Improved Particle Swarm Optimization Based MPPT

for PV Systems under Partial Shading Conditions

Eduardo Avila1, Nataly Pozo1, Marcelo Pozo1, Gabriel Xavier Dominguez
Salazar2 Facultad de Ingeniería en Ciencias Aplicadas
1
Departamento de Automatización y Control Electrónico Universidad Técnica del Norte
Industrial Ibarra, Ecuador
2
Departamento de Energía Eléctrica [email protected]
Escuela Politécnica Nacional
Quito, Ecuador
{eduardo.avila, nataly.pozo, marcelo.pozo,
gabriel.salazar}@epn.edu.ec

Abstract— In this work, an improved Particle Swarm depicted on Fig.1 [5]. In fact, the uneven distribution of solar
Optimization (PSO) algorithm is employed to perform the irradiance over the PV modules of the same string may lead to
Maximum Power Point Tracking (MPPT) of a 1.2 MW a power loss as high as 70% [6]. The main reason for this
Photovoltaic farm under Partial Shading Conditions (PSC). High excessive power loss is that traditional MPPT algorithms
efficiency, low complexity and reduced tracking time are the generally reach a Local Maximum Power Point (LMPP)
main features achieved by integrating the ability to ensure a decreasing the overall efficiency of the system. Thus, in order
Global Maximum Power Point (GMPP) operation provided by to overcome this issue and force the system to operate in the
PSO and the fast convergence of Perturb and Observe (P&O) Global Maximum Power Point (GMPP), both extension of
algorithm. In addition, Sliding-Mode Control (SMC) is used on a
classical MPPT [7], [8] and Evolutionary Computing
cascade scheme on the DC-DC power converters in order to
achieve a suitable energy conversion. Thus, the design of the
Algorithms (ECA) [9], [10] have been discussed in the
SMC is achieved by using the space state representation of the literature.
power converters. Finally, photovoltaic system performance is
analyzed for different solar radiation profiles. The obtained
results show the proper dynamic and steady-state responses.

Keywords— Maximum power point tracking, partial shading,

particle swarm optimization, photovoltaic system, sliding-mode
control.

I. INTRODUCTION
Nowadays, photovoltaic systems have become an
environmentally-friendly solution to minimize the carbon-
dioxide emissions caused by the high fossil oil dependency.
Figure 1. P-V curve of a PV partially shaded array
[1]. High reliability, high life cycle, low operational and
maintenance costs are the main advantages of PV systems [2]. Among ECA-based MPPT methods, Fuzzy Logic (FL),
Nevertheless, the power generated by the PV modules mainly Artificial Neural Networks (ANN), Genetic Algorithms (GA),
depends on the solar irradiance and temperature. As these Grey Wolf Optimization (GWO) and Particle Swarm
environment conditions vary over the time, the P-V Optimization (PSO) are the main techniques employed for
characteristic curve exhibits a nonlinear relationship with a MPPT on PV systems under PSC. Even though these methods
unique Maximum Power Point (MPP) under uniform always ensure a GMPP operation, unlike the modified classical
irradiation. Therefore, the implementation of a Maximum MPPT algorithms, the high complexity and the high time
Power Point Tracking (MPPT) technique to enhance the response are their main drawbacks [11].
efficiency of the system is required [3]. Perturb and Observe
(P&O), Hill Climb (HC), Incremental Conductance (IC), In this case, in order to provide a high efficiency, fast
Fractional Open-Circuit Voltage (FOCV), Fractional Short- convergence and low complexity MPPT algorithm for PV
Circuit Current (FSCI), etc. are common strategies employed systems under PSC, a hybrid PSO-P&O MPPT technique is
for MPPT on photovoltaic systems under uniform irradiation proposed by integrating the ability to ensure a GMPP operation
[4]. provided by PSO and the fast convergence of P&O algorithm.
Thus, the PSO-P&O MPPT algorithm provides the reference to
Nevertheless, the performance of the traditional MPPT the voltage controller. Meanwhile, in most applications voltage
algorithms decreases significantly under Partial Shading regulation is achieved by means of a single control loop that
Conditions (PSC) due to the appearance of multiple local modifies the duty cycle of the control signal, in this study a
maxima on the P-V characteristic curve of the system, as cascade control scheme with a superimposed voltage loop and

