8 shows the flowchart for searching mechanism by value pbest, the new best value of particle i must be updated. After
PSO
based tracker. PSO is initialized as a set of random solutions; that, select the particle that has optimal value among all particles
each particle is considered to be a potential solution (fitness) as Gbest. After evaluation of all particles, the velocity and position of
which moves in the search space following the optimal particle. each particle is modified based on (1) and (3). After finishing all
Firstly appropriate variables have to be selected for the search. iterations, the algorithm stops and gives the optimal solution,
The total number of particles is defined as n, in this work; the par- which represents the optimum value of the duty cycle correspond-
ticles are defined as the duty cycles of boost converter. The initial ing to global MPP (Liu et al., 2012).
generated samples are applied to the PV system and the corre-
sponding system voltage and current has been sensed. These val- 5. Differential evolution based tracker
ues used for calculating PV power that represents fitness
function. If current calculated power is more than stored optimal DE has been suggested by Storn and Price. It is one of the best
population based techniques used for global optimization because
of its simplicity and efficiency. Basically, a population of n particle
Start
is required in DE and little iteration needed to reach the optimal
Initiate a population n of particles, w, d min, solution. DE begins by initial random population and then imple-
d max and iter max ments mutation, crossover, and selection operations for modifying
the population during the evolution toward the optimum solution
i=1 (Salam and Saad, 2013). It implements mutation and crossover
processes for getting a trial vector dui for each target vector. Next,
Output the duty cycle di for converter and measure the a selection process is performed among trial vector and target vec-
corresponding PV voltage and current tor. There are many mutation strategies; the one used in this work
is called DE/best/bin. The process of operation of DE technique can
be summarized as following.
Calculate PV power (PPVi)
The duty cycle of boost converter is used as a target vector and
the PV power as the fitness function. The target vectors are placed
Pbest,i=di
on fixed locations over their corresponding feasible limits [dmin,
Next particle dmax]. The initial population of duty cycles is send to the boost con-
no
i=i+1 i>n verter. After that, the PV system current and voltage correspond to
yes each target vector has been measured. The measured values
iter=1 employed for calculating PV system power Pi that represents the
fitness value. After the initialization step finished, the highest PV
power is selected as Pbest and corresponding duty cycle di is saved
Specify G best
as the best one Dbest. Next, two different population members are
randomly selected. The, a mutation factor F is used to weight the
i=1 difference between the selected target vector and the weighted dif-
ference is added into the best duty cycle to generate mutated par-
Update the particle velocity and ticle or so called donor vector dvi. The mutation process is given by
duty cycle using (1) and (3)
Salam and Saad (2013);
dv i ¼ Dbest þ F  ðdr1
dr2 Þ ð7Þ
di < d min && di >d max no
where indexes r1, and r2 represent mutually different randomly
yes integers; F denotes the scaling factor.
In case if a one element of a mutant vector breaks the boundary
Output the duty cycle di for converter and measure the constraint, the element can be reset based on the following:
corresponding PV voltage and current

dui ¼ dmax ! if dui greater than dmax
dui ¼ ð8Þ
dui ¼ dmin ! if dui smaller than dmin
Calculate PV power
(new PPVi) After the mutation process, the trail vector dui is generated by
making use of a binomial crossover operation on the target vector
yes di and the mutant vector dvi. The condition employed in crossover
new PPVi > old Pbest,i =di process has been explained based on the following;
PPVi


no dv i if rand 6 C r
Next particle dui ¼ ð9Þ
no di
i=i+1 i>n
yes Cr is the crossover control parameter in rang of [0 1].
Next iteration After crossover, the selection stage is reach. It has been per-
no formed via comparing target vector di with trial vector dui. Each
iter>iter max iter=iter+1
trial vector dui is assessed through PV system and Pui have to be
yes
measured. However, the value of dui may remain as di after cross-
Output the duty cycle over and Pui is also equal to Pi. Therefore, only the Pui which does
of the G best not have the same duty cycle as Pi measured again from the input
of boost converter. This process helps to reduce the searching time.
End When all values of the Pui are obtained, it is compared to Pi. From
comparison, the duty cycle which corresponds to higher value of
Fig. 8. Flowchart for searching mechanism by PSO
based tracker. power is used as the next target vector according to the following;
178 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186

