Research on MPPT control of PV system based on PSO algorithm

Ze Cheng, Hang Zhou, Hongzhi Yang

School of Electrical Engineering and Automation, Tianjin University, Tianjin, 300072
E-mail: chengze@tju.edu.cn

Abstract: Partially shaded conditions are common in all kinds of PV systems. The mismatch phenomenon not only
affects the power generated by solar system, but also leads to security problem. Still, under partially shaded conditions,
the P-V curve of PV arrays will have the characteristics of multi-summit, which makes the maximum power point
tracking very difficult to accomplish. In this paper, the PSO algorithm is applied to the MPPT control algorithm in two
different PV structure. Based on establishing the multi-peak model of the PV array under partially shaded condition, the
multi-peak MPPT algorithm is designed. Through the introduction of modular structure, the distributed MPPT algorithm
is presented. The effectiveness of the algorithm and the advantages and disadvantages of two different algorithms are
examined using the MATLAB. At last, the hardware circuit of the control system with man-made partially shaded
condition is constructed. The effectiveness and preciseness of the MPPT system is proved by the results of the
experiments.
Key Words: Maximum Power Point Tracking, Particle Swarm Optimization, PV System, AC Module

caused by partial shade. Each panel has independent MPPT

1 INTRODUCTION
Photovoltaic power generation industry is the world's
fastest growing high-tech industry in the 20th century.
Solar energy has the advantages of pollution-free energy
conversion process, low maintenance cost. However, the
major disadvantage of the PV source is the ineffectiveness
during low insolation periods or partially shaded
conditions. Partial shaded condition is popular and complex
in all kinds of PV systems, especially in BIPV. When the
solar illumination angle and intensity change with the
season or time, some obstacles such as trees, telephone
poles, cables the obstacles may produce partial shadow on
the solar panel. The mismatch phenomenon caused by
partial shade will affect the power output of solar systems
and lead to the security problem. Besides, under partially
shaded conditions, when the entire array does not receive
uniform insolation, the P–V characteristics get more
complex, displaying multiple peaks, as a result the
conventional maximum power point tracking method will
fail.
Some methods [1-5] are presented by some researchers
regarding to the Global tracking method under partial
controllers, which is the most competitive structure of the
PV system in the future. In this structure, each module
needs independent control, which in essence is a
multi-variable control problem. Using the characteristics of
multi-variable optimization of PSO algorithm, a PSO based
distributed maximum power point tracking algorithm is
designed.
2 DESIGN OF PSO-BASED MULTI-PEAK
MPPT
2.1 The reasearch of characteristics of PV array with
multi-peak
As a result of the occurrence of the phenomenon of partial
shade, the IV characteristics of each module are different,
so they have different maximum power point. When these
modules connect in series with the same branch current, the
formation of the overall PV series features a multi-peak
form.
I
MPP

shading, but these methods are limited to the improvement

Insolation
Change

of conventional algorithms, or use lengthy calculations, and

the discussion of the array structure is not deep enough.
Considering the outstanding performance of multi-peak
function optimization and global search of the particle
swarm optimization algorithm, this paper presents a
PSO-based maximum power tracking algorithm, which V
solve the problem caused by partially shaded condition. 
This paper also introduces a new type of PV structure called Figure 1. I-V characteristics under different light
AC module, which does not have the problem of power loss In this paper, the simulation of the IV and PV
characteristics of array with 1000 modules is presented
This work is supported by National Basic Research Program of China using Matlab. In accordance with the illumination of the
(973 Program) (No. 2009CB219700) and Tianjin science and technology arrays which are divided into three different groups, the
supported program(09ZCGYGX01100)

