ICSET 2008

Evaluation of Genetic Algorithm based Solar

Tracking System for Photovoltaic Panels
Syamsiah Mashohor, Khairulmizam Samsudin, Amirullah M. Noor and Adi Razlan A. Rahman
AbstractThe maximum power supplied by a Photovoltaic
(PV) panels system change over time. It depends on
environmental factors such as the solar irradiation and the
temperature of these panels. The average solar energy harvested
by the conventional solar panels during the course of the day,
is not always maximized. This is due to the static placement
of the panel which limits their area of exposure to the sun. In
practice, there are three possible approaches for maximizing the
solar power extraction in medium and large scale PV systems
are sun tracking, maximum power point (MPP) tracking or
combination of both. In this paper, a Genetic Algorithm (GA)
has been proposed utilizing sun tracking approaches to maximize
the performance of PV panels. Literature suggested that the PV
panels could produce maximum power if the panels have angle of
inclination zero degree to the sun position. This work evaluate the
best combination of GA parameters to optimize a solar tracking
system for PV panels in terms of azimuth angle and tilt angle.
Simulation results demonstrated the ability of the proposed GA
system to search for optimal panel positions in term of consistency
and convergence properties. It also has proved the ability of
the GA-Solar to adapt to different environmental conditions and
successfully track sun positions in nding the maximum power
by precisely orienting the PV panels.
Index Termsgenetic algorithm, solar tracking, PV panels
I. INTRODUCTION
The photovoltaic (PV) technology is expanding due to
the growing demand for renewable energy mainly due to
the depletion of fossil fuel. Solar panel is the fundamental
energy conversion component of photovoltaic (PV) systems
[1]. PV technology offers safe and clean energy sources [2].
The abundance of solar energy throughout the whole year in
Malaysia due to the geographic location near the Equator line
provides strong reason for the implementation of an efcient
PV energy system. Studies show that solar panels constitute
a large portion (57%) of the total cost to install PV energy
system [3]. As the solar panels are relatively expensive, much
research work has been conducted to improve the utilization
of solar energy. Physically, the power supplied by the panels
depends on many extrinsic factors, such as insolation levels
(incident of solar radiation), temperature and load condition.
To ensure the economical viability of this energy system,
the development of an adaptive system that always obtains
the maximum power from these solar panels is essential. The
average solar energy intercepted by the conventional solar
panels, during the course of the day, is not always maximized.
This work was funded by Universiti Putra Malaysia under the New
Appointed Lecturer Grant Scheme, grant no. 5315958.
Dept. of Computer and Communication Systems Engineering, Universiti
Putra Malaysia, 43400 Serdang, Selangor, Malaysia (phone: +603 89464348;
fax: +603 86567127; e-mail: syamsiah@eng.upm.edu.my)
This is due to the static placement of the panel which limits
their area of exposure from the sun [4]. An automatic sun
tracking system using a number of Light Detecting Resistors
(LDRs) is proposed in [4, 5]. The principle idea is to sense the
sun light using LDR and the differential signal representing
angular error of the panel is used to rotate the panel in such
a way that the angular error is minimized. The usage of LDR
is to ensure that the solar panel is always aligned towards
the direction of the sun. However, a cloudy day and shading
condition around the PV panels may reduce the intensity of
light and effect the performance of the tracking system. Proper
placement of LDR in the system and the correct quantity of
LDR are also important in order to provide correct positioning
information for the solar panel at all time.
Researcher in [6, 7] has presented the design and
implementation of a computer-controlled dual-axis sun
tracking system to obtain high precision positioning of the PV
panel. The control of dual axis tracking system is complex
due to nonlinear dynamics and unavailability of the model
parameters. A PC-based fuzzy logic control algorithm utilizing
the knowledge of the system behavior is designed in order to
achieve the control objectives. Implementation of fuzzy logic
reasoning is limited to the predetermined conditions which
may not satisfy all conditions in the PV panel environment.
Therefore, sun tracking only occurs when these conditions are
met, which will not necessarily maximize the solar energy
absorption.
In this work, GA based solar tracking is proposed to
overcome the limitation of current method. GA is an
evolutionary algorithm which can adapt itself to the different
conditions, not only suitable for solar farm but also applicable
for harvesting energy for aerospace industries application,
electric vehicles, communication equipment and mobile
robots.
