IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 27, NO.

4, DECEMBER 2012

1027

A Particle Swarm Optimization-Based Maximum

Power Point Tracking Algorithm for PV Systems
Operating Under Partially Shaded Conditions
Yi-Hwa Liu, Member, IEEE, Shyh-Ching Huang, Jia-Wei Huang, and Wen-Cheng Liang

AbstractA photovoltaic (PV) generation system (PGS) is becoming increasingly important as renewable energy sources due
to its advantages such as absence of fuel cost, low maintenance
requirement, and environmental friendliness. For large PGS, the
probability for partially shaded condition (PSC) to occur is also
high. Under PSC, the PV curve of PGS exhibits multiple peaks,
which reduces the effectiveness of conventional maximum power
point tracking (MPPT) methods. In this paper, a particle swarm
optimization (PSO)-based MPPT algorithm for PGS operating under PSC is proposed. The standard version of PSO is modified
to meet the practical consideration of PGS operating under PSC.
The problem formulation, design procedure, and parameter setting method which takes the hardware limitation into account are
described and explained in detail. The proposed method boasts the
advantages such as easy to implement, system-independent, and
high tracking efficiency. To validate the correctness of the proposed method, simulation, and experimental results of a 500-W
PGS will also be provided to demonstrate the effectiveness of the
proposed technique.
Index TermsMaximum power point tracking (MPPT), partially shaded condition (PSC), particle swarm optimization (PSO),
photovoltaic (PV) generation system (PGS).

I. INTRODUCTION
HE ever-increasing world energy demand and growing
concern about environmental issues have generated enormous interest in the utilization of renewable energy sources.
Among them, the photovoltaic (PV) generation system (PGS) is
an established technology and has rapid growth in recent years.
The advantages of PGS include absence of fuel cost, low maintenance requirement, and environmental friendliness. However,
due to the high investment cost on the PGS, it is vital to make
the most of the available solar energy [1][5]. A major challenge in using a PGS is to tackle its nonlinear currentvoltage
(IV) characteristics, which result in a unique maximum power
point (MPP) on its powervoltage (PV) curve. Since the power

Manuscript received May 3, 2012; revised August 10, 2012; accepted

September 5, 2012. Date of publication September 29, 2012; date of current
version November 16, 2012. This work was supported by the National
Science Council of Taiwan and Taiwan Power Company under contract NSC
100-3113-p-110-004. Paper no. TEC-00185-2012
The authors are with the Department of Electrical Engineering, National
Taiwan University of Science and Technology, Taipei 106, Taiwan (e-mail:
yhliu@mail.ntust.edu.tw; d9707921@mail.ntust.edu.tw; d9707203@mail.
ntust.edu.tw; m9907113@mail.ntust.edu.tw).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TEC.2012.2219533

generated from a given PV module mainly depends on solar

insolation and panel temperature. As these quantities vary with
time, it is essential to develop a maximum power point tracking (MPPT) algorithm to extract maximum power from the PV
module at real time. Over the past decades, many MPPT algorithms have been proposed. These methods include perturb and
observe (P&O), incremental conductance, short-circuit current,
open-circuit voltage, fuzzy logic control, and ripple correlation
approaches. Some modified techniques which aim to minimize
the hardware requirement or to improve the performance have
also been proposed. These methods mentioned earlier are effective and time tested under uniform solar insolation [6][8].
In the PGS, multiple PV modules are generally interconnected
in series and/or parallel to create a system with the desired voltage and loading current capacity. Therefore, partially shaded
condition (PSC) is sometimes inevitable because some parts of
the module or the PGS may receive less intensity of sunlight due
to clouds or shadows of trees, buildings, and other neighboring
objects. PSC can have a significant impact on the power output
of PGS, depending on the system configuration, shading pattern,
and the bypass diodes incorporated in the PV modules. The effect of PSC on PGS has been analyzed in several publications.
Under PSC, PV modules belonging to the same string experience different insolation. The resulting PV characteristic curve
becomes more complex and exhibits multiple peaks [9][11].
The presence of multiple peaks reduces the effectiveness of the
conventional MPPT algorithms, which assumes a single MPP
on the PV curve. The reason is that these methods are based
on the hill-climbing principle of moving the next operating
point (OP) in the direction in which power increases. If the
PV curve is not unimodal, these methods may only reach a
local MPP. Since the occurrence of PSC being quite common,
there is a need to develop a suitable MPPT algorithm that can
track the global maximum power point (GMPP) under these
conditions. Some researchers have worked on GMPP tracking schemes for PGS operating under PSC [12][26]. In [12],
a new MPPT technique which is able to operate under PSC
is presented. To find the GMPP, the voltage factors of all the
MPPs have to be previously assessed once. Therefore, the proposed method is system dependent. Kobayashi et al. [13] and
Ji et al. [14] propose a two-stage method to track the GMPP.
In the first stage, the OP of the PGS is moved to the vicinity
of the GMPP using the load line, and in the second stage, it
converges to the GMPP. However, these methods cannot obtain
the GMPP if the GMPP lies on the left side of the load line.
Line search algorithms are utilized in [15] and [16] to find the

