Energy 157 (2018) 1063e1078

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

An enhanced moth-flame optimizer for solving non-smooth economic

dispatch problems with emissions
Asmaa A. Elsakaan a, *, Ragab A. El-Sehiemy a, **, Sahar S. Kaddah b, Mohammed I. Elsaid b
a
Electrical Engineering Department, Faculty of Engineering, Kafrelsheikh University, Kafrelsheikh, Egypt
b
Electrical Engineering Department, Faculty of Engineering, Mansoura University, Mansoura, Egypt

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

Article history: This paper proposes an Enhanced Moth-Flame Optimization (EMFO) algorithm for solving the non-
Received 16 August 2017 convex economic dispatch (ED) problem with valve point effects and emissions. It determines the
Received in revised form optimal generation schedule of generating units by minimizing both fuel cost and emission simulta-
19 May 2018
neously while the system constraints are achieved. The moth-flame optimization (MFO) is a recent
Accepted 14 June 2018
Available online 18 June 2018
nature-inspired method, which is based on the navigation mechanism called transverse orientation of
Moths in space. The EMFO combines the merits of the traditional MFO and levy flight by concentration
the search space. The usage of Le
vy-flight has the prominent properties to increase the diversity of
Index terms:
Moth-flame optimization
population. The effectiveness of the proposed EMFO method is proven on using 10 benchmark functions
Multi-objective optimization and 3 standard test systems consisting of 6, 40 and a large scale 80 generating units with non-convex fuel
Economic dispatch cost functions. The capability of the proposed algorithm is verified also for single and multi-objective
Valve-point effects studied cases and its results are compared with several well-known previous techniques. The results
Emissions confirm the high performance of the proposed EMFO method for finding the optimal economic gener-
Power loss ation scheduling at acceptable low emission levels.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction traditional quadratic power-cost equation [3].

In real power systems, practical features such as valve point
1.1. Motivation effects, ramp rate limits, prohibited operating zones, and multiple
fuel options are usually encountered. Neglecting these features may
Economic dispatch (ED) is a major operational issue in modern lead to inaccurate solutions of the ED problem [4,5]. On the other
power systems. It aims to determine the best schedule of gener- hand, considering these practical features converts the ED problem
ating units output to reduce the total production costs while to a complex optimization problem in which the cost function is
satisfying constraints of total load demand as well as respecting non-convex and non-smooth. In this subject, the need to develop
individual resource capacity limits [1]. Theoretically, to make the efficient solving methods to such problem becomes the concern of
cost function of a generator relatively easy to solve, it can be rep- power system operators and planners. Therefore, the field of the
resented by a quadratic convex/linear function [2]. In the conven- developed optimization is one of the major ways to solve this
tional formulation of ED problem, it was assumed that the cost problem.
function is a quadratic polynomial and convex. In reality, when a
steam valve starts to open, there is a rippling effect on a unit's
power-cost curve. To model the unit's power-cost curve with the 1.2. Literature review on solution methods
valve-point effects, a sinusoidal function has been added to the
Classical optimization methods are developed to solve convex
ED problems such as lambda iteration method [6], dynamic pro-
gramming [7], gradient method [8], base-point participation factor
* Corresponding author.
method [9] etc. To avoid the problems associated with the usage of
** Corresponding author. Intelligent Systems Research Group (ISRG), Kafrelsheikh
University, Kafrelsheikh, Egypt
classical optimization methods, various heuristic and meta-
E-mail addresses: [email protected], eng.asmaa_ehab@yahoo. heuristic techniques are used to solve these problems. Simulated
com (A.A. Elsakaan), [email protected] (R.A. El-Sehiemy). annealing [10] is used to find a global optimal solution for solving

https://doi.org/10.1016/j.energy.2018.06.088
0360-5442/© 2018 Elsevier Ltd. All rights reserved.
1064 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