c
978-1-4244-5182-1/10/$26.00 2010 IEEE 887
 illumination Ȝ = 1, the illumination Ȝ = 0.1 and no shade.
The simulation results are shown in Figure 3.
Figure 2. Layout of photovoltaic arrays
(a) (b)
Figure 3.(a) for the PV array voltage curve (b) total power output curve
As can be seen from the figure, as a result of
series-connected modules with different illumination, the
voltage output has multi-peak character, so that the total
output power has also produced a number of extreme points
in which the conventional single-peak maximum power
tracking algorithm fail.
2.2 Particle swarm optimization applications in MPPT
Particle swarm optimization has parallel processing, good
robustness characteristics, and high probability of finding
the global optimal solution and higher computational
efficiency than the traditional random method. Its greatest
advantage lies in its simple realization, easy and fast
convergence. Due to its good performance in the
multi-peak function optimization, particle swarm
optimization is applied to the PV array under partially
shaded condition.
First of all, particle swarm optimization initializes the
variables randomly in a given space. The number of
decision variables determines the dimension of space. Each
optimization problem is to search the solution space of a
particle, each particle runs at a certain speed in the search
space, the speed of particles is in accordance with its own
flight experience and flight experience of other examples
with dynamic adjustments. In the optimization space, each
particle has decided to adapt the objective function value,
and recorded their own best position Pi found so far, and the
entire group of all particles found in the best position Pg.
Velocity update formula and the PgǃPi calculation formula
is as follows.
Vi k +1 = ωVi k + c1r 1( Pi k − X ik ) + c 2 r 2( Pgk − X ik )
888 2010 Chinese Control and Decision Conference
X ik +1 = X ik + Vi k +1 (2)
­ Pi f ( X i ) ≥ f ( Pi )
Pi = ® (3)
¯ Xi f ( Xi ) < f ( Pi )
Pg = max { f ( P0 ), f ( P1 ), ⋅⋅⋅, f ( Pm )} (4)
Where I= 1,2 ... M, M is the number of groups; m is the
number of particles; N is dimension for the problem space;
c1, c2 are constants greater than 0 for, individual and
overall acceleration speed toward the most optimal solution
respectively. According to the experience, set the value as
1.0 ~ 2.0; r1 ~ U (0,1), r2 ~ U (0,1); Ȧ is inertia weighted
factor; f is the objective function.
In this paper, particle swarm optimization is applied to the
maximum power tracking control. The whole algorithm is
divided into two parts: a part for the particle swarm
optimization process, the other part for maximum power
tracking. In the PSO algorithm, each optimization problem
has a potential solution (that is, maximum power point
voltage) in the search space; all the particles have a decision
to adapt the objective function value, which defines the
array’s total power function as the objective function, while
 the variable is an array of output voltage. Concrete steps are
as follows:
1) Initialization, as well as the determination of the
objective functions.
The use of particles is equivalent to 10 population size, the
evolution is set to 80 generations, and initialize the learning
factor, weighted factors and maximum speed, respectively.
Finally the location and speed of each particle will be
initialized.
Particle Swarm initialization settings are as follows:
Pop size = 10; part size = 2;
Max gen = 80; w max = 0.9; w min = 0.4;
V max = 2; c1 = 2; c2 = 2;
2) Evaluation of the population x (t)
After the initialization of various parameters, the objective
function of fitness for each particle will be calculated. The
objective function is the total array output power, using the
system of two modules as an example, the expression of
fitness function as follows:
fit=I *(PVprog(i1 ,Sun1 ,T1 )+PVprog(i2 , Sun2 , T2 ))
(5)
3.8*Sun - I +2.2*10-8
PVprog(I , Sun, T )=1.1103* log( )
2.2*10-8
-0.2844*I (6)
PVprog(I , Sun, T ) is function of the output characteristics,
where temperature is set to 25 °.
By calling a function to calculate the fitness function, which
consists of every module’s current and voltage function to
sum up the total power output.
3) Comparison to determine individual and overall fitness
as well as the extreme point.
Compare a single particle’s current fitness value with
historical fitness value of the best, if the current value is
(1)
greater, then update a single particle Pbest. Each particle is
compared to determine the best fitness values Gbest.
4) Update particle velocity and location. According to the
formula (1), (2), update the location and the speed of
particles.
5) Check the end conditions, if met, the end of optimization,
output is the optimal solution; Otherwise, t = t +1, go to 2,
the end of optimizing the conditions for the evolution of
maximum generations.
Particle Swarm Optimization flow chart shown in Figure 6.
2.3 Simulation results and analysis
In this paper, the case for two modules of the particle
swarm MPPT algorithm is simulated using Matlab. The
parameters of the two modules are set to: Sun1 = 1KW/m2,
T1 = 25; Sun2 = 0.8 KW/m2, T2 = 23 °
Two-dimensional characteristics of the multi-peak curve
and particle swarm optimization results are shown.
Figure 4. The two peaks characteristic of multi-module
Figure 5. Two-Dimensional Multi-Particle Swarm MPPT simulation
results
As we can be seen from the simulation results:
two-dimensional multi-peak curve of the overall maximum
is 102W, Multi-particle swarm optimization finds the
maximum power point of the power value as 102.6322W,
error with the theoretical value of 102W is very small,
which shows that the algorithm maximum power point
tracking has very good accuracy. The time for optimizing is
0.078s, which means the convergence of the algorithm is
very fast, and the algorithm can also optimize the algorithm
parameters to further improve the algorithm convergence
speed. Multi-particle swarm algorithm for maximum power
tracking can be achieved under the partial shaded condition
of photovoltaic arrays for global optimization.
Start
Input system parameters , Set variable limit value
Initiate particle Speed and Position
Decide objective function, calculate total power of
two module
Evaluate the fitness value of each particle
Search Pbest and Gbest
Update the speed and position of particle
based on (3-1) and (3-2)
Recalculate the total power of array 䯸 re-evaluate
fitness value of each particle Determine whether
update Pbest and G best according to 䯴 3-3䯵
NO Check whether
condition satisfied
YES
Output optimal solution
End
Figure 6. MPPT control Based on Particle Swarm Algorithm
2.4 PSO-based design of distributed MPPT
Independent photovoltaic power generation components
and network structure is shown in Figure 7.It is a trend of
development in the future, it adopted an independent
optimization for each module, and then runs for grid
generation, which not only greatly improves the overall
power output, boosts the transmission efficiency, but also
overcomes a very good partial shade phenomena, such as
mismatch problem of power loss. Modular structure,
however, is just in its infancy, so there are still many
problems. Independent components grid generation system
needs distributed MPPT control, which makes each module
install an independent controller, and then uses
conventional MPPT algorithm. This will greatly increase
the cost and the centralized control is not easy. In this paper,
particle swarm optimization based-on distributed MPPT
control algorithm is presented, multi-module MPPT control
is converted to a system of multi-variable optimization
problem, the use of particle swarm optimization has
achieved the distributed maximum power tracking control.
2010 Chinese Control and Decision Conference 889
 3 SIMULATION AND RESULT

