Renewable Energy 95 (2016) 367e380

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

A reliable methodology based on mine blast optimization algorithm

for optimal sizing of hybrid PV-wind-FC system for remote area in
Egypt
Ahmed Fathy
Electrical Power & Machine Dept., Faculty of Engineering, Zagazig University, Egypt

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, a reliable methodology incorporated mine blast algorithm (MBA) is applied to solve the
Received 26 August 2015 optimal sizing of a hybrid system consisting of photovoltaic modules, wind turbines and fuel cells (PV/
Received in revised form WT/FC) to meet a certain load of remote area in Egypt. The main objective of the optimal sizing process is
8 April 2016
to achieve the minimum annual cost of the system with load coverage. The sizing process is performed
Accepted 16 April 2016
optimally based on real measured data for solar radiation, ambient temperature and wind velocity
recorded by the solar radiation and meteorological station located at national research institute of as-
tronomy and geophysics, Helwan city, Egypt. Three other meta-heuristic optimization techniques, par-
Keywords:
PV/WT/FC system
ticle swarm optimization, cuckoo search and artificial bee colony are applied to solve the problem and
Fuel cell the results are compared with those obtained by the proposed methodology. A power management
Wind turbine strategy that regulates the power flow between each system component is also presented. The obtained
Mine blast algorithm results show that; applying the proposed methodology will save about 24.8% in the annual total cost of
the proposed system compared with PSO, 8.956% compared with CS and 11.5576% compared with ABC.
The proposed algorithm based on MBA is candidate for solving the presented optimization problem of
optimal sizing the hybrid PV/WT/FC system.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction and required less maintenance cost [2]. In order to design efficient,
reliable and economic hybrid system; the optimal sizing must be
With the predicted end of fossil resources, it is required to considered. The design of such hybrid system is very complex
provide alternative energy clean sources such as renewable energy problem as it is required to model each component of the system
sources; additionally there are about 1.5 billion people in the world [3]. Many studies in the field of optimal sizing of the hybrid systems
still living in small villages in remote areas [1]. They need an have been reported [4e16]. The previous methods in the field of the
electrical energy for living requirements with difficulties in optimal sizing of hybrid system can be classified into three main
extension the utility grid so stand-alone energy system should be categories; Basic technical methods [4,5], simulation programs
taken place to supply such loads located in remote areas. One of the [6e8] such as HOMER [9] and HOGA [10] programs and meta-
common hybrid renewable energy systems comprises photovoltaic heuristic optimization algorithms such as particle swarm optimi-
(PV) modules and wind turbines (WT) with the usage of storage zation (PSO) [11], tabu search (TS), simulated annealing (SA), har-
system like fuel cell (FC). In this system the extra power generated mony search (HS) as described in Refs. [12,13], artificial bee colony
from the PV/WT system is fed to electrolyzer which produces (ABC) [14], multi-objective artificial bee colony (MOABC) algorithm
hydrogen stored in hydrogen tank. In the case of failure of PV/WT [15] and Cuckoo search algorithm [16]. A delay perturbation esti-
system in supplying the load; the FC acts as a secondary power mation with signal compensation techniques have been presented
source to plug the deficit of the generating units (PV/WT) and to in Ref. [17] for designing PI controller to control the pitch of wind
supply the required power to the load. The main advantages of this turbine. In Ref. [18] a freshwater pinch analysis has been used to
hybrid system can be summarized as; it is environmental friendly modify photovoltaic with reverse osmosis system (RO) system to
minimize the use of fresh water outsources.
This work presents proposed steps incorporating recent meta-
E-mail address: [email protected]. heuristic optimization technique, mine blast algorithm (MBA), in

http://dx.doi.org/10.1016/j.renene.2016.04.030
0960-1481/© 2016 Elsevier Ltd. All rights reserved.
368 A. Fathy / Renewable Energy 95 (2016) 367e380

Fig. 2. Two diode model of PV cell [19].

The PV cell can be modeled either by single diode model or double

diode model, in this analysis a double diode model described in Ref.
[19] is used and shown in Fig. 2.
Fig. 1. The configuration of the proposed hybrid system. Referring to Fig. 2, the output current extracted from the PV
module can be expressed as follows:
solving the optimal sizing of hybrid PV/WT/FC system to meet a 8 0 0 1 1
< Vmod
N þ ImodNp$Rs
certain load and achieve the minimum annual cost of the overall Imod ¼ Np Iph  Io1 @exp@ s
A  1A
system. The analysis is performed based on measured data for solar : a1 VT1
radiation, ambient temperature and wind velocity recorded by the 0 0 1 19 ( )
Vmod Imod $Rs = Np
solar radiation and meteorological station of national research N þ N Ns Vmod þ Imod Rs
@
 Io2 exp @ s p
A 1 A 
institute of astronomy and geophysics located at Helwan city, south a2 VT2 ; Rp
of Cairo, Egypt. The station is located above the sea level over a hell
top of about 114 m at latitude 29.87
N and longitude 31.30
E. Three (1)
other intelligent optimization algorithms in the previous works,
particle swarm optimization (PSO) [11], artificial bee colony (ABC) where Iph is photon current, Io1 and Io2 are the reverse currents of
[14] and cuckoo search (CS) [16], are applied to solve the problem diode 1 and diode 2 respectively, Vmod is the module voltage, a1 and
and the results are compared to those obtained by MBA. A power a2 are the ideality factors of each diode, Ns is the number of series
management methodology that regulates the power flow between cells, Np is the number of parallel cells (Np ¼ 1), VT1 and VT2 are the
each system component is also presented. The obtained results thermal voltages of each diode, Rs is the series resistance of the PV
show that the presented methodology based on MBA algorithm is cell and Rp is the parallel resistance of the cell. The power extracted
the most reliable, the most efficient and the fastest one compared from the PV array consisting of Npv connected modules at time t can
with the others. be expressed as follows:

PPV ðtÞ ¼ Npv Imod ðtÞVmod ðtÞ (2)

2. Mathematical model of hybrid system

In this section the mathematical model of the proposed hybrid

renewable system consisting of PV array, wind turbines, fuel cells, 2.2. Wind turbine
electrolyzer, hydrogen tank and inverter is presented. The config-
uration of this system is shown in Fig. 1. This system under study is The power generated from the wind turbine is related on the
a stand-alone system for load in a remote area. The PV array and wind speed, it can be expressed in the following formula [20]:
wind turbine are connected to DC bus to feed AC load via inverter;

8
> 0 if V < VCutin or V > VCutout
>
>  
>
>
>
<P PWTo  PWTmax
WTmax ðVðtÞ  Vr Þ if Vr < V  VCutout
Pw ðtÞ ¼ VCutout  Vr (3)
>
>  
>
>
>
> VðtÞ  VCutin 3
: PWTmax if VCutin  V  Vr
Vr  VCutin

the extra power from the PV/WT is fed to the electrolyzer which where PWT-max is the rated power of the wind turbine, Vcut-in is the
produces hydrogen, the hydrogen is stored H2 tank which feeds the cut-in wind speed, Vcut-out is the cut-out wind speed, Vr is the
fuel cell. The FC is connected to DC bus as in case of PV/WT failure nominal wind speed, V (t) is the wind speed at time t and PWTo is the
the load is supplied from it. turbine power at cut-out wind velocity. The total power extracted
from wind turbines at time t are:
2.1. PV array
PWT ðtÞ ¼ Nw Pw ðtÞ (4)
The PV module comprises a number of series cells to obtain the
required voltage. The power obtained from the PV module is equal where Nw is the number of wind turbines and Pw(t) is the output
to the multiplication of the terminal voltage and the output current. power from one wind turbine at time t. The model of wind turbine
 A. Fathy / Renewable Energy 95 (2016) 367e380 369

used in this work is 7.5 kW Bergey wind power's BWC Excel-R/48

model [20]. The Bergey wind turbine model has rated capacity of
7.5 kW and provides output voltage of 48 V.

2.3. Fuel cell

The main advantages of fuel cells compared with other storage

systems are they have highly efficiency and environmental friendly.
The most commonly used FC technologies are solid oxide fuel cells
(SOFCs) and polymer electrolyte membrane, or Proton exchange
membrane fuel cells (PEMs). In order to calculate the amount of
electric energy extracted from the FC, it is important to estimate the
amount of generated energy from the FC per 1 kg hydrogen which
can be expressed as following [11]: Fig. 3. The power flow direction in case of Pg ¼ PLoad/xInv.

H2 heating value ð kWh =m3Þ

EgFC ¼ H2used ðkgÞxFC (5)
H2 denisty ðkg=m3Þ  Mode of balance generation

where Hused
2 is the amount of hydrogen input to FC in Kg, xFC is the FC In this mode the amount of power generated is equal to the load
efficiency, H2 heating value is equal to 3.4 kWh/m3 in the standard power which means there is no extra power to charge FC. There-
condition and H2 density is 0.09 kg/m3. In this work; the used fuel fore; the direction of power in this case is a unique from source to
cell is Ballard one [21]. load as shown in Fig. 3. The total power generated from the hybrid
PV/WT, Pg, is transferred to load via inverter, the power from source
2.4. Electrolyzer to electrolyzer, Pg_Elec, is zero, the power from electrolyzer to tank,
PEl-tank, is zero, the power from tank to fuel cell, PT_FC, is zero and
Electrolyzer is an electrochemical device utilizing the electrical the power from fuel cell to converter, PFC_Inv, is zero. In this mode
current to dissociate water into hydrogen and oxygen. It consists of one can derive the following equations:
cells connected in series. In order to estimate the amount of
hydrogen produced by the electrolyzer; the amount of energy used PLoad
Pg ¼ Pg ¼ PPV þ PWT ¼ (7)
to produce 1 kg hydrogen must be calculated. The mathematical Inv
xInv
formula of the produced hydrogen from the electrolyzer can be
expressed as follows [11]:
Etank ðtÞ ¼ Etank ðt  1Þ (8)
H heating value 3:4 ðkWh =m3Þ
Energy 1kgH2 ¼ xelec 2 ¼ xelec where Etank (t) and Etank (t1) are the energy stored in tank at time t
H2 denisty 0:09 ðkg=m3Þ
and t1 respectively and xInv is the inverter efficiency.
(6.a)
 Mode of over generation
Pg Elec
H2 produced in ðkgÞ ¼ (6.b)
Energy 1kgH2 In this mode the generated power is greater than the load power
therefore there is power excess exploited in charging the fuel cell;
Where xelec is the electrolyzer efficiency and Pg_Elec is the power therefore the power in this mode has two directions as shown in
transferred from the PV/WT system to electrolyzer. In this study Fig. 4; the first one is from the PV/WT system to load via inverter
Avalence electrolyzer type is used [22]. while the second one is from source to FC via electrolyzer and H2
tank. The mathematical formula of powers in this mode can be
2.5. Hydrogen tank expressed as follows:

Hydrogen tank is a device used to store either compressed Pg ¼ Pg  Pg (9)

Elec Inv
hydrogen gas or liquid. The amount of hydrogen stored in the tank
is based on the case of operation which may be one of three pro-
posed cases which are: mode of balance generation, mode of over
generation, mode of over demand. These three modes of operation
are explained in the next section.

2.6. Inverter

Inverter is a power electronic device which converts DC signal to

AC one. It has been used with the hybrid system in case of supplying
AC load.