Nomenclature Bij ; B0i and B00 Transmission loss coefficients

r1 ; r2 ; r3 and r4 Random parameter which conforms to a
Pit Real power generation of the generation unit i uniform distribution
fi ðPit Þ Cost function of the generation unit i Mi ith moth
NG Number of generation buses Fj jth flame
ai bi and ,ci , Smooth cost coefficients of ith
generator S Spiral function
di and ei non-smooth cost coefficients of the valve- Di Distance from ith moth to jth flame
point impact of ith generator n Number of intervals in the time horizon
Nf Maximum number of flames m The shape of the logarithmic spiral constant
Pimin Minimum limit of power generation l Current number of iteration
Pimax Maximum limit of power generation Xitþ1 The ith moth or solution vector Xit at iteration t
ai bi and gi Smooth coefficients of the ith unit emission 4 The dot product
function G Standard Gamma function
Titer Maximum number of iterations
; hi and di non-smooth emission coefficients of ith
generator Ei ðPit Þ Amount of emission from unit i from producing
PL Total active power losses power Pit
Pd Load demand active power sign[rand 0.5] Takes only three values 1, 0, and 1

dynamic economic dispatch problem. It uses search technique with for solving different constrained optimization problems. It is inte-
probabilistic approach to deal with individual solution to get local grated with levy flights to reduce the computational times for
optima solutions. The major drawbacks of this algorithm are highly complex problems [28]. In Ref. [29], the MFO algorithm was
choosing an appropriate control parameter and speed of operation. proposed to forecast the annual power load essential for an electric
Tabu search [11] has a disadvantage of short term memory struc- power system with the least squares support vector machine
ture. GA [12] is used to solve ED problems which it is encoded of the (LSSVM). MFO was used by Ref. [30] to present better results for
actual parameters then various genetic operators (reproduction, simultaneous optimization of a two-area hydro-thermal system
crossover and mutation) are applied until finding near global under a deregulated scenario. MFO was proposed and applied to
optimal solution. machine learning to find the optimal feature combination by using
The traditional ED model reduces the total cost from units the wrapper-based feature selection mode as a searching method
neglecting the impact of emission even though the minimum cost [31]. Ref. [32] proposed MFO to solve the optimal power flow (OPF)
causes lower emissions. But, the unit emission characteristic is not problem, reduce fuel cost, and active and reactive power loss
as same as the fuel combustion characteristic [13]. minimization. This technique was a better performance comparing
Economic emission dispatch was developed in Ref. [14] for the to flower pollination algorithm (FPA) and particle swarm optimi-
effective handling of practical constraints. In Ref. [15], GA, micro GA zation (PSO). In Ref. [33], MFO was applied to solve the OPF problem
(MGA), and evolutionary programming (EP) are compared to solve in the interconnected power system with many objective functions
the ED problem. Then, EP is used to better solve the economic and the results were better than those obtained by artificial bee
emission dispatch problem. Reference [16] considers the unit fuel colony (ABC) and other meta-heuristics. In Ref. [34], MFO was used
cost as one objective and unit emission as another objective. It uses to various non-convex, nonlinear optimum power flow problems
non-dominated sorting GA to solve the multiple objective optimi- with five single objective functions. it was effective with optimum
zation problem and to get optimal solutions. Reference [17] pro- value with rapid and smooth convergence. In Ref. [35], MFO was
poses a mathematical model to deal with wind power generation in used to solve the optimal power flow (OPF) problems. The objective
the ED problem. Reference [18] discusses the impact on total cost functions are minimizing the fuel cost, improving the voltage
saving for power generation after adding wind power generation profile, enhancing the voltage stability and minimizing active and
by solving the ED problem with wind power generation using a reactive power losses. The results were better compared with
quantum genetic algorithm. Flower Pollination Algorithm (FPA) and particle swarm optimizer
Several methods have been proposed to solve the ED with valve (PSO). In Ref. [36], MFO was used to deal with optimal reactive
point and emission such as a novel parallel hurricane optimization power dispatch (ORPD) problem to minimize total power loss and
algorithm [19], Chaotic Krill Herd algorithm [20], improved genetic voltage deviation. MFO was used to solve economic load dispatch
algorithm [21], Taguchi method [22], particle swarm optimization (ELD) problem with highly complex constraint and applied to three
[23,24], Group Search Optimizer [25] and artificial bee colony al- different standard case study without valve point effect including
gorithm [26]. The comparisons among the published results for ED ramp rate limits (RRL) and prohibited operating zones (POZ) and
with valve point effect are provided in Ref. [26]. Although these with valve point effect loading. This technique provides optimal
methods have given fair solutions to ED with valve point effect, the solution and faster convergence for optimization algorithms [28].
search efficiency isn't high and a huge amount of computation is
vy-flights, a random technique, have relevant role in meta-
Le
consumed. heuristic algorithms in exploration and exploitation. A diverse
Recently, Seyedali Mirjalili [27] has proposed a method known range of both natural and artificial phenomena are now being
as the Moth-Flame Optimization MFO algorithm in 2015. It is a described in terms of Le
vy statistics [37]. Many studies have shown
meta-heuristic optimization algorithm which is a novel nature- that flight behavior of many animals and insects has revealed the
einspired. It has an ability to avoid local optima and get global typical characteristics of Le
vy-flight [38e40]. Le
vy-flights maxi-
optimal solution that makes it appropriate for practical applications mize the efficiency of resource searches and converge faster in
 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078 1065

