You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/324738157

An Overall Distribution Particle Swarm Optimization MPPT Algorithm for


Photovoltaic System under Partial Shading

Article  in  IEEE Transactions on Industrial Electronics · April 2018


DOI: 10.1109/TIE.2018.2829668

CITATIONS READS

3 362

5 authors, including:

Hong Li Wenzhe Su
Beijing Jiaotong University Beijing Jiaotong University
75 PUBLICATIONS   330 CITATIONS    5 PUBLICATIONS   6 CITATIONS   

SEE PROFILE SEE PROFILE

Jinhu Lu Xinghuo Yu
Academy of Mathematics and Systems Science, Chinese Academy of Sciences RMIT University
229 PUBLICATIONS   14,200 CITATIONS    762 PUBLICATIONS   17,224 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Call for Papers ESREF 2018 Special Issue on Reliability in transportation applications View project

Photovoltaic system and maximum power point tracking algorithm View project

All content following this page was uploaded by Hong Li on 02 May 2018.

The user has requested enhancement of the downloaded file.


This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

An Overall Distribution Particle Swarm


Optimization MPPT Algorithm for
Photovoltaic System under Partial Shading
Hong Li, Senior Member, IEEE, Duo Yang, Wenzhe Su,
Jinhu Lü, Fellow Member, IEEE and Xinghuo Yu, Fellow Member, IEEE

such as solar irradiance and temperature, the maximum power


Abstract—Solar photovoltaic (PV) systems under Partial point (MPP) of a PV panel will vary along with these external
Shading Conditions (PSC) have a non-monotonic P-V factors. Therefore, effective MPPT technique is of great
characteristic with multiple local maximum power points, importance for efficiency improvement of PV power
which makes the existing Maximum Power Point Tracking generation systems.
(MPPT) algorithms unsatisfactory performance for Global In a PV system, each PV array consists of many PV panels
Maximum Power Point Tracking (GMPPT), if not invalid.
connected in series and parallel to exert high voltage and high
This paper proposes a novel Overall Distribution (OD)
MPPT algorithm to rapidly search the area near the global current, for purpose of increasing output power of PV arrays, as
maximum power points, which is further integrated with shown in Fig.1. There are two types of diodes in a solar power
the Particle Swarm Optimization (PSO) MPPT algorithm to system, which are bypass diodes and blocking diodes for
improve the accuracy of MPPT. Through simulations and different purpose. Blocking diodes prevent backflow of
experimentations, the higher effectiveness and accuracy of electricity. Bypass diodes are used to inhibit hot-spot heating
the proposed OD-PSO MPPT algorithm in solar PV systems effect and reduce the power loss due to shading. This situation
is demonstrated in comparison to existing two Artificial is defined as PSC where each panel may receive and experience
Intelligence (AI) MPPT algorithms. different solar irradiations and temperatures at the same time
Index Terms—Maximum power point tracking, solar PV
[5-9]. When a PV array is under PSCs, its P-V (power-voltage)
systems, particle swarm optimization, partial shading
conditions characteristic curve will have multiple peaks. Among all the
peaks, there is only one Global Maximum Power Point
I.INTRODUCTION (GMPP), other peaks are called Local Maximum Power Points

I N recent years, the rapid consumption of fossil fuels has (LMPPs). In the literature, many MPPT algorithms [11-29]
resulted in severe environmental pollution, making have been proposed to track the GMPP of PV systems under
renewable energy sources a more favorable option in PSCs.
Blocking diode
generating electricity due to their inexhaustibleness and I +
Panel 11 Panel 1N
environmental friendliness. Solar energy is considered one of
the most important renewable energy sources and has been PV cell
V
widely used in photovoltaic power generation [1-5]. However,
the low efficiency and high cost of PV systems impedes the Panel M 1 Panel M N
Bypass diode
development of PV power generation. Since the output −
characteristics of PV panels are influenced by external factors, Fig. 1. Structure of the PV array
It is well understood that the traditional MPPT algorithms,
Manuscript received October 3, 2017; revised December 1, 2017 and
February 8, 2018; accepted April 2, 2018. This work was supported in
such as Perturbation and Observation (P&O) algorithm,
part by the National Key Research and Development Program of China Incremental Conductance (IC) algorithm and Constant Voltage
under Grant No. 2016YFB0800401, by the General Programs of the (CV) algorithm, cannot well track the GMPP under PSC [2]. To
National Natural Science Foundation of China under Grants 51577010, solve the GMPPT problem, many other algorithms have been
51777012, 61621003, 61532020, by the Fundamental Research Funds
for the Central Universities under Grant 2017JBM054 (Corresponding
proposed. The algorithm presented in [10] is an improvement
author: Hong Li). of the IC algorithm, which can find the GMPP by identifying all
Hong Li and Wenzhe Su are with the School of Electrical Engineering, LMPPs and GMPP, but it requires a wide range of information
Beijing Jiaotong University, Beijing, China (e-mail: hli@bjtu.edu.cn; about the PV arrays, such as open-circuit voltage and
17117418@bjtu.edu.cn)
Duo Yang is with the China Academy of Railway Sciences
short-circuit current of each PV panel. The GMPPT algorithm
Corporation Limited, Beijing, China (e-mail: yangduo@bjtu.edu.cn). proposed in [5] is to scan the P-V curve by increasing the output
Jinhu Lü is with the School of Automation Science and Electrical voltage of PV array step by step, which is helpful to improve
Engineering, State Key Laboratory of Software Development the transient response of the PV system and increase the
Environment, and Beijing Advanced Innovation Center for Big Data and
Brain Machine Intelligence, Beihang University, Beijing 100083, China,
tracking speed. However a high-precision, fast-speed digital
also with the Academy of Mathematics and Systems Science, Chinese processor is required for the implementation of this algorithm,
Academy of Sciences, Beijing 100190, China, and also with the which will increase the cost of the overall PV system. The
University of Chinese Academy of Sciences, Beijing 100049, China proposed MPPT algorithm in [11] is to first estimate the
(e-mail: jhlu@iss.ac.cn).
Xinghuo Yu is with RMIT University (Royal Melbourne Institute of
possible GMPP according to the open-circuit voltage of the PV
Technology), Melbourne, Australia (e-mail: x.yu@rmit.edu.au). array and then use the P&O-based MPPT algorithm to track the

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

GMPP. The MPPT algorithm in [11] has good dynamic MPPT algorithm is proposed in Section III, followed by its
tracking performances, however, due to its algorithmic nature operating principle. In Section IV MPPT simulation results
this algorithm could be invalid for GMMP under complex PSCs. using the proposed OD-PSO MPPT and other AI-based
In [12-16], the regions of all MPPs are firstly calculated to algorithms are presented, compared and analyzed.
decrease the tracking time, which however requires excessive Experimental results are provided in Section V. Finally,
information about the structure of PV array. These values are conclusions are drawn in section VI.
difficult to obtain or predict in real-world engineering practices.
Furthermore, MPPT can be considered an optimization II.CHARACTERISTICS OF PV ARRAY UNDER PSC
problem which can be solved by using AI-based MPPT Suppose a M × N PV array is composed of M PV panels
algorithms [17-26]. Popular AI-based MPPT algorithms used connected in parallel and N PV panels connected in series to
for MPPT include particle swarm optimization algorithm, obtain high current and high voltage in an electric circuit, as
firefly algorithm and ant colony optimization algorithm. shown in Fig. 1. A PV panel is composed of a number of PV
Compared with the traditional MPPT methods, the main cells, which can be represented as an equivalent circuit, as
advantage of AI-based MPPT algorithms is that these shown in Fig. 2.
algorithms can find the GMPP under PSCs without a global np ns Rs / n p




