Novice Insights

Innovative Global Maximum Power Point

Tracking based on Genetic Algorithm
Vlad BORZA
Abstract
This paper is about studying Genetic Algorithms (GA) and the influence brought by
changing several parameters of a GA such as: number of genes and random number
generation function to optimize the systems response. The algorithm is applied to a
photovoltaic system with four solar panels and partial shading phenomenon in order to
find the global maximum power point (GMPP). Results are given for the conditions
mentioned above.

Introduction

Nowadays it seems like everything is pushed to the limit and the same thing is happening
with photovoltaic (PV) energy. It is desired to build an algorithm that harvests as much energy
as possible from a PV system.
The PV system is made of four solar panels connected in series to increase the voltage but
there are problems when the panels are irradiated differently and begin to consume power so a
bypass diode is placed in parallel with each panel. This doesnt solve the partial shading effect
[1] which creates multiple maxima in the power voltage (P-V) representation.
This problem is solved by the GA implemented in Matlab. The basic idea of the algorithm is
to sweep the voltage range in a random manner and then focusing on the part with the highest
power. This algorithm has the advantage of applicability on different systems with variable
numbers of panels and no previous knowledge about power-voltage characteristics.
In this paper we will study the influence of gene number, random function and number of
iterations on the GA functionality to find the optimal solution.
A full schematic of the system is present in Fig. 1 and is divided in three blocks that will be
discussed in the following two chapters. Then simulation and experimental results are given
next and in the end we will give results for different parameters of the GA.

System Description

This chapter describes the circuit tested in Simulink. A macro model of the real circuit is
used because if a converter had been used it will result in large simulation time, and a lot of
convergence problems. This macro model acts in the same way as a normal converter would
and is much simple.
1

Novice Insights

Figure 1: System schematic

2.1.

Solar Panel Array Block

As previously said the PV system is made of four solar panels (Fig. 1 - orange) connected in
series and each panel has a bypass diode which shunts the panel in case of low irradiance.
The irradiance is controlled by the constant value located at right side of each panel. The
main characteristic of this system is the power-voltage representation (Fig. 2) plotted for
different values of short-circuit current (Table 1).
Fig.

Color

a)
b)
c)

Grey
0.5-0.4-0.3-0.2
0.5-0.5-0.3-0.3
0.5-0.5-0.1-0.1

Black
0.5-0.5-0.5-0.5
0.5-0.5-0.4-0.4
0.5-0.5-0.5-0.5

Table 1: Values for short-circuit current [A]

Figure 2: Power-Voltage characteristics for different shading patterns

2

Novice Insights
Using a simple Perturb and Observe (P&O) algorithm [2] the first LMPP will be found
and the rest will remain undiscovered resulting in severe power loss in most cases. This is the
reason why a GA [3] is used instead of a simple P&O algorithm (which can be used on systems
with single solar panel) or using a converter for each panel like in [4] (which is expensive).
Another characteristic is current-voltage (I-V) plotted in Fig. 3: without shading in black and
with the first shading pattern used for P-V characteristics in grey.

Figure 3: Current-Voltage characteristics

2.2.

Converter and Feedback Loop Block

The converter is in fact a controlled current source (equivalent in a certain way with the
inductor from a Buck Converter). At the output is a DC voltage source (a battery).
The converter has a feedback loop with a PI controller to command the input voltage in
order to sweep the P-V characteristic to find the GMPP [5-6]. At the input called Vref will be
connected the output of the GA block. All this is shown in Fig. 1 colored in green.

Genetic Algorithm Block

This is a software block that implements the GA using an Embedded Matlab Function
available in Simulink library. As it can be seen in Fig. 1 (yellow) the GA block has three inputs:
voltage and current received from the converter block and sampling time (used for simulate a
discrete appellation of the algorithm) and outputs the set point value for the voltage loop.
A GA operates within a population of individuals where each individual has a specific
genome and in this case it has 3 chromosomes: Vin, direction and step (Fig. 4). Where x 10
12 giving a total number of genes equal to 17, 18 and 19 respectively. A discus is made in
chapter 4 to determine which is the optimal number of genes.

Figure 4: Structure of an individual

In Fig. 5 is presented the flowchart of the proposed GA and the operations are explained and
shown in Fig. 6.
3

Novice Insights

Figure 5: Genetic Algorithm flow chart

The algorithm is based on the principle of natural selection between randomly chosen
individuals of a population and optimizes the solution by applying a series of genetic operations
(Fig. 6) such as: selection, crossover and mutation. Some functions are proposed for these
operations:
a) selection is accomplished with Roulette Wheel method which gives a higher chance
of selection to the best individuals depending on their fitness function (in our case the power);
b) crossover is done in binary format with single point crossover function and to create
new offspring a random number smaller then the number of genes sets the point of crossover;
c) mutation provides to the algorithm new values that normally would not be generated in
crossover operation. This operation is bitwise and complements certain bits with the probability
named Mutation Rate which is 0.1.