order to obtain optimum solution in uncertain environments. Levy of both SO2 and CO2 can be modeled as quadratic polynomial
flights mechanism is helpful to escape and avoid local optima. The functions. NOx emissions are more difficult to model since they
application includes the finding of the optimal values of control come from different sources and their production is associated with
variables to minimize the objective function. Because of the merits several factors, such as boiler temperature and air content. One

vy-flight, a hybrid moth-flame optimization algorithm with
of Le approach to represent NOx emissions is using a combination of

vy-flight, to combine their effective merits, is introduced in this
Le polynomial and exponential terms [42]. The total pollutants emis-
work. sion over the dispatch period can be expressed as:

1.3. Paper contributions and salient EMFO features X

n X
NG  
minF2 ¼ Ei Pit
t¼1 i¼1
 This paper is concerned with developing an enhanced version of

vy-flight technique.
X
n X
NG  2  
MFO that is dependent on Le ¼ ai þ bi Pit þ gi Pit þ hi exp di Pit ton=h (3)
 The exploration and exploitation processes is improved by using t¼1 i¼1
the proposed enhanced MFO.
 In the economical regard, the EMFO algorithm is applied to non- The multi-objective optimization can be transformed into a
smooth cost function ED framework. single objective function using weighting factors. The aggregated
 The proposed EMFO enhances the solution quality by enhancing objective function is expressed as follow:
exploration and exploitation processes.
X
n X
NG   X
n X
NG  
 Developing single and multi-objective frameworks of moth- minF3 ¼ w1 fi Pit þ w2 Ei Pit (4)
flame optimizer for solving of ED problem with considering t¼1 i¼1 t¼1 i¼1
valve-point concerns with equality and inequality operational
constraints. w1 and w2 are the weight factors that are reflects the priority
 Prove the efficiency of the proposed an enhanced MFO with factors. The factors w1 and w2 equal 1 and 0, respectively for

vy-flight compared with several well-known techniques in
Le minimizing fuel costs only. The factors w1 and w2 equal 0 and 1,
benchmark functions, small and large-scale practical power respectively for minimizing ecological emissions only. When both
systems. objectives have the same priority or simultaneously optimized, the
weighting factors equal 1 to represent the multi-objective model.

1.4. Paper organization 2.2. Constraints

The organization of the current paper is ordered as follow: 1) Equality Constraints

Section 2 presents the problem formulation; Section 3 presents the
proposed Moth-flame algorithm, and section 4 presents application The power balance requirement is expressed as equality
of MFO for solving economic dispatch problem. section 5 presents
the main findings of this research work.

2. Problem formulation

The proposed method aims to minimize multi-objective func-

tion such as, the total cost and emission in a given power network.

2.1. Economic dispatch problem objectives

In conventional ED, it was assumed quadratic cost function for

generating unit. Traditional ED problem is formulated as a con-
strained optimization framework as:
Fig. 1. Transverse orientation.
X
n X
NG   n X
X NG  2
minF1 ¼ fi Pit ¼ ai þ bi Pit þ ci Pit (1)
t¼1 i¼1 t¼1 i¼1

In practical problems, the valve-point loading effects and mul-

tiple fuel options are considered to model the accurate cost func-
tions. The fuel cost function considering the impact of valve-point
loadings of the generator is expressed as:

X
n X
NG  2    

minF1 ¼ ai þ bi Pit þ ci Pit þ di  sin ei Pimin  Pit 
t¼1 i¼1
(2)
The ecological emissions are considered very important in the
power generation industry, because of their effects on the envi-
ronment. The most important emissions are SO2 , CO2 and NOx .
These emissions can be modeled through functions that associate
emissions with power production for each unit [41]. The emissions Fig. 2. Logarithmic spiral, space around a flame, and the position with respect to t.
1066 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

Fig. 3. Flowchart of the EMFO for solving the ED problem.

constraint of economic subject as in Eq (5). In Eq (5), the total

generated power equals to the total power demand plus trans- X
N X
N X
N
mission power losses. PL ¼ Pi Bij Pj þ B0i Pi þ B00 (6)
i¼1 j¼1 i¼1

X
NG
Pit  PL ¼ Pd (5) 2) Inequality Constraints
i¼1

The power loss is formulated using Eq. (6) as [9]: Each generating unit is preserved with upper and lower
 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078 1067

Table 1
Results of benchmark functions.