scan of P-V curve, which requires less sampling points in the S
I +
process and thus entails lower computational costs. PSO 
algorithm, has been widely used for MPPT due to its simplicity n 
s ns Rsh / n p V
in mathematical structure and implementation [22-24]. In [23], np I g  −
the problems of non-convergence and failure to capture the
Fig. 2. Equivalent circuit model of a PV Panel
GMPP in PSO algorithm have been pointed out, and a GMPPT
method was proposed based on power closed-loop, which According to Fig. 2, the I-V characteristics of a PV panel can
however could increase the complexity of the control design. In be expressed by (1) [28, 29], where A is the ideal diode factor, k
[18], an improved firefly MPPT algorithm was proposed to is the Boltzmann’s constant, T is the temperature in Kelvin, q is
increase the MPPT speed by simplifying the movement rules of the electron charge, Rs and Rsh are the equivalent series and
fireflies in the algorithm. In recent years, simulated annealing shunt resistances, and Isc is the saturation current of the diode. V
algorithm and ant colony algorithm have also been proposed for and I are the output voltage and current of PV array.
the MPPT [25, 26] in PV systems. In [36, 37], improved Java   q (V + I ⋅ Rs ⋅ nns )   n V / n + IR
  (1)
I= n p I g − n p I sc exp  p
 ) − 1 − p s s

MPPT algorithms were proposed without algorithm-specific  n AkT  R


  s
  sh
parameters.
According to the experimental results in [25, 26], suitable Under PSC, the shaded panels cannot produce power and
initial values are very important for improving the tracking become a load, which may create hot-spot heating effect and
speed of the AI-based MPPT algorithms. The determination of even cause damage to the entire PV panel [28]. To solve the
a small region containing the GMPP, with which the initial above problem, a PV panel is normally connected in parallel to
values of AI-based MPPT algorithms can be assigned, is thus a bypass diode, as shown in Fig. 1. Therefore, the shaded panels
important to increase the tracking speed [24, 28, 29]. Several can be bypassed and then be protected by the anti-paralleled
strategies in finding this region are documented in [24, 28, 29] diode under PSC, which results in the multiple peaks in P-V
where however the detailed information of each PV panel, such characteristic curve of PV array [8]. Furthermore, the existing
as the open-circuit voltage and short-circuit current, is needed. results have shown that the multiple peaks in P-V characteristic
This may reduce the feasibility of implementing these are related to the PV panels connected in series in the PV array
algorithms into hardware, due to the necessity of under PSCs [8]. Therefore, in this paper, a PV array composed
pre-implementation hardware testing. of three panels in series is chosen as the testbed to validate the
Having well understood the drawbacks of these proposed OD-PSO MPPT algorithm. Three cases with different
aforementioned existing MPPT algorithms, in this paper, we solar irradiance of the PV array are illustrated in Fig. 3. The
propose a novel OD-PSO MPPT method, which does not irradiances are shown in TABLE I. Case III is used as a control
require any hardware information of PV systems and is able to group where the irradiance is identical for all solar panels. The
search and find the GMPP accurately and rapidly. Particularly, P-V characteristics under these three cases shown in Fig. 3 are
the OD MPPT algorithm is employed to locate the vicinity of simulated in PSIM and the simulation results are shown in Fig.
the GMPP region, which facilitates the setting of initial values 4.
S S S
that will be imported to the PSO MPPT algorithm. After Panel 1 Panel 1 Panel 1
obtaining the initial values, the PSO MPPT algorithm will only S S S
need to find the GMPP within a very small search region, Panel 2 Panel 2 Panel 2
enabling it to find the exact GMPP in a rapid manner. Therefore, S S S
the integration of the OD MPPT algorithm and PSO MPPT Panel 3 Panel 3 Panel 3
algorithm is capable of quickly and accurately tracking and Case I Case II Case III
identifying the GMPP of PV panels in diverse PSCs. Fig. 3. The PV array composed of three panels in series
The remainder of this paper is organized as follows: the
output characteristics of the PV arrays and mathematical
modeling under PSC are introduced in Section II. The OD-PSO

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

TABLE I random number of particle i at the (j+1)th iteration, γ is the scale


IRRADIANCES IN THIS STUDY
parameter is a random number.
Irradiance of PV panels (W/m2) According to the OD algorithm and (2), particles di j can be
Situation
Panel 1 Panel 2 Panel 3 set in a certain area around the GMPP directly due to Cauchy
Case I 1000 500 500 distribution. Furthermore, this region can reduce adaptively and
Case II 1000 600 300 OD MPPT algorithm will end until the region is smaller than ɛ
Case III 1000 1000 1000
(ɛ>0).
It is obvious that in Case I, PV array receives two different To introduce the OD MPPT algorithm, a small PV power
radiation intensities, the irradiance of the shaded PV panels is generation system with a PV array and a buck converter is taken
only 400W/m2. There is a voltage difference between the panel as an example, which is shown in Fig. 5. In this experiment, we
1 and panel 2 (or panel 3), which will turn on the bypass diodes model the inverter as a resistive load, without using an actual
of panel 2 and panel 3 [8]. Thus, there will be two MPPs on the inverter to simplify the experimental setup and loosen the
P-V characteristic of the PV array, as shown in Fig. 4. In this assumptions. Modeling the inverter as a resistive load is to
case, the GMPP is located at GP1. In Case II, three PV panels of account for the power transmitted into the DC/AC inverter,
the PV array receive three different irradiances, which leads to which can sufficiently reflect the mechanism of the inverter for
three MPPs on the P-V characteristic curve and the GMPP is the purpose of this study [26, 27]. For this PV power generation
now located at GP2, as shown in Fig. 4. In case III, PV array system, the duty cycle (d) of buck converter is chosen as the
works without any shading, so in Fig. 4 there is only one power particle of the OD MPPT algorithm. Term Rj Cauchy radius
peak on the P-V characteristic curve, namely GP3. represents the distribution region of the particles in the OD
400 Case I GP3 MPPT algorithm at jth iteration. Rj is between 0 and 1. The
300 Case II positions of particles are updated by Rj, Cauchy and
Case III
P (W)

200 GP2 GP1 mppt_gbestj. The position of particle di j , Cauchy radius R j and
100 the Cauchy distribution random number Cauchy of the OD
MPPT algorithm, can be defined by (2). Term mppt_gbestj
0 20 40 60
V (V) means the existing best position of particles in the OD MPPT
Fig. 4. P-V characteristic curves of the PV array under PSC algorithm within j iterations. The best position of particles
It can be seen that both the number of the MPPs and the peak indicates that the fitness value of the position is the maximum
values of the P-V characteristic differ from case to case. one among these particles. The fitness function of the PV power
Therefore, it is very important to identify the GMPP for PV generation system can be expressed as (3).
systems under PSCs.
+
I Q L
III.THE PRINCIPLE OF OD-PSO MPPT ALGORITHM V
C1 D C2
Although the AI-based MPPT algorithms can be used to
accurately capture the GMPP. They are sensitive to initial -
condition. Optimizing the initial values using a variety of
measures [25, 26] can improve the performance of the AI-based d
MPPT algorithms. In the following, an OD MPPT algorithm
will be proposed to shrink the search region of GMPP, which Fig. 5. A small PV power generation system
obtains the initial particles for the AI-based MPPT algorithms.
P(di j +1 )= V ⋅ I (3)
Therefore the two algorithms are able to complement each
other and achieve a better MPPT performance. Terms V and I are the output voltage and current of PV array.
In general, the P-V characteristic curve of PV array and the
A. OD MPPT Algorithm relationship among the parameters in the OD MPPT algorithm
The OD algorithm is an intelligent algorithm which was first is depicted in Fig. 6.
proposed in 2012 [30]. OD algorithm has been successfully di j +1 mppt _ gbest
j

applied to solve the short-term optimization dispatch problems j +1


P(di )
P (W)