Authorized licensed use limited to: UNIVERSITY TENAGA NASIONAL. Downloaded on November 22,2023 at 09:41:35 UTC from IEEE Xplore. Restrictions apply.
an inner current loop is proposed in order to improve the where and are the number of strings and columns
dynamic response and the disturbance rejection of the system respectively, and are the cell nominal irradiance and
[12]. Usually, classical PI regulators are employed for voltage ambient temperature respectively, and and are the
and current regulation due to their simplicity, high efficiency, temperature coefficients related to and [14]. In
low cost and ease of implementation. Nevertheless, on recent addition, the nominal operation temperature of the cell is
years Sliding-Mode Control (SMC) based schemes have been used to calculate as follows:
used as SMC allows large signal stability, excellent dynamic
response, simple and intuitive implementation, high robustness 20 /0.8 (3).
against parameter variations and external perturbations [13].
Moreover, SMC is usually employed for regulating magnitudes B. Maximum Power Point Tracking
with fast dynamic response and narrow bandwidth. For these The PSO algorithm has an exploratory and a convergence
reasons, SMC is proposed for current regulation on the inner phase. In the first phase, agents called particles are randomly
loop and a classical PI for voltage control on the superimposed initialized in a given space. Then, an objective function is
loop. evaluated for each particle and the best performance of the
This paper is organized as follows. First, the PV swarm is stored in memory ( ) as well as the best
connection scheme is described. Then the selection and design performance of each agent ( ). In the convergence phase,
of the power converters are explained in detail and finally as particles move through the space with a certain speed ( )
the main contribution of this work, an improved PSO-P&O determined by and as depicted as follows [15]:
based MPPT algorithm with a SMC in a cascade scheme is (4),
proposed.
where and are the velocity of the particle “ ” in the
II. PHOTOVOLTAIC GENERATION SYSTEM instant and 1 respectively, is the current position of
In order to accurately analyze the PSC effect on PV the particle “ ”, and are acceleration coefficients, and
systems, real irradiation data and the connection scheme from and are random coefficients between zero and one.
a 1.2 MW PV generation system installed on Salinas de Ibarra Furthermore, the next position of each particle x ) is
city are employed. Moreover, the specifications of the calculated as follows:
“YL250P-29b” PV modules installed on this PV system are 1 1
detailed in Table I. (5),

TABLE I. TECHNICAL DATA OF YL250P-29B PV MODULE

Parameter Symbol Value Unit

Nominal power 250
Open circuit voltage 38.4 V
Short-circuit current 8.79 A
Voltage at MPP 30.4 V
Current at MPP 8.24 A

At the present time, The PV generation system is

composed by 10 PV arrays of 480 modules distributed on 24
parallel strings of 20 modules each one. Furthermore, every
set of 5 PV arrays is connected to a DC/AC converter. Thus,
two inverters are in-charge of the integration and
synchronization with the distribution network. So, in order to
improve the efficiency of the system under PSC, the series
connection of a DC/DC power converter to each one of the ten
PV arrays is proposed.
III. POWER CONVERSION TOPOLOGIES AND CONTROL
STRATEGIES
A. Mathematical Model of the Photovoltaic Array
The output power of the PV array depends on the solar
irradiation (G) and the operating temperature of the PV
modules ( ). Thus, the mathematical model for the current (I)
and voltage (V) of the PV array can be described as follows:
Figure 2. Flowchart of conventional PSO-based MPPT algorithm
/ (1),
In MPPT applications, the objective function of the
(2), algorithm is to maximize the output power of the PV system.

Authorized licensed use limited to: UNIVERSITY TENAGA NASIONAL. Downloaded on November 22,2023 at 09:41:35 UTC from IEEE Xplore. Restrictions apply.
Moreover, the position of the particles is defined as the fast tracking ability of the P&O algorithm is employed as
reference voltage of the PV array ( ) and the velocity is depicted below:
defined as a small disturbance or variation of . Likewise,
& (6),
represents the nearest voltage value of each particle to the
GMPP and defines the voltage value that produces the where & is the voltage variation provided by the typical
maximum power [16]. Thus, particles are moved from their P&O algorithm depicted on Fig 4.
initial positions to the GMPP. Nevertheless, this process (Fig.
2) implies a huge amount of iterations for each particle until
GMPP is reached. Therefore, the convergence time of PSO
algorithm is slow despite of being the faster among ECA
methods. For these reasons, in this case a hybrid PSO-P&O
MPPT algorithm is proposed on Fig. 3 in order to improve the
dynamic response of the system.

