PSO and GA-based maximum power point tracking

for partially shaded photovoltaic systems

Afef Badis, Mohamed Nejib Mansouri Anis Sakly
Electronics and Microelectronics Laboratory (EµE) Industrial Systems Study and Renewable Energy (ESIER)
Faculty of Sciences, University of Monastir The National Engineering School of Monastir (ENIM),
Monastir, Tunisia University of Monastir
Monastir, Tunisia

Abstract— Under partial shading (PS) conditions, particular type of MPPT gaining visibility in the literature is
photovoltaic (PV) systems are popularly known to suffer from soft-computing based MPPT algorithms, such as Particle
low-energy efficiency. Therefore, an effective MPPT algorithm Swarm Optimization (PSO) [7], [8]. It has proven good
should be used to detect the unique global peak as the maximum
power point (MPP), and avoid any local maxima in order to performance and reduced steady-state oscillations under
mitigate the effect of PS. To date, various MPPT techniques have various shading shapes. Another work that combines PSO
been developed to reliably track the MPP under all with a conventional method, Incremental Conductance
circumstances and reduce the energy losses due to PS. Usually, (IncCond), is reported in [7]. In the above case, PSO is
conventional methods such as Perturb and Observe (P&O) and activated only under PS. Otherwise, the conventional MPPT
the Incremental Conductance (IncCond), fail to extract the global tracks the MPP. This technique guarantees convergence to the
MPP of the PV panel if the PV generator is partially shaded. To
overcome this problem, Evolutionary Algorithms (AEs), namely global MPP but the use of six particles increases the tracking
the Particle Swarm Optimization (PSO) and Genetic Algorithm time and complexity. References [10] and [11] combine fuzzy
(GA) are studied, simulated and compared under the same logic MPPT with P&O. The whole P-V curve is scanned and
software. the global MPP is stored by the fuzzy logic while perturbing
and observing with exhibiting a fast converging speed and
Keywords— Evolutionary Algorithms, Maximum power point
small steady-state oscillations. Genetic Algorithm (GA) was
tracking, Partial shading conditions, Photovoltaic system.
proposed to show that P&O and IncCond are trapped at the
I. INTRODUCTION local peak when GA reaches the global MPP successfully in

M aximum power point (MPPT) is an integral part of [6], [12].

grid connected photovoltaic (PV) systems. Nowadays, In this paper, some of the most prolifically published
several (MPPT) methods which are based on hill climbing, methods such as PSO and GA will be discussed. We intend to
have been reported such as the Perturb and Observe (P&O) implement and compare both of them during PS conditions in
[1]-[3]. order to evaluate them. Simulations were done with two series
Normally, with no shading, and with a single power peak in standard PVG type KC200GT where the first is totally sunny,
the PV characteristic, it is simple to track the maximum power and the other is under shade.
using the conventional MPPT methods and usually they can The remainder of the paper is organized as follows. In
achieve 99% or higher tracking efficiency [1], [4]. However, Section II, the circuit model of the PV cell is briefly described
when the PV array experiences partial shading (PS) and the characteristics of the PV array under PS conditions are
conditions, extracting the maximum becomes a challenging represented. In section III, the MPPT algorithms, used to
task because of the non-linearity of the Power-Voltage (P-V)
curve, which exhibits multiple local peaks. Thus, it is
imperative to use a suitable MPPT technique which tracks the
unique global maximum point (MPP) of the shaded PV array
effectively, fast and smoothly [5]. In the latter case,
conventional MPPT techniques become ineffective and
inaccurate because they cannot discriminate between global
and local peak and they can be trapped on the first local peak
[6], [17].
The P-V characteristic of PV modules is shown in Fig. 1,
where the global and the local MPPs are indicated. There is a Fig. 1. Example of the Power-Voltage characteristics of a PV array
large amount of publications targeting this issue, and a under partial shading conditions.