for hydro thermal power systems [30, 31]. In this paper, the OD R j ⋅ Cauchy
MPPT algorithm is employed to rapidly find the rough search
region around GMPP. The mathematical expression of the OD V (V)
MPPT algorithm is given in (2). Fig. 6. The P-V characteristic of the PV array
 R j +1= α ⋅ R j The OD MPPT algorithm starts by initializing the duty
 j +1 cycles of the buck converter, which are noted as di j (i = 1, 2, N ) .
= di mppt _ gbest j + R j +1 ⋅ Cauchyij +1
 (2) The corresponding power at each duty cycle can be calculated
Cauchyi =−γ ⋅ tan(π ⋅ r1 )
j +1
by (3). The maximum fitness value will be saved and its
= i 1, 2,...N ,Cauchy ∈ [−2, 2],α ∈ (0,1) corresponding duty cycle is set as lbest j
. According to the
Term di is the position of the ith particle at the (j+1)th
j +1
definition of mppt_gbest , mppt_gbestj need to satisfy the
j

iteration, Rj is the Cauchy radius in jth iteration, α is the following relation:


shrinking coefficient, Cauchyij +1 is the Cauchy distribution P (mppt _ gbest j ) ≥ P (lbest
k
) (k=1, 2, …, j) (4)

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

This generation of duty cycles will be updated to track the Iteration 1: OD MPPT algorithm sends these three initial
GMPP, and the updated duty cycles di j +1 (i = 1, 2, N ) , can be particles to the buck converter and the corresponding output
calculated by (2). The new duty cycles are distributed around power of PV array will be calculated. As shown in Fig. 8 (a), x12
mppt_gbestj within a region of radius R (R ∈ [0,1]), which is the best particle of the present particles. Term x12 is denoted
follows the Cauchy distribution. R is reduced by a shrinking as lbest
1
, and according to the definition in (4), lbest1
is also
coefficient α. The above process is repeated until the radius R<ɛ
(ɛ>0) is satisfied. mppt_gbest . 1

In the OD MPPT algorithm, Cauchy distribution, a Iteration 2: In accordance with the obtained mppt_gbest1,
continuous probability distribution, is selected to update the particles of the next generation are generated by (2). Particles
particles [30]. The Cauchy distribution function can be are distributed as shown Fig. 8 (b). It is obvious that x32 is lbest
2
.
expressed as (5) [30]. Comparing lbest
2
with mppt_gbest1,it can be found that lbest 2
is
1 π the best particles until now. Since P(lbest ) > P(mppt _ gbest1 ) ,
2
F= (Cauchy ) (arctan(Cauchy − x0 ) / γ + ) (5)
π 2 mppt_gbest2 will be replaced by lbest 2
and Cauchy radius Rj will
Term x0 is the position parameter used to define the
decrease according to (2). If the shrinking coefficient α is set as
distribution peak position, and γ is the scale parameter.
0.75 to meet the tracking performance, the Cauchy radius will
Cauchy distribution has its own probability density function,
be R2=0.45·0.75=0.3375 in this iteration.
which is given as (6) [30]. In this paper, x0 is chosen as 0.
Iteration 3: Repeat the process in iteration 2, and generate
1 particles of the third generation, which are distributed as shown
f (Cauchy; x0 , γ ) = (6)
Cauchy − x0 2 in Fig. 8 (c). It is obvious that x23 is the best particle at iteration
π ⋅ γ [1 + ( ) ]
γ 3, so lbest
3
is set as x23 . Since P(lbest 3
) > P (mppt _ gbest 2 ) ,
Because the position of particles di in the OD MPPT j
mppt_gbest3 is replaced by lbest 3
. And the Cauchy radius will
algorithm is a linear change of Cauchy, di j has the same further shrink to R =0.3375·0.75.
3

distribution rule as Cauchy. The probability density ρ of Iteration 4: The fourth generation particles obtained by the
Cauchy distribution is shown in Fig. 7. same process are shown as Fig. 8 (d). From Fig. 8 (d), it can be
0.7
0.6
γ =0.5 seen that x24 is the best particle in iteration 4. Therefore, lbest4
is
γ =1
0.5 set as x24 . Since P(lbest
4
) < P (mppt _ gbest 3 ) , mppt_gbest4 is the
γ =2
0.4
ρ

0.3 same as mppt_gbest3. And the Cauchy radius now is


0.2 R4=0.2531·0.75<ɛ (ɛ=0.2).
0.1
Interation1 Interation 2
0 di j mppt _ gbest j d i j+1 GMPP x32 = mppt _ gbest 2
300 300
d LMPP x12
x22
P (W)

Fig. 7. Cauchy distribution function diagram under different γ 200 P (W) 200
It can be seen that the probability is more concentrated
100 x11 x12 = mppt _ gbest1
100 x11 x12
toward to the mppt_gbestj when γ is smaller and the probability x31 x31
of particles in Cauchy distribution is more dispersed when γ is 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
greater. [30]. Therefore, the closer to the mppt_gbestj, the d d
(a) (b)
higher the probability density of particles [30]. In the algorithm, Interation 3 x 3 = mppt _ gbest 3 Interation 4 x 3 = mppt _ gbest 4
the particles distributed near the mppt_gbestj can be used to 300
2
2 300
2

improve the accuracy of GMPPT and particles far from the x22 x 3 x24 x14
P (W)
P (W)

mppt_gbestj can help the OD MPPT algorithm avoid tracking 200 x13 200 x13
the LMPPs. Therefore in theory, the OD MPPT algorithm has 100 x12 x33 100 x34 x33
the ability to find the GMPP. 0 0
To better interpret the proposed OD MPPT algorithm, a 0 0.2 0.4 0.6 0.8 1
d
0 0.2 0.4 0.6 0.8 1
d
non-monotonic P-d curve with two MPPs is given as an (c) (d)
example, which is shown in Fig. 8. And the operating principle, Fig. 8. Operating principle of OD MPPT Algorithm
i.e., the iteration steps are given as follows: According to the OD MPPT algorithm, the mppt_gbest4 at
First, let N=3 for the OD MPPT algorithm. In general, the final iteration and other two particles will be considered as
GMPP will not occur at d=0 or d=1. Three initial particles are the optimal particles and mppt_gbest4 will be regarded the Gbest
1
.
given as follows: x1=0.05, x2=0.5, x3=0.95. In Fig. 8, x1, x2, x3 The flow chart of the OD MPPT algorithm is shown in Fig. 9.
are marked by dot, triangle and star, respectively. In simulation In addition, in order to prevent the missing of GMPP, a
and experimentation, Cauchyij (i = 1, 2,3) is produced by (2). But stagnation coefficient b is added to the proposed OD MPPT
now, in order to facilitate the description of the proposed algorithm, which is used to define the shrinking time of Cauchy
algorithm, Cauchyij (i = 1, 2,3) are replaced by three fixed values, radius [30]. The increased values of stagnation coefficient are
i.e., 0.03, -0.08, 0.6 respectively. The initial value of Cauchy marked in different colors as shown in Fig. 9. Using Fig. 8 (a)
radius R1 is usually set as half of interval of the initial particles. as an example, the duty cycles di1 (i = 1, 2,3) is updated by (2).
So initial value of R1 is chosen as ((0.95-0.05)/2=)0.45. The particles generated at the new generation di2 (i = 1, 2,3) may