0885-8969/$31.00 2012 IEEE

1028

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 27, NO. 4, DECEMBER 2012

GMPP. These methods compare the values of measured power

at two OPs and then determine the OP movement. These techniques are similar to the P&O method with variable step size;
the only difference here is that the step size is determined by
the Fibonacci sequence in [15], and by the dividing rectangles
(DIRECT) in [16]. These approaches too cannot guarantee to
find the GMPP under all conditions. Patel and Agarwal [17] and
Renaudineau et al. [18] also propose two-stage methods to track
the GMPP. In [17], a global stage is used to find the regions of
local MPPs, while the local stage employs P&O to locate the
GMPP. In [18], a scanning process is first utilized to detect the
regions which contain the GMPP. After the scanning process, a
P&O algorithm with variable step is used to find the GMPP. The
tracking speeds of these methods are limited because almost all
local MPPs should be found and compared to obtain the GMPP.
Lei et al. [19] propose a sequential extremum seeking control
(ESC)-based MPPT algorithm which is based on approximate
modeling and analysis for the characteristics of PV modules under variable PSCs. The proposed MPPT algorithm is based on
segmental search with consistent definition of searching range,
which can achieve better computational efficiency than sweeping search. However, this method possesses steady-state errors
and is system dependent. In [20], a novel MPPT algorithm is
proposed using artificial neural network (ANN) and fuzzy logic
with polar information controller. The ANN with three layer
feed-forward is trained once for several PSCs to determine the
GMPP voltage; therefore, it is system dependent. Moreover,
this method uses insolation and temperature as inputs to obtain GMPP, while these information are often not available in
PGS. Another option to deal with PSC is to use intelligent PV
modules or alternating current modules [21][23]. In [21], a distributed MPPT for PGS is proposed and analyzed. A switching
converter dedicated to each module and performing the MPPT
is used. In [22], several PV modules with each coupled to its
own dcdc boost converter are cascaded on the same dc bus
and interfaced to the grid by means of a dcac inverter. Each
PV generator is equipped with a second control law aiming at
limiting the converters output voltage. In these configurations,
each PV module is treated as one unit that tracks its own MPP;
therefore, the PGS can always track the GMPP. However, these
methods incur extra hardware and cost. Moreover, a good MPPT
algorithm should also be implemented in each unit to track the
MPP when a module is shaded. Gao et al. [23] propose a PV
system which adopts the parallel configuration at the individual
cell level so that every cell in the PV module can achieve its MPP
under PSC. Since the input voltage of this configuration is very
low, this may increase the difficulty of designing an appropriate
power converter. Moreover, the proposed configuration is only
suitable for low power applications. Chen et al. [24] present
a novel MPPT method based on biological swarm chasing behavior to increase the MPPT performance. This method is only
applicable when the entire module is under uniform insolation
conditions, hence PSC is not considered. In [25], an adaptive
perceptive particle swarm optimization (APPSO)-based MPPT
algorithm is presented. Also, Miyatake et al. [26] attempted
to approach the GMPP using the particle swarm optimization
(PSO) algorithm. In these investigations, the authors try to re-

Fig. 1.

Equivalent circuit of the PV cell.

alize centralized MPPT control of the modular (multimodule)

PGS. These MPPT algorithms have good performance under
various PSC; however, these methods are only suitable for systems that consist of multiple converters. However, for PGS,
the use of one central high-power single-stage electronic converter is very common for economical reasons and the relative
simplicity of the overall system. Ishaque et al. [27] present an
improved PSO-based MPPT algorithms for PGS, the advantages
of using PSO in conjunction with the direct duty cycle control
are discussed in detail. However, no system design guidelines
and practical design considerations are provided in these papers.
This paper aims to develop an accurate and systemindependent MPPT algorithm for centralized-type PGS operating under PSC. The PSO method has been successfully employed to solve different engineering optimization problems
[28][33]. According to these investigations, the PSO method
is a simple and effective metaheuristic approach that can be
applied to optimization problems having many local optimal
points. Consequently, it will be adopted in this paper to realize
the MPPT algorithm which is suitable for centralized PGS under
PSC. In this paper, the standard version of PSO will be modified
to meet the practical consideration of PGS under PSC. Detailed
design procedures which take the hardware limitation into account will be presented first, and a 500-W prototype will be
implemented to demonstrate the validity of the proposed MPPT
algorithm. Experimental results show that the proposed MPPT
technique can obtain the GMPP in all the test cases no matter
where the GMPP locates. The tracking efficiencies in all test
cases are higher than 99.5%. The proposed MPPT algorithm is
simple, accurate and system independent, and can be realized
using a low cost digital signal controller (DSC).
II. CHARACTERISTICS OF THE PGS UNDER PSC
A. Basic Characteristics of a PV Cell
A PV cell can be represented by an electrical equivalent onediode model as shown in Fig. 1. This model contains a current
source Ig , a diode D, and a series resistance RS , which represents the resistance inside each cell and in the connection
between the cells. The net current IPV is the difference between
the photocurrent Ig and the diode current ID :




q(Vpv + Ipv RS )
Ipv = Ig IS exp
1
(1)
nkT
where n is the diode ideality factor, k is Boltzmanns constant,
q is the electron charge, T is the temperature in Kelvin, RS is