Start and cross-pollination are considered the two main types of pollina-
tion. Self-pollination occurs when pollen from one flower fertilize
Initiate a population n of host nests, F, Cr, dmin, same flower. On the other side, cross-pollination takes place when
dmax and itermax pollen grains are transferred from a different plant to flower. From
another point of view, the flower attempt to spread their pollen
i=1 through various methods. The first method is abiotic pollination
where the wind helps the pollen to transfer. The second method
Output the duty cycle di for boost converter and measure is biotic pollination which take place through insects and animals.
the corresponding PV voltage and current According to biological evolution point, the main goal of flowers
pollination is the survival of the fittest. Moreover, the optimum
Calculate PV power (Pi) reproduction of plants in terms of numbers in addition to the fit-
test should be considered as an important target. This may be con-
Next nest no sidered as an optimizing process of plants species (Abdelaziz et al.,
i=i+1 i>n 2016). In order to design and develop the FPA algorithm, the fol-
yes lowing four rules must be taken into account as following:
iter=1
1. Biotic and cross-pollination are regards as a global pollination
Specify Pbest and Dbest process, and pollen-carrying pollinators may fly a long distance
that obeys Lévy flights;
i=1 2. A biotic and self-pollination have been considered as local
pollination;
Create the donor vector dvi using () 3. Flower constancy is regards as an equivalent to a reproduction
probability which is proportionate to the similarity of two flow-
ers involved;
no 4. The switching of local pollination and global pollination has
di <d min && di >d max Limitation
been controlled via switch probability.
yes
Create the donor The mathematical models can be obtained based on previous
vector dui using () rules as fellows;
The global pollination step as first rule and the flower constancy
Output the duty cycle di for boost converter and measure as third rule may be converted in a mathematical equation as fol-
the corresponding PV voltage and current low (Yang, 2012; Abdelaziz et al., 2016).

xtþ1
i ¼ xti þ cLðkÞðg 
xti Þ ð11Þ
Calculate PV power (Pui)
where xtþ1
i denotes pollen i or the solution vector xi at iteration t
yes and g denotes best solution of boost converter duty cycle. c denotes
Pui>Pi di=dui a scaling factor that used for controlling the step size. LðkÞ repre-
no sents Lévy flights-based step size, that corresponds to the strength
Next particle no
i=i+1 i>n of the pollination. Insects can fly over a long distance with different
yes distance steps; this is drawn from a Lévy distribution (Yang, 2012;
Next iteration Abdelaziz et al., 2016).
no
iter>iter max iter=iter+1 kCðkÞ sinðp2kÞ 1
L ðs  s0 > 0Þ ð12Þ
yes p S1þk
Output the duty cycle Here CðkÞ denotes gamma function.
of the D best
For the local pollination, both second and third rules may be
mathematically represented as follow (Abdelaziz et al., 2016).
End
xtþ1
i ¼ xti þ 2 ðxtj
xtk Þ ð13Þ
Fig. 9. Flowchart for searching mechanism by DE-based tracker.
where xtj , xtk are pollen from different flowers of the same plant spe-

cies. If xtj and xtk come from the same species or are taken from the
dui if f ðdui Þ P f ðdi Þ
diþ1 ¼ ð10Þ same population, this equivalently becomes a local random walk if 
di otherwise:
is drawn from a uniform distribution in [0,1] (Yang, 2012; Abdelaziz
The iterations have been continued until a termination criterion et al., 2016).
is achieved. The searching mechanism of DE-based tracker is Flower pollination activities can tack place at all scales, both
shown in Fig. 9. local and global. The switch probability or proximity probability
can be effectively used in order to switch between the local and
global (Yang, 2012; Abdelaziz et al., 2016). For the extraction of
6. Flower pollination algorithm based tracker the optimal duty cycle corresponding to global MPP of PV system
has been occurs. The flow chart of the proposed FPA technique
FPA has been considered a modern population-based optimiza- has been shown in Fig. 10.
tion method (Yang, 2012). Such technique based on behavior of the FPA algorithm was concluded that a population size is n = 3.
flower pollination. As known the pollination is a natural Also, maximum number of iterations is taken as 10. While, the
physiological operation of mating in plants. It associated with probability switch value was p = 0.8 (Yang, 2012; Abdelaziz et al.,
transferring pollen via pollinators like insects. Self-pollination 2016), in this paper the probability switch has been taken as
 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186 179

Start

Initialize flower population n , switching probability

p, d min, d max and max generations iter max

i=1

Output the duty cycle di for boost converter and measure

the corresponding PV voltage and current

Calculate PV power (Pi)

no
i=i+1 i>n
yes
Specify Pbest and Dbest in the initial population

iter=1

Iter<iter max
no
Output the duty cycle
of the D best yes
i=1
End
yes
i>n
no
yes no
Global Pollination Is rand< p Local Pollination

Create the donor vector dvi using ()

no
di <d min && di >d max Limitation

yes
Create the population
vector dui

Output the duty cycle di for boost converter and measure

the corresponding PV voltage and current

Calculate PV power (Pui)

