International Conference on Renewable Energies and Power Quality (ICREPQ’14)

Cordoba (Spain), 8th to 10th April, 2014

Renewable Energy and Power Quality Journal (RE&PQJ)
ISSN 2172-038 X, No.12, April 2014

Comparison of Artificial Bee Colony Algorithm with other Algorithms used for
Tracking of Maximum Power Point of Photovoltaic Arrays

B. Babar and A. Crăciunescu

Faculty of Electrical Engineering

University „Politehnica” of Bucharest
Splaiul Independenței 313, 060046 – București, Romania
e-mail: [email protected], [email protected],

Abstract. Maximum power point tracker (MPPT) is one of and Observe algorithm, Genetic Algorithm and Fuzzy
the key components of a solar electricity generation systems. It is Logic algorithm.
used to extract the maximum power produced by an array of
Photovoltaic cells. In this paper Artificial Bee Colony algorithm 2. Artificial Bee Colony algorithm
is analysed and implemented for tracking the maximum power
point of Photovoltaic arrays. Furthermore the results obtained
from ABC algorithm are compared with that of Perturb and
The artificial bee colony algorithm (ABC) was
Observe method, Fuzzy Logic Control and Genetic Algorithms. introduced by Karaboga in 2005 [6] and a modified
version was introduced in by Akay and Karaboga in 2010
[7]. In a bee swarm there is a very efficient natural task
Key words: management mechanism that actively changes by the
condition of the surroundings. Foraging behaviour is one
of the very important parts of bee swarms. It is guided by
Artificial Bee Colony algorithm, Photovoltaic arrays,
the factors such as task management, division of labour,
Maximum Power Point Tracking.
grouping of bees according to their statuses and the
quantity and quality of the available food. The two main
1. Introduction management criteria for the foraging are recruitment for
exploring food sources and the abandonment after their
An increase in the use of solar electric energy has been depletion.
observed in recent years. The reason of this focus on
photovoltaic is due of their simple structure, noiseless Naturally there are three groups of bees in a swarm and
operations and little maintenance. More over these systems they differ with each other due to the tasks handled by
can be designed from very small power generating systems them. Scouts bees have the job of founding new food
to very large Mega-Watt systems. The main drawbacks of sources; they do their job by randomly exploring the area
Photovoltaic systems include lower efficiency and a high either by internal motivation or by some external clue.
capital cost. Currently the efficiency of photovoltaic Employed bees exploit the already discovered food
systems is in the range of 9÷17%. In addition to that, source until the food source is depleted. Onlooker bees
characteristics of photovoltaic are non-linear and mainly wait in the hives and they select a food source after
depend on the environmental conditions. To overcome the getting motivated by watching dances performed by the
issues of high installation costs the efficiency of these employed bees in the dancing area. Dancing area can be
systems has to be increased. One of the main techniques to considered as the main information sharing area of the
increase PV energy harvesting is by keeping the bee hives. In this area the employed bees return with the
photovoltaic to extract power at its maximum point. In Fig. nectar and perform different types of dances, while at the
1 P-V and I-V typical characteristic diagrams of same time the onlooker bees watch these dances.
Photovoltaic systems are shown. These characteristics are Different types of dances are performed for example
mainly dependent on the environmental conditions and the round, tremble and waggle dance. The type of dance and
maximum power point coordinates change with the its intensity provides a measure of the feasibility of the
temperature and solar irradiance. In the last time, in food source [8].
literature are proposed different intelligent algorithms for
actively seeking of the maximum power point [1÷5]. In The three groups of bees in a hive, as explained before,
this work the Artificial Bee Colony algorithm used for perform specialized task for maximizing the amount of
maximum power point tracking is analysed and compared nectar. In the algorithm, half of the bees are always
with other algorithms used for the same purpose: Perturb employed bees while half of them are onlooker bees.
Every food source is exploited by a single employed bee, fitnessi
Pi  , (4)
which means for every employed bee there is one food SN