978-1-4673-9768-1/16/$31.00 ©2016 IEEE

extract the maximum power from the PV source partially continuously the power converter operation under randomly
shaded, are discussed. Section IV closes the paper with changing weather conditions [13].
simulation results which are presented and discussed. Finally, TABLE I. PARAMETERS OF SINGLE PV MODULE USED.
the conclusion and scope of future work are presented in
Section V. Parameters Value
Maximum power (Pm) 200.1 W
II. MODELING OF THE PHOTOVOLTAIC SYSTEM UNDER PS Nominal open circuit voltage (Voc) 32.9 V
CONDITIONS Maximum power voltage (Vm) 26.3 V
Nominal short circuit current (Iscn) 8.21 A
A. Photovoltaic generator modeling Maximum power current (Im) 7.61 A
Ns 54
Among numerous models of PV arrays available in the Rs 0.001 Ω
literature, the single-diode five-parameter model, as depicted in Rsh 5 Ω
the Fig. 2, is the most commonly adopted method since it
overcomes the long computational time and promotes a III. MPPT CONTROL FOR PV ARRAY OPERATING UNDER
compromise between effectiveness and simplicity [5]. PARTIAL SHADING CONDITIONS
A. P&O Algorithm
The P&O is the most widely applied method due to its
simplicity and generic nature [3]. As its name indicates, P&O
works by perturbing the PV voltage (or PV current)
periodically and observes the PV power to increase or to
decrease. If the perturbation produces the increase of the PV
power, the direction of the perturbation is maintained
(meaning that the array’s operating point makes towards the
Fig. 2. Single-diode model of a PV module with Ns series cells. MPP). Otherwise, when the PV power decreases, the system is
moving away from the MPP. Then, the direction of the
The PV model is mathematically modeled using (1) shown perturbation is reversed. The derivative of power in function
below: of voltage is zero when the maximum power point (MPP) is
reached [3]. In spite of its low equipment demand and its
 V pv + Rsg I pv  uncomplicated structure, picking up the MPPT by using P&O
  V pv + Rsg I pv
I =I − I e N sVT −1 − (1) becomes ineffective when the amount of the solar irradiation
pv ph 0  Rsgh
 incident on one or more modules of the PV array is unequal
 