3.1 Modeling and simulation

Figure 7. Independent electricity grid structure

The idea of PSO-based distributed MPPT control algorithm

is that regard the multi-module’s voltage output as a
multi-dimensional variables and the system's total power
output as the objective function. Through the use of particle
swarm optimization, the maximum total output power is
found, that is, each module finds the maximum power point
at the same time, and then solar panels control the output
voltage at maximum power point by conventional constant
voltage method. This realizes the distributed MPPT of PV
generation system and the group control of multi-module.
The steps and multi-peak optimization of Particle Swarm Figure 10. Simulation diagram system
Algorithm MPPT are similar, which needn’t repeat here.
In this paper, simulation mode of the PSO-based distributed
The main difference is that the objective function is
MPPT control system is presented using Matlab / Simulink
different. The objective function of distributed MPPT
environment, which verify the feasibility of the algorithm.
algorithm is as follows:
The entire control system includes the following
fit=i1*PVprog(i1,Sun1,T1)+i2 *PVprog(i2 ,Sun2,T2) components: solar cell modules, Boost type DC / DC
i1,i2 are the current output ,respectively. module, PWM pulse signal modules, PSO-based
Two-dimensional system P-V characteristics and distributed particle swarm optimization MPPT module.
simulation results are as follows: The simulation results are shown below.