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

be distributed as Fig. 10 (a) and may also be distributed as Fig. rapidly and do not need the information about the structure of
10 (b). In Fig. 10 (a), the mppt_gbest2 will change to x32 , then PV array, such as open-circuit voltage and short-circuit
the Cauchy radius remains unchanged and di2 (i = 1, 2,3) will be currents of each PV panel. This feature can solve the problem
mentioned in the Introduction for the improved AI-based
updated by (2). If the distribution of the three particles is shown
MPPT algorithms. Due to the tracking effectiveness and
as Fig. 10 (b), the mppt_gbest2 will not change to xi2 . Term structural simplicity of the PSO MPPT algorithm, it is selected
mppt_gbest2 remains the same as mppt_gbest1. The stagnation to be integrated with OD MPPT algorithm which is branded
coefficient b subtracts 1 and di2 (i = 1, 2,3) is updated OD-PSO MPPT algorithm in this paper.
accordingly. If the stagnation coefficient decreases to 0, the B. OD-PSO MPPT Algorithm
Cauchy radius is calculated as R2=0.75·R1 and the stagnation
coefficient b will be set to its original value. If the stagnation PSO algorithm is proposed in 1995 by the Dr. Kennedy and
coefficient b does not decrease to 0, the Cauchy radius remains Dr. Eberhart [32]. Later, PSO algorithm has been widely used
in optimal control designs, and PSO MPPT algorithm is firstly
unchanged and di2 (i = 1, 2,3) is updated according to (2).
proposed in [28]. The mathematical expression of PSO MPPT
Stagnation coefficient is used to find the best particles and algorithm is given in (7).
ensure the location of GMPP can be found at a certain Cauchy
vi = ω vi + c1r2 ( Pbesti − xi ) + c2 r3 (Gbest − xi )
 k +1 k k k
radius. Normally, the stagnation coefficient is a positive  k +1 (7)
k k +1
integer , which can be set to 3 to 5. A larger stagnation  xi = xi + vi

coefficient leads to a larger number of iterations, which will Term k is the iteration number, ω is the inertia weight, r2 and
deteriorate the fluctuations in the output power of the PV r3 are random numbers within [0, 1], and c1 and c2 are the
system. Therefore, there is also a trade-off between the tracking cognitive coefficient and social coefficient, respectively. Term
speed and tracking accuracy in OD MPPT algorithm. vik represents the offset vector of the particle i at kth iteration
Start
which is usually limited below a maximum vmax [28], and xik
Initializing:= j
di (i 1, 2,=
5), R 0.45,
= b 3
represents the position vector of the particle i at kth iteration. In
Update P(dinj )= V ⋅ I , i= 1, 2, 5
OD MPPT each iteration, the position and the offset of the particles are
Update lbest
j updated by the personal best (Pbesti) position of the particle and
the global best (Gbest) position of the particle swarm,
j
P(lbest ) > P(mppt _ gbest j −1 )?
YES
respectively.
NO The position and offset of the particles are updated as follows
mppt _ gbest j = mppt _ gbest j −1
b= b − 1 [28]. From (7), it is obvious that the position of particles xi k is
NO
b=0? mppt _ gbest j = lbest
j changed by the offset vik to move closer to the GMPP. The PSO
j +1
YES
MPPT algorithm enables its particles to converge to the GMPP
R= 0.75 ⋅ R j
b=3 when Pbesti=Gbesti. In order to avoid the particles of PSO MPPT
j +1
algorithm converging to an LMPP, the maximum offset vmax is
=di mppt _ gbet + R j +1 ⋅ Cauchy
= j +1
j
i ,i 1, 2, 5
always restricted to a small value [5, 33] and according to [5],
NO
R j +1 < ε ? vmax can be estimated by vmax=1/Voc, where Voc is the
YES open-circuit voltage of PV array. That is to say the particle
Output the newest generation of
j +1
duty cycles d i (i = 1, 2, 5) approaches the GMPP at vmax or a slower speed, which
decreases the tracking speed of PSO MPPT algorithm
end
significantly. Therefore, although PSO MPPT algorithm has
Fig. 9. The flow chart of OD MPPT algorithm high accuracy on tracking the GMPP [8], the tracking speed of
It should be noticed that, according to Rj+1=α·Rj and PSO MPPT algorithm is not satisfactory. With the initial
=di mppt _ gbest j + R j +1 ⋅ Cauchy , R j +1 ⋅ Cauchy will gradually
j +1
particles obtained by the OD MPPT algorithm, the PSO MPPT
reduce since α<1, and the generated particles will move algorithm can further be used to track the GMPP quickly and
closer to the obtained mppt_gbest in each iteration This accurately.
means di j +1 will decrease over iterations and the GMPPT speed In the OD-PSO MPPT algorithm, the OD MPPT algorithm is
first used to rapidly obtain the particles, which are within a
the will gradually slow down.
small region that contains the GMPP. The obtained particles
x32 = mppt _ gbest 2 x12 = mppt _ gbest 2
300 300 will be used as the initial particles for PSO MPPT algorithm,
x12
P (W)

and finally the GMPP will be captured by PSO MPPT


P (W)

200 200
x12 x12 algorithm. Since the GMPP of the PV array will change under
x11 x22 x11 x22 x32
100 1 100 different irradiances, the OD-PSO MPPT algorithm needs to
x
3 x31
0 0 restart to track the new GMPP. In order to verify the dynamic
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
d d performances of the OD-PSO MPPT algorithm, the criterion
(a) (b) for judging the extent of solar irradiance variation can be
Fig. 10. Operating principle of OD MPPT Algorithm expressed by (8).
From the above analysis, it is obvious that the OD MPPT
algorithm can find the small region containing the GMPP

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Pn +1 − Pn The parameters of PSO MPPT algorithm are ω=0.4, c1=0.2,


≥δ (8) c2=2, vmax = 0.01, to ensure it converges to either LMPPs or
Pn
GMPP [32]. The selection of cognitive and social parameters
Term Pn denotes the n output power of PV array, δ is the
th
are adopted and modified from [34]. According to [5], the
threshold value and set as 0.02 in this paper according to [5]. sampling time of V and I needs to be longer than the transient
The flow chart of the OD-PSO MPPT algorithm is shown in response time of the PV system. The transient response time
Fig. 11. The flow chart of OD MPPT algorithm is the same as means the time duration that the converter stabilizes after a new
Fig. 9. Further, the term jmax is used to set the maximum duty cycle is implemented. And the transient response time is
iteration number of the PSO MPPT algorithm for eliminating determined by the main circuit parameters. In this paper, the
the steady-state oscillation. Of course, too small jmax may also transient response time is within 1ms, and therefore the
lead to accuracy problems. The choice of jmax requires a sampling time is set to be 5ms.
trade-off between accuracy and suppression of steady-state In order to compare the tracking performance of the
oscillations. proposed OD-PSO MPPT algorithm with other AI-based
Start MPPT algorithms, the firefly MPPT algorithm in [18], denoted
OD MPPT algorithm OD MPPT as firefly MPPT algorithm in this paper, and the MPPT
algorithm based on P&O combined with PSO in [24], denoted
Initializing: di j +1 (i = 1, 2 5),=j 1,=
w 0.4,=
c1 0.2,=
c2 2
as P&O-PSO MPPT algorithm in this paper, are implemented
PSO MPPT Update P(di j +1 )= V ⋅ I for the comparison in this paper. The simulation results with the
algorithm
Find and Update Pbesti , Gbest proposed OD-PSO MPPT algorithm, the firefly MPPT
algorithm and the P&O-PSO MPPT algorithm under different
Calculate vij +2and di j +2
PSCs are carried out as follows:
j > jmax NO
j= j + 1 Case I: The irradiance of PV panels is given in Table I. The
YES P-V characteristic curve is shown in Fig. 4. From Fig. 4, it can
Output mppt _ gbest jmax