3. Operation scheme of hybrid PV/WT/FC system

In this section a strategy of power management that shows the

direction of power flow between source and load is presented. The
system operation mode can be divided into three main categories: Fig. 4. The power flow direction in case of Pg < PLoad/xInv.
370 A. Fathy / Renewable Energy 95 (2016) 367e380

PEl tank ¼ xelec Pg Elec (10)

Etank ðtÞ ¼ Etank ðt  1Þ þ PEl tank ðtÞ (11)

 Mode of over demand

In this operation mode the generated power is less than the load
power, the PV/WT system fails in supplying the load; therefore the
FC generates power to help the hybrid system in meeting the de- Fig. 6. MBA schematic diagram including exploration and exploitation processes [24].
mand as shown in Fig. 5. The following equations describe the
power flow in such mode:
represents the annual total cost of the hybrid system, x ¼ [NPV, NW,
 
PLoad NFC, NElec, NTank, NInv] and u ¼ [CCPV, CM C M C M C M
PV, Cw, Cw , CFC, CFC, CElec, CElec,
PFC ¼  Pg (12) C M C
Conv
xInv Inv CTank, CTank, CInv]. The importance of constraints is dividing the
design space into feasible and infeasible domains. In case of feasible
PFC domain all constraints are satisfied while in case of one constraint is
Inv
PTank ¼ (13) violated; the infeasible domain is taken place. The difference be-
FC
xFC
tween the equality and inequality constraints functions is obtained
via their names as the equality constraints are satisfied for the
Etank ðtÞ ¼ Etank ðt  1Þ  PTank FC ðtÞ (14)
design and control variables such that a function called h(x, u) is
equal to zero. On the other hand; inequality constraints are satisfied
in case of a function called gðx; uÞ is greater than or equal to 0 [23].
To derive the objective function; the system annual cost can be
4. The proposed constraint objective function divided into three parts; capital cost, replacement cost and opera-
tion maintenance cost. The total cost of each component can be
In this work the optimal sizing of the PV/WT/FC is performed for expressed as follows:
minimizing the overall annual cost of the system. The novelty of
   n 
M ð1 þ iÞ  1
this work is proposing efficient steps incorporating MBA algorithm
for determining the optimal size of PV/Wind/FC system based on CtPV ¼ NPV CPV
C
þ CPV n (17.a)
ið1 þ iÞ
real measured data such that minimizing the total annual cost and
covering the load; additionally a power management strategy    n 
which regulates the direction of power flow in the proposed system M ð1 þ iÞ  1
CtW ¼ NW CW
C
þ CW (17.b)
is also presented. The general formula of optimization problem can ið1 þ iÞn
be described as follows:
0  
Minimize f ðx; uÞ (15) B
j¼ n
nFC
1 !
B C X 1
Subjected to constrains CtFC ¼ NFC B R
BCFC þ CFC 1þ
@ j¼1 ð1 þ iÞjnFC
xmin  x < xmax & umin  u < umax (16.a)
1
gðx; uÞ  0 (16.b)   C
n
M ð1 þ iÞ  1
C
þ CFC C (17.c)
ið1 þ iÞ n C
hðx; uÞ ¼ 0 (16.c) A

where f(x,u) is the objective function to be minimized, x is the

vector of design variables, u is the vector of control variables, g(x,u) 0  
represents inequality functional constraints and h(x,u) represents j¼ n
1 !
B nElec
X
equality functional constraints. In this work the objective function B C 1
CtElec ¼ NElec B R
BCElec þ CElec 1þ
@ j¼1 ð1 þ iÞjnElec

1
  C
ð1 þ iÞn  1 C
M
þ CElec C (17.d)
ið1 þ iÞn C
A

   
ð1 þ iÞn  1
CtTank ¼ NTank CTank
C M
þ CTank n (17.e)
ið1 þ iÞ

CtInv ¼ NInv CInv

C
(17.f)
Fig. 5. The power flow direction in case of Pg > PLoad/xInv.
Fig. 7. (a) The flow chart of the MBA algorithm (b) The proposed steps incorporated MBA.
372 A. Fathy / Renewable Energy 95 (2016) 367e380

900

800

700
Load power (kW)

600

500

400

300

200
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Time (hours)

Fig. 8. The average hourly load demand.

1200

1000

800
Solar radiation (W/m2)

600

400

200

0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Time(hours)

Fig. 9. The average hourly solar radiation (W/m2).

40

35

30
Ambient temperature (C)

25

20

15

10

5
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Time (hours)

Fig. 10. The average hourly ambient temperature (
C).

The objective function represents the overall system annual cost life time of the system, nFC is the life time of fuel cell and nElec is the
can be written as follows: life time of electrolyzer. It is observed that there is no replacement
cost for PV/WT, H2 tank as their life time is assumed to be equal to
f ðx; uÞ ¼ CtPV þ CtW þ CtFC þ CtElec þ CtTank þ CtInv (18) the system life time which is 20 year. While the life time of FC and
electrolyzer are assumed to be 5 years. The proposed parametric
where CC denotes the capital cost, CM denotes the maintenance constraints are:
cost, CR denotes the replacement cost, i is the interest rate, n is the
 A. Fathy / Renewable Energy 95 (2016) 367e380 373

40

35

30

25
Wind speed (m/s)

20

15

10

0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Time (hours)

Fig. 11. The average hourly wind speed (m/s).