since the P-V curve becomes distorted, exhibiting multiple
where Rshg = N s ⋅Rsh and Rsg = Ns ⋅Rs peaks (one global and several local) [5].
Vpv module voltage of PV cell (V) B. Evolutionary Algorithms
Ipv output current of PV cell (A) Since the P&O algorithm fails to optimally harvest the
Iph photo current at short circuit (A)
I0 diode reverse saturation current (A)
Rsg series resistance of the module (Ω)
Rshg shunt resistance of the module (Ω)
Rs series resistance of the cell (Ω)
Rsh shunt resistance of the cell (Ω)
VT thermal voltage (KT/q)
N diode ideality factor
K Boltzman’s constant (Joules/Kelvin)
T temperature (Kelvin)
Ns number of cells in series
Np number of cells in parallel
G1 irradiation of totally illuminated cells (W/m2)
G2 irradiation of shaded illumination cells (W/m2)
The characteristics of the solar module are enumerated in
Table I [6].
Fig. 3 depicts the series association of two PV modules
under Matlab/SimPowersystem software when one of them is
partially shaded. Fig. 3. MATLAB equivalent circuit of the series association of two
An accurate MPPT process is required in order to adjust PVG.
maximum power generated by the PV arrays, its usefulness Invented by James Kennedy and Russell Eberhart in 1995,
diminishes rapidly under PS conditions [14]. However, the PSO algorithm is inspired by the behavior of bird and fish
multi-extreme optimization problem can be solved by using flocks [16]. Over a number of iterations, a swarm of particles
Evolutionary Algorithms (EAs) which are characterized by (each of them represents a candidate solution) follows a
their good robustness and fast computation capability. Among simple behavior: the position of each particle is adjusted
the EA methods, PSO and GA algorithms are very suited for according to the best position of the neighboring particle as
MPPT process due to their simple structures, easy well as the best position in the whole population whose value
implementation and their ability to locate the global peak of is the closest to the target (fitness function) [14]. Fig. 4 shows
the PV curve regardless of environmental variations [14]. the potential movement of particles in the optimization
problem [14]. In this method, each individual keeps tracking
1) Genetic Algorithm
the two best current positions, moves gradually to better
Originally introduced by Holland [15], GA is an
region with a velocity, which determines the direction and the
optimization method which resembles biological evolutionary
step of search. Finally, it reaches the best position of the whole
process.
search space [9].
Through this method, a number of candidate solutions
called chromosomes is assessed. Then, they undergo
recombination or are passed through a mutation process.
Finally, each member of the population is tested for a fitness
function which is used to reflect the goodness of each of them
[6].
GA relies on three fundamental operators namely the
selection, crossover and mutation. In the selection process,
chromosomes with higher fitness values are likely to be
chosen for inclusion in the next generation's population. The
crossover operator combines two parents in order to produce
offspring by means of (2) and (3):
child ( k )=r⋅ parent ( k )+(1−r )⋅ parent ( k +1) (2)
child ( k +1)=(1−r )⋅ parent ( k )+ r⋅ parent ( k +1) (3) Fig. 4. Movement of particles in the optimization algorithm.
with: r is a random number r∈[ 0,1] .
The velocity and position updating equations for the PSO
In order to get a faster convergence, mutation operator is process are calculated as follows [9]:
used. It maintains genetic diversity from one generation to the
next and attempts to achieve some stochastic variability of GA φi = wφi + c1 r1 ( pbest i − xi ) + c 2 r2 ( gbest − xi ) (6)
[6]. xi = xi + φ i (7)
However, the mutation operator is not considered in this
MPPT algorithm regardless its impact on the response where w is the inertia weight, c1 and c2 are the cognitive
convergence and reducing oscillations. In this case, and social learning rate respectively, r1, r2 ∈ [0,1], Pbesti is the
reinitialization of the GA process is imperative in order to personal best position of the i th particle, and Gbest is the global
seek for the new MPP after every solar irradiation change. In best position of the group.
fact, if the following conditions are satisfied, the initial
3) Application of PSO for MPPT
population is reset whenever a solar irradiation is detected:
Identically to the GA method, only one current sensor Ipv is
V ( k +1)−V ( k ) <ΔV (4) used from the PV modules in order to apply the PSO process
in MPPT. Typically for any EA, the initial population is made
Ppv ( k +1)− Ppv ( k ) up of Np individuals randomly generated, which are applied
>ΔP (5) one after another.
Ppv ( k )
In the present work, 4 individuals are considered. The initial
The parameters used are: position of this population are grouped and represented by the
4-dimensional vector below:
Number of population N p =4 ;
x=[ 0.2 0.4 0.6 0.8]∗2Voc (8)
Crossover probability: 0.9;
The fitness function to be maximized presents the generated
ΔV =0.4 and ΔP = 0.15 . PV power. It is developed as in (9):
2) Proposed MPPT algorithm based on PSO
 Fitness =V pv *I pv (9)

Initially, each member of the initial population is evaluated

using the fitness function which is sorted decreasingly. Then,
it undergoes changes as the PSO process progresses by
updating the the perturbation and the duty cycle using (6) and
(7). The relation between the output voltage and the duty ratio
of the boost converter is written as:
xi
α (i ) = 1 − (10)
2Voc

Fig. 6. MATLAB/Simulink simulation model for the PV system.

Then, the algorithm transmits four updated duty cycles α (i )

(i = 1, 2, 3, 4) to the power converter in a sequential order and
evaluates the power.
Those steps are repeated until the final solution is obtained
and the maximum power is reached and PSO keeps tracking
the MPP even if a sudden shadow occurs. Fig. 5 shows the
complete flowchart of the proposed method.

IV. SIMULATION RESULTS

Fig. 6 depicts the Matlab/Simpowersystem model of the PV
system used in this work. Two series PV arrays are connected
to a resistive load via a Dc/Dc boost converter. The converter
is designed with the following specification:
L = 18mH C1 = 470 μ F C2 = 220μ F R = 50Ω

The choice of the sampling time for the MPPT controller is

very critical since at each iteration of the program, four new
particles are considered. In this case, the change in the duty
ratio from the previous one should be small. Thus, due to this
choice, when the duty cycle is slightly modified, the particles
will have to search a reduced area of the PV curve and the
MPP will be tracked with diminished fluctuations in the
operating point. Hence, some amount of energy will be gained
during the searching process. In this study, it is equal to 20µs
when the switching frequency is chosen equal to 20 kHz.
To evaluate the performance of the proposed PSO, the
different input data, parameters adopted for the PSO algorithm
are: Np=4, c1=c2=2.05, w =0.73. Then, a comparison is made
with the GA. A very challenging scenario is imposed to the