LIU et al.: PARTICLE SWARM OPTIMIZATION-BASED MAXIMUM POWER POINT TRACKING ALGORITHM

1029

III. OVERVIEW OF THE PSO ALGORITHM

PSO is a swarm intelligence optimization algorithm developed by Eberhart and Kennedy in 1995, which is inspired by
the social behavior of bird flocking and fish schooling. PSO is
a global optimization algorithm for dealing with problems on
which a point or surface in an n-dimensional space represents
a best solution. In this algorithm, several cooperative agents
are used, and each agent exchanges information obtained in its
respective search process. Each agent, referred to as a particle,
follows two very simple rules, i.e., to follow the best performing
particle, and to move toward the best conditions found by the
particle itself. By this way, each particle ultimately evolves to
an optimal or close to optimal solution.
The standard PSO method can be defined using the following
equations [34]:
Fig. 2.
curve.

3s2p system under three different PSCs. (a) IV curve. (b) PV

vi (k + 1) = wvi (k) + c1 r1 (pb est,i xi (k))

+ c2 r2 (gb est xi (k))
xi (k + 1) = xi (k) + vi (k + 1)

(2)
(3)

i = 1, 2, . . . , N

Fig. 3. IV and PV curves of the 3s2p system shown in Fig. 2. (a) IV

curve. (b) PV curve.

the equivalent series resistance, and Is is the saturation current,

respectively.
B. Effect of PSC on PGS
A PV module consists of several PV cells connected in parallel to increase current and in series to produce a higher voltage.
Several PV modules are then connected in series/parallel to form
a PGS. Under PSC, the PV curve of PV module will display
multiple MPPs because of the bypass diodes. The characteristics of a PV module under PSC with bypass diodes connected
at module terminal can be explained as follows. Under PSC, the
shaded cells behave as a load instead of a generator and create the hot spot. Therefore, bypass diodes of these shaded cells
will conduct to avoid this problem. Since the shaded cells are
bypassed, multiple peaks in the PV curve will be presented.
As an example, a simple PV module with six cells organized
in three serial groups of two parallel cells as shown in Fig. 2
is considered. This configuration is named 3s2p connection.
When this system is under different shading patterns as shown
in Fig. 2, the resulting IV curves are shown in Fig. 3(a). These
characteristic curves can then be utilized to obtain the PV characteristic curves as shown in Fig. 3(b). From Fig. 3, it can be
observed that the GMPP can occur in either the lower or higher
voltage range, depending on the type of shading pattern. This
phenomenon makes it difficult to directly apply the conventional
MPPT algorithms.

where xi is the position of particle i; vi is the velocity of particle

i; k denotes the iteration number; w is the inertia weight; r1 and
r2 are random variables uniformly distributed within [0,1]; and
c1 , c2 are the cognitive and social coefficient, respectively. The
variable pb est,i is used to store the best position that the ith
particle has found so far, and gb est is used to store the best
position of all the particles.
The flowchart of a basic PSO algorithm is illustrated in Fig. 4.
From Fig. 4, the operating principles of a basic PSO method can
be described as follows:
Step 1 (PSO Initialization): Particles are usually initialized
randomly following a uniform distribution over the
search space, or are initialized on grid nodes that
cover the search space with equidistant points. Initial
velocities are taken randomly.
Step 2 (Fitness Evaluation):Evaluate the fitness value of
each particle. Fitness evaluation is conducted by supplying the candidate solution to the objective function.
Step 3 (Update Individual and Global Best Data): Individual and global best fitness values (pb est,i and gb est )
and positions are updated by comparing the newly
calculated fitness values against the previous ones,
and replacing the pb est,i and gb est as well as their
corresponding positions as necessary.
Step 4 (Update Velocity and Position of Each Particl)e: The
velocity and position of each particle in the swarm
are updated using (1) and (2).
Step 5 (Convergence Determination): Check the convergence criterion. If the convergence criterion is met,
the process can be terminated; otherwise, the iteration number will increase by 1 and goto step 2.

1030

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 27, NO. 4, DECEMBER 2012

Fig. 5.

Fig. 4.

Flowchart of a standard PSO [34].