Methods Fun. F1 F2 F3 F4 F5 F6 F7 F8 F9 F 10

EMFO Ave. 0.0000 0.2047 0.0000 0.0000 0.0000 0.0000 0.002 0.8999 0.4101 0.0499
Std. 0.0000 0.9124 0.0000 0.6389 0.0000 0.0000 0.0051 0.0061 0.9500 0.5012
MFO Ave. 0.0000 0.2187 0.0000 0.0000 0.0000 0.0000 0.0024 0.9987 0.4125 0.0523
Std. 0.0000 0.9987 0.0000 0.6457 0.0000 0.0000 0.006 0.0069 0.9501 0.5267
PSO [47] Ave. 0.2148 0.2858 0.1502 0.3443 0.0461 0.7212 0.0817 1.0000 0.3548 0.5917
Std. 0.2663 0.0867 0.1436 0.1023 0.0706 0.5303 0.0635 0.0094 0.6283 0.9783
GSA [48] Ave. 0.003 0.0000 0.0289 0.1821 0.0000 0.3944 0.0017 1.0000 0.0000 0.0000
Std. 0.0224 0.0084 0.0374 0.0981 0.0000 0.2328 0.0058 0.0026 0.329 0.0000
BA [49] Ave. 0.7518 1.0000 1.0000 0.9059 1.0000 1.0000 1.0000 0.0000 0.6155 0.9443
Std. 1.0000 0.4826 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 0.4541
FPA [50] Ave. 0.024 0.5394 0.0034 0.1571 0.0031 0.0153 0.0138 1.0000 0.5894 0.7708
Std. 0.0388 0.3095 0.0008 0.2651 0.0054 0.0561 0.014 0.0021 0.6006 1.0000
SoM [51] Ave. 1.0000 0.5152 0.2049 1.0000 0.7049 0.754 0.0000 1.0000 0.9074 1.0000
Std. 0.399 0.1338 0.0508 0.2571 0.4865 0.3097 0.0000 0.0000 0.5564 0.2696
FA [52] Ave. 0.1407 0.3064 0.0419 0.2882 0.0095 0.041 0.0445 1.0000 0.8299 0.6937
Std. 0.0845 0.0000 0.0079 0.0000 0.0053 0.0223 0.0274 0.0009 0.1451 0.4449
GA [53] Ave. 0.3485 0.4172 0.1049 0.6959 0.1056 0.1892 0.3625 1.0000 1.0000 0.8864
Std. 0.3714 0.0764 0.0362 0.0174 0.0836 0.3023 0.1503 0.0016 0.0000 0.2961
SSA [54] Ave. 0.0000 0.2272 0.0000 0.0000 0.0000 0.0000 0.0028 1.0000 0.4254 0.0598
Std. 0.0000 1.0000 0.0000 0.6556 0.0000 0.0000 0.007 0.0071 0.9502 0.5279

boundary of each unit as: Step 2. Checking the equality and inequality constraints for the
initialized moths using equations (5) and (7). For violated so-
Pimin  Pit  Pimax (7) lutions repeat step 1 otherwise evaluate the fitness function.
Step 3. Evaluate the fitness function

For all the moths, it was assumed that there is a matrix to store
the corresponding fitness values as follows:
3. Proposed optimization methodology
M ¼ ½ M1 M2 … MN T (9)
3.1. Basics of moth-flame optimizer
The second evaluation index of the MFO algorithm is flames
which is a matrix similar to the moth matrix. This index is
Moth-Flame Optimizer (MFO) was first presented in Ref. [27]. It
expressed as:
is a recent nature-inspired method which is based on the naviga-
tion mechanism called transverse orientation of Moths in space.
F ¼ ½ F1 F2 … FN T (10)
Fig. 1 shows a conceptual model of transverse orientation. Trans-
verse orientation for navigation uses a constant angle by Moths The moths are actual search agents that move on the search
with the moon which is thousands of miles away. It means to fly in space, while the flames are the best position moths. The dimension
straight direction. Moths applied this transverse mechanism for of M and F matrices are no of population  no of control variables.
artificial lights that is nearer than Moon, they tricked as they do not
keep fixed angle with flame so fly around flames in a Logarithmic Step 4. Generation of new solution via moth-flame
spiral way and finally converges towards the flame. This Logarith-
mic spiral way is the exploration area and it refers to exploit the The generation of updated a logarithmic spiral of each moth
optimum solution. The logarithmic spiral, space around the flame, which is based on the difference between the moths and flames
and the position considering different t on the curve are shown in solutions as described in Fig. 2. Equation (11) updates the moths
Fig. 2. with respect to a flame as:
 