Figure 11. MPPT distributed simulation results

As waveform shown, solar module output voltage is set fast

Figure 8. Two-dimensional system p-v characteristics
Figure 9. Two-dimensional MPPT simulation results
We can see from the map: the two modules’ theoretical
maximum power value are: Pmax1 = 61W, Pmax2 = 48W,
the theoretical value of the total power is 109W. The
algorithm found the maximum power value of 109.1394W.
Optimization time is 0.094s. Simulation results show that
PSO-based distributed MPPT algorithm can quickly and
accurately find the maximum power of each module, and
the system achieved a true sense of the maximum power
output.
to the module’s maximum power point voltage value. The
steady-state does not shock, and the error is very small, and
the output power reached module’s maximum power output
of the theoretical value, which means that particle swarm
optimization makes module accurately find the maximum
power point voltage, and through the adoption of Constant
voltage method, solar module is stable in the maximum
power point.
3.2 Simulation results and analysis
In order to verify the feasibility of the above algorithm, this
paper establishes the modular structure of maximum power
tracking control system which consists of solar panels,
Boost converter, AVR single-chip control module, data
acquisition module. Physical map of the hardware system is
shown below.

890 2010 Chinese Control and Decision Conference


Figure 12. Hardware circuit diagram objects
4 CONCLUSION
In this paper, in the case of partial shading of PV array the
output characteristics presents a number of extreme points,
as well as the problem independent components and
network structure of the distributed power generation
MPPT control, so that the conventional MPPT algorithm
fail. With the consideration of the good performance of the
particle swarm optimization in multi-peak function
optimization, global optimization, as well as multi-variable
optimization, the particle swarm algorithm is applied to the
maximum power tracking control algorithm.
The analysis of the two algorithms is shown in the
following three aspects.
1. The total maximum power tracking, the two algorithms
in this article is better than the previous single-peak
conventional MPPT algorithm to find large values of the
maximum power, which is proved by simulation
experiments in this article, MPPT algorithm distributed
unevenly in the light of the circumstances to find a value
greater than the maximum power MPPT algorithm for
multi-peak type. For a large number of photovoltaic
modules generating system, the difference of the two values
found will be very large.
2. To consider the applicability of conventional single-peak
MPPT algorithm and multi-peak MPPT algorithm, they are
suitable for series-parallel structure of traditional
photovoltaic power generation the algorithm needs a small
number of sensors, and only needs a high-power DC-DC to
achieve MPPT tracking. And distributed MPPT algorithm
is applicable to the modular structure of the photovoltaic
power generation system with non-uniform illumination
modules.
3. Response speed and adaptability to environment
changes.
In this paper, response speed to the two algorithms is not
very different, but with the increase of dimension the
response speed will be slightly lower. These two algorithms
in the environment under conditions of mutation have a
better tracking capability.
