G Model

ASOC 2876 1–11 ARTICLE IN PRESS

Applied Soft Computing xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Applied Soft Computing

journal homepage: www.elsevier.com/locate/asoc

1 Artificial bee colony based algorithm for maximum power point

2 tracking (MPPT) for PV systems operating under partial shaded
3 conditions
4 Q1 Abou Soufyane Benyoucef a,∗ , Aissa Chouder b , Kamel Kara a , Santiago Silvestre c ,
5 Oussama Ait Sahed a
a
6 SET Laboratory, Electronics Department, Blida 1 University, BP 270 Blida, Algeria
b
7 Electrical Engineering Department, Faculty of Technology, University of M’sila, BP 166 Ichbilia, Algeria
c
8 Electronic Engineering Department, Universitat Politècnica de Catalunya, C/ Jordi Girona 1-3, Campus Nord UPC, 08034 Barcelona, Spain
9

10
27 a r t i c l e i n f o a b s t r a c t
11
12 Article history: Artificial bee colony (ABC) algorithm has several characteristics that make it more attractive than other
13 Received 22 June 2014 bio-inspired methods. Particularly, it is simple, it uses fewer control parameters and its convergence is
14 Received in revised form 3 November 2014 independent of the initial conditions. In this paper, a novel artificial bee colony based maximum power
15 Accepted 24 March 2015
point tracking algorithm (MPPT) is proposed. The developed algorithm, does not allow only overcom-
16 Available online xxx
ing the common drawback of the conventional MPPT methods, but it gives a simple and a robust MPPT
17
scheme. A co-simulation methodology, combining Matlab/SimulinkTM and Cadence/PspiceTM , is used to
18 Keyword:
verify the effectiveness of the proposed method and compare its performance, under dynamic weather
19 Maximum power point tracking
20 Photovoltaic systems
conditions, with that of the Particle Swarm Optimization (PSO) based MPPT algorithm. Moreover, a lab-
21 Partially shaded conditions oratory setup has been realized and used to experimentally validate the proposed ABC-based MPPT
22 Boost converter algorithm. Simulation and experimental results have shown the satisfactory performance of the proposed
23 Artificial bee colony algorithm approach.
24 Particle Swarm Optimization © 2015 Published by Elsevier B.V.
25 Co-simulation Simulink/Pspice
26 Q3 DSP implementation

28 1. Introduction (MPP) is achieved by adjusting the operating point of the PV array 40

using a DC–DC converter. 41

29Q4 Photovoltaic (PV) energy sources are becoming a mature Most MPPT techniques are based on the assumption that all cells 42

30 technology while their applications are spreading, ranging from in the same module and all modules in the same string receive 43

31 supplying small electronic devices to large power plants con- the same irradiance. Perturb and observe (P&O) and incremental 44

32 nected to medium and low voltage grids. However, some problems conductance are the most popular algorithms implemented in com- 45

33 remain a challenge for PV systems as improving overall efficiency, mercial charge regulators and grid connected inverters and can 46

34 maximize the available power, minimize the return period of the accurately track the MPP under uniform illuminating conditions 47

35 installation cost, detection and diagnosis of faults, etc. [1–4]. Several [5–7]. However, PV modules are often subject to partial shade con- 48

36 research works and specialized laboratory reports have addressed ditions (PSC), being the real problem responsible of most output 49

37 the issue of low yields and power losses in PV systems facilities and power reduction and mismatch [8–10]. When PV array is operating 50

38 all have approved the use of a power optimizer known as maximum under this conditions, the power voltage curves are characterized 51

39 power point tracking (MPPT). Typically, the maximum power point by the apparition of multiple local peaks which are due to the acti- 52

vation of bypass diodes avoiding shaded cells from damage [11–13]. 53

In such case, conventional MPPT algorithms may miss the goal by 54

converging to a local maximum rather than the global one, giving 55

a rise to a significant loss in the output power and consequently 56

Q2 ∗ Corresponding author. Tel.: +213 698131514; fax: +213 27669217. to a low yield of the overall system. In order to cope with the 57
E-mail addresses: benyoucefsoufi[email protected] (A.S. Benyoucef), effects of shading on the P–V curves, a number of improvements 58
[email protected] (A. Chouder), [email protected] (K. Kara),
of conventional MPPT algorithms have been proposed. Some of 59
[email protected] (S. Silvestre), [email protected]
(O.A. Sahed). them are topology based and need extra additional power circuits 60

http://dx.doi.org/10.1016/j.asoc.2015.03.047
1568-4946/© 2015 Published by Elsevier B.V.

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
2 A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx

61 to perform global MPPT (GMPPT) [14–16]. Thus, the overall effi- In this work, a novel ABC-based MPPT with direct control 127

62 ciency is reduced and the total cost will be increased. Some other method for PV systems working under PSC is proposed. With this 128

63 techniques are algorithm based, such as fuzzy logic with polar con- MPPT control scheme, the duty cycle is adjusted directly by the 129

64 troller [17], sequential extremum seeking control [18], dividing algorithm without the need of using a linear controller. The pro- 130

65 rectangle (DIRECT) search control [19]. These techniques are costly, posed MPPT scheme advances the state-of-the art through the 131

66 time-consuming processes and need a complex hardware for their following contributions: 132