140 8000

120 7000

6000
Wind turbine output power (W)

100
PV module output power (W)

5000
80

4000

60
3000

40
2000

20
1000

0 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Time (hours) Time (hours)

(a) (b)
Fig. 12. The output power from one PV module and one wind turbine.

assumed to be that at any time the summation of the generated

min max
NPV  NPN  NPV (19.a) power from the PV/WT and the power supplied by the FC is greater
than or equal the demand and expressed as follows:
min
NW max
 NW  NW (19.b)  
gðx; uÞ ¼ Pg Inv ðtÞ þ PFC Inv ðtÞ  Pload ðtÞ  0 (20)

min max Since the solar radiation, ambient temperature, wind speed are
NFC  NFC  NFC (19.c) varied due to climate variability; the parameters of PV array and
wind turbines must be well designed to keep up with these fluc-
min max tuations; the hybrid system parameters involved in the design
NElec  NElec  NElec (19.d)
process is the numbers of each component to cover the demand
power overall the year not their electrical specifications.
min max
NTank  NTank  NTank (19.e)

min max 5. Mine blast optimization algorithm (MBA)

NInv  NInv  NInv (19.f)
In this study; the minimum limits of the design variables are Mine blast algorithm is one of the most modern meta-heuristic
selected to guarantee complete hybrid renewable system able to optimization algorithms. It has been proposed by Sadollah et al.
supply the load and the maximum limit of each component is [24]. The basic idea of MBA is motivated from the observation of a
calculated are calculated based on assumption that each power mine bomb explosion; in this process thrown pieces of shrapnel
resource needs to satisfy the load solely. The calculated values are; have been generated and collided with other mine bombs inside
Nmax
PV ¼ 510, Nw
max
¼ 150, Nmax max max
FC ¼ 530, NElec ¼ 660, NTank ¼ 670, the mine field and near the explosion area helping in exploring it.
Nmax
Inv ¼ 200. The main aim of the MBA algorithm is to evaluate the most
The proposed functional inequality functional constraints is explosive mine located at optimal location resulting in most
374 A. Fathy / Renewable Energy 95 (2016) 367e380

Fig. 13. A complete model of the proposed PV/WT/FC Simulink model.

Table 1
The optimal sizing obtained by PSO algorithm [11].

Proposed hybrid system NPV NW NFC Nelec Ntank NInv PV cost ($) Wind cost ($) FC cost ($) Elec cost ($) Tank cost ($) Inv. cost ($) Total cost ($)

PV/FC 510 e 530 660 670 99 313140.0 e 5404086.98 4186122.07 996245.213 79,200 10978794.27
Wind/FC e 150 199 229 300 39 e 3050199.86 2029081.71 1452457.50 1452457.50 31,200 7015841.13
PV/Wind/FC 137 123 150 184 202 37 83941.28 2492781.43 1527266.62 1164102.43 300132.12 29943.36 5598167.26

Table 2
The optimal sizing obtained by CS algorithm [16].

Proposed hybrid system NPV NW NFC Nelec Ntank NInv PV cost ($) Wind cost ($) FC cost ($) Elec cost ($) Tank cost ($) Inv. cost ($) Total cost ($)

PV/FC 506 e 522 591 650 193 310,684 e 5322515.86 3748482.03 966506.550 154,400 10502588.448
Wind/FC e 142 184 214 291 38 e 2887522.54 1876135.85 1357318.36 432697.548 30,400 6584074.3158
PV/Wind/FC 133 119 82 165 143 34 81,662 2419825.22 836104.024 1046530.51 212631.441 27,200 4623953.2103

Table 3
The optimal sizing obtained by ABC algorithm [14].

Proposed hybrid system NPV NW NFC Nelec Ntank NInv PV cost ($) Wind cost ($) FC cost ($) Elec cost ($) Tank cost ($) Inv. cost ($) Total cost ($)

PV/FC 337 e 521 537 571 189 206,918 e 5312319.47 3405981.13 849038.831 151,200 9925457.4418
Wind/FC e 132 181 232 290 38 e 2684175.88 1845546.68 1471485.33 431210.614 30,400 6462818.5189
PV/Wind/FC 135 123 100 140 160 38 82,890 2501163.89 1019639.05 887965.287 237909.304 30,400 4759967.5368

Table 4
The optimal sizing obtained by the proposed MBA algorithm.

Proposed hybrid system NPV NW NFC Nelec Ntank NInv PV cost ($) Wind cost ($) FC cost ($) Elec cost ($) Tank cost ($) Inv. cost ($) Total cost ($)

PV/FC 329 e 513 533 565 194 202000.5 e 5229816.98 3379472.87 839464.216 155465.7 9806220.3607
Wind/FC e 128 172 202 287 36 e 2602837.21 1753779.17 1281207.05 426749.815 28,800 6093373.2655
PV/Wind/FC 130 108 90 125 133 32 79,820 2196143.90 917675.148 792826.149 197762.109 25,600 4209827.3116

causalities (objective function). The initial point of MBA algorithm pieces, Ns, which act as individuals inside a population. The location
is called shot point which is represented by xf0 where f is the of exploding mine by the shrapnel pieces is calculated as follows:
number of shot points; this first shot produces number of shrapnel
 A. Fathy / Renewable Energy 95 (2016) 367e380 375

7
x 10
1.6

1.5 PSO solution

Cuckoo search
ABC solution
1.4 MBA solution
Function Values

1.3

1.2

1.1

0.9
0 10 20 30 40 50 60 70 80 90 100
Number of Iterations

Fig. 14. A comparison between algorithms responses for PV/FC sizing.

x 10
11

10.5
PSO solution
Cuckoo serach
10
ABC solution
MBA solution
9.5
Fitness function ($)

8.5

7.5

6.5

6
0 10 20 30 40 50 60 70 80 90 100
Iteration number

Fig. 15. A comparison between algorithms responses for WT/FC sizing.

x 10
13
PSO solution
Cuckoo search
12
ABC solution
MBA solution
11

10
Fitness function ($)

4
0 10 20 30 40 50 60 70 80 90 100
Iteration number

Fig. 16. A comparison between algorithms responses for PV/WT/FC sizing.