At Pui>Pi Update the solution

in the population di=dui
Next population
i=i+1

Next iteration
Find the current best solution
iter=iter+1
di=dui

Fig. 10. Flowchart for searching mechanism by FPA-based tracker.

p = 0.2. The probability switch is to control the algorithm will do a
global search (cross-pollination) and the other part will do local
search (self-pollination). When p = 0.2, means that 20% of the sim-
ulation the algorithm will do local search, and 80% will do global
search. For The lower and upper bounds [0 1] of the solution vector
are taken as the boundary of the duty cycle.
7. Analysis and discussion
In order to analyze the tracking performances of PFA, PSO and
DE
based trackers and compare between them, extensive MATLAB
simulation studies under different PSC patterns has been per-
formed. The DC/DC boost converter is operating in continuous
180 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186
conducted current mode with switching frequency of 30 kHz. The
input inductance is 1 mH and output capacitor is 47 lF with 60 O
resistive loads. Three different studied global MPPT algorithms
has been employed in simulation as a controller feeding the con-
verter with suitable duty cycle for studying and comparing perfor-
mance of BIPVPS during PSC. The power against voltage (P
V
curves) of BIPVPS under uniform irradiance and three different par-
tial shading patterns are illustrated in Fig. 11(a). Moreover, for the
purpose of validation the three MPPT techniques, initially, the
BIPVPS is operated under uniform condition T = 25 °C,
G = 1000 W/m2 as uniform irradiance in Table 2. Fig. 11(b) shows
the BIPVPS output power for the three different MPPT techniques.
From this figure, it can be seen that, all studied optimization tech-
niques caught the MPP of 1220 W with high tracking efficiency
under uniform irradiance as shown in Fig. 11(b). The detailed sim-
ulation of such case under study has been shown in Fig. 12. The fig-
ure shows the PV power, the duty cycle, the PV voltage and the PV
current for the three techniques DE, FPA and PSO based trackers.
Based on Fig. 12, the convergence time for DE is 0.30 s, while it is
0.20 s for FPA and it is 0.38 s for PSO. This means that using FPA
based tracker decreases the tracking time by 33.33% compared
with DE and by 47.36% compared with PSO.
The obtained simulation results for the second PSC pattern #2
are illustrated in Fig. 11(b). Under this case of study, there are
two peaks and at the second one on P
V curve, the global MPP
of 755.4 W is located. Using the P–V characteristics of the PV mod-
ule under study under partial shading as pattern #2 in Table 1, it is
easy to check the matching between MPP corresponding to the
irradiance level and those results are obtained by the DE, FPA
and PSO
based MPPT algorithm. The PV power is almost very close
to the maximum power produced by the PV module. However the
FPA based tracker has the priority of the speed to reach the MPP
than the DE and PSO based trackers. The effectivity of the FPA
Fig. 11. (a) P
V curve (b) the BIPVPS output power for PSO, FLC and DE techniques under uniform irradiance, second, third and forth partial shading patterns.
mean to save the cost of implementation and process time. The
detailed simulation results for BIPVPS with different MPPT tech-
niques under the second shading pattern have been illustrated in
Fig. 13. The analysis of the simulation results depends on conver-
gence time to catch the global MPP and also on the tracking effi-
ciency. As shown in Fig. 13, the convergence time for DE is
0.38 s, while it is 0.34 s for FPA and it is 0.485 s for PSO. This means
that using FPA based tracker decreases the tracking time by 10.52%
compared with DE and by 30% compared with PSO. Also it can be
noted that, the three optimization based trackers have the ability
to catch the global MPP with high tracking efficiency.
To assess the robustness of the proposed algorithm in seeking
the global MPP, BIPVPS has been subjected to another non-
uniform irradiance pattern as #3 in Table 2 and the temperature
is kept constant at 25 °C. As shown from Fig. 11(a), the global
MPP of 486.8 W is the located at the third point on P–V curve from
four points of peaks as shown from Fig. 11(b). It is clearly observed;
that the DE, FPA and PSO based MPPT is not trapped by the local
maximum power of the P–V characteristics and has converged
accurately to the global MPP. From Fig. 11(b), one can derive that,
three optimization techniques FPA, DE and PSO can ignore the first
and second local MPP and have an excellent capability to achieve
global MPP with very high tracking efficiency. The detailed simula-
tion results for BIPVPS with the three global MPPT techniques
under second shading pattern are shown in Fig. 14. As shown in
Fig. 14, the convergence time for DE is 0.41 s, while it is 0.39 s
for FPA and it is 0.41 s for PSO. This means that using FPA based
tracker decreases the tracking time by 4.87% compared with DE
and PSO. Also it can be noted that, the three optimization based
trackers have the ability to catch the global MPP with high tracking
efficiency.
Another case of study has been taken in order to validate the
MPPT-trackers. In this case, the system is tested with the forth
 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186 181