67 implementation.
68 Recently, bio-inspired (BI) methods such as Artificial Neural Net- • Excellent tracking capability with a good accuracy. 133
69 works (ANN), evolutionary computation and swarm intelligence • No requirement of knowledge about the characteristics of the PV 134
70 algorithms have emerged as powerful optimization algorithms for array. 135
71 solving complex problems and providing optimum solutions. Par- • The use of just two control parameters, allowing great flexibility 136
72 allel, dynamic, decentralized, asynchronous and self-organizing and simplicity. 137
73 behavior of the nature inspired algorithms is best suited for soft
74 computing applications [20,21]. Indeed, BI-inspired techniques The rest of this paper is organized as follows. Section 2 presents 138
75 were used in several applications as an efficient tool to solve com- an overview of partial shading on PV modules and its impact on 139
76 plex optimization problems and design sophisticated controllers MPP location. Section 3 briefly describes the principle of the ABC 140
77 [22–25]. algorithm and how it is applied to MPPT in PV systems. Section 141
78 The effectiveness of BI-inspired methods in handling complex 4 provides simulation and experimental results of the proposed 142
79 nonlinearities, such that encountered in PV array behavior, and approach. A comparison of the proposed MPPT algorithm with the 143
80 their implementation simplicity make them very attractive to solve PSO-based MPPT method is also presented in this section. Finally, 144
81 the MPPT problem of PV systems, especially in the case of partial the drawn conclusions from this work are provided. 145
82 shading and module mismatches [26,27]. Artificial Neural Net-
83 works (ANN), are one of the bio-inspired methods that was used
84 in MPPT techniques. Typically, they were used to estimate the 2. PV systems working under partial shading conditions 146

85 MPP with respect to the randomly changing weather conditions

86 [28], and to improve the P&O and IC algorithms [29]. It is known A PV array is formed of several PV modules connected in series 147

87 that, using ANN with larger number of hidden nodes leads to more and/or parallel and the total power is a combination of the power 148

88 accurate results, but at the expense of longer computational time derived from each PV module. Fig. 1(a) presents a PV array formed 149

89 and complex hardware implementation. Evolutionary computation by two series connected PV modules. If one of the PV modules is 150

90 techniques, such as Genetic algorithm (GA) [30] and Differential shaded, it acts as a load instead of a power source. In long term 151

91 Evolution (DE) [31], have been also proposed to deal with the MPPT conditions, the shaded PV module will be damaged due to hotspots 152

92 problem. Commonly, the parameters setting of these algorithms phenomenon. Hence, the bypass diodes are added to protect the PV 153

93 are achieved using trial-and-error process which leads to a large modules from self-heating during partial shading conditions [12]. 154

94 computational time [32]. Under uniform insolation, the bypass diodes are reverse biased and 155

95 Particle Swarm Optimization (PSO) and Ant Colony Optimiza- have no impact. When the PV module is shaded, the bypass diode 156

96 tion (ACO) are the most swarm intelligence techniques used in across the PV module is forward biased and the current passes 157

97 the development of MPPT schemes [33,34]. The PSO convergence through the diode instead of the PV module as shown in Fig. 1(a) 158

98 significantly depends on the initial position of the agents which and (b). 159

99 gives a poor convergence rate in some situation. In [35] the However, the bypass diode transforms the P–V curves into more 160

100 authors have used the ant colony based optimization technique. complicated shape characterized by multiple peaks as it is shown 161

101 Though the authors have demonstrated the applicability of the in Fig. 1(c). So, in order to extract the maximum power from the 162

102 method using simulation, it is difficult to be implemented and PV array, the system should be operated at the global maximum 163

103 realized in hardware platform. In addition, both of PSO and ACO power point (Gmpp). If not it is the case, up to 70% of power could 164

104 have five parameters to be determined, which make them inflex- be lost [51]. 165

105 ible and complex. In [36] the authors have used fuzzy logic to Thus, in order to achieve optimal energy harvesting from the PV 166

106 dynamically adapt some important parameters in PSO and ACO array, an intelligent and efficient MPPT method should be used. 167

107 approaches. Better results than those of the original methods have
108 been obtained, but the developed MPPT algorithms are complex 3. ABC optimization applied to direct control of MPPT 168

109 and time-consuming.

110 Artificial bee colony (ABC) is a relatively new member of swarm 3.1. Fundamentals of ABC optimization algorithm 169

111 intelligence techniques. It was proposed by Karaboga [37], based on

112 foraging behavior, learning, memorizing and information sharing The artificial bee colony algorithm is a swarm based 170

113 characteristics of honeybees. Several research papers have already meta-heuristic algorithm that was introduced for solving multidi- 171

114 evaluated the performances of ABC algorithm against different mensional and multimodal optimization problems. The algorithm 172

115 approaches like PSO, ACO and several other optimization tech- is specifically based on the model proposed by Tereshko and Loen- 173

116 niques [38–41]. Their finding states that the main advantages of the garov [52–54] for the foraging behavior of honeybee colonies. 174

117 ABC algorithm are its simplicity [42], light computing complexity, In the ABC algorithm, the artificial bees are classified into three 175

118 high solutions accuracy [38,43,44], convergence independent of the groups: employed bees, onlooker bees and scouts. A bee that is 176

119 initial conditions [45] and its ability to deal with local minima [37]. currently searching for food or exploiting a food source is called 177

120 At present, the ABC algorithm has been successfully applied in dis- an employed bee. A bee waiting in the hive for making decision 178

121 tinct fields of science such as electrical engineering: in power flow to choose a food source is named as an onlooker. Employed bees 179

122 optimization and optimal sizing of photovoltaic systems [46,47], whose food sources cannot be improved through a predetermined 180

123 control engineering: in enhancement of control algorithms [48], number of trials become scouts and their food sources are aban- 181

124 image processing: in block matching techniques [49], mechanical doned. The number of food sources is equal to the number of 182

125 engineering: in optimization of design approaches [50] and many employed bees and also equal to the number of onlooker bees. 183

126 others. Analogously, in the optimization context, the position of a food 184

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx 3

Fig. 1. Operation of PV array: (a) under uniform irradiation, (b) under PSC, (c) the resulting I–V and P–V curves for (a) and (b).