376 A. Fathy / Renewable Energy 95 (2016) 367e380

Table 5
The statistical parameters of the used optimization algorithms.

Proposed hybrid system Statistical parameters ($) PSO [11] CS [16] ABC [14] The proposed MBA

PV/FC Mean 1.198eþ07 1.05eþ07 9.935eþ06 9.891eþ06

Std. dev. 7.852eþ05 6.769eþ05 4.745eþ05 4.073eþ05
Best 1.17eþ07 1.05eþ07 9.93eþ06 9.806eþ06
Worst 1.57eþ07 1.16eþ07 1.03eþ07 1.281eþ07
Wind/FC Mean 7.392eþ06 6.601eþ06 6.466eþ06 6.308eþ06
Std. dev. 8.425eþ05 8.283eþ05 8.062eþ05 7.548eþ05
Best 7.01eþ06 6.58eþ06 6.46eþ06 6.093eþ06
Worst 1eþ07 1.07eþ07 6.78eþ06 9.796eþ06
PV/Wind/FC Mean 6.206eþ06 4.63eþ06 4.782eþ06 4.542eþ06
Std. dev. 1.607eþ06 1.575eþ06 1.111eþ05 1.05eþ05
Best 5.6eþ06 4.62eþ06 4.76eþ06 4.21eþ06
Worst 1.26eþ07 9.45eþ06 5.59eþ06 1.052eþ07

1000 800 500

800 400
600
Pg Conv (Kw)

Pg Elec (kW)
Pg (kW)

600 300
400
400 200
200
200 100

0 0 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Time (hours) Time (hours) Time (hours)

1000 500 2000

800 400
PElec tank (kW)

1500
PFC Conv (kW)

PTank FC (kW)
600 300
1000
400 200

500
200 100

0 0 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Time (hours) Time (hours) Time (hours)

Fig. 17. The optimal PV/WT/FC system power flow performance with time.

700
PV/WT generated power
Load power
600 Fuel cell generated power

500

400
Power (kW)

300

200

100

0
0 20 40 60 80 100 120 140 160 180 200
Time (Hours)

Fig. 18. The generated power from PV/WT/FC system and the load power for the first 200 h.
 A. Fathy / Renewable Energy 95 (2016) 367e380 377

Table 6
Comparison between the meta-heuristic algorithms in terms of optimal solution and cost saving after 50 runs.

Proposed system Algorithm Total number of iterations Total cost ($) Convergence time (s) Cost saving ($) % Cost saving

PV/FC PSO 100 11700896.99 21.66 NA NA

CS 100 10502588.448 20.52 1,198,309 10.24117
ABC 100 9925457.4418 13.68 1,775,440 15.17353
MBA 100 9806220.3607 17.1 1,894,677 16.19258
Wind/FC PSO 100 7015841.13 23.94 NA NA
CS 100 6584074.3158 21.66 431766.8 6.15417
ABC 100 6462818.5189 49.02 553022.6 7.882485
MBA 100 6093373.2655 18.24 922467.9 13.14836
PV/Wind/FC PSO 100 5598167.26 20.52 NA NA
CS 100 4623953.2103 51.3 974,214 17.40237
ABC 100 4759967.5368 28.5 838199.7 14.97275
MBA 100 4209827.3116 15.96 1,388,340 24.7999

6 6
x 10 (a) x 10 (b)
14 10
After one run After one run
After fifty run 9 After fifty run
12

8
Total cost ($)

Total cost ($)

10
7
8
6

6
5

4 4
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Iteration No. Iteration No.

6 6
x 10 (c) x 10 (d)
5.8 11
After one run After one run
5.6 10
After fifty run After fifty run
9
5.4
Total cost ($)

Total cost ($)

8
5.2
7
5
6
4.8 5

4.6 4
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Iteration No. Iteration No.

Fig. 19. The obtained optimal solution for PV/Wind/FC system after one run and fifty runs (a) PSO (b) CS (c) ABC (d) MBA.

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
0 exploration and exploitation processes. The exploration process is
u f
um conducted when the exploration factor (m), defined by the user, is
f f B u ðnþ1Þ C f
xðnþ1Þ ¼ xeðnþ1Þ þ exp@  t f Axn n ¼ 0; 1; 2; 3; …: greater than the iteration number index (k). In this process the
dðnþ1Þ value of xfe(nþ1) is calculated by the following equation:

(21)
xfeðnþ1Þ ¼ dfnþ1 cosðqÞ; n ¼ 0; 1; 2; …: (23)
where xfe(nþ1)
is the location of exploding mine bomb, f
m(nþ1) , f
d(nþ1)
are the direction and the distance of the generated thrown shrapnel where
pieces in each iteration respectively and xfn is the previous solution.
The value of xfe(nþ1) can be calculated as follows:
dfnþ1 ¼ dfn ðjrandnjÞ2 (24)
xfeðnþ1Þ ¼ dfn randcosðqÞ (22) The exploitation process is taken place if m < k; in this process
the algorithm is encouraged to converge the optimal solution by
where rand is a random number in range [0, 1], q is the angle of gradually reducing the initial distance of shrapnel pieces through a
shrapnel pieces which is constant value and equal to 360/Ns. In MBA reduction constant defined by the user which is called a. The
algorithm there are two processes for searching the solution; the reduction in the initial distance is calculated as follows:
378 A. Fathy / Renewable Energy 95 (2016) 367e380

Table 7
Comparison of the optimal solution obtained via the methods under study after several runs for sizing PV PV/Wind/FC system.