be seen that the output voltage of the PV array for Case I at the
Is irradiation changed?
NO GP1 is 49.86V and the power of the PV array is 159.94W.
YES The simulation results for Case I are shown in Fig. 12. In Fig.
Fig. 11. The flow chart of OD-PSO MPPT algorithm 12 (a), the output waveforms of PV system using the proposed
From the above analysis, it can be seen that the OD-PSO OD-PSO MPPT algorithm and the firefly MPPT algorithm are
MPPT algorithm has the following advantages: shown. With proposed OD-PSO MPPT algorithm, it takes
1) It can improve the tracking speed of MPPT compared with 0.210s, 8 iterations to successfully track the GMPP.
AI-based MPPT algorithm. Particularly, 5 iterations are used for OD MPPT algorithm and
2) It can rapidly find the small region that contains the GMPP 3 iterations for PSO MPPT algorithm. With the firefly MPPT
without the information about the structure of PV array, algorithm, it takes 0.206s, 8 iterations. In Fig. 12(b), the output
which demonstrates a higher accuracy in finding GMPP. waveforms of PV system using the proposed OD-PSO MPPT
algorithm and the P&O-PSO MPPT algorithm are
IV. SIMULATION RESULTS demonstrated. It takes 0.286s, 15 iterations, to acquire the
In order to validate the MPPT performance of the proposed GMPP with the P&O-PSO MPPT algorithm. Therefore, the
OD-PSO MPPT algorithm, the simulation platform is proposed OD-PSO MPPT algorithm has almost the same
established as depicted in Fig. 5 using PSIM and MATLAB. tracking speed with the firefly MPPT algorithm and has the
The duty cycle d of the buck converter is taken as the particle in better MPPT tracking speed comparing with the P&O-PSO
the proposed OD-PSO MPPT algorithm. The circuit parameters MPPT algorithm.
of buck converter are as follows: C1=50μF , C2=22μF , OD-PSO MPPT algorithm OD-PSO MPPT algorithm
Firefly MPPT algorithm in [18] P&O-PSO MPPT algorithm in [24]
L=600μH,R=1.7Ω, and switching frequency is 20kHz. The 200 200
P (W)

P (W)

parameters of the PV module are shown in TABLE II. In the 100 160
100
simulation, ɛ is set as 0.035, and γ is set as 0.1 for OD MPPT 0
158
0.22 0.26
0
algorithm, by referring to [30]. The initial number of particles is 4 4
I (A)

I (A)

5, and the initial particles of OD-PSO MPPT algorithm are 2 2


uniformly distributed within [0, 1], which are chosen as x1=0.1, 0
60
0
60
x2=0.3, x3=0.5, x4=0.7, x5=0.9. And the jmax is 10 for PSO MPPT
V (V)

40 40
V (V)

algorithm. 20 20
0 0
TABLE II
PARAMETERS FOR THE PV MODULES 0.8 0.8
d

0.4 0.4
d

Parameters Values 0
0 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4
Maximum power(Pmax) 101.3W Time(s) Time(s)
Open circuit voltage(Voc) 20.2V
(a) (b)
Maximum power voltage(Vmp) 16.7V
Fig. 12. The output waveforms of the PV system in Case I. (a) the
Short circuit current(Isc) 6.719A
OD-PSO MPPT algorithm and the firefly MPPT algorithm in [18]; (b) the
Maximum power current(Imp) 6.074A
OD-PSO MPPT algorithm and P&O-PSO MPPT algorithm in [24].
Current temperature coefficient -0.379%/°C
Voltage temperature coefficient 0.0396%/°C

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Here, the tracking time is defined as the time from 0 to the shown. With the proposed OD-PSO MPPT algorithm, it takes
moment when the fluctuation of P is less than 5%. In this paper, 0.185s, 7 iterations, for acquiring the GMPP, wherein 5
we implemented a threshold value of 5% to indicate the system iterations are used for OD MPPT algorithm and 2 iterations for
has reached its steady state. PSO MPPT algorithm. With the firefly MPPT algorithm, it
In order to thoroughly compare the tracking performances of takes 0.206s, 8 iterations. In Fig. 13(b), the output waveforms
these three MPPT algorithms, a variable Pave is introduced to of PV system using the proposed OD-PSO MPPT algorithm
represent the average output power of the PV system with and the P&O-PSO MPPT algorithm are shown. It takes 0.266s
MPPT algorithm during time duration T. Pave can be expressed with 13 iterations for successfully tracking the GMPP with the
in (9): P&O-PSO MPPT algorithm.
T Num Additionally, according to the simulation results in PSIM,
∫ pdt ∑ p ⋅ ∆t i Pave is 114.12W, 108.74W and 113.9W, while the MPPT
Pave
= 0
= i =1
(9) efficiency is 99.83%, 99.73% and 99.82%, using the proposed
T Num ⋅ ∆t
the OD-PSO MPPT algorithm, the firefly MPPT algorithm and
Term pi is the instantaneous power, Δt is a constant sampling
the P&O-PSO MPPT algorithm respectively, which is given in
time, Num is the total sampling number. Term T, the tracking
TABLE III. It is obvious that the proposed OD-PSO MPPT
time is defined as the duration from the starting of the MPPT
algorithm also has the best overall performance with the highest
algorithm to the moment when the GMPP is captured.
Pave and fastest tracking speed in Case II. Furthermore, the
Further, the tracking performances of these three MPPT
power fluctuation with the proposed OD-PSO MPPT algorithm
algorithms can also be described by MPPT efficiency η, which
is much lower than that with the firefly MPPT algorithm and
can be calculated by (10).
the P&O-PSO MPPT algorithm, as shown in Fig. 13. Therefore,
P the PV system has a better dynamic performance with a higher
η = 1 × 100% (10)
P2 output power.
Term P1 means the output power in stable mode of the PV Please note that the proposed OD-PSO MPPT method is able
system under OD-PSO MPPT algorithm. P2 is the theoretical to rapidly and accurately track and find the GMPP in various
maximum output power of the PV system under certain PSC. PSCs, regardless of the GMPP position. In this study, without
Since the simulation data in PSIM can be exported, Pave and the loss of generality, we have implemented two case studies to
the MPPT efficiency can be calculated using (9) and (10). Pave test the functionality of the proposed method. The GMPPs in
and the MPPT efficiency with the proposed OD-PSO MPPT the case studies are designed to locate at the rightmost peak and
algorithm, the firefly MPPT algorithm and the P&O-PSO the middle peak in Case I and Case II respectively.
MPPT algorithm in can also be obtained. According to the The dynamic waveforms of the PV system using the
simulation results, Pave is 134.17W, 129.4W and 134W, while proposed OD-PSO MPPT algorithm are given in Fig. 14. It is
the MPPT efficiency is 99.82%, 99.75% and 99.80% three obvious that the proposed algorithm can accurately capture the
MPPT methods respectively, as given in Table III. It is obvious GMPP that varies with different solar irradiances.
that the proposed OD-PSO MPPT algorithm has the best TABLE III
SIMULATION RESULTS OF MPPT
overall performance considering Pave and the tracking speed.
MPPT Tracking
Case II: The irradiance of PV panels is given in Table I. The Algorithms Pave
efficiency
Iterations
time
P-V characteristic curve is shown in Fig. 4. From Fig. 4, the OD-PSO 134.17(W) 99.82% 8 0.210(s)
voltage at the GP2 is 33.85V and the maximum power is Case
Firefly 129.4(W) 99.75% 8 0.206(s)
131.71W. I
OD-PSO MPPT algorithm P&O-PSO 134(W) 99.80% 15(PSO) 0.286(s)
OD-PSO MPPT algorithm
Firefly MPPT algorithm in [18] OD-PSO 114.12(W) 99.83% 7 0.185(s)
P&O-PSO MPPT algorithm in [24]
150 150 Case
100 Firefly 108.74(W) 99.73% 8 0.206(s)
P (W)

100 II
P (W)

135
50 125 50 P&O-PSO 113.9(W) 99.82% 13(PSO) 0.266(s)
0.22 0.26
0 0
4 4 200
I (A)

I (A)

2
I (A) P (W)

2 100
0 0
60 60 0
V (V)
V (V)