185 source represents a candidate solution of the optimization prob- by scout bee and a new solution is randomly searched. The new 222

186 lem and the nectar amount of a food source corresponds to the solution xi will be generated using Eq. (1). 223

187 quality (fitness) of the associated solution. It is clear, from the above explanation, that there are three con- 224

188 At the initialization phase, the ABC generates a randomly trol parameters in the basic ABC: The number of candidate solutions 225

189 distributed initial population of SN solutions. Each solution is pro- which is equal to the number of employed and onlooker bees SN, 226

190 duced within its limits according to the equation below: the value of “limit” and the maximum cycle number MCN. 227

j j j j
A detailed pseudo-code of the ABC algorithm for one- 228
191 xi = xmin + rand[0, 1](xmax − xmin ) i = 1, 2, . . ., SN, dimensional optimization problem is given below: 229
1: Begin
192 j = 1, 2, . . ., D (1) 2: Generate the initial solutions xi , i = 1, 2, . . ., SN using Eq. (1)
j j 3: Evaluate the fitness (fiti ) of the population
193 where xmin and xmax represent respectively the minimum and the 4: Set cycle to 1
194 maximum of the parameter j and D is the number of optimization 5: Repeat
195 parameters. 6: For each employed bee {
196 After initialization, the population of the solutions is subject to Produce new solution vi using Eq. (2)
Calculate the value fiti
197 repeated cycles C = 1, 2, . . ., MCN, of the search processes of the
Apply the greedy selection process}
198 employed bees, the onlooker bees and the scout bees. 7: Calculate the probability values Pi for the solutions xi by Eq. (3)
199 For each cycle, every employed bee produces new solution vij 8: For each onlooker bee { 230

200 according to Eq. (2) and then evaluates its fitness fiti . Select a solution xi depending on pi
Produce new solution vi by using Eq. (2)
201 vij = xij + ij (xij − xkj ) (2) Calculate the value fiti
Apply the greedy selection process}
202 where k ∈ {1, 2, . . ., SN} and j ∈ {1, 2, . . ., D} are randomly cho- 9: If there is an abandoned solution for the scout
203 sen indexes. Although k has to be different from i. ij is a random Then replace it with a new solution which will be randomly produced
using Eq. (1)
204 number between [−1, 1].
10: Memorize the best solution achieved so far
205 After the information is shared by the employed bees, each 11: cycle = cycle + 1
206 onlooker finds new solution vij within the neighborhood of xi by 12: until cycle = MCN
207 using Eq. (2), based on the probability Pi defined as:
fiti
208 Pi = SN (3) 3.2. Application of ABC to the MPPT problem 231
n=1
fitn

209 where fiti is the fitness value of the solution xi . The ABC-based optimization described in Section 3.1 is now 232

210 The value of each generated candidate solution vij that is not applied with a slight change made in the scout bees’ phase, to real- 233

211 within the boundary of the allowed search space is updated so that ize the MPPT algorithm for photovoltaic generation system (PGS) 234

212 it will be within this space. operating under PSC with direct control technique. 235

213 The fitness of each new produced candidate solution vij is com- To realize the direct control ABC-based MPPT, each candidate 236

214 pared with that of its old one. If the new solution has an equal solution is defined as the duty cycle value d of the DC–DC converter, 237

215 or better fitness than the old solution, it replaces the old one in the so the optimization problem has one parameter to be optimized 238

216 memory. Otherwise, the old one is retained in the memory. In other (D = 1). Thus, Eqs. (1) and (2) become: 239

217 words, a greedy selection mechanism is employed in the selection

218 operation between the old and the candidate one. di = dmin + rand[0, 1](dmax − dmin ) (4) 240
219 At the end of each search cycle, if the fitness of a solution can-
220 not be improved and the predetermined number of trials, which
221 is called “limit”, is exhausted, then the solution will be abandoned new di = di + i (di − dk ) (5) 241

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
4 A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx

242 The fitness of each solution (duty cycle) is chosen as the gener- Table 1
Parameters of the used PV modules (BP SX 80).
243 ated power Ppt of the PGS. Then, Eq. (3) becomes:
Parameter Value
Ppvi
244 Pi = SN (6) Maximum power 80 W
n=1
Ppvi Open circuit voltage 22.1 V
Maximum power voltage 17.6 V
245 To evaluate the duty cycles, the digital controller successively
Short circuit current 4.8 A
246 outputs the PWM signal according to the value of di , and then the Maximum power current 4.6 A
247 PV voltage Vpvi and current Ipvi can be measured and the corre- Temperature coefficient −0.080 V/◦ C
248 sponding power (Ppvi ) of each duty cycle di can be calculated. It Configuration 2s1P (2 bypass diodes)
249 should be noted that in order to acquire correct samples, the time
250 interval between two successive duty cycle evaluations (Ts) has to
251 be greater than the power converter’s settling time. The search process of a new MPP will be executed again when- 300

252 In the original ABC algorithm, at the end of each search cycle the ever the inequality, given above, is satisfied. This ensures that the 301

253 abandoned solution that has not improved its fitness over a prede- MPPT algorithm can always find the global MPP under various 302

254 termined number of cycles (limit) is replaced by a new randomly working environmental conditions. 303

255 chosen solution. If this strategy is used in the MPPT algorithm, the The flowchart of the proposed ABC-based MPPT algorithm of PV 304

256 stopping criterion stops the search process before the limit cycle be systems is shown in Fig. 2. 305

257 reached, because the power Ppv remains unchanged (premature

258 convergence). Thus, a very important phase that gives ability to the 4. Results and discussion 306

259 algorithm to escape from local minima is eliminated. In this work

260 a different strategy for the scout bees’ phase is proposed to avoid The 160-W prototyping system, given in Fig. 3, is implemented 307

261 that problem, which consists in replacing the duty cycle giving less and used to evaluate, by simulation and experimentally, the per- 308

262 power at the end of each search cycle, by another randomly chosen formance of the proposed MPPT algorithm. 309