IV. APPLICATION OF PSO TO MPPT

The PSO method described in Section III is now applied
to realize the MPPT algorithm for PGS operating under PSC,
wherein the PV curve exhibits multiple local MPPs. Due to the
uniqueness of this problem, the standard version of PSO will be
modified to meet the practical consideration of PGS under PSC.
Detailed design procedures which take the hardware limitation
into account will be presented in the following. Fig. 5 depicts
the block diagram of the proposed system. From Fig. 5, the
presented system consists of a series-connected PV module, a
dcdc converter and a digital controller in which the proposed
MPPT algorithm is implemented. In this paper, a simple boost
converter is used to interface the voltage from the PV mod-

Block diagram of the proposed system.

ule to the load. The implementation of this part of circuit is

conventional, therefore, will not be discussed further here.
Fig. 6 shows the flowchart of the proposed PSO-based MPPT
technique. In Fig. 6, the steps which are different from the
standard PSO will be marked in different colors. From Fig. 6,
the main blocks are described in detail in the following:
Step 1 (Parameter Selection): In the proposed system, the
particle position is defined as the duty cycle value d
of the dcdc converter, and the fitness value evaluation function is chosen as the generated power PPV
of the whole PGS. From the algorithm point of view,
a larger number of particles result in more accurate MPP tracking even under complicated shading
patterns. However, a larger number of particles also
lead to longer computation time. Therefore, a tradeoff
should be made to ensure good tracking speed and accuracy. According to the literature, there exist at most
m MPPs in the PV curve for PV modules consist m
series connected PV cells [9]. Consequently, the particle number N is chosen as the number of the series
connected cells in the PGS.
Step 2 (PSO Initialization): In PSO initialization phase, particles can be placed on fixed position or be placed
in the space randomly. Basically, if there is information available regarding the location of the GMPP
in the search space, it makes more sense to initialize the particles around it. According to [17], the
peaks on the PV curve occur nearly at multiples of
80% of the module open voltage VOC m o dule , and the
minimum displacement between successive peaks is
also nearly 80% of VOC m o dule . Therefore, the particles are initialized on fixed positions which cover
the search space [Dm in , Dm ax ] with equal distances
in this paper. Dm ax and Dm in are the maximum and
minimum duty cycle of the utilized dc-dc converter,
respectively.
Step 3 (Fitness Evaluation): The goal of the proposed MPPT
algorithm is to maximize the generated power PPV .
After the digital controller output, the PWM command according to the position of particle i (which
represents the duty cycle command), the PV voltage

LIU et al.: PARTICLE SWARM OPTIMIZATION-BASED MAXIMUM POWER POINT TRACKING ALGORITHM

1031

VPV and current IPV can be measured and filtered using digital finite impulse response filters. These values
can then be utilized to calculate the fitness value PPV
of particle i. It should be noted that in order to acquire
correct samples, the time interval between successive
particle evaluations has to be greater than the power
converters settling time.
Step 4 (Update Individual and Global Best Data): If the fitness value of particle i is better than the best fitness
value in history pb est,i , set current value as the new
pb est,i . Then, choose the particle with the best fitness
value of all the particles as the gb est . This step is
similar to step 3 of the standard PSO method.
Step 5 (Update Velocity and Position of Each Particle): After
all the particles are evaluated, the velocity and position of each particle in the swarm should be updated.
In conventional PSO method, the update is performed
using (1) and (2), in which the parameters w, c1 , and
c2 are constants. In this paper, the parameters w, c1 ,
and c2 are set as variables to speed up the convergence. Therefore, (1) can be rewritten as
vi (k + 1) = w(k)vi (k) + c1 (k)r1 (pb est,i xi (k))
+ c2 (k)r2 (gb est xi (k)).

(4)

In (3), the first term w(k)vi (k) is utilized to keep

the particle moving in the same direction it was originally heading; therefore, it controls the convergence
behavior of PSO. To accelerate convergence, the inertia weight shall be selected such that the effect of vi (k)
fades during the execution of the algorithm. Thus, a
decreasing value of w with time is preferable. A very
common choice is to initially set the inertia weight
to a larger value for better exploration and gradually
reduce it to get refined solutions. In this paper, a linearly decreasing scheme for w is used, as shown in
(4) [35]
w(k) = wm ax

k
(wm ax wm in )
kM AX

(5)

In (4), wm in and wm ax are the lower and upper

bounds of w, and kM AX is the maximum allowed
number of iterations. Similarly, the cognitive and social parameters can be modified. In (3), the values
of c1 and c2 can affect the search ability of PSO by
biasing the direction of a particle. Choosing c1 > c2
would bias sampling toward the direction of pb est , i,
while in the opposite case, c1 < c2 , sampling toward
the direction of gb est would be favored. In this paper,
these two parameters are defined as linearly decreasing and linearly increasing functions, respectively.
c1 (k) = c1,m ax
Fig. 6.