This paper will discuss the development of GA-based solar
tracking system (GA-Solar) by orienting PV panel on two axis
(azimuth and tilt angles) to obtain maximum power. In the
following section, the methodology of this system is discussed
which comprise of system description, GA architecture and
solar tracking simulator. Then, the results of the simulation
are discussed and nally, the conclusion is presented.
II. METHODOLOGY
A. System Description
This paper is a preliminary work for the implementation of
sun tracking using our specially-tailored GA for maximizing
269
978-1-4244-1888-6/08/$25.00 c 2008 IEEE
PV panel
180 degree
180 degree
Surface
Figure 1. The geometry of the PV panel for GA-Solar system
performance of PV systems. The details of GA architecture
that is implemented in this simulation is discussed in [8]. The
geometry of the PV panel that will be explored by the GA
technique is shown in gure 1. The azimuth and tilt angle can
varies from 0 to 180 degrees which can fulll any location
condition either the PV panel is installed on a high building
or on the ground. If both tilt and azimuth angle is 90 degrees,
the PV panel is normal to the ground surface.
This work assumes that the orientation search by GA-Solar
is done during the rst time calibration, right after a PV
system has been installed. GA-Solar would only require the
power measurement from the PV panel to assist the GA search
of the optimal position. After acquiring the maximum power
value from the PV system, GA-Solar will only start searching
again after power obtained from PV systems fell below some
threshold value. Subsequent search would be much faster as
the GA would start searching from the best individual.
B. GA Architecture
In this work, the GA is used to track the best tilt and
azimuth angle of the PV panel that generate the maximum
power from the sun. For this work, real integer coding has
been implemented considering the position of the sun in the
sky. The tilt and azimuth angle are both ranged from 0 to
180 degrees. Every individual represents a combination of
both position parameters which describes position of the PV
panel in relation to the sun. The domain of search for these
parameters is large (180180=32400) which is suitable for
the GA to explore.
The tness function in this work is evaluated from the power
value measured from the PV panel according to the selected tilt
and azimuth angle. There should be only one maximum power
value at one time assuming that the panel angle of inclination
is zero degrees to the sun, while the minimum value is 0W.
Iteratively, the whole population for the next generation is
formed by selected individuals from the parents and offsprings
in current generation. These individuals are ranked based on
their tness performance and only the top ttest individual are
selected for a new population. The population will perform GA
activities such as selection, mutation and crossover in every
generation until the termination requirements are fullled. The
GA search will be terminated if an ideal individual which
gives the maximum power value has been found or maximum
generation has been reached.
Table I
THE VALUES OF GA PARAMETERS INVOLVED IN EXHAUSTIVE SEARCH
GA parameter Values
Population size 30, 50, 100, 200
Maximum generation 50, 100, 150, 200, 250
Probability of crossover 0.1 to 1.0
Probability of mutation 0.001 to 0.01
C. Solar Tracking Simulator
In this work, to facilitate the development of the GA-
Solar system a solar tracking simulator has been developed.
The initial orientation of the panel are generated randomly.
The GA is used to search for the best tilt and azimuth
angle that gives the maximum power value (10W) that has
been randomly generated into a Look-Up Table (LUT). The
simulator also imitate environment factors such as insolation
level and temperature that degrade the performance of PV
system. As there is only one source of light which is from
the sun, we use a bi-variate normal distribution function to
initialize the LUT. The maximum power value and the angular
position parameters that have the maximum power value are
randomly picked and changed every time a new simulation
starts.
Figure 2 illustrate four generated LUT to simulate the power
obtained by a PV system oriented to a specic tilt and azimuth
angle ([tilt,azimuth]). From gure 2(a) and gure 2(b), a
maximum power of 10W from a PV system can be obtained by
orienting the panel to [90,90] and [0,0] position respectively.
Orienting the panel to [180,90] and [120,80] positions would
obtain the maximum power of 6.5W and 4.9W from LUT of
gure 2(c) and gure 2(d) respectively.
D. Experimental setup
To maximize the capability of GA to position the PV panel
to obtain the maximum power, the values of GA parameters
must be chosen carefully. Hence, an exhaustive search has
been conducted on all combination of GA parameters as listed
in table I to nd the best combination of GA parameters. The
exhaustive search is possible due to the availability of the solar
tracking simulator which is able to simulate the two angular
axis positioning of PV panel very fast. The total combination
of GA parameters are 2000 combinations ( 4 x 10 x 10 x 5) and
each combination is repeated for consistency evaluation. The
maximum power generated by the PV system is xed at 10W
to simplify the performance comparison between different GA
parameters setting. The effect of every GA parameter in this
GA-Solar also can be assessed based on this simulation.