source available. All the food sources are considered as

 i 1
fitnessi
viable solutions for a particular problem, which means,
there are as many solutions as there are employed bees. 4) In such a probabilistic selection scheme, as the
The steps taken by the algorithm is explained below: fitness of solutions increases, the number of
1) In the beginning the population is initialised in the onlookers attracted towards them also increases. This
range encompassing the pre-defined upper limit and is the positive feedback feature of ABC. When one
the lower. All the employed bees are allocated a complete search cycle is complete, the algorithm
position in the search space, bearing a particular searches for the exhausted food sources by checking
solution. The initialisation is performed by using the the counters and comparing it with a pre-set value
following equation: called the “limit”. If the values are the same the food
source is abandoned and a new food source is
(1) searched by the scout bees and replaced with the
xij  x min
j  rand (0;1)( x max
j  x min
j ), abandoned source. This is the negative feedback
feature of ABC.
Where i=1,2,3 …, SN and j=1,2,3 …, D. SN is the 5) The above three steps are repeated and the possible
population of the colony and D is the number of solution sets are optimized till the criteria of the
parameters to be optimised by the algorithm. After maximum number of cycles is met or the pre-defined
the initialisation, changes are brought in these error value is reached.
positions until the maximum number of cycles is By following the above steps a global minima is achieved
reached or some error criteria are met. 2) As by the algorithm.
explained earlier, each food source is associated with
one employed bee, after the initialisation each 3. The objective function
employed bee is allotted a food source. After this
allotment the bee makes changes the position of the Swarm intelligence algorithm like Particle Swarm
food source depending on local information and Optimization, ABC algorithm etc or evolution algorithms
explores surrounding food source. Once it finds a like Genetic algorithms works on minimizing an
source it evaluates its quality. The neighbouring food objective function. But, when we consider seeking a
sources are defined by the following equation. maximum power point in PV arrays the scenario changes.
The problem becomes a maximization problem. There
tij  xij   ij ( xij  xkj ), (2) are two solutions for solving this problem, the first is to
bring a change in the whole algorithm and change it to
The food source ti in the neighbourhood of xi is found
seek for maximization, or to bring a change in the
by changing one parameter of xi. In the above
objective function. For MPPT, the power was chosen as
equation j is a random integer in the range of [1, D]
the objective function which is shown in the equation
while and k = {1,2,3 …, SN} is a randomly chosen
below:
index that is different than i. and φ ij is a randomly
distributed number in the domain [-1, +1]. It can be    q(V  IRs   V  IRs 
f min ( I ,V )   I ph  I Dsat exp    1  R   V , (4)
seen from equation (2) that as the difference between    am.k .T   p 
xij and xjk decreases an optimal value is reached.
When these two values become equal, there are no The above equation is the equation for the output power
more changes in the position. After finding ti a fitness of the PV cells. The minus in the beginning of the
value is assigned to each ti by the following equation changes all the produced values of power into
equations: negative values. The parameters I and V are varied over
the range defined in the ABC algorithm and the solutions
1  (1  fi) if fi  0 are fed into the algorithm and minimization process
fitnessi   (3)
1  abs( fi) if fi  0, begins. In the MATLAB m-file equation (4) was written
as:
Where, fi is the cost value of the solution ti.
Following that a greedy selection is applied to the ObjVal = (Ipv – Io*(exp((x2+x1*Rs)/Vt/Ns/a)-1)-
(5)
values obtained for xi and ti, a better value between (x2+x1*Rs)/Rp*x2;
the two is selected according to their fitness.
3) When all the employed bees have completed their where x1 is the current and x2 is the voltage.
searches and the first cycle is completed, the
information regarding the quantity of the nectar 4. Results and analysis
(solution) is shared with the onlooker bees in the
dancing area. An onlooker bee makes a probability on For tracking the maximum power point, the artificial bee
the quality of nectar (by the fitness values). In the colony algorithm was implemented in MATLAB. For
ABC algorithm a roulette wheel scheme is used for testing the algorithm, the Kyocera KC200GT PV panel
such probabilistic values as shown in the equation was modelled in MATLAB. The specifications for this
below: PV panel are given in the Table I.
 Table III. Pmax given by different algorithms