Flowchart of the proposed algorithm.

c2 (k) = c2,m in +

k
(c1,m ax c1,m in ) (6)
kM AX
k
kM AX

(c2,m ax c2,m in ). (7)

1032

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 27, NO. 4, DECEMBER 2012

TABLE I
PARAMETERS OF THE UTILIZED PV MODULE

TABLE II
SPECIFICATIONS OF THE IMPLEMENTED BOOST CONVERTER

TABLE III
PARAMETER SETTINGS OF THE IMPLEMENTED ALGORITHM

In (5) and (6), c1,m in , c1,m ax and c2 ,m in , c2,m ax are

the lower and upper bounds of c1 and c2 , respectively.
It should be noted that due to the stochastic nature
of PSO, the calculated new position x1 (k + 1) may
become very far away from the previously outputted
position xN (k). Since x(k) represents the duty cycle
command of the utilized dcdc converter, this sudden
duty cycle change will result in very large voltage
stress on the power switch. One method to deal with
this condition is to sort the obtained particle positions
according to the last outputted duty cycle command
in advance, and output the nearest particle first. This
will reduce the voltage stress significantly.
Step 6 (Convergence Determination): Two convergence criteria are utilized in this paper. If the velocities of
all particles become smaller than a threshold, or if
the maximum number of iterations is reached, the
proposed MPPT algorithm will stop and output the
obtained gb est solution.
Step 7 (Reinitialization): Typically PSO method is used to
solve problems that the optimal solution is time invariant. However, in this application, the fitness value
(global maximum available power) often changes
with environments as well as loading conditions. In
such cases, the particles must be reinitialized to search
for the new GMPP again. In this paper, the following
constraint is utilized to detect the insolation change
and shading pattern changes. The proposed PSO algorithm will reinitialize the particles whenever the
following condition as shown in (7) is satisfied:
|PPV ,new PPV ,last |
P (%).
PPV ,last

(8)

V. SIMULATION AND EXPERIMENTAL RESULTS

To verify the correctness of the proposed MPPT method, a
500-W prototyping circuit is implemented from which simulations and experiments are carried out accordingly. The parameters of the utilized PV module are listed in Table I. The
specifications of the implemented boost converter are listed in
Table II. In this paper, the simulations are made using MATLAB.
According to the design guideline illustrated in Section IV, the
parameter settings of the implemented PSO-based MPPT algorithm are listed in Table III. The effect of PSC on PV module
characteristic curves is simulated by arbitrary set the insolation
of the series connected PV cells. In our simulation, cell temperature is assumed to be constant at 25 C. Unshaded cells are con-

TABLE IV
SIMULATION RESULTS OF THE TWO PSO-BASED MPPT ALGORITHM

sidered fully illuminated at 1000 W/m2 . Insolation on shaded

cells is considered uniform and varies from 0 to 1000 W/m2
with a step of 100 W/m2 . In order to verify the effectiveness
of the proposed algorithm, the Monte Carlo method is adopted
in this paper [36]. Two PSO-based MPPT algorithms are tested
for 1000 different shading patterns and the simulation results
are presented in Table IV. In Table IV, the method denoted as
variable-PSO represents the proposed method, and the method
named constant-PSO utilizes the following parameter: w = 1.0,
c1 = 2, and c2 = 2, which remains constant in the whole MPPT
process. In Table IV, the tracking efficiency is defined as
=

PO
100%
PM AX

(9)

where PO is the averaged output power obtained under steady

state and PM AX is the maximum available power of the PV module under certain shading pattern. From Table IV, the proposed
method requires lesser iteration to converge, and the averaged
tracking efficiency is higher.
A prototyping circuit using the parameters shown in
Tables I and II is also constructed. In the prototyping system, the
PSO-based MPPT algorithm is realized using the low-cost DSC
dsPIC33FJ16GS502 from Microchip Corporation. The execution time of the whole procedure is less than 80 s when the CPU
speed is set as 40 MIPS. The implemented algorithm allocates
5.632 kB (35%) of program memory. The proposed algorithm
is validated using the Chroma 62150 H-600 S Solar Array Simulator in SAS mode [37]. Using the SHADOW I-V CURVE
SIMULATION function, the user is able to program various
IV curves which can be used to verify the MPPT capability of
the proposed algorithm. Three different patterns are used to test
the correctness of the proposed method; experimental results of
these shading patterns are shown in Fig. 7 through Fig. 13. Fig. 7
shows the IV and PV curves of the shading pattern 1. In Fig. 7,
the measured MPPT tracking efficiency is 99.6%. Fig. 8 shows

LIU et al.: PARTICLE SWARM OPTIMIZATION-BASED MAXIMUM POWER POINT TRACKING ALGORITHM

Fig. 7.

Fig. 8.

Fig. 9.

1033

IV and PV curves of shading pattern 1 (GMPP at right).

Fig. 10.

IV and PV curves of shading pattern 2 (GMPP at middle).

Fig. 11.

Measured waveforms of shading pattern 2.

Fig. 12.

IV and PV curves of shading pattern 3 (GMPP at left).

Measured waveforms of shading pattern 1.