The best GA parameter obtained from the exhaustive search
is then used to simulate 1000 GA-Solar search considering
the inuence of different environmental conditions and sun
positions. The maximum power generated by the PV system
will vary randomly between zero and 10W due to the
environmental factors.
270
0
1
2
3
4
5
6
7
8
9
10
0
30
60
90
120
150
180
0
30
60
90
120
150
180
2
4
6
8
10
Azimuth [Degree]
Tilt [Degree]
(a)
0
2
4
6
8
10
0
30
60
90
120
150
0
30
60
90
120
150
180
0
2
4
6
8
10
Azimuth [Degree]
Tilt [Degree]
(b)
0
2
4
6
8
10
0
30
60
90
120
150
180
0
30
60
90
120
150
180
0
2
4
6
8
10
Azimuth [Degree]
Tilt [Degree]
(c)
0
2
4
6
8
10
0
30
60
90
120
150
180
0
30
60
90
120
150
180
0
2
4
6
8
10
Azimuth [Degree]
Tilt [Degree]
(d)
Figure 2. Generated Look-Up Table for solar tracking simulator using bi-variate normal distribution function
III. RESULTS AND DISCUSSIONS
A. Optimization of GA
The goal of the GA-Solar is to obtain maximum power from
PV system at all the time and in any condition. Therefore,
selected values of GA parameters that will be used in the future
simulation or real-environment experiments must consistently
give the maximum power to optimize the utilization of the
solar energy. The selected GA parameters should also require
less generation to produce the ndings as each measurement
of tness function requires positioning of the two axis of
the PV panel. To evaluate each combination of the GA
parameters, the consistency and convergence properties of
the GA-Solar aligning the PV panel to obtain the maximum
power is considered. Consistency is dened as the frequency
of nding the maximum power, while the convergence is
dened as the least number of generation required to nd the
maximum power. Each GA parameters combination simulation
is repeated 10 times.
The consistency in nding the maximum power is the
most important feature of the GA-Solar due to adaptability
to different conditions, PV panel material and system designs.
0
2
4
6
8
10
0 400 800 1200 1600 2000
F
r
e
q
u
e
n
c
y
GA with different parameters
69%
Figure 3. Consistency of GA-Solar in nding the maximum power
In the exhaustive search simulation, the best parameters give
consistent best results in nding the maximum power point
from the Look-Up-Table that emulate power measurement
from PV panels that are exposed to the sun. As shown in
gure 3, 69% of the simulation results give good consistency
in nding the maximum power, which is 10W in all 10
sets. With regard to this ndings, the consistency of nding
maximum power in all 10 sets and convergence (sum of
generation) properties of these combination of GA parameters
are analyzed in order to choose the best GA parameter values.
The GA parameters are analyzed in sequence of population,
271
Table II
THE NORMALIZED CONSISTENCY AND CONVERGENCE FOR DIFFERENT
POPULATION SIZE
Population size Consistency Convergence
30 0.58 1.00
50 0.92 0.59
100 1.00 0.39
200 0.94 0.40
maximum generation, probability of crossover and probability
of mutation.
The population size have signicant impact on embedded
system memory footprint. Table II summarized the consistency
and convergence for different population size. Based on the
table, the population size of 100 gives the most consistent
results in nding the maximum power in every set of
simulation and the results are found in the least convergence
value.
The maximum generation determine the limit of time
that GA will spend to search for the best solution.
There is no signicant difference of consistency and
convergence properties of maximum generation parameter for
the population size of 100. However, a maximum generation of
50 is chosen since the population size of 100 is already large
enough to give enough time for the GA to nd the solution.
It would also ensure that the GA would be terminated if the
duration taken to obtain the optimal position for maximum
power is too long.
Detail analysis on the probability of crossover shows that
the probability of 0.7 has the most consistency and the least
convergence values compared to other probability values. With
this value of probability, the crossover operation will happen
moderately frequent to produce better offsprings for the next
generation. For the probability of mutation, the value of 0.001
is selected based on the least convergence required to nd the
same consistency as the other values. The value of 0.001 gives
enough occurrence of mutation for the generation to evolve
but excessive probability of mutation may increase the time to
compute.
B. Inuence of Environmental Factors
In the simulation of 1000 different environmental
conditions, maximum power from PV systems may varied
caused by various environmental factors such as cloudy
condition, raining condition, dusty PV panels, shading by other
objects and temperature of the PV systems. Hence, the ability
of the GA-Solar to adapt and nd the best panel position are
tested in these conditions.