Figure 4. Flowchart of P&O algorithm

Moreover, in order to detect a change in the PSC and reset

the algorithm, the following condition must be fulfilled.
/ ∆ (7),

where and are the actual and the previous

output power of the PV system , respectively. And ∆ is the
maximum admissible variation of output power before a PSC
change is considered.
C. DC-DC converters
The aim of the DC-DC power converter is to attain the
GMPP of operation of the PV array and provide a suitable
integration with the DC-AC converter. Therefore, smooth input
and output currents and the ability to step up or step down the
output voltage with a wide conversion ratio are the main
features required for the DC-DC converter [17]. In this case,
Ćuk power converter is selected and connected as detailed on
Figure 3. Flowchart of proposed hybrid PSO-P&O MPPT algorithm Fig. 5.
Exploratory phase is equal for the conventional PSO and
the proposed hybrid PSO-P&O MPPT algorithm. The main
difference is on the convergence phase. On conventional PSO
algorithm, all particles should converge to GMPP. Therefore,
the reaching time of the GMPP is increased due to the large
amount of iterations. In contrast, on hybrid PSO-P&O MPPT Figure 5. Cuk converter scheme
algorithm only the particle with the best performance ( ) is where is the operating voltage of the PV array, ∆ is the
used and the other ones are not considered. This means that control signal, and are the voltages on the capacitors
from this moment, the whole swarm is reduced to a single and , and are currents on inductors and
particle. In this context, the acceleration coefficient from (4) respectively. Moreover, the mathematical representation of the
should be set to zero. Now, in order to reach the GMPP, the Ćuk power converter, obtained from the analysis of the

Authorized licensed use limited to: UNIVERSITY TENAGA NASIONAL. Downloaded on November 22,2023 at 09:41:35 UTC from IEEE Xplore. Restrictions apply.
 ∗
behaviour of the system during the ON and OFF states of the where, is the current reference provided by the
semiconductor , can be described as steady state as superimposed voltage control loop.
following [17]:
3) Equivalent control signal
(8),
An operation of the system over the sliding surface implies
where: that the derivative of the sliding surface and the value of the
discontinuous control signal are zero. Thus, the equivalent
: Space state vector
control signal can be found under these conditions.
: Derivative of x
The derivative of the sliding surface is expressed as
: Input of the power converter follows:
σ ∗ 0 (12),
1 ∆
0 0 0
where the equation for is derived from (8) and defined as:
∆ 1 1
0 0 1 ∆ 1
(13).
; 0 (9).
1 ∆ ∆
0 0 0 By replacing (13) in (12) follows:
0
1 1 1 ∆ 1
0 0 ∗ 0 (14).

D. Cascade Control Scheme Now, the sliding condition (∆ ) is applied on (14) and
the equivalent control signal is found and depicted as:
The proposed control scheme for PV system is depicted in
Fig. 6. As previously mentioned, a cascade scheme with a ∗
1 (15).
classical PI regulator on the superimposed loop is used to
control the PV array voltage. Besides, a SMC strategy is used
4) Discontinuous control signal
on the inner loop to regulate the current on the inductor .
The discontinuous control signal is commonly composed
by a sign function along with a positive constant ) as
detailed below [20]:
σ (16).
Thus, the SMC law is:

∆ 1 ∗ σ (17).
Figure 6. Cascade control scheme for the PV system
E. Sliding-Mode Control (SMC) 5) Lyapunov stability criterion
1) General description The following conditions should be fulfilled in order to
satisfy the Lyapunov stability criterion.
The SMC has a hitting and an existence condition. The
hitting condition of the SMC modifies the dynamics of a 0 0
system from an initial condition to an operating point over σ 0 (17),
sliding surface by means of a discontinuous control signal
σ 0
( ). Besides, the existence condition keeps the dynamics of
the system over the sliding surface when external perturbations where is the Lyapunov function defined as [21]:
appear. In this case, an equivalent control signal ( ) fulfills
1
the existence condition. Thus, SMC control law (∆) composed σ (18).
by the equivalent and the discontinuous signals is depicted as 2
[18], [19]: Moreover, the mathematical expression for σ is
depicted below:
∆ (10).
σ σ∙σ 0 (19).
2) Sliding surface
Now, by substituting (14) on (19):
The sliding surface ( ) for the current regulation in the
inductor is defined as: ∗
1 ∆ 1
σ∙ 0 (20).
∗
σ (11),
Replacing (17) in (20) and simplifying follows:

Authorized licensed use limited to: UNIVERSITY TENAGA NASIONAL. Downloaded on November 22,2023 at 09:41:35 UTC from IEEE Xplore. Restrictions apply.
 σ∙ σ 0 (21).