40 40 4
20 20 2
0 0 0
0.8 0.8 60
V (V)

0.4 0.4
d

30
d

0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4


0
1
Time(s) Time(s)
d

0.5
(a) (b)
0
Fig. 13. The output waveforms of the PV system in Case II. (a)
OD-PSO MPPT algorithm and the firefly MPPT algorithm in [18]; (b)
OD-PSO MPPT algorithm and P&O-PSO MPPT algorithm in [24]. Fig. 14. The dynamic waveforms of the PV system using OD-PSO
MPPT algorithm
The simulation results for Case II are shown in Fig. 13. In Fig.
13 (a), the output waveforms of PV system using the proposed
OD-PSO MPPT algorithm and the firefly MPPT algorithm are

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

V. EXPERIMENTAL RESULTS the proposed method being implemented into hardware.


In order to verify the feasibility and effectiveness of the TABLE IV
MAIN EXPERIMENTAL DEVICES
OD-PSO MPPT algorithm, an experiment is also conducted.
The experiment platform is built, as shown in Fig. 15. The Devices Types Parameters
parameters of the circuit in the experiment are the same as those PV simulator TC.P.32 1000V/13A
Oscilloscope DPO 4054B 500MHz
in the simulation and the sampling time of V and I is 0.075s. Resistive Load BT2-014 0-10.73Ω
The OD-PSO MPPT algorithm is implemented into a single Digital Signal Processor MC56F8245 60MHz
chip microprocessor, MC56F8245. In this paper, the PV array
is replaced by a PV simulator for higher controllability, which Case I: Fig. 16 shows the output power, current and voltage
can edit the P-V curve and the I-V curve for Case I and Case II. curves of the PV system in Case I using the proposed OD-PSO
As mentioned before, the inverter is replaced by a resistive load MPPT algorithm and using the firefly MPPT algorithm and the
in this experiment. The main devices in experiment are P&O-PSO MPPT algorithm, respectively.
provided in TABLE IV. In the experiment, ɛ is set as 0.025, and To compare the experimental results, Pave, MPPT efficiency
γ is set as 0.1 for OD MPPT algorithm. The parameters of PSO and the tracking time are obtained from the experimental
MPPT algorithm are ω=0.4, c1=0.2, c2=2, jmax = 10 and vmax = waveforms and the experimental data are exported from
0.035. The experimental results for Case I and Case II are given oscilloscope. From Fig. 16, it takes in 1.64s, 4 iterations to
in Fig. 16 and Fig. 17, respectively. reach the GMPP in Case I using the proposed OD-PSO MPPT
algorithm, wherein, 3 iterations are for OD MPPT algorithm. It
takes 2.56s, 6 iterations using the firefly MPPT algorithm and
4.53s, 12 iterations using the P&O-PSO MPPT algorithm in
this case, respectively. It should be noticed that, in addition to
Oscilloscope
Upper computer the sampling time and the number of particles, the time
consumption for each iteration and the tracking time in
experiment is actually also influenced by the speed of the
employed digital controller and the program statements written
for the MPPT algorithms.
PV simulater Load Using the same comparison standards as in the simulation,
Pave of the PV system with the proposed OD-PSO MPPT, the
Fig. 15. Experimental platform firefly MPPT algorithm and the P&O-PSO MPPT algorithm
It is noteworthy that a longer sampling time in the are obtained by (8). Pave in Case I with the OD-PSO MPPT, the
experiment is chosen due to the slow response speed of PV firefly MPPT algorithm and the P&O-PSO MPPT algorithm is
simulator used in our laboratory and the time cost for the buck 112.58W, 126.07W and 133.08W, respectively.
converter and PV simulator to reach harmonious synergy. The Case II: Fig. 17 shows the output power, current and voltage
superiority of the proposed OD-PSO MPPT algorithm over curves of the PV system in Case II using the proposed OD-PSO
established ones in terms of a faster tracking speed and more MPPT algorithm, the firefly MPPT algorithm and the
accurate GMPP location can be noticed in the simulation results. P&O-PSO MPPT algorithm, respectively.
The experiment is for the purpose of verifying the feasibility of From Fig. 17, it can be seen that 2.08s, 4 iterations, is spent
OD PSO P&O PSO

P(100W / div)

I (2.5 A / div)

t (1s/ div) t (1s/ div) t (1s/ div)


Number of Iterations Number of Iterations
Number of Iterations V (20V / div)
1 23 12 3456 1 2 3 4 5 6 7 8 9101112
1.64s 2.56s 4.53s
(a) (b) (c)
Fig. 16. The output waveforms of the PV system in Case I. (a) the OD-PSO MPPT algorithm (b) the firefly MPPT algorithm (c) the P&O-PSO
MPPT algorithm
P&O PSO
OD PSO
P(50W / div)

I (2.5 A / div)

t (1s/ div) t (1s/ div) t (1s/ div)


123 4 Number of Iterations 1 2345 Number of Iterations 1234567 Number of Iterations V (20V / div)
2.08s 2.21s 2.74s
(a) (b) (c)
Fig. 17. The output waveforms of the PV system in Case II. (a) the OD-PSO MPPT algorithm (b) the firefly MPPT algorithm (c) the P&O-PSO
MPPT algorithm