From the simulation ndings, the GA-Solar has performed
very well when it is able to nd all the maximum power values
initiated randomly in every set. The initial maximum power
values generated and the GA-Solar solutions that have been
found are shown in gure 4. From the gure, it is obvious that
every maximum power initialized randomly in the look-up-
table, overlapped with the GA-Solar solution. Detail analysis
shows that GA-Solar is able to accurately track the correct
0
2
4
6
8
10
0 200 400 600 800 1000
P
o
w
e
r

[
W
]
Simulation runs
Initialized
GA-Solar
Figure 4. Comparison of maximum power initialized by the solar tracking
simulator with GA-Solar search solution
0
2
4
6
8
10
12
14
16
18
20
22
0 200 400 600 800 1000
G
e
n
e
r
a
t
i
o
n

G
a
i
n
Simulation runs
Figure 5. Generation gain for GA-Solar search
position of the PV panel in nding the maximum power 99.99
percent of the time.
Figure 5 shows the analysis on the generation gain required
by the GA-Solar to nd the best solutions in every set of
simulation. In average, the GA-Solar found the best solutions
after the third generation and the fastest is after the rst
generation while the slowest is after the 21
st
generation. The
standard deviation is 1.55 indicates that the generation gain
required by the GA-Solar is nearly similar in any conditions. In
hardware implementation, short generation gain is important
for the system due to the repositioning of the solar panel.
The speed of motor to position the PV panel also has to be
considered in overall system performance.
IV. CONCLUSION
This paper described the simulation of GA-based solar
tracking system for PV panels. The solar tracking simulator
developed is an important tool to simulate and evaluate the
performance of the GA-solar system. The best value for GA
parameters have been selected according to the simulation
results. The most signicant difference of frequency and
convergence occur in different population size, followed
by different probability of crossover. For various values of
probability of mutation, only the convergence yields the
signicant difference. From the observation, there is no
272
signicant difference in consistency neither the convergence
properties when the maximum generation is varied. The
GA parameter that will be used in future simulation
and hardware implementation are population size of 100,
maximum generation of 50 and probability of crossover
and mutation are 0.7 and 0.001 respectively. The simulation
ndings also show that the GA-Solar can be implemented
to maximize the power harvested by the PV panel systems
considering the various inuence of environmental factors and
different sun positions.
Future work will concentrate on
the hardware implementation to prove that the GA-Solar is
capable of maximizing the power obtained by the PV panel in
real-environment experiments and posses the ability to adapt
to different installation location and PV system design. The
hardware setup is nearly completed and later, more factors
that normally affect the solar panel system performance will
be added to the GA module.
REFERENCES
[1] H. S.-H. Chung, K. K. Tse, S. Y. R. Hui, C. M.
Mok, and M. T. . Ho, A novel maximum power point
tracking technique for solar panels using a SEPIC or
Cuk converter, IEEE Transactions on power electronics,
vol. 18, no. 3, pp. 717724, 2003.
[2] T. H. et al, Evaluation of neural network based real time
maximum power- tracking controller goes PV system,
IEEE Transactions on Energy Conversion, vol. 10, no. 3,
1995.
[3] J. Enslin, Renewable energy as an economic energy
source for remote areas, in Renewable Energy, vol. 1,
pp. 243248, 1991.
[4] B. Koyuncu and K. Balasubramanian, A microprocessor
controlled automatic sun tracker, IEEE Transactions on
Consumer Electronics, vol. 37, no. 4, pp. 913917, 1991.
[5] O. Bingol, A. Altinta, and Y. Oner, Microcontroller based
solar-tracking system and its implementation, Journal of
Engineering Sciences, vol. 12, pp. 243248, 2006.
[6] H. A. Yousef, Design and implementation of a
fuzzy logic computer-controlled sun tracking system, in
Proceedings of the IEEE International Symposium on
Industrial Electronics, ISIE99, vol. 3, pp. 10301034,
1999.
[7] A. Yazidi, F. Betin, G. Notton, and G. A. Capolino,
Low cost two-axis solar tracker with high precision
positioning, in First International Symposium on
Environment Identities and Mediterranean Area,
ISEIMA06, pp. 211216, 2006.
[8] S. Mashohor, J. R. Evans, and T. Arslan, Image
registration of printed circuit boards using hybrid
genetic algorithm, in IEEE Congress on Evolutionary
Computation, 2006.
273