If , then:

σ∙ σ 0 (22).
Thus, |σ| is positive definite if and only if 0.
IV. SIMULATION RESULTS
In order to verify proper steady state and dynamic-
responses of the PV generation system under PSC, Simulink®
from Matlab was used to carry out a simulation of one of the
ten PV arrays of “Salinas de Ibarra” PV farm for the parameter Figure 8. PV system response under partial shading conditions
values detailed on Table II. Moreover, the integral squared error (ISE) index was found
using (23) in order to compare the performance of PSO and
TABLE II. PARAMETER VALUES EMPLOYED ON SIMULATION PSO-P&O algorithms.
Component Parameter Symbol Value Units
(23).
Nominal power 120 kW
PV Array Parallel Strings 24 - Table III exposes that PSO-P&O outperforms the use of the
Modules per string 20 - conventional PSO MPPT algorithm.
Capacitor 47 uF
Ćuk Inductor , 30 mH TABLE III. ISE UNDER DIFFERENT PARTIAL SHADING CONDITIONS
Converter Capacitor 47 uF Time Integral Square Error Settling Time
Study Range
Capacitor 200 uF Case
(ISE) [p.u.] [s]
Number of particles 300 - [s] PSO PSO-P&O PSO PSO-P&O
A 0-0.4 146.02 61.73 0.22 0.10
Position of the first
300 V B 0.4-0.8 139.63 56.58 0.21 0.05
particle
Hybrid C 0.8-1.2 108.25 52.05 0.22 0.07
Position of the last D 1.2-1.6 77.83 46.17 0.22 0.05
PSO-P&O 600 V
particle E 1.6-2.0 70.83 32.96 0.21 0.14
MPPT
Acceleration F 2.0-2.4 70.47 28.66 0.22 0.07
Algorithm 0.5 -
Coefficient G 2.4-2.8 61.33 23.87 0.25 0.10
Voltage Variation
Provided by P&O & 0.25 V In addition, in order to provide a better understanding of the
GMPP tracking process of the PV system under PCS, Fig. 9
Irradiance profile is modified over the time considering shows the voltage curve corresponding to the “C” study case.
PSC on the strings of the PV generation system and the P-V
characteristic was found for each study case as detailed on
Fig.7.

Figure 7. P-V characteristic curves under partial shading conditions

The results from the GMPPT process are detailed on Fig. 8

and Fig. 9. Notice that the proposed MPPT algorithm
successfully reaches the GMPP of operation in all study cases
as shown on Fig.8.
Figure 9. GMPP tracking process under PSC for “C” study case