Mi ¼ S Mi ; Fj ¼ Di eMi r1 cosð2pr1 Þ þ Fj (11)
3.2. Implementation of MFO for ED problem
The distance from ith moth to jth flame is calculated as follows:
In the proposed MFO, each one of moths represents a solution  
and population of moths is used to finding the best solution (flame) Di ¼ Fj  Mi  (12)
of ED problem. The main steps of proposed MFO method are pre-
sented below:

Step 1. Initialization: A population of NP moths is represented Step 5. Reduction of flames number

by X ¼ ½X1 ; X2 ; X3 …:XNp  T where each moth
Xi ¼ ½Pi1 ; Pi2 ; Pi3 ; …:PiNp  represents power output of generating Number of flames is reduced by increasing the number of iter-
units. Mapping the control variables from each group of moth ations. The following formula is utilized in this subject:
into ED data is employed using Eq. (8) as:  
Nf  1
  flam no ¼ round Nf  l* (13)
T
Pij ¼ Pjmin þ rand* Pjmax  Pjmin (8)

vy-flight to MFO.
Step 6. Introducing Le
1068 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

Fig. 4. Sample of search space and convergence curve of benchmark test functions.
 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078 1069

Table 2
Description of different cases of different tests.

Objectives Case No Description

Single objective Case 1 Minimization of generation fuel cost

Case 2 Minimization of generation fuel cost with valve point effect
Case 3 Minimization of total emission
Multi-objectives Case 4 Minimization of generation fuel cost and emission
Case 5 Minimization of generation fuel cost with valve point effect and emission

Table 3
Results of the minimum cost and power generation compared with global optimization methods for 6-unit system.

Methods P G1 ðMWÞ P G2 ðMWÞ P G3 ðMWÞ P G4 ðMWÞ P G5 ðMWÞ P G6 ðMWÞ Total P G ðMWÞ P loss ðMWÞ Cost ($/hr) Emission (ton/hr)

Case 1 Proposed EMFO 446.1124 172.1872 262.0157 137.0125 164.9874 87.987 1270.3022 7.3022 15374.57 1766.05
Proposed MFO 446.4561 173.3145 263.0371 138.9772 165.1246 87.8682 1274.7777 11.7777 15434.31 1775.29
PSO [58] 447.497 173.322 263.475 139.059 165.476 87.128 1276.010 12.958 15450.000 1781.127
GA [58] 474.807 178.636 262.209 134.283 151.904 74.181 1276.030 13.022 15459 1879.096
NPSO-LRS [59] 446.960 173.394 262.344 139.512 164.709 89.016 1275.940 12.936 15450 1776.029
PSO-LRS [59] 447.444 173.343 263.365 139.128 165.508 87.170 1275.950 12.957 15450 1780.751
NPSO [59] 447.473 173.101 262.680 139.416 165.300 87.976 1275.950 12.947 15450 1778.769
MTS [60] 448.128 172.808 262.593 136.961 168.203 87.330 1276.023 13.021 15450.060 1780.149
BSA [57] 447.490 173.331 263.456 139.060 165.480 87.141 1275.958 12.958 15449.9 1781.065
HS [61] 449.381 173.53 263.524 132.049 167.262 90.262 1276.008 13.08 15449 1782.975
HHS [61] 447.496 173.314 263.445 139.055 165.475 87.125 1275.91 12.95 15449 1781.0097
Case 2 Proposed EMFO 445.8124 170.7569 260.9878 139 164.0145 85 1265.572 2.5716 15423.51 1760.025
Proposed MFO 446.0678 170.9874 261.3278 139.0012 164.1245 85.1128 1266.6215 3.6215 15441.15 1762.638
CSA [55] 447.477 173.223 263.379 138.952 165.412 87.002 1275.447 12.447 15443.08 1780.329
ICS [62] 447.616 173.580 262.758 139.121 165.643 86.666 1275.380 12.392 15442.265 1780.005
BBO [63] 447.400 173.239 263.316 138.001 165.410 87.080 1275.446 12.446 15443.096 1778.593
SOH-PSO [64] 438.21 172.58 257.42 141.09 179.37 86.88 1275.55 12.55 15446.02 1741.753
Case 3 Proposed EMFO 436.1271 173.3819 262.6189 139.0027 168 89.1078 1268.238 5.2384 15348.28 1735.914
Proposed MFO 436.2054 173.8751 262.6442 140.3511 168.0035 89.34772 1270.42702 7.42702 15377.669 1739.0971
Case 4 Proposed EMFO 447.8645 173.1278 263 138.8745 165.3478 87.0145 1275.229 12.2291 17220.62 1780.419
Proposed MFO 447.9987 173.3861 263.1048 138.9601 165.4273 87.3223 1276.1993 13.1993 17235.13092 1782.0322
Case 5 Proposed EMFO 446.5247 173.1025 263.7124 138.62541 166.29872 87.6748 1275.939 12.938 17393.28 1777.9141
Proposed MFO 446.6682 173.5861 263.7041 138.9601 166.4102 87.9223 1277.251 14.251 17417.1009 1779.974