No. of runs Total cost based on PSO ($) Total cost based on CS ($) Total cost based on ABC ($) Total cost based on MBA ($)

100 5600587.3932 4623953.0064 4759967.5368 4209826.7180

200 5597990.1878 4623952.9044 4759353.5368 4209826.6773
300 5593975.4975 4623943.0139 4759764.1902 4209785.4143
400 5595877.1934 4623922.6211 4759743.8555 4209730.9008
500 5596376.6023 4623922.5192 4759723.5208 4209725.3477
600 5595255.7478 4623912.4248 4759703.1862 4209715.1513
700 5594632.4476 4623912.3228 4759682.8515 4209704.9550
800 5592741.9629 4623912.2208 4759701.1527 4209704.1317
900 5592512.9062 4623912.1189 4759682.8515 4209703.8855
1000 5592418.6530 4623912.0169 4759662.5168 4208976.9524

MBA are shown in Fig. 7(a) while the proposed steps incorporated
f
f dn1 MBA algorithm is shown in Fig. 7(b).
dn ¼   (25)
exp ak
6. Simulation results
In this process the location of exploded mine bomb is calculated
The analysis of the optimal sizing of hybrid system is applied to
using Eq. (25) while the distance and shrapnel pieces are calculates
cover electric load located near to the national research institute of
by the following equation:
astronomy and geophysics (NRIAG) located at Helwan city, south
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi of Cairo, Egypt. This area of is located at latitude 29.87
N and
 2  2
dfnþ1 ¼ xfðnþ1Þ  xfðnÞ þ F fðnþ1Þ  F fðnÞ ; n ¼ 0; 1; 2; …:: longitude 31.30
E. The average hourly load power is given in Fig. 8.
These data are measured from 1 January 2014 to 31 December
(26) 2014.
The solar radiation, ambient temperature and wind speed of the
f f location are measured by the solar radiation and meteorological
Fðnþ1Þ  FðnÞ
mfðnþ1Þ ¼ f f
(27) station located in the NRIAG. The data are recorded each ten minute
xðnþ1Þ  xðnÞ per day. The average hourly data for solar radiation, ambient tem-
perature and wind speed are calculated and illustrated in
where F denotes the value of fitness function at location x. Fig. 6 Figs. 9e11.
shows the two processes of searching the global point in MBA al- The parameters of the proposed system components are tabu-
gorithm, exploration and exploitation processes. The main steps of lated in Table A in appendix; the measured data are fed to PV/WT

6 PSO responses 6 CS responses

x 10 x 10
14 12
After 100 run After 100 run
12 After 200 run After 200 run
10
After 300 run
Total cost ($)

Total cost ($)

After 300 run

10 After 400 run After 400 run
After 500 run 8 After 500 run
8

6
6

4 4
0 20 40 60 80 100 0 20 40 60 80 100
Iteration number Iteration number
6 ABC responses 6 MBA responses
x 10 x 10
5.8 12
After 100 run After 100 run
5.6 After 200 run After 200 run
10
After 300 run After 300 run
Total cost ($)

Total cost ($)

5.4
After 400 run After 400 run
5.2 After 500 run 8 After 500 run

5
6
4.8

4.6 4
0 20 40 60 80 100 0 20 40 60 80 100
Iteration number Iteration number

Fig. 20. The responses of the methods under study after several runs.
 A. Fathy / Renewable Energy 95 (2016) 367e380 379