Authorized licensed use limited to: UNIVERSITY TENAGA NASIONAL. Downloaded on November 22,2023 at 09:41:35 UTC from IEEE Xplore. Restrictions apply.
 On the exploratory phase, the PV array voltage is modified in PV array under partially shaded conditions," in Proc. 2011 IEEE
from the value of the first particle ( x 300 V) to the value of Transactions on Power Electronics, Vol.26, pp.1001-1009.
the last one (x 600 V) in steps of one Volt in order to [7] K. Sundareswaran, S. Palani, and V. Vigneshkumar, "Development of a
hybrid genetic algorithm/perturb and observe algorithm for maximum
acquire the power output corresponding to each one of the 300 power point tracking in photovoltaic systems under non-uniform
particles. In this case, the particle that provides the greater insolation," in Proc. 2015 IET Renewable Power Generation, pp. 757-
power output (30.77 kW) is x , corresponding to 508 V. 765.
Notice that the fast tracking on the exploratory phase cannot [8] K. Soon Tey, and S. Mekhilef, "Modified incremental conductance
provide the exact GMPP (31.8 kw at 495V). For this reason, algorithm for photovoltaic system under partial shading conditions and
P&O algorithm was implemented in the convergence phase in load variation," in Proc. 2014 IEEE Transactions on Industrial
Electronics, pp. 5384 – 5392.
order to reach the GMPP in a short period of time despite of
[9] M. Mao, Q. Duan, Z. Yang, and P. Duan, "Modeling and global MPPT
the presence of small variations in the radiation profile. for PV system under partial shading conditions using modified artificial
fish swarm algorithm," in Proc. 2016 IEEE International Symposium on
Systems Engineering (ISSE), pp. 1-7.
[10] S. Mohanty, B. Subudhi, and P. Ray, "A grey wolf optimization based
MPPT for PV system under changing insolation level," in Proc. 2016
IEEE Students’ Technology Symposium (TechSym), pp. 175-179.
[11] C. Huang, L. Wang; R. Yeung, Z. Zhang, H. Chung, and A. Bensoussan,
"A Prediction Model Guided Jaya Algorithm for the PV System
Maximum Power Point Tracking," 2017 IEEE Transactions on
Sustainable Energy, in press.
[12] T. Marlin, Process Control, Designing Processes and Control Systems
for Dynamic Performance, 2nd ed., Ed. New York: McGraw-Hill, 1995,
pp. 457
[13] F. Ahmad, A. Rasool, E. Ozsoy, A. Sabanovic, M. Elitas, "Design of a
robust cascaded controller for cuk converter," in Proc. 2016 IEEE
Figure 10. Current regulation with SMC for “C” study case International Power Electronics and Motion Control Conf., pp. 80-85.
[14] J. Cabezas, and L. Ortega, "Design and implementation of a hybrid
Finally, the current regulation provided by the SMC for the power electronics system for lithium polymer battery charging," Eng.
“C” study case is detailed on Fig. 10. Notice that the current of thesis, Dept. of Elec. Eng. EPN, Quito, 2016
the PV array (blue segmented line) successfully tracks the [15] Z. Cheng, H. Zhou, and H. Yang, "Research on MPPT control of PV
reference provided by the superimposed voltage control loop system based on PSO algorithm," in Proc. 2010 Chinese Control and
(pink line) and stabilizes on the current value that allows a Decision Conf., pp. 887-892.
[16] K. Ishaque, Z. Salam, M. Amjad, and S. Mekhilef, "An improved
GMPP of operation (green line). particle swarm optimization (PSO)–based MPPT for PV with reduced
steady-state oscillation," in Proc. 2012 IEEE Transactions on Power
V. CONCLUSIONS Electronics, Vol. 27, pp. 3627-3638.
The proposed hybrid PSO-P&O MPPT algorithm has [17] M. Shadmand, A. Shirkhani, and M. Pasupuleti, "Design and
performance comparison of LQR nonzero setpoint, sliding mode, and
probed to ensure the GMPP of operation under PSC in all the current mode digital controllers for cuk converter " in Proc. 2013 4th
study cases with higher computational efficiency, lower ISE IEEE International Symposium on Power Electronics for Distributed
(63% lower) and faster convergence time (100 milliseconds Generation Systems (PEDG), pp. 1-6.
faster) than conventional PSO algorithm. Furthermore, [18] L. Malesani, L. Rosetto, G.Spiazzi, and P. Tenti, "Performance
proposed PI-SMC controller on a cascade scheme exhibits fast optimization of cuk converters by sliding-mode control," in Proc. 1995
IEEE Transactions on Power Electronics
and accurate dynamic response, large signal stability, and high
robustness against irradiance level changes. [19] X. Dominguez, O. Camacho, P. Leica and A. Rosales, "A fixed-
frequency Sliding-mode control in a cascade scheme for the Half-bridge
Bidirectional DC-DC converter," 2016 IEEE Ecuador Technical
REFERENCES Chapters Meeting (ETCM), pp. 1-6.
[1] V. Phimmasone, Y. Kondo, N. Shiota, and M. Miyatake, "The [20] V. Utkin, "Sliding mode control of DC/DC converters," in Proc. 2013
demonstration experiments to verify the effectiveness of the improved Journal of the Franklin Institute, Vol. 350, pp. 2146–2165.
PSO-based MPPT controlling multiple photovoltaic arrays," in Proc. [21] V. Utkin, J. Guldner, and J. Shi, Sliding Mode Control in
2013 IEEE ECCE Asia Downunder, pp. 83-92. Electromechanical Systems, CRC Press, New York, NY, USA, 1999.
[2] P. Chen, B. Yan, Ch. Liu, S. Wang, and Y. Liu, "A comparative study
on MPPT for photovoltaic generation systems," in Proc. 2015 IEEE 2nd
International Future Energy Electronics Conference (IFEEC), pp. 1-6.
[3] L. Bouselham, B. Hajji, and H. Hajji, "Comparative study of different
MPPT methods for photovoltaic system," in Proc. 2015 3rd International
Renewable and Sustainable Energy Conference (IRSEC), pp. 1-5.
[4] B. Subudhi, and R. Pradhan, "A comparative study on maximum power
point tracking techniques for photovoltaic power systems," in Proc. 2013
IEEE Transactions on Sustainable Energy, pp. 89-98.
[5] W. Yunliang, and B. Nan, "Research of MPPT control method based on
PSO algorithm" in Proc. 2015 4th International Conference on Computer
Science and Network Technology (ICCSNT), pp. 698-701.
[6] Y. Ji, D. Jung, J. G. Kim, J. H. Kim, T. Lee, and C. Won, "A real
maximum power point tracking method for mismatching compensation

Authorized licensed use limited to: UNIVERSITY TENAGA NASIONAL. Downloaded on November 22,2023 at 09:41:35 UTC from IEEE Xplore. Restrictions apply.