studied partial shading pattern Table 2; pattern No. 4. As shown
from Fig. 11, the global MPP of 456.1 W is the located at the second
point on P–V curve from four points of peaks. From Fig. 11(b), the
three optimization techniques FPA, DE and PSO based trackers can
ignore the first local MPP and have an excellent capability to
achieve global MPP with very high tracking efficiency.
The detailed simulation results for PV system with different
MPPT techniques under the second shading pattern are shown in
Fig. 15. As shown in Fig. 15, the convergence time for DE is
0.42 s, while it is 0.36 s for FPA and it is 0.40 s for PSO. This means
that using FPA based tracker decreases the tracking time by 14.28%
compared with DE and by 10% compared with PSO. Also it can be

Fig. 12. Detailed simulation results for BIPVPS under uniform radiation (a) DE
based tracker (b) FPA
based tracker and (c) PSO
based tracker.

Fig. 13. Detailed simulation results for BIPVPS system under second shading pattern #2 (a) DE
based tracker (b) FPA
based tracker and (c) PSO
based tracker.
182 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186

noted that, the three optimization based trackers have the ability
to catch the global MPP with high tracking efficiency.
8. Comparative study
For achieving the fairness of comparison, the initial population
has been identical for FPA, DE and PSO algorithms under all studied
cases. The performance of the FPA and DE–based trackers is the
best copamred with PSO based tracker in terms of tracking time
and number of iterations for all different studied shading patterns.
The updating process of the duty cycle for all iterations is illus-
trated in Fig. 16. A comparison between different studied MPPT
techniques is listed in Table 4. From this table, one must see the
superiority of FPA and DE–based trackers over the other technique.

Fig. 14. Detailed simulation results for BIPVPS system under third shading pattern #3 (a) DE
based tracker (b) FPA
based tracker and (c) PSO
based tracker.

Fig. 15. Detailed simulation results for BIPVPS system under third shading pattern #4 (a) DE
based tracker (b) FPA
based tracker and (c) PSO
based tracker.
 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186 183

Pattern #2 Pattern #3 Pattern #4

DE DE DE
1 1 1

Duty cycle

Duty cycle

Duty cycle
0.5 0.5 0.5

0 0 0
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
FPA FPA FPA
1 1 1
Duty cycle

Duty cycle

Duty cycle
0.5 0.5 0.5

0 0 0
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
PSO PSO PSO
1 1 1
Duty cycle

Duty cycle

Duty cycle
0.5 0.5 0.5

0 0 0
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
Number of Iterations Number of Iterations Number of Iterations
Fig. 16. Update process of particles and learners positions.

the FPA-based tracker has been worked will and with higher speed Table 5
tracking. However, the three trackers have the ability for tracking Quality indexes employed for evaluating the performance of FPA.

global MPP of BIPVPS. Metric Abbreviation Formula

The FPA algorithm has main merits over other techniques such Pnr
Relative error RE P
PPVt
i¼1 PVe;i
 100%
as its investigation via inclusive pollination along with manipula- Pnr
P PVt
Mean absolute error MAE jPPVe;i
PPVt j
tion through local pollination for keeping alive the tracking pro- i¼1
nr
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pnr
cess. In the other hand, many researchers try to improve the Root mean square error RMSE 2
ðPPVe;i
PPVt Þ
i¼1
tracking speed of other techniques by construction of random duty rP
nr
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Standard deviation STD nr
 PVe Þ2
cycles with increasing the complexity. Moreover, the FPA has only i¼1
ðPPVe;i
P
nr
two parameters, but other techniques use more number of param- Efficiency – PPVe;i
 100%
eters leading to increase the computational burden. P PVt

Table 4
Comparison of different studied MPPT techniques.