Fig. 5. Flowchart of the proposed PSO method. Fig. 7. Illumination variation of the first PV generator G1.
system: it is exposed to partial shadow.
At t=0.25s, partial shading occurs in G1 while the
temperature is maintained constant at 25°C. The scenario of
the illumination variation, given through Fig.7, can be divided
into four sequences: the value of G1 is varying from
1000W/m2 to 600W/m² and then to 1100W/m2 while
maintaining the value of G2 equal to 600W/m². Consequently,
the P-V curves and Current-Voltage (I-V) curves are
characterized by two maxima as shown respectively in Fig. 8
and Fig. 9. Among these, the first maximum is the local peak
while P2 (261.2 W) is the global peak for the green curve.
Results carried out to track the MPP of the proposed PV
system using GA and PSO are presented and discussed below.
Fig. 10. Influence of the illumination variation of the G1 at fixed temperature
Fig. 10 portrays the comparative convergence profiles of 25°C on the PV system.
output power of the partially shaded PV array.
climb to this peak until G1 changes to 1100W/m² when the
power achieves 250W.
B. PSO
PSO algorithm results are displayed in column 4 of Table
II. Fig. 10 shows how the PSO optimizes the power. It is
observed that the global MPP is detected by the PSO, too.
Nevertheless, its performance is better than GA in detecting
the global peak. In fact, PSO is able to maintain very rapidly
the operating point of the PV systems at the MPP (254.9W),
hence, improving the amount of energy effectively extracted,
i.e., increasing the performance of the PV system. It is noticed
Fig. 8. P-V characteristics in different sequences of G1 scenario that the fluctuation in the convergence profile of output power
variation. while adopting the PSO is less than using GA. Furthermore,
the steady final output power, yielded by the PSO method, is
higher than the final output power of the GA and closer to the
actual global MPP value.

TABLE II. ACTUAL RESULTS BY COMPRARISON WITH RESULTS

OBTAINED BY USING GA AND PSO (G1=1000W/M², G2=600W/M²)

Results
Variables
Actual GA PSO
Peak power, Pm (W) 261.2 249.9 254.9
Voltage at peak power, Vm (V) 55.96 57.7 53.67
Current at peak power, Im (A) 4.69 4.33 4.74
Time to converge (s) -- 0.141 0.071