Irradiation ABC P&O GA FL

(W/m2) W W W W
1000 206.08 200.10 205.00 199.00
800 163.61 160.40 161.17 157.10
600 118.63 119.30 118.46 91.09
400 71.51 45.55 75.42 41.31
200 33.56 36.70 31.09 10.92

It is worth mentioning here that for Perturb and Observe

algorithm a time of 4.5 seconds was used, while for other
algorithms on the average 0.3 seconds was used. The data
from the Table III can be better visualized in graphic
form. In Figure 2 the power correlation graph of the used
algorithms is shown. It can be seen that ABC algorithm
Fig. 1. Photovoltaic panel I-V and P-V characteristic curves; performs better that other algorithms.
maximum power point is shown on these curves.

Table I. Specifications of KYOCERA KC200GT panel

Variable Value
Maximum Power, Pmax 200 Watts (+10% / -5%)
Maximum Power Voltage, Vmpp 26.3 V
Maximum Power Current, Impp 7.61 A
Open Circuit Voltage, Voc 32.9 V
Short Circuit Current, Isc 8.21 A
Temperature Coefficient, KVoc -1.23 x 10-1 V/0C
Temperature Coefficient, KIsc 3.18 x 10-3 A/0C

The results of output power maximum searched by the

Artificial Bee Colony algorithm are given in Table II. The
irradiance G and the temperature T were set at 1000 W/m2
and 250C, respectively. The simulation was performed Fig. 2. Power correlation graph comparing performances of
ABC, P&O, GA and FL algorithms in seeking
10times to check the reliability of the algorithm.
MPP at different irradiations.
Table II. ABC algorithm results for 1000 W/m2 and 250C
5. Conclusion
Run Number Output Value of Pmax
1 206.087
The Artificial Bee Colony algorithm can be considered as
2 195.189
one of the main competitive algorithms for the maximum
3 189.110
point tracking of the Photovoltaic arrays. In comparison
4 205.008
with Perturb & Observe and Fuzzy Logic algorithms,
5 199.028 ABC Perturb & Observe and Fuzzy Logic algorithms,
6 205.008 ABC algorithm give better results. It gives similar or
7 203.606 superior results in comparison with Genetic algorithm.
8 204.848
9 205.008 Acknowledgement
10 200.675
This work was realized through the Partnership program
In the above sample of outputs the best value is 206.087, in priority domains - PN II, developed with support from
the worst value is 189.11 and the average of the above data ANCS CNDI - UEFISCDI, project no. PN-II-PT-PCCA-
is 201.36 which has an error of 0.0068%. The average time 2011-3.2-1670.
for one run was 0.2 seconds.
References
Along with Artificial Bee Colony algorithm, Perturb and
[1] S. Farhad, M. Saffar-Avval, Younessi-Sinaki, “Efficient
Observe Algorithm, Genetic Algorithm and Fuzzy Logic
design of feedwater heaters network in steam power plants
Algorithm were applied for the same PV panel. A using pinch technology and exergy analysis”. International
comparison of Artificial Bee Colony algorithm with Journal of Energy Research, 32 (2008), pp. 1–11.
Perturb and Observe algorithm, Genetic Algorithms and [2] UNDP. World energy assessment 2000 – energy and the
Fuzzy Logic algorithm is shown in Table III. challenge of sustainability. New York: UNDP; 2000
(ISBN 9211261260).
[3] S. Bilgen, K. Kaygusuz and A. Sari “Renewable energy for [13] Won CY, Kim DH, Kim SC, Kim WS, Kim HS. “A new
a clean and sustainable future Energy Sources”, Part A: maximum power point tracker of photovoltaic arrays using
Recovery, Utilization, and Environmental Effects, 26 (12) fuzzy controller”. In: Proceedings of 25th annual IEEE