Shading Pattern Technique Power, W Tracking Speed Global Power W

sec.
Pattern #1 DE 1220 0.30 1220
FPA 1220 0.20
PSO 1220 0.38
Pattern #2 DE 755.4 0.38 755.4
FPA 755.4 0.36
PSO 755.4 0.485
Pattern #3 DE 486.8 0.41 486.8
FPA 486.8 0.39
PSO 486.8 0.41
Pattern #4 DE 456.1 0.42 456.1
FPA 456.1 0.30
PSO 456.1 0.34
Parameter DE FPA PSO
Tracking speed Fast Very fast Moderate
Efficiency High High High
Computational Complexity Moderate Less High
Exploration process No Yes Yes
Exploitation Process No Yes No
Number of tuning parameters 2 2 5
184 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186

Based on the previous discussion, the advantages of FPA algo-
rithm may be written as: Robustness, Reliability, Simplicity of its
structure and programming, Effectivity to catch the global MPP
under uniform or partial shaded patterns and high convergence
speed. FPA based tracker is far superior to the PSO and DE in terms
of tracking speed, the tracking time decreased with average
17.218% compared with DE 22.6% compared with PSO in all differ-
ent studied partial shading patterns of 4S3P configuration
In order to measure the performance of the FPA algorithm, it is
necessary to define a set of metrics that can be useful for the eval-
uation. Table 5 enlists different functions that measure the perfor-
mance quality of each optimization technique. Such metrics
include; the Relative Error RE, Mean Absolute Error (MAE), Root
Mean Square Error (RMSE), standard deviation (STD) and Median.
These metrics measures the values of the true PV power (PPVt)
obtained from the Simulink model illustrated in Fig. 6 and the

Fig. 17. Simulink model of the complete PV system integrated with global MPPT for 3S configuration.

Table 6
Performance evaluation of DE under different cases of shadow patterns for 3S configuration.

Case Global power, Best power, Worst power, Average Median, STD Successful rate Average RE MAE RMSE Efficiency
W W W power W (%)
1 152.38 152.38 152.38 152.38 152.38 0.00 100 0.00 0.00 0.00 100
2 132.2 132.2 99.10 128.892667 132.2 9.922 90 2.501765 3.30733 10.4587 97.49823
3 99.18 99.18 67.18 95.9426667 99.18 9.202 86.66667 3.26409894 3.23733 9.75484 96.7359
4 130.4 130.4 90.55 126.311333 130.4 11.6676 83.33333 3.13548057 4.08867 12.3633 96.86452
5 84.00 84.00 59.43 82.381 84.00 6.0582 93.33333 1.92738095 1.619 6.27079 98.07262
6 45.98 45.98 32.15 44.9513333 45.98 3.4201 90 2.23720458 1.02867 3.5714 97.7628
7 84.00 84.00 43.59 81.2406667 84.00 9.8496 86.66667 3.28492063 2.75933 10.2288 96.71508
8 45.98 45.98 42.98 45.6926667 45.98 0.8625 90 0.62490938 0.28733 0.90908 99.37509
9 36.61 36.61 20.98 34.5436667 36.61 5.2683 86.66667 5.64417737 2.06633 5.65904 94.35582

Table 7
Performance evaluation of FPA under different cases of shadow patterns for 3S configuration.

Case Global power, Best power, Worst power, Average Median, STD Successful rate Average RE MAE RMSE Efficiency
W W W power W (%)
1 152.38 152.38 152.38 152.38 152.38 0.00 100 0.00 0.00 0.00 100
2 132.2 132.2 132.20 132.20 132.2 0.00 100 0.00 0.00 0.00 100
3 99.18 99.18 69.55 98.1923333 99.18 5.31875 96.66667 0.99583249 0.98767 5.40967 99.00417
4 130.4 130.4 92.55 127.876667 130.4 9.4414 93.33333 1.93507157 2.52333 9.77283 98.06493
5 84.00 84.00 84.00 84.00 84.00 0.00 100 0.00 0.00 0.00 100
6 45.98 45.98 33.27 45.5563333 45.98 2.2815 96.66667 0.92141511 0.42367 2.32052 99.07858
7 84.00 84.00 84.00 84.00 84.00 0.00 100 0.00 0.00 0.00 100
8 45.98 45.98 43.33 45.8916667 45.98 0.4757 96.66667 0.19211251 0.08833 0.48382 99.80789
9 36.61 36.61 36.61 36.61 36.61 0.00 100 0.00 0.00 0.00 100
 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186 185

Table 8
Performance evaluation of PSO under different cases of shadow patterns for 3S configuration.