263 value. This system consists of two PV module connected in series, a 310

264 The application procedure of the proposed method can be DC–DC converter and a digital controller in which the MPPT algo- 311

265 divided in four phases: rithms under test are implemented. In this paper, a simple boost 312

converter is used to interface the voltage from the PV modules to 313

266 - Initialization phase the load. The parameters of the used PV module are listed in Table 1. 314

267 Set the algorithm parameters by including the number of can-

268 didate solutions SN, the maximum cycle number MCN and the 5. Simulation results 315
269 sampling time Ts. Then, a randomly distributed initial SN duty
270 cycles are generated using Eq. (4) and evaluated. In this section, a co-simulation methodology combining Mat- 316

lab/Simulink and Pspice has been adopted to assess the feasibility 317
271 di (i = 1, 2, . . ., SN) and the effectiveness of the proposed MPPT algorithm. The PV array 318

and the boost converter were implemented in Pspice environment 319

272 - Employed bees phase while the MPPT algorithms were implemented in Matlab/Simulink 320
273 The employed bees’ duty cycle are updated using Eq. (5) and environment. Comparisons of the proposed algorithm against 321
274 then evaluated. Then, the greedy selection process is applied and PSO-based MPPT, under dynamic weather conditions, were also 322
275 the probability value Pi for each duty cycle di using Eq. (6) is performed. Fig. 4 shows the implemented Simulink model, and 323
276 calculated. Fig. 5 shows the Pspice circuit of the boost converter and the PV 324
277 - Onlooker bees phase array, where two bypass diodes protecting eighteen cells in each 325
278 The roulette wheel selection is used to recruit the onlooker bees module are considered. 326
279 for local searching around the chosen duty cycle value which is
280 depending on the calculated probability Pi . The new duty cycles
5.1. Parameterization of the ABC algorithm 327
281 are evaluated and then the greedy selection process is applied.
282 - Scout bees phase
In the proposed MPPT algorithm, the number of candidate 328
283 Eq. (4) is used to replace the duty cycle giving less power with
solutions (SN) influences the convergence speed and the tracking 329
284 the new randomly chosen duty cycle. The fitness of the generated
performance of the algorithm. More of bees makes easier to find the 330
285 duty cycle is then evaluated.
global MPP with a good accuracy, but more time will be required. A 331

fewer number of bees gives a better convergence speed, but the 332
286 At the end of each search cycle, the algorithm memorizes the
convergence rate could be unsatisfactory. Therefore, a trade-off 333
287 best solution achieved so far and repeats the procedure from the
between fast convergence speed and the convergence rate should 334
288 employed bees phase until the maximum cycle number (MCN) is
be achieved when choosing the number of candidate solutions SN. 335
289 reached or until the power value remains unchanged within a spec-
Fig. 6 shows the relationship between the number of candidate 336
290 ified number of cycles.
solutions (SN) and the convergence rate for all cases investigated 337
291 Usually, the real working environment of the PV system is
in the simulations. The relationship between the number of candi- 338
292 always changing due to the varying weather conditions, and as a
date solutions (SN) and the average convergence time, when Ts is 339
293 result, the global MPP is always changing. This requires the MPPT
chosen equal to 0.05 s, is shown in Fig. 7. Since our objective for the 340
294 algorithm to have the ability to search for a global MPP for the
MPPT algorithm is to get fast convergence with a high convergence 341
295 new weather condition. For this purpose, the search process has
rate, SN has been chosen equal to 3. The maximum cycle number 342
296 to restart with a total re-initialization whenever the weather con-
is set to 30 cycles; the largest number reached in these tests is 15 343
297 ditions are changed. Here, we use the following strategy to detect
cycles. 344
298 these changes:
According to these obtained results, the parameters values of 345
|Ppvnew − Ppvlast | the implemented ABC-based MPPT algorithm are: SN = 3 bees and 346
299 ≥ Ppv (%) (7)
Ppvlast MCN = 30 cycles. 347

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
G Model
ASOC 2876 1–11 ARTICLE IN PRESS
A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx 5

Fig. 2. Flowchart of the proposed ABC-based MPPT with direct control technique for PV system.

Fig. 3. System block diagram.

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
6 A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx

Fig. 4. Implemented Simulink model.

Fig. 5. Pspice schematic of PV array and boost converter.

348 5.2. Power tracking with various shading patterns: simulation curve. For SP2 there are two peaks, while for SP3 and SP4 there are 357

349 study three peaks. 358

For each of the above four cases, the ABC-based MPPT algorithm 359

350 A simulation study was carried out to validate the ability of the is executed 1000 times. The ideal power values and the average 360

351 proposed ABC-based MPPT to track the global MPP under steady values of the extracted power are shown in Table 3. It is shown 361

352 and transient shading patterns. Since it is very difficult to test all that the proposed ABC-based MPPT is able to successfully track 362

353 the non-uniform irradiance conditions, only some shading patterns the global MPP. In fact, the extracted power is very close to the 363

354 are considered in the study. The four used shading patterns (SPs) are ideal power in all cases. The distribution of the average values 364

355 listed in Table 2. For SP1, the irradiance on all the PV sub-modules is of the extracted power from the PV modules, over the shading 365

356 uniform; as a result, only one peak appears in the P–V characteristic patterns, is shown in Fig. 8. The results obtained indicate that 366

Fig. 6. Relationship between the number of candidate solutions (SN) and conver- Fig. 7. Relationship between the number of candidate solutions (SN) and the average
gence rate. convergence time.