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

for reaching the GMPP in Case II using the proposed OD-PSO VI. CONCLUSION
MPPT algorithm; where, 3 iterations are for OD MPPT In this paper, an OD-PSO MPPT algorithm has been
algorithm and 1 iteration is for PSO MPPT algorithm. It takes proposed for a PV power system to track the GMPP under
2.21s, 5 iterations using the firefly MPPT algorithm and 2.74s 7 PSCs. The OD-PSO MPPT algorithm can rapidly find the small
iterations using the P&O-PSO MPPT algorithm in Case II, region that contains the GMPP without detailed information
respectively. about the PV array and thus has a fast MPPT speed. In addition,
Average power Pave of the PV system with the proposed the OD-PSO MPPT algorithm can make the PV system output
OD-PSO MPPT and the firefly MPPT algorithm and the more power and have lower power fluctuation in comparison
P&O-PSO MPPT algorithm can be calculated by (8). Pave in with the existing MPPT algorithms. Both simulation and
Case II with the OD-PSO MPPT algorithm, the firefly MPPT experimental results have validated the effectiveness and
algorithm and the P&O-PSO MPPT algorithm is 110.85W, superiority of the proposed OD-PSO MPPT algorithm over
106.58W and 111.63W, respectively. Power fluctuation of the established MPPT methods.
OD-PSO MPPT algorithm is noticeably smaller than that with
the firefly MPPT algorithm and the P&O-PSO MPPT REFERENCES
algorithm, which aligns with simulation results. [1] D. C. Huynh, T. N. Nguyen, M. W. Dunnigan and M. A. Muller,
The numerical results for the experiment on Pave, number of “Dynamic particle swarm optimization algorithm based maximum
iterations and tracking time are shown in Table V. It can be power point tracking of solar photovoltaic panels.” in 2013 IEEE Int.
Symp. Ind. Electron Conf., May 2013, pp. 1-6.
found that the experimental results are consistent with the [2] S. Lyden and M. E. Haque, “Comparison of the perturb and observe and
simulation results. With the proposed OD-PSO MPPT simulated annealing approaches for maximum power point tracking in a
algorithm, the PV system has a faster GMPP tracking speed, a photovoltaic system under partial shading conditions.” in 2014 IEEE
higher output power and better dynamic performances. Energy Convers Congr.Expo.Conf., Sept. 2014, pp. 2517-2523.
[3] S. Yu, L. Zhang, H. Iu, T. Fernando, and Kit Po Wong. "A DSE-Based
It should be noticed that according to Figs. 16-17, the Power System Frequency Restoration Strategy for PV-Integrated
tracking times are all longer than 2s in the experiment with all Power Systems Considering Solar Irradiance Variations." in IEEE
three MPPT methods. This is caused by the fact that the Transactions on Industrial Informatics., vol. 13, no. 5, pp. 2511-2518
sampling time is set as 0.075s in experiment, which however is Oct. 2017.
[4] K. Ishaque, Z. Salam, M. Amjad, S. Mekhilef, “An improved particle
0.005s in simulation. Normally, the tracking time in Figs. 16-17 swarm optimization (PSO) based MPPT for PV with reduced
can satisfy most of the requirements in practical applications steady-state oscillation.” IEEE Trans. Power Electron., vol. 27, no. 8,
where the change rate of solar irradiance is considered pp. 3627-3638, Aug. 2012.
relatively slow [35]. On the other hand, the tracking speed of [5] M. A. Ghasemi, H. M. Forushanis and M. Parniani, “Partial shading
detection and smooth maximum power point tracking of PV arrays
OD-PSO MPPT can be improved significantly by setting a under PSC.” IEEE Trans. Ind. Electron., vol. 31, no. 9, pp. 6281-6292,
smaller sampling time, however the cost of digital processors Sept. 2016.
will also be increased for faster calculation speeds and high [6] H. Renaudineau, F. Donatantonio, J. Fontchastagner, G. Petrone, G.
performances. It is worth noting that in the PSO method the Spagnuolo, J. P. Martin and S. Pierfederici, “A PSO-based global
MPPT technique for distributed PV power generation.” IEEE Trans.
current levels shift quickly which may increase the stress on the Ind. Electron., vol. 62, no. 2, pp. 1047-1058, Feb. 2015.
switches and decrease the control bandwidth to achieve higher [7] M. Killi and S. Samanta, “An adaptive voltage-sensor-based MPPT for
accuracy. In this study, the use of OD method in the proposed photovoltaic systems with SEPIC converter including steady-state and
OD-PSO MPPT algorithm, to some extent, mitigates the drift analysis.” IEEE Trans. Ind. Electron., vol. 62, no. 12, pp.
7609-7619, Dec. 2015.
computational stress caused by PSO. This can be observed in [8] Y. H. Liu, S. C. Huang, J. W. Huang and W. C. Liang, “A particle
Fig.16 and Fig.17 below, where the current does not show swarm optimization-based maximum power point tracking algorithm
significant fluctuations. According to Fig. 16(a) and Fig. 17(a), for PV systems operating under partially shaded conditions.” IEEE
the peak values of dv/dt using OD-PSO MPPT algorithm are Trans. Energy Convers., vol. 27, no. 4, pp. 1027-1035, Dec. 2012.
[9] Q. Zhu, X. Zhang, S. S. Li, C. Liu and H. Ni, “Research and test of
5.85×10-2 V/ns in Case I and 5.26×10-2V/ns in Case II and the power loop based dynamic multi-peak MPPT algorithm.” IEEE Trans.
peak values of di/dt are 9.43A/μs in Case I and 10.59A/μs in Ind. Electron., vol. PP, no. 99, pp. 1-1, July 2016.
Case II, respectively. Comparing with the rating parameters of [10] Y. H. Ji, D. Y. Jung, J. G. Kim, J. H. Kim, T. W. Lee and C. Y. Won, “A
the MOSFET selected in the experiment [38], the switching real maximum power point tracking method for mismatching
compensation in PV array under partially shaded conditions.” IEEE
stresses of the MOSFET are enough for realizing all the three Trans. Power Electron, vol. 26, no. 4, pp. 1001-1009, April 2011.
MPPT algorithms in this paper. [11] H. Patel and V. Agarwal, “Maximum power point tracking scheme for
TABLE V PV systems operating under partially shaded conditions.” IEEE Trans.
EXPERIMENTAL RESULTS OF THREE MPPT ALGORITHMS Ind. Electron, vol. 55, no. 4, pp. 1689-1698, April 2008.
MPPT Tracking [12] T. L. Nguyen and K. S. Low, “A global maximum power point tracking
Algorithms Pave Iterations scheme employing DIRECT search algorithm for photovoltaic system.”
efficiency time
OD-PSO 112.85(W) 97.28% 3 1.64(s) IEEE Trans. Ind. Electron., vol. 57, no. 10, pp. 3456-3467, Oct. 2010.
Case [13] K. S. Tey and S. Mekhilef, “Modified incremental conductance
Firefly 126.07(W) 97.79% 6 2.56(s) algorithm for photovoltaic system under partial shading conditions and
I
P&O-PSO 133.08(W) 98.18% 12(PSO) 4.53(s) load variation.” IEEE Trans. Ind. Electron., vol. 61, no. 10, pp.
OD-PSO 110.85(W) 98.21% 4 2.08(s) 5384-5392, Oct. 2014.
Case [14] P. Lei, Y. Y. Li and J. E. Seem, “Sequential ESC-based global MPPT
Firefly 106.58(W) 97.60% 5 2.21(s) control for photovoltaic array with variable shading.” IEEE Trans.
II
P&O-PSO 111.63(W) 98.18% 7(PSO) 2.74(s) Sustain. Energy, vol. 2, no. 3, pp. 348-358, July 2011.
[15] K. Chen, S. Tian, Y. H. Cheng and L. B. Bai, “An improved MPPT
controller for photovoltaic system under partial shading condition.”
IEEE Trans. Sustain. Energy, vol. 5, no.3, pp. 978-985, July 2014.