Zoomed version of Fig. 8.

the measured waveforms of the shading pattern 1. From Fig. 8,

the proposed MPPT algorithm converges after eight iterations.
Fig. 9 shows the zoomed version of the first six iterations of
Fig. 8. From Fig. 9, the standard deviation of the found solution
decreases steadily; therefore, the proposed algorithm can obtain
a converged solution. Fig. 9 also shows that after the sorting process, the first particle V1 of the nth iteration is always closet to
the last particle V5 of the (n 1)th iteration. Similarly, Figs. 10
and 12 show the characteristic curves of shading pattern 2 and
3, and Figs. 11 and 13 show the measured waveforms of shading
pattern 2 and 3, respectively. From these experimental results,
the proposed method can successfully deal with PSCs no matter
where the GMPP locates. Figs. 14 and 15 show the dynamic
tracking capability of the proposed algorithm. Fig. 14 shows the
changing sequence of the test shading patterns. In Fig. 14, the
utilized three shading patterns are the same as those in Figs. 7,
10 and 12. In Fig. 14, each shading pattern lasts 25 s. Fig. 15
shows the measured waveforms for the dynamic tracking test.
From Fig. 15, the proposed algorithm can successfully detect the

shading pattern changes and reinitialize the MPPT process accordingly. Due to limited space, more test results are presented
at https://sites.google.com/site/pvtestrecord/pvtest_record, in
which the test results are recorded as video files.
To compare the performance of the proposed method with
other MPPT techniques, the methods proposed in [16] and [17]
are also implemented. The convergence criteria (dcr in [16]
and perturbation step in [17]) for all the three methods are all
set as 0.3%. The simulations and experiments are conducted
with the same shading patterns as shown in Figs. 7, 10 and 12.
Table V summarizes the obtained results. From Table V, all these
algorithms can successfully track the GMPP with similar MPPT
tracking efficiency. Comparing to the method proposed in [17],
the proposed algorithm requires fewer search steps. Although
the required iteration number of the proposed method is higher
than [16], only the proposed method can successfully obtain the
GMPP in all the 1000 simulation cases.

1034

Fig. 13.

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 27, NO. 4, DECEMBER 2012

Measured waveforms of shading pattern 3.

are also utilized to experimentally validate the correctness of the

proposed system. According to the experimental results, the proposed method can obtain the GMPP in all the test cases no matter
where the GMPP locates. Experimental results also show that
the proposed method can successfully detect the shading pattern changes and reinitialize the MPPT process. The tracking
efficiencies in five test cases are all higher than 99.5%. The proposed technique boasts the following advantages: 1) Comparing
to other GMPP searching methods, the tracking efficiency of the
PSO-based MPPT algorithm is very high. 2) PSO-based method
is a good candidate for MPPT algorithms, as it is easy to implement and converges to the desired solution in a reasonable
time. 3) The proposed method requires knowledge only of the
number of the series cells; therefore, it is system independent.
It is well known that the choice of PSO parameters may have
some impact on optimization performance; this aspect will be
investigated in the future work.
REFERENCES

Fig. 14.

Shading pattern change sequence.

Fig. 15.

Measured waveforms of dynamic tracking tests.

TABLE V
PERFORMANCE COMPARISON WITH OTHER MPPT METHODS

VI. CONCLUSION AND DISCUSSION

The main purpose of this paper is to develop an accurate and
system-independent MPPT algorithm for centralized-type PGS
operating under PSC. The standard version of PSO is modified
to meet the practical consideration of PGS operating under PSC.
The problem formulation, design procedure, and parameter setting method which takes the hardware limitation into account
are described and explained in detail. According to the simulation results of 1000 test cases, the proposed method can reach
the GMPP in less than 27 iterations, and the averaged tracking
efficiency is higher than 99.9%. Four different shading patterns