In order to show how accurate an MPPT works, efficiency is

Fig. 9. I-V characteristics in different sequences of G1
evaluated in Table III. It indicates how close to MPP the
scenario variation. MPPT is and can be defined as:
P
A. GA η = (11)
MPPT , P PMPP
The results yielded by the GA to harvest the global MPP
under PS conditions for the first portion (G2=1000W/m² and
TABLE III. MPPT EFFICIENCY ( η )
G2=600W/m²) are presented in column 3 of Table II. Fig. 10 MPPT , P
shows how genetically the power is optimized. It is obvious Efficiency
that the GA tracks successfully the global MPP unlike the Irradiation variations
GA PSO
conventional methods. However, it doesn't reach exactly the G1=1000W/m² and G2=600W/m² 0.956 0.975
maximum power (249.9W). Then, when G1=600W/m², the G1=600W/m² and G2=600W/m² 0.966 0.994
operating point changes, it enters the vicinity of P2 (230.9W) G1=800W/m² and G2=600W/m² 0.946 0.985
G1=1100W/m² and G2=600W/m² 0.944 0.97
when it should be close to 239W as depicted in Fig. 8. It will
It is obvious that both methods track accurately the global [14] K. Ishaque, Z. Salam, M. Amjad, and S. Mekhilef, “An improved
particle swarm optimization (PSO)-based MPPT for PV with reduced
MPP and can achieve more than 95% efficiency under PS steady-state oscillation,” IEEE Trans. Power Electron., vol. 27, no. 8,
conditions. pp. 3627-3638, 2012.
[15] J.H. Holland, “Outline for a logical theory of adaptive systems,” J.
V. CONCLUSION Assoc. Comput. Mach., vol. 3, pp. 297-314, 1962.
[16] J. Kennedy, and R.C. Eberhart, “Particle swarm optimization”, Proc.
A PV array has been simulated under Matlab/Simpowersystem IEEE Int. Conf. Neural Networks, pp. 1942-1948, 1995.
environment. GA and PSO techniques are applied in order to [17] Y.H. Liu, S.C. Huang, J.W. Huang, and W.C. Liang, “A particle swarm
detect the global MPP when different amounts of solar optimization-based maximum power point tracking algorithm for PV
irradiation are incident on the individual module. The results systems operating under partially shaded conditions,” IEEE Trans.
Energy Convers., vol. 27, no. 4, pp. 1027-1035, 2012.
obtained are promising and show that both of them can solve
the multi-extreme MPP problem, associated with the
conventional methods. Moreover, it is proven that PSO
process mitigate more efficiently power losses caused by PS
with lesser fluctuation and higher steady final output power.
Hence, PS conditions are more elegantly handled by the
proposed PSO method than the GA. The authors are
improving the proposed method and intending to apply the
other variants of evolutionary optimization techniques for
detecting the MPP of a partially shaded PV array.
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, June 2007.
[2] N.A. Ahmed, and M. Miyatake, “A Novel Maximum Power Point
Tracking for Photovoltaic Applications Under Partially Shaded
Insolation Conditions,” Electric Power System Research, vol. 78, no. 5,
pp. 777-784, 2008.
[3] D. Sera, T. Kerekes, R. Teodorescu, and F. Blaabjerg, “Improved
MPPT method for rapidly changing environmental conditions,” Proc.
IEEE Int. Ind. Electron. Symp., vol. 2, pp. 1420-1425, July 2006.
[4] D.P. Hohm, and M.E. Ropp, “Comparative study of maximum power
point tracking algorithms,” Proc. IEE Photovoltaic’s: Res. Appl., vol.
11, no. 1, pp. 4762, January 2003.
[5] E. Koutroulis, and F. Blaabjerg, “Overview of Maximum Power Point
Tracking Techniques for Photovoltaic Energy Production Systems,”
Electric Power Components and Systems, vol. 43(12), pp. 1329-1351,
2015.
[6] Y. Shaiek, M. Ben Smida, A. Sakly, and M.F. Mimouni, “Comparison
between conventional methods and GA approach for maximum power
point tracking of shaded solar PV generators,” Solar Energy, vol. 90,
pp. 107-122, 2013.
[7] J. Young-Hyok, J. Doo-Yong, K. Jun-Gu, K. Jae-Hyung, L. Tae-Won,
and W. Chung-Yuen, “A real maximum power point tracking method
for mismatching compensation in PV array under partially shaded
conditions,” IEEE Trans. Power Electron., vol. 26, no. 4, pp. 1001-
1009, April 2011.
[8] D.C. Huynh, T.M. Nguyen, M.W. Dunnigan, and M.A. Mueller,
“Global MPPT of Solar PV Modules using a Dynamic PSO Algorithm
under Partial Shading Conditions”, Preceedings of IEEE Conference on
Clean Energy and Technology (CEAT), pp 134-139, 2013.
[9] J. Shi, W. Zhang, Y. Zhang, F. Xue and T. Yang, “MPPT for PV
systems based on a dormant PSO algorithm,” Electric Power Systems
Research, vol. 123, pp. 100-107, 2015.
[10] C.S. Chin , P. Neelakantan , H.P. Yoong, and K.T.K. Teo, “Maximum
power point tracking for PV array under partially shaded conditions,”
Proc. CICSyn, vol. 30, pp. 72-77, 2011.
[11] B.N. Alajmi, K.H. Ahmed, S.J. Finney, and B.W. Williams, “A
photovoltaic systems in microgrids,” IEEE Trans. Ind. Electron., vol.
60, no. 4, pp. 1596-1606, 2013.
[12] S. Hadji, J.P. Gaubert, and F. Krim, “Genetic algorithms for maximum
power point tracking in photovoltaic systems,” In Power Electronics
and Applications (EPE 2011), Proceedings of the 2011-14th European
Conference on IEEE, pp. 1-9, August 2011.
[13] K. Ishaque, and Z. Salam, “A review of maximum power point tracking
techniques of PV system for uniform insolation and partial shading
condition,” Renew. Sustain. Energy Rev., vol. 19, pp. 475-488, 2013.