Case Global Best Worst Average Median, STD Successful Average RE MAE RMSE Efficiency
power, W power, W power, W power W rate (%)
1 152.38 152.38 152.38 152.38 152.38 0.00 100 0.00 0.00 0.00 100
2 132.2 132.2 98 127.7533333 132.2 11.3382 86.66666667 3.363590519 4.446666667 12.17902131 96.63640948
3 99.18 99.18 68.6 94.01533333 99.18 11.1045 76.66666667 5.207367077 5.164666667 12.24678298 94.79263292
4 130.4 130.4 91.89 123.667 130.4 14.0648 73.33333333 5.163343558 6.733 15.59334206 94.83665644
5 84.00 84.00 45.71 79.81900 84.00 10.9518 86.66666667 4.977380952 4.181 11.72270262 95.02261905
6 45.98 45.98 30.18 44.22100 45.98 4.628 83.33333333 3.825576338 1.759 4.951013028 96.17442366
7 84.00 84.00 45.15 79.97233333 84.00 11.5312 80.0000000 4.79484127 4.027666667 12.21437637 95.20515873
8 45.98 45.98 16.58 43.76766667 45.98 7.3235 86.66666667 4.811512252 2.212333333 7.650352715 95.18848775
9 36.61 36.61 21.05 34.36333333 36.61 5.2198 76.66666667 6.136756806 2.246666667 5.682810924 93.86324319

Table 9
Comparison results between DE, FPA and PSO for 3S configuration.

Algorithm STD Successful rate (%) Average RE MAE RMSE Efficiency

DE 6.250033333 89.62962963 2.513326382 2.043778 6.579550872 97.48667362
FPA 1.946371975 98.14814815 0.449381299 0.44700 1.998537844 99.5506187
PSO 8.462422222 83.33333333 4.253374308 3.41900 9.137822444 95.74662569

Fig. 18. Comparison results between DE, FPA and PSO for 3S configuration.