[1] H. J. Chiu, Y. K. Lo, C. J. Yao, T. P. Lee, J. M. Wang, and J. X. Lee, A

modular self-controlled photovoltaic charger with interintegrated circuit
(I2C) Interface, IEEE Trans. Energy Convers., vol. 26, no. 1, pp. 281
289, Mar. 2011.
[2] G. Farivar and B. Asaei, A new approach for solar module temperature
estimation using the simple diode model, IEEE Trans. Energy Convers.,
vol. 26, no. 4, pp. 11181126, Dec. 2011.
[3] P. Thounthong, Model based-energy control of a solar power plant with
a supercapacitor for grid-independent applications, IEEE Trans. Energy
Convers., vol. 26, no. 4, pp. 12101218, Dec. 2011.
[4] A. Maki and S. Valkealahti, Power losses in long string and parallelconnected short strings of series-connected silicon-based photovoltaic
modules due to partial shading conditions, IEEE Trans. Energy Convers., vol. 27, no. 1, pp. 173183, Mar. 2012.
[5] O. Lopez-Lapena, M. T. Penella, and M. Gasulla, A closed-loop maximum power point tracker for subwatt photovoltaic panels, IEEE Trans.
Ind. Electron., vol. 59, no. 3, pp. 15881596, Mar. 2012.
[6] D. P. Hohm and M. E. Ropp, Comparative study of maximum power
point tracking algorithms, Proc. IEE Photovoltaics: Res. Appl., vol. 11,
no. 1, pp. 4762, Jan. 2003.
[7] V. Salas, E. Olas, A. Barrado, and A. Lazaro, Review of the maximum
power point tracking algorithms for stand-alone photovoltaic systems,
Sol. Energy Mater. Sol. Cells, vol. 90, no. 11, pp. 15551578, Jul. 2006.
[8] T. Esram and P. L. Chapman, Comparison of photovoltaic array maximum
power point tracking techniques, IEEE Trans. Energy Convers., vol. 22,
no. 2, pp. 439449, Jun. 2007.
[9] H. Patel and V. Agarwal, MATLAB-based modeling to study the effects
of partial shading on PV array characteristics, IEEE Trans. Energy
Convers., vol. 23, no. 1, pp. 302310, Mar. 2008.
[10] Y. J. Wang and P. C. Hsu, Analytical modelling of partial shading and
different orientation of photovoltaic modules, IET Renew. Power Gener.,
vol. 4, no. 3, pp. 272282, May 2010.
[11] E. V. Paraskevadaki and S. A. Papathanassiou, Evaluation of MPP voltage
and power of mc-Si PV modules in partial shading conditions, IEEE
Trans. Energy Convers., vol. 26, no. 3, pp. 923932, Sep. 2011.
[12] G. Carannante, C. Fraddanno, M. Pagano, and L. Piegari, Experimental performance of MPPT algorithm for photovoltaic sources subject to
inhomogeneous insolation, IEEE Trans. Ind. Electron., vol. 56, no. 11,
pp. 43744380, Nov. 2009.
[13] K. Kobayashi, I. Takano, and Y. Sawada, A study of a two stage maximum
power point tracking control of a photovoltaic system under partially
shaded insolation conditions, Sol. Energy Mater. Sol. Cells, vol. 90,
no. 1819, pp. 29752988, Nov. 2006.
[14] Y. H. Ji, D. Y. Jung, J. G. Kim, J. H. Kim, T. W. Lee, and C. Y. Won, 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. 10011009, Apr. 2011.
[15] N. A. Ahmed and M. Miyatake, A novel maximum power point tracking
for photovoltaic applications under partially shaded insolation conditions,
Electric Power Syst. Res., vol. 78, no. 5, pp. 777784, May 2008.