Fig. 7. Pareto front obtained by MFO for case 4.

Fig. 5. Convergence characteristic of 6-unit system for cases 1 and 2.

Fig. 6. Convergence characteristic of 6-unit system for case 3. Fig. 8. Pareto front obtained by MFO for case 5.
1070 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

vy-flight has the prominent properties to increase the di-

Le The above algorithm is stopped when the number of iterations
versity of population, sequentially, which can make the algorithm reaches the predefined value. The flowchart of the proposed moth-
effectively jump out of the local optimum. It is formulated using flame algorithm is given in Fig. 3.
(14) after the position updating as follows [43e45]:
4. Applications

Xitþ1 ¼ Xit þ r2 sign½rand  0:54LevyðbÞ (14)

4.1. Application to benchmark test functions
Le
vy -flight are a kind of random walk in which the steps are
determined by the step lengths, and the jumps conform to a Le
vy Ten benchmark functions are introduced to verify the effect of
distribution as follows [45,46]: proposed EMFO. The mathematical formulation of ten tests func-
tions are presented in Table A1. The proposed MFO algorithm is
developed in ten benchmark functions. To explore the performance
r ∅

vyðxÞ ¼ 0:01* 3
Le (15) of the proposed MFO algorithm, some of the recent and well-
jr4 j1=b known algorithms in the literature are chosen PSO [47], gravita-
tional search algorithm (GSA) [48], BA [49], FPA [50], States of
Matter algorithm (SoM) [51], FA [52], GA [53] and Salp swarm al-
1=b
Gð1 þ bÞ  sinðp  b=2Þ gorithm (SSA) [54]. The population size and maximum iterations
f¼ ; ð0  b  2Þ
Gððð1 þ bÞ=2Þ  b  2ðb1Þ=2 are set 30 and 1000 for benchmark functions. As shown in Table 1,
(16) the EMFO algorithm provides the best results on a ten test func-
tions. The experiments are redone on some of the test functions
with 2 variables and using 100 moths over 1000 iterations. Sample
Step 7. Stopping criteria of results is provided in Fig. 4. The results are consistent with those

Table 4
Results of the minimum cost and power generation compared with global optimization methods for 40-unit system.

Case 1 Case 3

U P max EMFO MFO MBFA [64] OGSA [64] EMFO MFO IABC [41] ABCDP [41] ABC [41] MBFA [64]