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx 7

Fig. 8. Power extracted by the ABC-based MPPT algorithm for the four different
shading patterns.
Fig. 9. Shading pattern SP1 to shading pattern SP2.
367 the effectiveness of the proposed method is not affected by the
368 initial conditions of the search process. The ability to find the
369 global MPP for the new weather conditions is very important. In
370 order to illustrate the tracking ability of the proposed ABC-based
371 MPPT algorithm under transient irradiance conditions, we have
372 considered three cases – Case 1: SP changes from SP1 to SP2; Case
373 2: SP changes from SP1 to SP3, and Case 3: SP changes from SP1 to
374 SP4. The power, the voltage, the current transient characteristics
375 and the corresponding duty cycle for Cases 1, 2 and 3 are shown
376 in Figs. 9–11, respectively. The sampling period Ts of the MPPT
377 algorithm is set to 0.05 s and Ppv (%) (the weather conditions
378 change detection) value is set to 2%. It can be seen that when the
379 shading pattern changes from a uniform condition to a partially
380 shaded condition at 8 s, the proposed MPPT algorithm can find the
381 global MPP for the new shading pattern.

382 5.3. Comparison of the proposed MPPT with PSO-based approach:

383 simulation study

384 This comparative study is performed with the PSO-based MPPT

385 approach proposed in [55]. The parameter settings of the imple-
386 mented PSO-based MPPT algorithm are listed in Table 4. The
Fig. 10. Shading pattern SP1 to shading pattern SP3.
387 detection parameter value of the weather conditions change
388 (Ppv (%)) is set to 2% for both algorithms.
389 First, the comparison is carried out on the basis of three statisti-
390 cal indicators, given in Table 5. Thus, both algorithms are executed
391 1000 times for each shading pattern given in Table 2. It can be
392 noticed that the ABC-based MPPT algorithm gives better perfor-
393 mance, in term of number of successful convergence, when the PV

Table 2
Considered shading patterns.

Pattern no. Shading pattern [Gpv11, Gpv12, Gpv21, Gpv22] (W/m2 )

SP1 [1000, 1000, 1000, 1000]

SP2 [1000, 500, 1000, 1000]
SP3 [1000, 700, 100, 1000]
SP4 [1000, 500, 100, 1000]

Table 3
Performance of the proposed ABC-based MPPT under various shading patterns.

Pattern no. Ideal output power (W) Output power obtained (W)

SP1 159.652 159.380

SP2 117.896 116.568
SP3 88.948 88.377
Fig. 11. Shading pattern SP1 to shading pattern SP4
SP4 74.140 73.976

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
8 A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx

Gpv22 (W\m2) Gpv21 (W\m2) Gpv12 (W\m2) Gpv11 (W\m2)

1000

500

0
1000

500

0
1000

500

0
1000

500
Cell Temperature (°c)

0
60

40

20
6 8 10 12 14 16 18 20
(Hour)
Time

Fig. 12. Fourteen hour typical Algerian daily profiles of irradiance and temperature (6.00 am–8.00 pm).

394 array working under PSC. However a slight difference with respect ance. This is more clearly highlighted in the bar-graph of Fig. 14, 412

395 to the accuracy and the average convergence value is reported in which it is reported the hourly array yield (YA ). The hourly array 413

396 where the ABC-based MPPT algorithm has a slow convergence yield (YA ) is defined as the hourly power output of the PV array 414
415
397 speed toward the GMPPT compared to the PSO-based MPPT algo- divided by the peak power of the installed PV array, it is given by:
398 rithm. n
399 The second comparison study is based on the real daily power P
n=1 PV array
/n
YA = 416
400 production of the PV array subject to non-uniform daily irradiation. PPV array peak
401 For this purpose, the simulation was carried out using real weather
402 profiles of typical Algerian sky. Fig. 12 gives the chosen temperature
403 and four irradiance daily profiles in order to be applied to each PV
150
404 sub-module. Pload~ABC
405 Fig. 13 reports the simulation results for the extracted power Pload~PSO

406 using the ABC- and the PSO-based MPPT controllers when irradi-
407 ance and temperature profiles shown in Fig. 12 are used. The data
Extracted power (w)

100
408 were recorded with a sampling period of 3 minutes. It is clearly seen
409 that for the whole profile, the extracted power in the case of the
410 ABC-based MPPT algorithm is greater than that of the PSO-based
411 MPPT algorithm when PV modules are receiving different irradi- 50

Table 4
PSO algorithm parameters used in the study. 0
6 8 10 12 14 16 18 20
Parameter Symbol Value Time (h)

Number of particles N 6
Fig. 13. Simulation results for the 14 h (6.00 am–8.00 pm) Algerian profiles
Maximum cycle number max generation 30
(extracted powers).
Inertia weight w 0.4
Cognitive coefficient c1 1.2
Social coefficient c2 1.6 90
ABC
80
PSO
Table 5
Comparison between the ABC- and the PSO-based MPPT algorithms under different 70
shading patterns. 60
Percentage (%)

Pattern no./ Number of Accuracy (%) Average

50
algorithm successful convergence
convergence (%) cycle 40
ABC 99.80 99.83 9.39 30
SP1
PSO 99.20 99.66 7.95
20
ABC 97.70 99.72 9.57
SP2
PSO 90.20 99.06 8.41 10
ABC 96.70 99.36 10.32
SP3 0
PSO 87.10 98.22 9.22 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Time (h)
ABC 95.60 99.78 10.27
SP4
PSO 78.40 98.56 9.17
Fig. 14. Hourly array yield.