REFERENCES
[1] G. Eason, B. Noble, and I. N. Sneddon, “On certain integrals
of Lipschitz-Hankel type involving products of Bessel
functions,” Phil. Trans. Roy. Soc. London, vol. A247, pp.
529–551, April 1955. (references)
[2] J. Clerk Maxwell, A Treatise on Electricity and Magnetism,
3rd ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73.
[3] E. Koutroulis and K. Kalaitrakis, “Development of a
Microcontroller-Based Photovoltaic Maximum Power Point
Tracking ControlSystem”, IEEE Trans. Power Elecfronics,
vol. 16,pp.46-54, Jan. 2001.
[4] Y. Yorozu, M. Hirano, K. Oka, and Y. Tagawa, “Electron
spectroscopy studies on magneto-optical media and plastic
substrate interface,” IEEE Transl. J. Magn. Japan, vol. 2, pp.
740–741, August 1987 [Digests 9th Annual Conf. Magnetics
Japan, p. 301, 1982].
[5] Tomokazu Mishima 㧘 Tokuo Ohnishi 㧘 Power
Compensation System for Partially Shaded PV Array Using
Electric Double Layer Capacitors[C] IEEE 㧘 Industrial
Electronics Society Conference㧘2002㧘4㧦3262 ~3267
[6] T. Shimizu㧘 M. Hirakata㧘T. Kamezawa㧘H. Watanabe㧘
Generationcontrol circuit for photovoltaic modules[J]. IEEE
Trans. On Power Electronic㧘2001㧘16(3)㧦293 ~300
[7] Tomokazu Mishima 㧘 Tokuo Ohnishi 㧘 A power
compensation strategy based on electric double layer
capacitors for a partially shaded PV array[C] The 15th
International Conference on Power Electronics and Drive
Systems㧘2003㧘2㧦858 ~863
[8] Toshihisa Shimizu㧘Osamu Hashimoto㧘and Gunji Kimura
㧘A Novel High Performance Utility Interactive Photovaltaic
Inverter System[J]. IEEE Trans㧘On Electronic㧘2003㧘
18(2)㧦704~711
[9] K.Kobayashi㧘I.Takano㧘and Y.Sawada㧘A study on a two
stage maximum power point tracking control of a
photovoltaic system under partially shaded insolation
conditions[C]. Power Engineering Society General Meeting
㧘2003㧘4㧦2612 ~2617
[10] Wenkai Wu㧘Pongratananukul N㧘Weihong Qiu, Rustom
K 㧘 DSP-based Multiple Peak Power Tracking for
Expandable Power System[C]㧘Applied Power Electronics
Conference and Exposition㧘2003㧘1㧦525~530
[11] Bodur.M㧘Ermis.M㧘Maximum power point tracking for
low power photovoltaic solar panels[C]㧘 Electrotechnical
Conference㧘1994㧘2㧦758 ~ 761
[12] Toshihiko Noguchi㧘Shigenori Togashi㧘Ryo Nakamoto㧘
Short-current pulse-based maximum power point tracking
method for multiple photovoltaic and converter module
system[J]. IEEE Trans. On Industrial Electronics㧘2002㧘
49(1)㧦217 ~222
[13] Bekker.B. 㧘Beukes㧘H.J㧘Finding an optimal PV panel
maximum power point tracking method[C]. 7th AFRICON
Conference in Africa㧘2004㧘2㧦1125~1129
[14] Miyatake, M.Inada, T.Hiratsuka, I.Hongyan Zhao㧘Control
characteristics of a Fibonacci search based maximum power
point tracker when a photovoltaic array is partially shaded[C].
Power Electronics and Motion Control Conference㧘2004㧘
2㧦816 ~ 821
[15] Solodovnik㧘E.V.㧘Shengyi Liu㧘Dougal㧘R.A.. Power
controller design for maximum power tracking in solar
installations[J]. Trans. On Power Electronics㧘2004㧘19(5)
㧦1295~1304
[16] Sera 㧘 D. 㧘 Kerekes 㧘 T., Teodorescu, R., Blaabjerg, F..
Improved MPPT Algorithms for Rapidly Changing
Environmental Condition[C]. 12th International Power
Electronics and Motion Control Conference 㧘 2006 㧦
1614~1619
2010 Chinese Control and Decision Conference 891
 [17] G. R. Walker 㧘 J. Xue 㧘 and P. Semia 㧘 PV String Australasian Universities Power Engineering Conference㧘
Per-Module Maximum Power Point Enabling Converters㧘 2003

892 2010 Chinese Control and Decision Conference