(2004), pp. 1119–1129. power electronics specialists conference; 1994. p. 396–
[4] P. Bhubaneswari, S. Iniyan, R. Goic..A review of solar 403.
photovoltaic technologies.. Renewable and Sustainable [14] Veerachary M, Senjyu T, Uezato K.” Neural-network-
Energy Reviews, 15 (2011), pp. 1625–1636. based maximum-power-point tracking of coupled-inductor
[5] H.L. Tsai et al, “Development of generalized photovoltaic interleaved- boostconverter-supplied PVsystem using
model using MATLAB/SIMULINK”. 978-988-98671-0-2. fuzzy controller”. IEEE Transactions on Industrial
World Congress on Engineering and Computer Science, Electronics 2003;50(4):749–58
WCECS’08, San Francisco (October 22–24, 2008). [15] Yacine Labbi and Djilani Ben Attous” Maximum
[6] D. Karaboga, An Idea Based On Honey Bee Swarm for Photovoltaic Power Tracking under Different Conditions
Numerical Optimization, Technical Report TR06, Erciyes Using a Genetic Algorithm and Particle Swarm
University, Engineering Faculty, Computer Engineering Optimization”. Journal of Electrical and Control
Department, 2005. Engineering. ISSN 2226-2881 (Print) ISSN 2226-289X
[7] A modified Artificial Bee Colony algorithm for real- (Online)
parameter optimization Original Research Article [16] Yusof, Y.; Sayuti, S.H.; Abdul Latif, M.; Wanik, M.Z.C.
Information Sciences, Volume 192, 1 June 2012, Pages 120- Modeling and simulation of maximum power point tracker
142 Bahriye Akay, Dervis Karaboga. for photovoltaic system. In Proceedings of Power and
[8] V. Tereshko, Reaction–diffusion model of a honeybee Energy Conference, Kuala Lumpur, Malaysia, 29–30
colony’s foraging behaviour, in: M. Schoenauer (Ed.), November 2004; pp. 88–93.
Parallel Problem Solving from Nature VI, Lecture Notes in [17] Youngseok J, et al. “Improved perturbation and
Computer Science, vol. 1917, Springer–Verlag, Berlin, observation method (IP&O) of MPPT control for
2000, pp. 807–816. photovoltaic power systems”. In: Proceedings of 31st
[9] http://www.kyocerasolar.com/assets/001/5195.pdf IEEE photovoltaic specialists conference; 2005.16]
[10] J.C.H. Phang, D.S.H. Chan and J.R. Phillips, “Accurate [18] Timothy, J. Ross. Fuzzy Logic with Engineering
analytical method for the extraction of solar cell model Applications, second ed. John Wiley & Sons Ltd., 2004.
parameters”. Electronics Letters, Volume 20, Issue 10, 10 [19] F. Chekired, C. Larbes, D. Rekioua and F. Haddad,
May 1984, p. 406 – 408, DOI: 10.1049/el:19840281. “Implementation of a MPPT fuzzy controller for
[11] Maximum photovoltaic power tracking for the PV array photovoltaic systems on FPGA circuit”. 1876–6102. 2011,
using the fractional-order incremental conductance method Elsevier Ltd. doi:10.1016/j.egypro.2011.05.062.
Original Research Article Applied Energy, Volume 88, [20] R. Ramaprabha and B.L. Mathur, “Genetic Algorithm
Issue 12, December 2011, Pages 4840-4847 Based Maximum Power Point Tracking for Partially
[12] Chia-Hung Lin, Cong-Hui Huang, Yi-Chun Du, Jian-Liung Shaded Solar Photovoltaic Array”. International Journal of
Chen. “Maximum power tracking for photovoltaic power Research and Reviews in Information Sciences (IJRRIS)
system: Development and experimental comparison of two Vol. 2, No. 1, March 2012, ISSN: 2046-6439.
algorithms”. Renewable Energy, Volume 35, Issue 10,
October 2010, Pages 2381-2387 Issam Houssamo, Fabrice
Locment, Manuela Sechilariu.