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx 9

Fig. 15. Experimental setup of MPPT system. Fig. 18. Corresponding array voltage, and current waveforms under uniform inso-
lation condition.

every Ts = 0.05 s. Figs. 15 and 16 show the hardware platform and 426

the PV array used in this experiment. 427

6.1. Power tracking with various shading patterns: experimental 428

study 429

Two tests were performed in order to verify the effectiveness of 430

the proposed MPPT algorithm under steady and transient shading 431

conditions. 432

In the first experiment, the PV array was subject to uniform 433

irradiance conditions where the corresponding I–V and P–V curves 434

Fig. 16. PV array with an artificial shade.

are recorded. These characteristics are shown in Fig. 17(a) and (b), 435

respectively. According to these figures, the corresponding volt- 436

age and current of the MPP in the unshaded case are 23.8 V and 437
417 where n is the samples number and PPV array peak is the peak power 3.66 A. The corresponding dynamic responses of the voltage and 438
418 of the installed PV array. the current of the PV array are given in Fig. 18. It can be seen that 439

the generated voltage and current in Fig. 18 are around 23.5 V and 440

419 6. Experimental results 3.6 A, presenting a good agreement with results shown in Fig. 17. 441

This result confirms the correct operation of the proposed MPPT 442

420 In the hardware implementation, both ABC- and PSO-based algorithm under uniform insolation conditions. 443

421 MPPT control programs were developed using C++ language and In the second experiment, the PV array was operated under uni- 444

422 compiled and downloaded on the eZdsp TMS320F28112 DSP board. form insolation condition until the instant 6 s when a part of the PV 445

423 The realized test bench has the same configuration given in Fig. 3. array was intentionally shadowed as shown in Fig. 16. The new 446

424 The boost converter was designed to be driven at a 20 kHz switch- recorded I–V and P–V curves are shown in Fig. 17. The behavior of 447

425 ing frequency, and the output voltage and current were sampled the MPPT algorithm before and during the partial shadow is shown 448

Fig. 17. I–V and P–V curves of the PV array under uniform insolation and partially shaded conditions.

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
10 A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx

The ABC algorithm has proven to be a good optimizer for the 477

MPPT problem, with great potential for future enhancement to 478

address the shortcoming observed in the slow convergence com- 479

pared to PSO-based approach. 480

References 481

[1] D.J. Swider, L. Beurskens, S. Davidson, J. Twidell, J. Pyrko, W. Prüggler, H. Auer, 482
K. Vertin, R. Skema, Conditions and costs for renewables electricity grid con- 483
nection: examples in Europe, Renew. Energy 33 (2008) 1832–1842. 484
[2] A. Elasser, M. Agamy, J. Sabate, R. Steigerwald, R. Fisher, M. Harfman- 485
Todorovic, A comparative study of central and distributed MPPT architectures 486
for megawatt utility and large scale commercial photovoltaic plants, in: IECON 487
2010 – 36th Annual Conference on IEEE Industrial Electronics Society, 2010, 488
pp. 2753–2758. 489
[3] D.G. Lorente, S. Pedrazzi, G. Zini, A. Dalla Rosa, P. Tartarini, Mismatch losses in 490
PV power plants, Sol. Energy 100 (2014) 42–49. 491
[4] S. Silvestre, A. Chouder, E. Karatepe, Automatic fault detection in grid connected 492
PV systems, Sol. Energy 94 (2013) 119–127. 493
[5] C.-H. Lin, C.-H. Huang, Y.-C. Du, J.-L. Chen, Maximum photovoltaic power 494
Fig. 19. Behavior of the new MPPT algorithm before and during the occurrence of tracking for the PV array using the fractional-order incremental conductance 495
partial shadow. method, Appl. Energy 88 (2011) 4840–4847. 496
[6] T. Esram, P.L. Chapman, Comparison of photovoltaic array maximum power 497
point tracking techniques, IEEE Trans. Energy Convers. 22 (2007) 439–449. 498
Table 6 [7] V. Salas, E. Olías, A. Barrado, A. Lázaro, Review of the maximum power point 499
Comparison between the ABC and the PSO based MPPT algorithms under different tracking algorithms for stand-alone photovoltaic systems, Sol. Energy Mater. 500
shading patterns. Sol. Cells 90 (2006) 1555–1578. 501
[8] A. Woyte, J. Nijs, R. Belmans, Partial shadowing of photovoltaic arrays with 502
Pattern no./ Number of Accuracy (%) Average different system configurations: literature review and field test results, Sol. 503
algorithm successful convergence Energy 74 (2003) 217–233. 504
convergence (%) cycle [9] A. Chouder, S. Silvestre, Analysis model of mismatch power losses in PV sys- 505
tems, J. Sol. Energy Eng. 131 (2009) 024504–024509. 506
ABC 98.50 99.22 10.1 [10] S. Armstrong, W.G. Hurley, A thermal model for photovoltaic panels under 507
SP1
PSO 97.50 98.53 8.9 varying atmospheric conditions, Appl. Therm. Eng. 30 (2010) 1488–1495. 508
[11] H. Patel, V. Agarwal, MATLAB-based modeling to study the effects of partial 509
ABC 96.00 98.52 10.3
SP2 shading on PV array characteristics, IEEE Trans. Energy Convers. 23 (2008) 510
PSO 89.5 98.26 9.1
302–310. 511

ABC 93.50 98.76 10.4 [12] S. Silvestre, A. Boronat, A. Chouder, Study of bypass diodes configuration on PV 512
SP3 modules, Appl. Energy 86 (2009) 1632–1640. 513
PSO 76.00 97.83 9.3
[13] J. Solórzano, M.A. Egido, Hot-spot mitigation in PV arrays with distributed MPPT 514
(DMPPT), Sol. Energy 101 (2014) 131–137. 515
[14] G. Velasco-Quesada, F. Guinjoan-Gispert, R. Pique-Lopez, M. Roman- 516
449 in Fig. 19. As can be seen in this figure, the generated PV power is Lumbreras, A. Conesa-Roca, Electrical PV array reconfiguration strategy for 517