LIU et al.: PARTICLE SWARM OPTIMIZATION-BASED MAXIMUM POWER POINT TRACKING ALGORITHM

[16] T. L. Nguyen and K. S. Low, A global maximum power point tracking

scheme employing DIRECT search algorithm for photovoltaic systems,
IEEE Trans. Ind. Electron., vol. 57, no. 10, pp. 34563467, Oct. 2010.
[17] H. Patel and V. Agarwal, Maximum power point tracking scheme for PV
systems operating under partially shaded conditions, IEEE Trans. Ind.
Electron., vol. 55, no. 4, pp. 16891698, Apr. 2008.
[18] H. Renaudineau, A. Houari, J. P. Martin, S. Pierfederici, F. M. Tabar, and
B. Gerardin, A new approach in tracking maximum power under partially
shaded conditions with consideration of converter losses, Sol. Energy,
vol. 85, no. 11, pp. 25802588, Nov. 2011.
[19] P. Lei, Y. Li, and J. E. Seem, Sequential ESC-based global MPPT control
for photovoltaic array with variable shading, IEEE Trans. Sustainable
Energy, vol. 2, no. 3, pp. 348358, Jul. 2011.
[20] Syafaruddin, E Karatepe, and T. Hiyama, Artificial neural network-polar
coordinated fuzzy controller based maximum power point tracking control
under partially shaded conditions, IET Renew. Power Gener., vol. 3, no. 2,
pp. 239253, Jun. 2009.
[21] N. Femia, G. Lisi, G. Petrone, G. Spagnuolo, and M. Vitelli, Distributed
maximum power point tracking of photovoltaic arrays: Novel approach
and system analysis, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp. 2610
2621, Jul. 2008.
[22] A. I. Bratcu, I. Munteanu, S. Bacha, D. Picault, and B. Raison, Cascaded DCDC converter photovoltaic systems: Power optimization issues, IEEE Trans. Ind. Electron., vol. 58, no. 2, pp. 403411, Feb. 2011.
[23] L. Gao, R. A. Dougal, S. Liu, and A. P. Iotova, Parallel-connected solar
PV system to address partial and rapidly fluctuating shadow conditions,
IEEE Trans. Ind. Electron., vol. 56, no. 5, pp. 15481556, May 2009.
[24] L. R. Chen, C. H. Tsai, Y. L. Lin, and Y. S. Lai, A biological swarm
chasing algorithm for tracking the PV maximum power point, IEEE
Trans. Energy Convers., vol. 25, no. 2, pp. 484493, Jun. 2010.
[25] S. R. Chowdhury and H. Saha, Maximum power point tracking of partially shaded solar photovoltaic arrays, Sol. Energy Mater. Sol. Cells,
vol. 94, no. 9, pp. 14411447, Sep. 2010.
[26] M. Miyatake, M. Veerachary, F. Toriumi, N. Fujii, and H. Ko, Maximum
power point tracking of multiple photovoltaic arrays: A PSO approach,
IEEE Trans. Aerospace Electron. Syst., vol. 47, no. 1, pp. 367380, Jan.
2011.
[27] K. Ishaque, Z. Salam, M. Amjad, and S. Mekhilef, An improved particle
swarm optimization (PSO)-based MPPT for PV with reduced steady-state
oscillation, IEEE Trans. Power Electron., vol. 27, no. 8, pp. 36273638,
Aug. 2012.
[28] L. F. Wang and C. Singh, Multicriteria design of hybrid power generation
systems based on a modified particle swarm optimization algorithm,
IEEE Trans. Energy Convers., vol. 24, no. 1, pp. 163172, Mar. 2009.
[29] A. M. A. Amin, M. I. E. Korfally, A. A. Sayed, and O. T. M. Hegazy,
Efficiency optimization of two- asymmetrical-winding induction motor
based on swarm intelligence multicriteria design of hybrid power generation systems based on a modified particle swarm optimization algorithm,
IEEE Trans. Energy Convers., vol. 24, no. 1, pp. 1220, Mar. 2009.
[30] J. S. Heo and K. Y. Lee, A multiagent-system-based intelligent reference
governor for multiobjective optimal power plant operation, IEEE Trans.
Energy Convers., vol. 23, no. 4, pp. 10821092, Dec. 2008.
[31] Q. Li, W. Chen, Y. Wang, S. Liu, and J. Jia, Parameter identification for
PEM fuel-cell mechanism model based on effective informed adaptive
particle swarm optimization, IEEE Trans. Ind. Electron., vol. 58, no. 6,
pp. 24102419, Jun. 2011.
[32] C. F. Juang, Y. C. Chang, and C. M. Hsiao, Evolving gaits of a hexapod
robot by recurrent neural networks with symbiotic species-based particle
swarm optimization, IEEE Trans. Ind. Electron., vol. 58, no. 7, pp. 3110
3119, Jul. 2011.
[33] H. M. Hasanien, Particle swarm design optimization of transverse flux
linear motor for weight reduction and improvement of thrust force, IEEE
Trans. Ind. Electron., vol. 58, no. 9, pp. 40484056, Sep. 2011.
[34] K. E. Parsopoulos and M. N. Vrahatis, Particle Swarm Optimization and
Intelligence: Advances and Applications. Hershey, PA: IGI Global,
2010.
[35] Y. Shi and R. C. Eberhart, Empirical study of particle swarm optimization, in Proc. Evolutionary Comput., 1999, pp. 19451950.

1035

[36] F. Iannone, G. Noviello, and A. Sarno, Monte Carlotechniques to analyse

the electrical mismatch losses in large-scale photovoltaic generators, Sol.
Energy, vol. 62, no. 2, pp. 8592, Feb. 1998.
[37] Data sheet of Solar Array Simulation 62050H-600S, CHROMA ATE
INC. (2012). [Online]. Available: http://www.chromaate.com/Product/
62150H-600S_1000S_Solar_Array_Simulator.htm

Yi-Hwa Liu (M01) received the Ph.D. degree in

electrical engineering from National Taiwan University, Taipei, Taiwan, in 1998.
He joined the Department of Electrical Engineering, Chang-Gung University, Taoyuan, Taiwan, in
2003. He is currently with the Department of Electrical Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan. His current research interests include the areas of power electronics
and battery management.

Shyh-Ching Huang was born in new Taipei, Taiwan,

in 1957. He received the B.S. degree in electronic
engineering from Chung Yuan University, Taipei,
Taiwan, in 1986, and the M.S. degree in electronic
engineering from the National Taiwan University of
Science and Technology, Taipei, Taiwan, in 2007, and
he is currently working toward the Ph.D. degree in
the Department of Electrical Engineering, in the same
university, where he focuses on switching power supply field.
His research interests include power electronics
converter design and renewable energy applications.

Jia-Wei Huang was born in Hsinchu, Taiwan, in

1987. He received the Bachelors and Ph.D. degrees both in electrical engineering from the National
Taiwan University of Science and Technology, Taipei,
Taiwan, in 2008 and 2012, respectively.
He is currently with the National Taiwan University of Science and Technology. His research interests
include photovoltaic system design and maximum
power point tracking technology.

Wen-Cheng Liang received the Bachelors and Masters degrees both from the Department of Electrical
Engineering, National Taiwan University of Science
and Technology, Taipei, Taiwan, in 2009 and 2012,
respectively.
He is currently with the National Taiwan University of Science and Technology. His current research
interest includes photovoltaic system.