estimated PV power (PPVe) obtained by the optimization algo- 9. Conclusion

rithms as illustrated in Fig. 17. In Table 5, nr is the number of data-
set groups (Runs of Simulink model). The parameters of each
algorithm are set as in the original references except the number
of iterations (10) and the size of population (3). As well as, each
algorithm has been executed 30 times.
Tables 6–8 show the performance evaluation for FPA, DE and
PSO respectively for different shadow cases of 3S configuration.
The comparison results between the algorithms are given in Table 9
and Fig. 18, which shows the RE, successful rate, MAE, RMSE, STD
and the efficiency. From this table, it is observed that FPA has the
lower RMSE and also the STD demonstrates that the results doesn’t
change along the iterative process, which indicated its stability
compared with DE and PSO. Employing FPA decreases the STD by
68.85% and 77% compared with DE and PSO respectively. Whereas
the RMSE is decreased by 69.15% and 78.13% compared with DE
and PSO respectively. The successful rate can be defined as the per-
centage of number of reaching correct global MPP into the total
number of iterations. Using FPA increased the successful rate by
9.5% and 17.78% compared with DE and PSO respectively.
This work presents a soft computing global MPPT technique
based on three different optimization techniques to improve the
efficiency of partially shaded BIPVPS system on which the power
against voltage curve exhibit multiple local peaks and only one glo-
bal MPP. A comprehensive assessment of the FPA is carried out
against DE and PSO with two different PV array configurations.
The first PV configuration (4S3P configuration) comprises of 12
panels, configured by three modules per string. The second one
(3S configuration) consists of three modules in series. The used
BIPVPS comprises global MPPT algorithms composed of the boost
converter and PV array have been simulated in MATLAB/Simulink
software for verifying tracking performance of studied algorithms.
The results clearly demonstrate that proposed trackers have high
accuracy and stability in extracting the global MPP in all the tested
cases no matter where the global MPP locates. For 4S3P configura-
tion; it is derived that FPA based tracker is far superior to the PSO
and DE in terms of tracking speed, the tracking time decreased
with average 17.218% compared with DE 22.6% compared with
186 A.A. Zaki Diab, H. Rezk / Solar Energy 157 (2017) 171–186
PSO in all different studied partial shading patterns. An extensive
statistical analysis are carried out with 3S configuration and the
results indicated that; employing FPA decreases the STD by
68.85% and 77% compared with DE and PSO respectively. Whereas
the RMSE is decreased by 69.15% and 78.13% compared with DE
and PSO respectively. In addition FPA increased the successful rate
by 9.5% and 17.78% compared with DE and PSO respectively.
References
Abdelaziz, A.Y., Ali, E.S., Abd Elazim, S.M., 2016. Optimal sizing and locations of
capacitors in radial distribution systems via flower pollination optimization
algorithm and power loss index. Int. J. Eng., Sci. Technol. 19 (1), 610–618.
Ahmed, J., Salam, Z., 2014. A maximum power point tracking (MPPT) for PV system
using Cuckoo Search with partial shading capability. Appl. Energy 119, 118–
130.
Ahmed, J., Salam, Z., 2015. A critical evaluation on maximum power point tracking
methods for partial shading in PV systems. Renew. Sustain. Energy Rev. 47,
933–953.
Bhandari, K.P., Collier, J.M., Ellingson, R.J., Apul, D.S., 2015. Energy payback time
(EPBT) and energy return on energy invested (EROI) of solar photovoltaic
systems: A systematic review and meta-analysis. Renew. Sustain. Energy Rev.
47, 133–141.
Chao, K.H., Lin, Y.S., Lai, U.D., 2015. Improved particle swarm optimization for
maximum power point tracking in photovoltaic module arrays. Appl. Energy
158, 609–618.
Chen, K., Tian, S., Cheng, Y., Bai, L., 2014. An improved MPPT controller for
photovoltaic system under partial shading condition. IEEE Transaction
Sustainable Energy 5 (3), 978–985.
Daraban, S., Petreus, D., Morel, C., 2013. A novel global MPPT based on genetic
algorithms for photovoltaic systems under the influence of partial shading. In:
IECON 2013–39th Annual Conference of the IEEE, 10–13 Nov. 2013, pp. 1490–
1495, 2013.
Dileep, G., Singh, S.N., 2017. Application of soft computing techniques for maximum
power point tracking of SPV system. Sol. Energy 141, 182–202.
Fathy, A., Rezk, H., 2016. A novel methodology for simulating maximum power
point trackers using mine blast optimization and teaching learning based
optimization algorithms for partially shaded photovoltaic system. J. Renew.
Sustain. Energy 8, 023503. http://dx.doi.org/10.1063/1.4944971.
Fathy, A., Rezk, H., 2017. Parameter estimation of photovoltaic system using
imperialist competitive algorithm. Renewable Energy. Available online 12 April.
Green, M.A., Emery, K., Hishikawa, Y., Warta, W., Dunlop, E.D., 2016. Solar cell
efficiency tables (Version 48). Prog. Photovoltaics Res. Appl. 24, 905–913.
Ishaque, K., Salam, Z., 2013. A deterministic particle swarm optimization maximum
power point tracker for photovoltaic system under partial shading condition.
IEEE Transaction Ind. Electron. 60 (8), 3195–3205.
Jiang, L., Maskell, D., Patra, J., 2013. A novel ant colony optimization-based
maximum power point tracking for photovoltaic systems under partially
shaded conditions. Energy Buildings 58, 227–236.
Karlis, A.D., Kottas, T.L., Boutalisb, Y.S., 2007. A novel maximum power point
tracking method for PV systems using fuzzy cognitive networks (FCN). Electric
Power Syst. Res. 3–4, 315–327.
Koad, Ramdan B.A., Zobaa, Ahmed F., 2017. A novel MPPT algorithm based on