450 around 20 W which is the maximum achievable power, according energy extraction improvement in grid-connected PV systems, IEEE Trans. Ind. 518
Electron. 56 (2009) 4319–4331. 519
451 to Fig. 17(b). [15] K. Kobayashi, I. Takano, Y. Sawada, A study of a two stage maximum power point 520
tracking control of a photovoltaic system under partially shaded insolation 521
conditions, Sol. Energy Mater. Sol. Cells 90 (2006) 2975–2988. 522
452 6.2. Comparison of the proposed MPPT with PSO-based approach:
[16] G.R. Walker, J.C. Pierce, Photovoltaic DC–DC module integrated converter for 523
453 experimental study novel cascaded and bypass grid connection topologies – design and optimi- 524
sation, in: Power Electronics Specialists Conference, 2006. PESC‘06. 37th IEEE, 525

454 In the last experiment a comparison between the proposed ABC- 2006, pp. 1–7. 526
[17] E. Syafaruddin, T. Karatepe, Hiyama, Artificial neural network-polar coordi- 527
455 based MPPT and PSO-based MPPT was carried out under three nated fuzzy controller based maximum power point tracking control under 528
456 shading patterns (SP1: 1 peak; SP2: 2 peaks; SP3: 3 peaks). This partially shaded conditions, IET Renew. Power Gener. 3 (2009) 239–253. 529

457 comparison was performed by executing, for each pattern, both [18] L. Peng, L. Yaoyu, J.E. Seem, Sequential ESC-based global MPPT control for pho- 530
tovoltaic array with variable shading, IEEE Trans. Sustain. Energy 2 (2011) 531
458 MPPT algorithms 200 times. The obtained results, given in Table 6, 348–358. 532
459 confirm the better performance of the ABC-based MPPT especially [19] N. Tat Luat, L. Kay-Soon, A global maximum power point tracking scheme 533

460 in terms of number of successful convergence. employing DIRECT search algorithm for photovoltaic systems, IEEE Trans. Ind. 534
Electron. 57 (2010) 3456–3467. 535
[20] M. Pedemonte, S. Nesmachnow, H. Cancela, A survey on parallel ant colony 536
461 7. Conclusion optimization, Appl. Soft Comput. 11 (2011) 5181–5197. 537
[21] S. Gupta, S. Bhardwaj, P.K. Bhatia, A reminiscent study of nature inspired com- 538
putation, Int. J. Adv. Eng. Technol. 1 (2011) 117–125. 539
462 In this work, a novel ABC-based MPPT by using direct duty [22] Y. Maldonado, O. Castillo, P. Melin, Particle swarm optimization of interval type- 540
463 cycle control method for PV systems under partially shaded is pre- 2 fuzzy systems for FPGA applications, Appl. Soft Comput. 13 (2013) 496–508. 541
[23] J.C. Cortes-Rios, E. Gomez-Ramirez, H.A. Ortiz-de-la-Vega, O. Castillo, P. Melin, 542
464 sented. The feasibility of the proposed algorithm has been verified
Optimal design of interval type 2 fuzzy controllers based on a simple tuning 543
465 by analysing various shading patterns on a PV system. The pro- algorithm, Appl. Soft Comput. 23 (2014) 270–285. 544
466 posed method has been compared with PSO-based MPPT algorithm. [24] O. Castillo, P. Melin, Optimization of type-2 fuzzy systems based on bio-inspired 545

467 Simulation results have shown that the proposed ABC-based MPPT methods: a concise review, Inf. Sci. 205 (2012) 1–19. 546
[25] O. Castillo, A. Martinez, A. Martinez, Evolutionary computing for topology 547
468 algorithm provides better tracking performance to find the global optimization of type-2 fuzzy systems, in: P. Melin, O. Castillo, E. Ramírez, J. 548
469 MPP under partially shaded and dynamic weather conditions than Kacprzyk, W. Pedrycz (Eds.), Analysis and Design of Intelligent Systems using 549

470 PSO-based MPPT algorithm. In addition, the proposed algorithm Soft Computing Techniques, Springer, Berlin Heidelberg, 2007, pp. 63–75. 550
[26] Z. Salam, J. Ahmed, B.S. Merugu, The application of soft computing methods for 551
471 requires only two control parameters and its convergence is not MPPT of PV system: a technological and status review, Appl. Energy 107 (2013) 552
472 dependent on the initial condition. Furthermore, it does not require 135–148. 553
473 knowledge about the characteristics of the PV array. Finally, the [27] K. Ishaque, Z. Salam, Syafaruddin, A comprehensive MATLAB Simulink PV sys- 554
tem simulator with partial shading capability based on two-diode model, Sol. 555
474 experimental results have demonstrated the ability of the algo- Energy 85 (2011) 2217–2227. 556
475 rithm to track, with a good accuracy, the MPP of the PV array under [28] M. Veerachary, N. Yadaiah, ANN based peak power tracking for PV supplied DC 557
476 partial shading conditions. motors, Sol. Energy 69 (2000) 343–350. 558

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047
 G Model
ASOC 2876 1–11 ARTICLE IN PRESS
A.S. Benyoucef et al. / Applied Soft Computing xxx (2015) xxx–xxx 11