system, the output power from one PV module and one wind tur- Figs. 14e16). It is shown that the MBA has the lowest mean,
bine are given in Fig. 12a and b respectively. A complete model of standard division and minimum values for all combinations
the proposed hybrid PV/WT/FC system is built in Matlab/Simulink compared to the others. To make sure of the validity of the ob-
library and given in Fig. 13. In order to ensure the validity and tained results by MBA; the power flow between each component
reliability of the proposed MBA in evaluating the best size of the of the optimal sizing of PV/WT/FC system described in Section 3 is
proposed hybrid system; three other meta-heuristic algorithms are given in Fig. 17.
programmed and applied, particle swarm optimization (PSO) [11], To ensure the validity of the proposed algorithm of MBA in
artificial bee colony (ABC) [14] and cuckoo search (CS) [16] to obtain finding the true optimal sizing of the hybrid PV/WT/FC system; the
the optimal size of the proposed system. The obtained results from power generated from the hybrid system sized by MBA and the
these algorithms are compared to those obtained by the proposed consumed power by the load for the first 200 h in the year are given
MBA. The analysis is performed for three suggested combinations; in Fig. 18.
the first one comprises PV with FC, the second one is WT with FC The first plot shown in Fig. 18 represents the generated power
and the last one is PV/WT/FC system. The combination which from the PV/WT; the second one represents the load power while
achieves the minimum annual cost with covering the load demand the third one is the FC generated power. It is clear that for periods at
is considered as the optimal solution. The controlling parameters of which the PV/WT system failed in supplying the load; the FC power
the used meta-heuristic algorithms are given in Table B in appen- is covering it such that the load is covered all time. From this figure
dix, as the maximum number of iterations is assumed to be 100 for one can get that the obtained optimal sizing based on the candidate
all algorithms. MBA technique and the suggested power flow management
In order to confident the validity of the proposed solution scheme are efficient and reliable in achieving the minimum annual
methodology compared with the others; each algorithm has been cost with load coverage. In order to clear the concept of reliability of
run for 50 times; each run is executed for 100 iterations. Table 1 the proposed solution algorithm; the studied algorithms are per-
shows the optimal sizing of the tree combinations, PV/FC, WT/FC formed one time and for fifty times, the results obtained from each
and PV/WT/FC, obtained by PSO algorithm. From this table one can one are compared. Table 6 shows such comparison in terms of cost
get that the minimum annual cost is obtained when PV/WT/FC saving and the convergence time of each algorithm to obtain the
combination is used, the cost of this combination is 5598167.26 optimal solution, each algorithm has been run for 50 times; each
$/year; this cost is obtained by using 137 PV module, 123 wind run is simulated for 100 iteration. The responses of the studied
turbine, 150 fuel cell, 184 electrolyzer, 202 tank and 37 inverter. algorithms after one run and after fifty runs are shown in Fig. 19.
Referring to the obtained results by PSO [11]; it is found that a very Referring to Table 6; the algorithm gives the largest total cost, PSO,
large number of required wind turbine which is not realistic is selected as a base case and the saving cost obtained via the others
sizing; therefore one can say that the PSO algorithm described in are compared to it. For example; in optimal sizing of PV/Wind/FC
Ref. [11] is failed in obtaining the correct optimal sizing of the system it is clear that the minimum cost is obtained via the pro-
hybrid system. posed algorithm after 15.96 s; and also in all cases, the MBA gives
Table 2 shows the optimal sizing of the three previous stated the minimum total cost after less convergence time; therefore one
combinations using CS algorithm; regarding to the obtained results can say that the proposed solution algorithm is reliable in obtaining
given in Table 4; one can derive that the optimal annual cost is the minimum cost with achieving proposed constrains with less
4623953.2103 $/year obtained from the third combination, this cost convergence time compared with others.
is achieved by using 133 PV module, 119 wind turbine, 82 fuel cell, In order to ensure the reliability of the proposed method based
165 electrolyzer, 143 tank and 34 inverter. on MBA algorithm; the performance of each method is compared
The optimal sizing of the proposed hybrid system based on ABC after several separate runs on the same optimization problem. This
algorithm is tabulated in Table 3. The optimal obtained annual cost comparison is given in Table 7. The results ensure that the proposed
is 4759967.5368 $/year obtained from also the third combination methodology is superior to the others under comparison. The re-
by using 135 PV module, 123 wind turbine, 100 fuel cell, 140 elec- sponses of each algorithm after performing 100 to 500 runs are
trolyzer, 160 tank and 38 inverter. shown in Fig. 20.
The MBA optimization algorithm is executed and the system
sizing based on three proposed combinations is tabulated in 7. Conclusion
Table 4. The final selected optimal annual cost is 4209827.3116 $
obtained from using 130 PV module, 108 wind turbine, 90 fuel cell, In this paper an optimal sizing of a hybrid system consisting of
125 electrolyzer, 133 tank and 32 inverter. photovoltaic module, wind turbine, fuel cell, electrolyzer, hydrogen
From the obtained results; one can derive that the MBA tech- tank and inverter is performed based on recent met-heuristic
nique is superior compared to PSO, CS and ABC algorithms as it is technique which is main blast algorithm (MBA) to meet a certain
saving in the cost about 24.8% compared to PSO algorithm, 8.956% load in remote area in Egypt. Measured data of solar radiation,
compared to CS algorithm and 11.5576% compared to ABC algo- ambient temperature and wind velocity are recorded by the solar
rithm. Therefore the MBA algorithm is valid, reliable and efficient radiation and meteorological station of national research institute
for solving the proposed optimization formula represented the of astronomy and geophysics located at Helwan city, Egypt. This
optimal sizing the hybrid PV/WT/FC system. A comparison be- station is above the sea level about 114 m. Besides MBA; three
tween the algorithms' responses for three combinations are given optimization techniques PSO, CS and ABC are performed and the
in Figs. 14e16. The obtained responses of the four meta-heuristic optimal results are compared to MBA. A power management
algorithms clarify an important derivation which is the proposed strategy that manages the power flow between each system
MBA technique not only successes in finding the optimal annual component is presented. The optimal sizing of the proposed hybrid
cost of the proposed hybrid system but also it converges to the system obtained by the proposed MBA technique saves about 24.8%
optimal solution faster than the other algorithms. The statistical in cost compared to PSO algorithm, 8.956% compared to CS algo-
parameters, mean, standard division, minimum, maximum, of rithm and 11.5576% compared to ABC algorithm. Additionally the
each algorithm in each proposed combination are given in Table 5. proposed MBA reaches to the optimal solution in time less than the
These statistical parameters are calculated from the optimal others. The obtained results encourage the usage of MBA in the field
annual cost obtained in each iteration for each algorithm (from of sizing the hybrid PV/WT/FC system than the other algorithm.
380 A. Fathy / Renewable Energy 95 (2016) 367e380

Appendix [2] N. Ahmed, M. Miyatake, A. Al-Othman, Power fluctuations suppression of