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

[16] M. Boztepe, F. Guinjoan, G. Velasco-Quesada, S. Silvestre, A. Algorithm in Highly Fluctuating Atmospheric Conditions. IEEE
Chouder, and E. Karatepe, “Global MPPT Scheme for photovoltaic Transactions on Industrial Informatics, vol. 13, no. 5, pp. 2406-2416,
String Inverters Based on Restricted Voltage Window Search May 2017.
Algorithm.” IEEE Trans. Ind. Electron., vol. 61, pp. 3302-3312, 2014. [37] C. Huang, Wang, L., Yeung, R. S. C., Zhang, Z., Chung, H. S. H., &
[17] T. T.. Yetayew, T. R. Jyothsna and G. Kusuma, “Evaluation of Bensoussan, A. (2017). A Prediction Model Guided Jaya Algorithm for
Incremental Conductance and Firefly algorithm for PV MPPT the PV System Maximum Power Point Tracking. IEEE Transactions on
application under partial shading condition. ” International Conference Sustainable Energy, vol. PP, no. 99, pp. 1-1, June 2017.
on Power Systems., Mar. 2016, pp. 1-6. [38] Fairchildsemi, FDB075N15A _datasheet, Feb. 2013. [Online].
[18] D. F. Teshome, C. H. Lee, Y. W. Lin and K. L. Lian, “A modified Available: http://pdf1.alldatasheet.com/datasheet-pdf/view/657966/
firefly algorithm for photovoltaic maximum power point tracking FAIRCHILD/FDB075N15A_F085.html.
control under partial shading.” IEEE Trans. Emerg. Sel. Topics Power
Electron., vol. 5, no. 2, pp. 661-671, June 2017. Hong Li (S’07-M’09 -SM’18) received her BSc,
[19] V. Phimmasone, T. Endo, Y. Kondo and M. Miyatake, “Improvement MSc, and PhD degrees from Taiyuan University
of the maximum power point tracker for photovoltaic generators with of Technology, South China University of
particle swarm optimization technique by adding repulsive force Technology, and Fernuniversität in Hagen,
among agents.” Elect. Mach. Syst. Int. Conf., Nov. 2009, pp. 1-6. Germany, in 2002, 2005, and 2009, respectively.
[20] C. Manickam, G. R. Raman, G. P. Raman, “A hybrid algorithm for Currently, she is a professor of Electrical
tracking of GMPP based on P&O and PSO with reduced power Engineering School, Beijing Jiaotong University,
oscillation in string inverters.” IEEE Trans. Ind. Electron., vol. 63, no. China. Her research interests include nonlinear
10, pp.6097-6106, Oct. 2016. modeling, analysis and its applications, EMI
[21] F. M. Li, F. Deng, S. Guo and X. Y. Fan, “MPPT control of PV system suppressing methods for power electronic system, wide band gap
under partially shaded conditions based on PSO-DE Hybrid algorithm.” power devices and applications. She is Associate Editor of IEEE
Proceedings of the 32nd Chinese Conrol Conference, July 2013, pp. Transactions on Industrial Electronics, Associate of the Chinese Journal
7553-7557. of Electrical Engineering, Vice Chairman of Electromagnetic
[22] M.Vysakh, M. Azharuddin, H. Vilas, K. Muralidhar, D. Paul, B. Jacob, Compatibility Specialized Committee in China Power Supply Society.
T. S. B. Sudhakar and N. Rajasekar, “Maximum power point tracking She has published 1 book, 30 journal papers, and 39 conference papers.
using modified PSO with Cuk converter.” in Advance Elect. Eng. Int. She has also applied 20 patents.
Conf., Jan. 2014, pp. 1-6.
[23] S. S. Li, X. Zhang, H. K. Zhang, W. Zhao and H. Ni, “Global MPPT Duo Yang was born in Hebei province, China, in
method based on power closed loop control and PSO algorithm.” Pro. 1991. She received the MSc degree in electrical
CSEE , vol. 34, no. 28, pp. 4809-4816, Oct. 2014 (in Chinese). engineering from Beijing Jiaotong University,
[24] K. L. Lian, J. H. Jhang and I. S. Tian, “A maximum power point Beijing, China, in 2017.
tracking method based on perturb and observe combined with particle She is currently working at the China
swarm optimization.” IEEE. J. Photovolt., vol. 4, no. 2, pp. 626-633, Academy of Railway Sciences Corporation
Mar. 2014. Limited, Beijing, China. Her research interest is
[25] K. Sundareswaran, V. Vigneshkumar, P. Sankar, S. P. Simon, P. S. R. power electronic technology in power systems
Nayak, and S. Palani, “Development of an improved P&O algorithm and renewable energies.
assisted through a colony of foraging ants for MPPT in PV system,”
IEEE Trans. Ind. Informat., vol. 12, no. 1, pp. 187–200, Feb. 2016.
[26] S. Lyden and Md. E. Haque, “A simulated annealing global maximum Wenzhe Su was born in Shanxi province, China,
power point tracking approach for PV modules under partial shading in 1993. He received the BSc degree in electrical
conditions,” IEEE Trans. Power Electron., vol. 31, no. 6, pp. 4171– engineering from Shanxi University, Shanxi,
4181, Jun. 2016. China, in 2016.
[27] K. Sundareswaran, S. Peddapti and S. Palani, “Application of random He is currently working toward the PhD
search method for maximum power point tracking in partially shaded degree in electrical engineering at the School of
photovoltaic systems.” IET Renewable Power Generation, vol. 8, no. 6, Electrical Engineering, Beijing Jiaotong
pp. 670-678, 2014. University, Beijing, China. His research interest
[28] K. Ishaque and Z. Salam, “A deterministic particle swarm optimization is power electronics technology in power
maximum power point tracker for photovoltaic system under partial systems and renewable energies.
shading condition.” IEEE Trans. Ind. Electron., vol. 60, no. 8,
pp.3195-3206, Aug. 2013. Jinhu Lü (M'03-SM'06-F'13) received the PhD
[29] S. K. Kollimalla and M. K. Mishra, “A novel adaptive P&O algorithm degree in applied mathematics from the
considering sudden changes in the irradiance.” IEEE Trans. Energy Academy of Mathematics and Systems Science,
Convers.,vol. 29, no. 3, Sept. 2014. Chinese Academy of Sciences, Beijing, China, in
[30] L. X. Li and B. H. Yu, “Research on short-term scheduling of 2002.
hybro-thermal system based on entire distribution optimization Currently, he is a Professor of the Academy of
thchniques”, Coumputer Appl. Softw., vol. 28, no. 7, pp. 240-242, Jul. Mathematics and Systems Science, Chinese
2011. Academy of Sciences. He was a Visiting Fellow
[31] X. J. Lei, “Swarm intelltgent optimization algorithms and theirs in Princeton University, USA from 2005 to 2006.
application.” Beijing, Science press, Aug. 2012, ch. 4, pp. 91-97. He is the author of two research monographs
[32] R. Eberhart and J. Kennedy, “A new optimizer using particle swarm and more than 100 international journal papers published in the fields of
theory.” In Proc. 6th Int. Symp. Micro Mach Human Sci. Conf., 1995, complex networks and complex systems, nonlinear circuits and systems,
pp. 39-43. with more than 5700 SCI citations and h-index 38, including 20 papers
[33] A. Kar and A. Kar, “A new maximum power point tracking algorithm with more than 100 SCI citations. He also has two authorized patents.
for PV modules under partial shading and rapidly varying illumimation.” He is now serving as the Chair of Technical Committee on Neural
2009 Annu. IEEE India Conf., Dec. 2009, pp. 1-4. Systems and Application and Secretary of Technical Committee on
[34] N. Pragallapati, T. Sen, and V. Agarwal. "Adaptive Velocity PSO for Nonlinear Circuits and Systems in the IEEE Circuits and Systems
Global Maximum Power Control of a PV Array Under Nonuniform Society. Prof. Lü served and is serving as Editors in various ranks for 11
Irradiation Conditions." IEEE Journal of Photovoltaics vol: 7, no. 2, SCI journals including the IEEE Transactions on Circuits and Systems
624-639, Mar. 2017. Part I: Regular Papers, IEEE Transactions on Circuits and Systems Part
[35] Overall Efficiency of Grid Connected Photovoltaic Inverters, European II: Brief Papers, IEEE Transactions on Neural Networks, IEEE
Standard EN 50530, Apr. 2010. Transactions on Industrial Informatics, PLOS Computational Biology,
[36] N. Kumar, Hussain, I., Singh, B., & Panigrahi, B. (2017). Rapid MPPT International Journal of Bifurcation and Chaos, and Asian Journal of
for Uniformly and Partial Shaded PV System by using JayaDE Control. Prof. Lü received the prestigious National Natural Science

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2829668, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS

Award twice from the Chinese government in 2008 and 2012, the 9th
Guanghua Engineering Science and Technology Award from the
Chinese Academy of Engineering in 2012, the 11th Science and
Technology Award for Youth of China and the Australian Research
Council Future Fellowships Award in 2009. Moreover, Prof. Lü attained
the National Natural Science Fund for Distinguished Young Scholars
and 100 Talents Program from the Chinese Academy of Sciences.

Xinghuo Yu (M'92-SM'98-F'08) received BEng


and MEng degrees in Electrical and Electronic
Engineering from the University of Science and
Technology of China, Hefei, China, in 1982 and
1984, and PhD degree in Control Science and
Engineering from Southeast University, Nanjing,
China in 1988, respectively.
He is an Associate Deputy Vice-Chancellor
and a Distinguished Professor of RMIT
University, Melbourne, Australia. He is also the
President of IEEE Industrial Electronics Society for 2018 and 2019. His
research interests include control systems, complex and intelligent
systems, and smart grids.
He received a number of awards and honors for his contributions,
including 2013 Dr.-Ing. Eugene Mittelmann Achievement Award of IEEE
Industrial Electronics Society and 2018 M. A. Sargent Medal of
Engineers Australia. He was named a Highly Cited Researcher by
Clarivate Analytics (formerly Thomson Reuters) in 2015, 2016 and
2017.

0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
View publication stats

You might also like