G1 114 110.8 114 114 114 112.897 112.999 114 114 113.999 114
G2 114 110.83 110.9817 110.8035 111.2821 114 114.012 114 114 114 114
G3 120 97.4 100 97.4002 120 119.31 119.210 119.999 119.999 120 120
G4 190 179.73 180 179.7333 189.6786 170 170.004 169.375 169.364 171.004 169.3671
G5 97 87.81 87.8072 87.8072 97 98 97.9995 97 97 96.9995 97
G6 140 140 142 140 140 127 127 124.255 124.253 127 124.263
G7 300 259.6 260 259.6004 259.2637 297.105 297.905 299.715 299.717 298.286 299.6931
G8 300 284.6 284.7002 284.6002 283.036 297.102 297.274 297.917 297.915 298.591 297.9093
G9 300 284.6 284.8796 284.6006 290.0689 297 297.100 297.265 297.257 296.967 297.2578
G10 300 130 135 130 130 135 135 130 130 134 130.0007
G11 375 94 168.7999 168.7999 101.2163 298.001 298.222 298.404 298.411 297.807 298.421
G12 375 94 170 168.7998 153.6064 296 296.002 298.026 298.023 296.168 298.0264
G13 500 214.76 215 214.7598 210.1281 433.987 434.731 433.561 433.547 434.731 433.559
G14 500 394.28 306.4201 304.5195 305.2207 421 421.008 421.728 421.738 421.008 421.736
G15 500 394.28 395 394.2794 297.6248 423.023 423.015 422.775 422.776 423.015 422.7884
G16 500 394.28 395 394.2794 212.8865 421.698 421.765 422.792 422.769 420.511 422.7841
G17 500 489.28 489.2122 489.2794 499.7336 441.398 441.381 439.411 439.425 441.381 439.4078
G18 500 489.28 489.2122 489.2794 489.4927 436.775 436.776 439.383 439.401 436.777 439.4132
G19 550 511.28 512 511.2795 511.9556 435.998 436.472 439.421 439.421 436.472 439.4111
G20 550 511.28 512 511.2795 515.1468 439.345 439.416 439.412 439.403 439.417 439.4155
G21 550 523.28 523.7248 523.2794 524.8624 436.795 436.997 439.452 439.455 436.957 439.4421
G22 550 523.28 523.7248 523.2794 516.3822 441.86 441.860 439.442 439.461 441.860 439.4587
G23 550 523.28 523.7248 523.2796 522.0437 438.425 438.519 439.776 439.766 438.519 439.7822
G24 550 523.28 523.7248 523.2794 526.5253 438.289 438.363 439.77 439.782 438.363 439.7697
G25 550 523.28 523.7248 523.2795 512.144 442 442.171 440.118 440.105 442.171 440.1191
G26 550 523.28 523.7248 523.2796 523.0461 441.523 441.378 440.108 440.097 441.355 440.1219
G27 150 10 10.8702 10.0001 131.8133 29.4587 29.0661 28.988 28.985 29.0661 28.9738
G28 150 10 10.8702 10.0002 150 27.8974 27.5502 28.997 28.997 27.5502 29.0007
G29 150 10 10.8702 10.0002 133.4239 31.4587 31.378 28.991 28.989 31.378 28.9828
G30 97 87.93 90 89.507 97 98 98 97 97 99 97
G31 190 190 200 190 187.6443 170.978 170.742 172.329 172.335 170.626 172.3348
G32 190 190 200 190 190 173.645 173.528 172.326 172.329 173.779 172.3327
G33 190 190 200 190 169.7297 173.482 173.482 172.334 172.334 173.212 172.3262
G34 200 164.8 164.9025 164.8026 190.9535 200 200 200 200 200 200
G35 200 194.22 164.9025 164.8035 176.4784 200 200 200 200 200 200
G36 200 200 164.9875 164.8292 164.6995 200 200 200 200 200 200
G37 110 110 111 110 99.1969 100.339 100.345 100.84 100.836 100 100.8441
G38 110 110 111 110 34.81 100.339 100.345 100.836 100.836 100 100.8346
G39 110 110 111 110 85.8264 100.339 100.345 100.836 100.836 100 100.8362
G40 550 511.28 515 511.2795 532.0793 438.456 438.028 439.405 439.42 438.028 439.3868

Total cost ($/hr) 120389.4 121308.6284 121415.653 120390.00 124795.1 124797.4 124772.1 124771.8 124793.7 124771.9

Emission(ton/hr) 359907.5 371140.9 356424.4971 453464.7 176482.9 176629.1 176682.3 176682.3 176633 176682.27
 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078 1071

Table 5
Results of the minimum cost and power generation compared with global optimization methods for 40-unit system (case 2).

Units P max EMFO MFO PSO PPSO CSA BBO IABC ABCDP IABC-LS [41] ABCDP-LS [41] ABC
[66] [66] [55] [67] [41] [41] [41]