stand-alone hybrid generation combining solar photovoltaic/wind turbine
and fuel cell systems, Energy Convers. Manag. 49 (10) (2008) 2711e2719.
[3] M. Castan ~ eda, A. Cano, F. Jurado, H. S ndez, Sizing optimization,
anchez, L. Ferna
Table A dynamic modeling and energy management strategies of a stand-alone PV/
The system component parameters. hydrogen/battery-based hybrid system, Int. J. Hydrogen Energy 38 (10) (2013)
3830e3845.
Parameter Value Parameter Value
[4] C. Wang, M. Nehrir, Power management of a stand-alone wind/photovoltaic/
I 0.05 CRFC ($) 2500 fuel cell energy system, IEEE Trans. Energy Convers. 23 (3) (2008) 957e967.
N 20 CMFC ($/year) 0.02 [5] M. Nehrir, C. Wang, Fuel Cell Modeling and Control: Distributed Generation
Vcut_in (m/s) 3.1 nFC (year) 5 Applications, IEEE Press and John Wiley & Sons, New York, 2009.
Vcut_out (m/s) 15.6 xFC 0.5 [6] R. Dufo-Lopez, J.L. Bernal-Agustín, Design and control strategies of PVediesel
systems using genetic algorithms, Sol. Energy 79 (1) (2005) 33e46.
Vr (m/s) 13.8 Prated
FC (kW) 3
[7] J. Agustín, R. Lo  pez, Simulation and optimization of stand-alone hybrid
PWT_max (kW) 7.5 CCElec ($) 2000
renewable energy systems, Renew. Sustain. Energy Rev. 13 (8) (2009)
PWTo (kW) 4.39 CRElec ($) 1500
2111e2118.
nW (year) 20 CMElec ($/year) 2 [8] G. Dalton, D. Lockington, T. Baldock, Feasibility analysis of stand-alone
CCW ($) 19,400 nElec (year) 5 renewable energy supply options for a large hotel, Renew. Energy 33 (7)
CRW ($) 0 xElec 0.9 (2008) 1475e1490.
CMW ($/year) 75 Prated
Elec (kW) 3 [9] Kaur D.T. Pooja, Optimal sizing of solar photovoltaic -wind hybrid system, Int.
Pmp (W) 120 CCTank ($) 1300 J. Innov. Res. Electr. Electron. Instrum. Control Eng. 3 (1) (2015) 99e103.
Vmp (V) 18.1 CRTank ($) 1200 [10] J.L. Bernal Agustín, R. Dufo Lo pez, Simulation and optimization of stand-alone
Imp (A) 6.6 CMTank ($/year) 15 hybrid renewable energy systems, Renew. Sustain. Energy Rev. 13 (8) (2009)
Voc (V) 22.2 nTank (year) 20 2111e2118.
Isc (A) 6.8 Erated [11] S.M. Hakimi, S.M. Moghaddas-Tafreshi, Optimal sizing of a stand-alone hybrid
Tank (kWh) 0.3
xPV 0.12 CCInv ($) 3000 power system via particle swarm optimization for Kahnouj area in south-east
of Iran, Renew. Energy 34 (2009) 1855e1862.
CCPV ($) 614 CRPV ($) 750
[12] A. Maleki, A. Askarzadeh, Comparative study of artificial intelligence tech-
CRPV ($) 0 nInv (year) 15
niques for sizing of a hydrogen-based stand-alone photovoltaic/wind hybrid
CMPV ($/year) 0 xInv 0.9
system, Int. J. Hydrogen Energy 39 (2014) 9973e9984.
nPV (year) 20 Prated
Inv (kW) 3 [13] A. Maleki, F. Pourfayaz, Optimal sizing of autonomous hybrid photovoltaic/
CCFC ($) 3000 wind/battery power system with LPSP technology by using evolutionary al-
gorithms, Sol. Energy 115 (2015) 471e483.
[14] A. Maleki, A. Askarzadeh, Artificial bee swarm optimization for optimum
sizing of a stand-alone PV/WT/FC hybrid system considering LPSP concept,
Sol. Energy 107 (2014) 227e235.
Table B [15] H. Nasiraghdam, S. Jadid, Optimal hybrid PV/WT/FC sizing and distribution
system reconfiguration using multi-objective artificial bee colony (MOABC)
The controlling parameters of PSO [11], ABC [14], CS [16] and the proposed MBA
algorithm, Sol. Energy 86 (10) (2012) 3057e3071.
algorithms.
[16] M. Rezaei, M. Ghanbari, Optimization of hybrid PV/Wind/FC system consid-
PSO controlling parameters ering reliability indices using cuckoo search algorithm, Indian J. Fundam. Appl.
No. of iterations Swarm size C1 C2 (Inertia weight) u Life Sci. 5 (S1) (2015) 3304e3320.
[17] R. Gao, Z. Gao, Pitch control for wind turbine systems using optimization,
100 60 2 2 0.7
estimation and compensation, Renew. Energy 91 (2016) 501e515.
CS controlling parameters [18] I.J. Esfahani, C. Yoo, An optimization algorithm-based pinch analysis and GA
No. of Number of Discovery rate of Beta vy multiplying
Le for an off-grid batteryless photovoltaic-powered reverse osmosis desalination
iterations nests alien eggs (b) coefficient system, Renew. Energy 91 (2016) 233e248.
100 7 0.25 1.5 0.1 [19] A. Fathy, Reliable and efficient approach for mitigating the shading effect on
photovoltaic module based on modified artificial bee colony algorithm,
ABC controlling parameters Renew. Energy 81 (2015) 78e88.
No. of iterations Colony size Food number Food source limit [20] M.J. Khan, M.T. Iqbal, Pre-feasibility study of stand-alone hybrid energy sys-
100 30 10 100 tems for applications in New found land, Renew. Energy 30 (6) (2005)
835e854.
The proposed MBA controlling parameters [21] J.C. Amphlett, R.M. Baumert, R.F. Mann, B.A. Peppley, P.R. Roberge, T.J. Harries,
No. of Number of shrapnel Reduction factor Number of function Performance modeling of the Ballard mark IV solid polymer electrolyte fuel
iterations pieces ( a) evaluations cell, J. Electrochem. Soc. 142 (1) (1995) 9e15.
100 50 1686.86037830281 10,000 [22] G. Saur, Wind-to-hydrogen Project: Electrolyzer Capital Cost Study, National
Renewable Energy Laboratory, 2008.
[23] P. Venkataraman, Applied Optimization With Matlab® Programming, second
References ed., John Wiley & Sons, Inc., Hoboken, New Jersey, 2009.
[24] A. Sadollah, A. Bahreininejada, H. Eskandar, M. Hamdi, Mine blast algorithm: a
new population based algorithm for solving constrained engineering opti-
[1] T. Ma, H. Yang, L. Lu, J. Peng, Technical feasibility study on a standalone hybrid mization problems, Appl. Soft Comput. 13 (5) (2013) 2592e2612.
solarewind system with pumped hydro storage for a remote island in Hong
Kong, Renew. Energy 69 (2014) 7e15.