559 [29] X. Jinbang, S. Anwen, Y. Cheng, R. Wenpei, Y. Xuan, ANN based on IncCond algo- [42] W.-f. Gao, S.-y. Liu, A modified artificial bee colony algorithm, Comput. Oper. 598
560 rithm for MPP tracker, in: 2011 Sixth International Conference on Bio-Inspired Res. 39 (2012) 687–697. 599
561 Computing: Theories and Applications (BIC-TA), 2011, pp. 129–134. [43] F. Kang, J. Li, Q. Xu, Structural inverse analysis by hybrid simplex artificial bee 600
562 [30] A. Messai, A. Mellit, A. Guessoum, S.A. Kalogirou, Maximum power point track- colony algorithms, Comput. Struct. 87 (2009) 861–870. 601
563 ing using a GA optimized fuzzy logic controller and its FPGA implementation, [44] A. Singh, An artificial bee colony algorithm for the leaf-constrained minimum 602
564 Sol. Energy 85 (2011) 265–277. spanning tree problem, Appl. Soft Comput. 9 (2009) 625–631. 603
565 [31] H. Taheri, Z. Salam, K. Ishaque, Syafaruddin, A novel maximum power point [45] E. Cuevas, F. Sención, D. Zaldivar, M. Pérez-Cisneros, H. Sossa, A multi-threshold 604
566 tracking control of photovoltaic system under partial and rapidly fluctuating segmentation approach based on artificial bee colony optimization, Appl. Intell. 605
567 shadow conditions using differential evolution, in: 2010 IEEE Symposium on 37 (2012) 321–336. 606
568 Industrial Electronics & Applications (ISIEA), 2010, pp. 82–87. [46] U. Kılıç, K. Ayan, Optimizing power flow of AC–DC power systems using 607
569 [32] A.K. Qin, V.L. Huang, P.N. Suganthan, Differential evolution algorithm with artificial bee colony algorithm, Int. J. Electr. Power Energy Syst. 53 (2013) 608
570 strategy adaptation for global numerical optimization, IEEE Trans. Evolut. Com- 592–602. 609
571 put. 13 (2009) 398–417. [47] A.F. Mohamed, M.M. Elarini, A.M. Othman, A new technique based on artificial 610
572 [33] F. Qiang, T. Nan, A complex-method-based PSO algorithm for the maximum bee colony algorithm for optimal sizing of stand-alone photovoltaic system, J. 611
573 power point tracking in photovoltaic system, in: 2010 Second International Adv. Res. 5 (2014) 397–408. 612
574 Conference on Information Technology and Computer Science (ITCS), 2010, pp. [48] O.A. Sahed, K. Kara, A. Benyoucef, Artificial bee colony-based predictive control 613
575 134–137. for non-linear systems, Trans. Instit. Meas. Control (2014). 614
576 [34] L. Yi-Hwa, H. Shyh-Ching, H. Jia-Wei, L. Wen-Cheng, A particle swarm [49] E. Cuevas, D. Zaldívar, M. Pérez-Cisneros, H. Sossa, V. Osuna, Block matching 615
577 optimization-based maximum power point tracking algorithm for PV systems algorithm for motion estimation based on artificial bee colony (ABC), Appl. 616
578 operating under partially shaded conditions, IEEE Trans. Energy Convers. 27 Soft Comput. 13 (2013) 3047–3059. 617
579 (2012) 1027–1035. [50] A. Şencan Şahin, B. Kılıç, U. Kılıç, Design and economic optimization of shell 618
580 [35] L.L. Jiang, D.L. Maskell, J.C. Patra, A novel ant colony optimization-based max- and tube heat exchangers using artificial bee colony (ABC) algorithm, Energy 619
581 imum power point tracking for photovoltaic systems under partially shaded Convers. Manag. 52 (2011) 3356–3362. 620
582 conditions, Energy Build. 58 (2013) 227–236. [51] S. Daraban, D. Petreus, C. Morel, A novel MPPT (maximum power point track- 621
583 [36] F. Valdez, P. Melin, O. Castillo, A survey on nature-inspired optimization algo- ing) algorithm based on a modified genetic algorithm specialized on tracking 622
584 rithms with fuzzy logic for dynamic parameter adaptation, Expert Syst. Appl. the global maximum power point in photovoltaic systems affected by partial 623
585 41 (2014) 6459–6466. shading, Energy 74 (2014) 374–388. 624
586 [37] D. Karaboga, An Idea Based on Honey Bee Swarm for numerical Optimiza- [52] V. Tereshko, Reaction-diffusion model of a honeybee colony’s foraging 625
587 tion. Technical Report-TR06, Erciyes University, Engineering Faculty, Computer behaviour, in: Proceedings of the 6th International Conference on Parallel 626
588 Engineering Department, 2005. Problem Solving from Nature, Springer-Verlag, London, UK, 2000, pp. 807– 627
589 [38] A. Banharnsakun, T. Achalakul, B. Sirinaovakul, The best-so-far selection in 816. 628
590 artificial bee colony algorithm, Appl. Soft Comput. 11 (2011) 2888–2901. [53] V. Tereshko, T. Lee, How information mapping patterns determine foraging 629
591 [39] S.-M. Chen, A. Sarosh, Y.-F. Dong, Simulated annealing based artificial bee behaviour of a honeybee colony, Open Syst. Inf. Dyn. 9 (2002) 181–193. 630
592 colony algorithm for global numerical optimization, Appl. Math. Comput. 219 [54] V. Tereshko, A. Loengarov, Collective decision-making in honeybee foraging 631
593 (2012) 3575–3589. dynamics, Comput. Inf. Syst. J. 9 (2005) 1–7. 632
594 [40] D. Karaboga, S. Okdem, C. Ozturk, Cluster based wireless sensor network routing [55] L. Chun-Liang, L. Yi-Feng, H. Jia-Wei, L. Yi-Hua, A PSO-based MPPT algorithm for 633
595 using artificial bee colony algorithm, Wirel. Netw. 18 (2012) 847–860. photovoltaic systems subject to inhomogeneous insolation, in: 2012 Joint 6th 634
596 [41] S. Samanta, S. Chakraborty, Parametric optimization of some non-traditional International Conference onSoft Computing and Intelligent Systems (SCIS) and 635
597 machining processes using artificial bee colony algorithm, Eng. Appl. Artif. 13th International Symposium on Advanced Intelligent Systems (ISIS), 2012, 636
Intell. 24 (2011) 946–957. pp. 721–726. 637

Please cite this article in press as: A.S. Benyoucef, et al., Artificial bee colony based algorithm for maximum power point tracking (MPPT)
for PV systems operating under partial shaded conditions, Appl. Soft Comput. J. (2015), http://dx.doi.org/10.1016/j.asoc.2015.03.047