particle swarm optimisation for photovoltaic systems. IEEE Transactions
Sustain. Energy 8 (2), 468–476. http://dx.doi.org/10.1109/TSTE.2016.2606421.
Kofinas, P., Dounis, A.I., Papadakis, G., Assimakopoulos, M.N., 2015. An Intelligent
MPPT controller based on direct neural control for partially shaded PV system.
Energy Build 90, 51–64.
Kumar, N., Hussain, B., Singh, B.K., Panigrahi, 2017. Rapid MPPT for uniformly and
partial shaded PV system by using JayaDE algorithm in highly fluctuating
atmospheric conditions. IEEE Transactions Ind. Informatics, doi: http://dx.doi.
org/10.1109/TII.2017.2700327.
Kumar, N., Hussain, I., Singh, B., Panigrahi, B.K., 2017. MPPT in dynamic condition of
partially shaded PV system by using WODE technique. IEEE Transactions
Sustain. Energy. http://dx.doi.org/10.1109/TSTE.2017.2669525.
Liu, Y. et al., 2012. A particle swarm optimization based maximum power point
tracking algorithm for PV systems operating under partially shaded conditions.
IEEE Transaction Energy Convers. 27 (4), 1027–1035.
Liu, C. et al., 2012. A PSO-based MPPT algorithm for photovoltaic systems subject to
inhomogeneous insolation. In: 6th International Conference on Soft Computing
and Intelligent Systems (SCIS) and 13th International Symposium on Advanced
Intelligent Systems (ISIS), pp. 721–726.
Ma, J. et al., 2013. Low–cost global MPPT scheme for photovoltaic systems under
partially shaded conditions. IEEE Int. Symp. Circuits Syst. (ISCAS), 245–248.
Mao, M., Zhang, Li, Duan, Q., Oghorada, O.J.K., Duan, P., 2017. A two-stage particle
swarm optimization algorithm for MPPT of partially shaded PV arrays. Int. J.
Green Energy. http://dx.doi.org/10.1080/15435075.2017.1324792.
Mirhassaniet, S.M. et al., 2015. An improved particle swarm optimization based
maximum power pointtracking strategy with variable sampling time. Electrical
Power Energy Syst. 64, 761–770.
Murphy, F., McDonnell, K., 2017. A feasibility assessment of photovoltaic power
systems in Ireland; a case study for the Dublin region. Sustainability 9, 302.
Phimmasone, V., Kondo, Y., Kamejima, T., Miyatake, M., 2010. Evaluation of
extracted energy from PV with PSO–based MPPT against various types of
solar irradiation changes. In: IEEE International Conf. on Electrical Machines
and Systems, Korea, pp. 487–492.
Rahmani, Rasoul, Yusof, Rubiyah, December 2014. A new simple, fast and efficient
algorithm for global optimization over continuous search-space problems:
Radial movement optimization. Appl. Math. Comput. 248 (1), 287–300. http://
dx.doi.org/10.1016/j.amc.2014.09.102.
Rezk, H., 2016. Performance of incremental resistance MPPT based proton exchange
membrane fuel cell power system. In: 2016 Eighteenth International Middle
East on Power Systems Conference (MEPCON), 27–29 Dec. 2016.
Rezk, H., 2016. A comprehensive sizing methodology for stand-alone battery-less
photovoltaic water pumping system under the Egyptian climate. J. Cogent Eng.
3.
Rezk, H., Dousoky, G.M., 2016. Technical and economic analysis of different
configurations of stand-alone hybrid renewable power systems – A case study.
Renew. Sustain. Energy Rev. 62, 941–953.
Rezk, H., Eltamaly, A.M., 2015. A comprehensive comparison of different MPPT
techniques for photovoltaic systems. Sol. Energy 112, 1–11.
Rezk, H., Fathy, A., Oct 2016. Simulation of global MPPT based on teaching–learning-
based optimization technique for partially shaded PV system. Electr. Eng. http://
dx.doi.org/10.1007/s00202-016-0449-3.
Rezk, H., Hasaneen, El-Sayed, 2015. A new MATLAB/Simulink model of triple-
junction solar cell and MPPT based on artificial neural networks for
photovoltaic energy systems. Ain Shams Eng. J. 6 (3), 873–881. http://dx.doi.
org/10.1016/j.asej.2015.03.001.
Rezk, H., Fathy, A., Abdelaziz, A.Y., July 2017. A comparison of different global MPPT
techniques based on meta-heuristic algorithms for photovoltaic system
subjected to partial shading conditions. Renew. Sustain. Energy Rev. 74, 377–
386. http://dx.doi.org/10.1016/j.rser.2017.02.051.
Salah, C.B., Chaabene, M., Ammar, M.B., 2008. Multi-criteria fuzzy algorithm for
energy management of a domestic photovoltaic panel. Renewable Energy 33,
993–1001.
Salam, Z., Saad, M.S., 2013. Evolutionary based maximum power point tracking
technique using differential evolution algorithm. Energy Building 67, 245–252.
Salam, Z., Ramli, Z., Ahmed, J., Amjad, M., 2015. Partial shading in building
integrated pv system: causes, effects and mitigating techniques. Int. J. Power
Electron. Drive Syst. 6 (4), 712–722.
Sarvi, M., Ahmadi, S., Abdi, S., 2015. A PSO-based maximum power point tracking
for photovoltaic systems under environmental and partially shaded conditions.
Prog. Photovolt: Res. Appl. 23, 201–214.
Seyedmahmoudian, M. et al., 2016. Efficient photovoltaic system maximum power
point tracking using a new technique. Energies 9, 147. http://dx.doi.org/
10.3390/en9030147. 2 March 2016.
Sridhar, R., Jeevananthan, S., Dash, S.S., Vishnuram, P., 2016. A new maximum
power tracking in PV system during partially shaded conditions based on
shuffled frog leap algorithm. J. Exp. Theor. Artif. Intell. http://dx.doi.org/
10.1080/0952813X.2016.1186750.
Sundareswaran, K. et al., 2015. Development of an improved P&O algorithm assisted
through a colony of foraging ants for MPPT in PV system. IEEE Trans. Industr.
Inf. http://dx.doi.org/10.1109/TII.2015.2502428.
Tey, K.S., Mekhilef, S., Yang, H., Chuang, M., 2014. A differential evolution based
MPPT method for photovoltaic modules under partial shading conditions. Int. J.
Photoenergy 2014, 1–10.
Yang, X.S., 2012. Flower pollination algorithm for global optimization
unconventional computation and natural computation. Lect. Notes Comput.
Sci. 45, 240–249.