Figure 6: GA operations
When the GA starts the entire population (20 individuals) is initialized with random
numbers generated by a random function. The function has to have a normal distribution with
the mean of 0.5 (if the range is between 0 and 1) because this influences the algorithm and in
chapter 5 we will compare the response of three functions.
Between the mutation and the replacing of old population, in GA flow chart (Fig. 5) we
apply an MPPT algorithm on parents shown in Fig. 6 where a new operating point is calculated
as follows:

Vin (i) Vin (i) direction ( j) step( j)

Vin ( j) Vin ( j) direction (i) step(i)

(1)

The algorithm runs until the number of iterations (15 iterations) is achieved. After the last
iteration it enters in idle state where the power is monitored and if there is a change the
algorithm starts all over again.
4

Novice Insights

Genetic Algorithm Optimization

In this chapter we will study two parameters of the GA: the number of genes and the random
function to see how these affects tracking the GMPP and also find an optimal variant of the GA.
As said in chapter 3 there will be studied three different numbers of genes: 17 (green), 18
(red) and 19 (blue). Fig. 7 presents the simulation results for these conditions with no partial
shading concluding that the higher number of genes does the GMPP tracking more accurate.

Figure 7: Power obtained with different number of genes

The other parameter studied is the random function and we present three random functions:
a) Matlabs function rand(): results are shown in Fig. 8 in blue;
b) The random function described by the equation (2) where a=8121, c=28411, m=524287,
x[0]=17862 and x[n] is the current number; the results are shown in Fig. 8 in green;
x[n ] (a x[n 1] c)%m

(2)

c) The last function is presented in equation (3) where a=36969, b=18000, x[n] current
number, z[n], w[n] - temporary results; results are plotted in Fig. 8 in red.

z[n] a z[n 1] z[n 1] 16

w[n] b w[n 1] w[n 1] 16
x[n] z[n] 16 w[n]
The results shown in Fig. 8 are with 19 genes and no partial shading condition.

Figure 8: Power obtained with different random functions

5

(3)

Novice Insights
For the experimental part the chromosome has 19 genes and the random function is
described in equation (2) because these values present the optimal results.

Simulation and Experimental Results

In this section we will show results for the simulated system and the experimental system
(implemented on TMS320F2808 DSP from Texas Instruments) with the shading patterns
present in Table 1. The first pattern is in Fig. 9.

a)
b)
Figure 9: First shading pattern in Table 1: a) Simulation results; b) Experimental results
It can be seen that the GA automatically detects partial shading and starts tracking the new
GMPP even though random numbers are used to converge to the final solution.
The results for pattern b) from Fig. 2 are shown in Fig. 10 and it can be seen that even
though the GMPPs are close to one another the GA depicts it and changes the operating point.

a)
b)
Figure 10: Second shading pattern in Table 1: a) Simulation results; b) Experimental results
6

Novice Insights
The algorithm provides good results (Fig. 11) with the shading pattern from Fig. 2.c) where
the two GMPP are far from one another. The tests were performed on a small scale PV system
built with strings of diodes in parallel with current sources. In this way precise and repetitive
irradiance level can be achieved in laboratory.

a)
b)
Figure 11: Third shading pattern in Table 1: a) Simulation results; b) Experimental results

References
[1] L.F.L. Villa, T.P. Ho, J.C. Crebier, and B. Raison A power electronics equalizer application for
partially shaded photovoltaic modules, IEEE Transactions on Industrial Electronics, vol. 60(2), pp.
1179-1190, 2013
[2] D. Petreus, T. Patarau, S. Daraban, C. Morel and B. Morley, A novel maximum power point tracker
based on analog and digital control loops, Solar Energy, vol. 85, pp. 588-600, 2011
[3] C. Larbes , S.M. At Cheikh, T. Obeidi and A. Zerguerras, Genetic algorithms optimized fuzzy logic
control for the maximum power point tracking in photovoltaic system, Renewable Energy, vol.
34(10), pp. 20932100, 2009
[4] 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 Transactions on
Industrial Electronics, vol. 55(7), pp. 2610-2621, 2008
[5] M. Bodur and M. Ermis, Maximum power point tracking for low power photovoltaic solar panels,
Electrotechnical Conference, 1994
[6] P.S. Shenoy, K.A. Kim, B.B. Johnson and P.T. Krein, Differential Power Processing for Increased
Energy Production and Reliability of Photovoltaic Systems, IEEE Transactions on Power Electronics,
vol. 28(6), pp. 2968-2979, 2013.

Biography
My name is Vlad Borza and Im from Abrud (ALBA). Im a student of Electronics,
Telecommunications and Information Technology faculty, final year. I followed the courses of
Applied Electronics specialization because I found it very challenging and it was what I planned
to do even from high school.
Vlad BORZA, IVth year student
Technical University of Cluj-Napoca
Faculty of Electronics, Telecommunications and Information Technology
26-28 G. Bariiu Street, 400027 Cluj-Napoca, ROMANIA
E-mail: vlady_borza@yahoo.com
Manuscript received on May 14, revised on October 23, 2013
7