G1 114 72.481 109.8725 113.116 111.601 112.0518 111.0465 110.802 110.799 110.799 110.799 110.7704
G2 114 103.0314 109.2125 113.01 111.781 111.4948 111.5915 110.8 110.799 110.799 110.803 110.7705
G3 120 83.2726 96.1226 119.702 118.613 97.5626 97.6077 97.4 97.399 97.399 97.4 97.3707
G4 190 182.3106 174.2874 81.647 179.819 179.8 179.7095 179.733 179.733 179.733 179.733 179.7038
G5 97 76.1669 89.4201 95.062 92.443 88.9834 88.306 92.7494 87.799 87.799 87.799 87.7705
G6 140 126.1346 137 139.209 139.846 140 139.9992 140 139.999 139.999 139.999 140
G7 300 258.8452 298.9987 299.127 296.703 299.9993 259.6313 259.599 259.599 259.599 259.599 259.5699
G8 300 297.1636 283.9102 287.491 284.566 284.9506 284.7366 284.599 284.599 284.599 284.599 284.5704
G9 300 290.8899 283.9874 292.316 285.164 284.9653 284.7801 284.599 284.599 284.599 284.599 284.5706
G10 300 274.8232 127 279.273 203.859 130.0006 130.2484 130 130 130 130 130
G11 375 356.9806 93 169.766 94.283 94 168.8461 168.799 94 94 94 94
G12 375 124.4054 94 94.344 94.09 94 168.8239 168.799 94 94 94 94
G13 500 493.3764 210.9897 214.871 304.83 214.7621 214.7038 214.759 214.757 214.759 214.759 214.7298
G14 500 344.9029 302.1253 304.79 304.173 304.5194 304.5894 394.279 394.278 394.279 394.279 394.2497
G15 500 372.3864 397.7896 304.563 304.467 394.2799 394.2761 304.519 394.276 394.279 394.279 394.2496
G16 500 345.4624 397.7896 304.302 304.177 394.2793 394.2409 394.279 394.278 394.279 394.279 394.2497
G17 500 422.6378 487.0125 489.173 489.544 489.2802 489.2919 489.279 489.279 489.279 489.279 489.2499
G18 500 434.4065 487.0214 491.336 489.773 489.2776 489.4188 489.279 489.279 489.279 489.279 489.2499
G19 550 461.3107 510.9874 510.88 511.28 511.2797 511.2997 511.279 511.279 511.279 511.279 511.2497
G20 550 434.3828 510.9196 511.474 510.904 511.2799 511.3073 511.279 511.279 511.279 511.279 511.2708
G21 550 545.2846 522.1023 524.814 524.092 523.3012 523.417 523.279 523.279 523.279 523.279 523.2729
G22 550 490.3572 523.1782 524.775 523.121 523.2928 523.2795 523.279 523.279 523.279 523.279 523.2737
G23 550 506.0639 522.2701 525.563 523.242 523.2892 523.3793 523.279 523.279 523.279 523.279 523.2747
G24 550 467.3109 522.1274 522.712 524.26 523.434 523.3225 523.279 523.279 523.279 523.279 523.2758
G25 550 488.1203 522.1265 503.211 523.283 523.2839 523.3661 523.279 523.279 523.279 523.279 523.2758
G26 550 486.9019 522.2701 524.199 523.074 523.281 523.4362 523.279 523.279 523.279 523.279 523.2697
G27 150 16.8002 10 10.082 10.8 10 10.05316 10 10 10 10 9.9912
G28 150 39.3475 10 10.663 10.742 10.0009 10.01135 10 10 10 10 9.9915
G29 150 23.6359 10 10.418 10.799 10.0014 10.00302 10 10 10 10 9.9909
G30 97 86.3295 90.0547 94.244 94.475 92.0666 88.47754 87.801 87.801 87.799 87.799 87.7723
G31 190 165.9924 188.4987 189.377 189.245 190 189.9983 190 190 190 190 189.9824
G32 190 174.5707 188.4987 189.796 189.995 190 189.9881 190 190 190 190 190
G33 190 184.057 188.4987 189.813 188.081 190 189.9663 190 190 190 190 190
G34 200 193.6668 196.2314 199.797 198.475 199.9998 164.8054 164.799 164.799 164.799 164.799 164.7731
G35 200 191.6152 196.2314 199.284 197.528 199.9999 165.1267 164.802 194.399 200 194.395 194.3642
G36 200 196.1763 200 198.165 196.971 200 165.7695 164.8 200 194.401 200 200
G37 110 90.0101 109.5478 109.291 109.161 110 109.9059 109.999 110 110 110 110
G38 110 37.5421 109.5478 109.087 109.9 110 109.9971 110 110 110 110 110
G39 110 89.4239 109.5478 109.909 109.855 110 109.9695 110 110 110 110 109.938
G40 550 471.4405 511.2758 512.348 510.984 511.2824 511.2794 511.279 511.279 511.279 511.279 511.3501

Total cost 121074.5 121407.6 122323 121788 121425.61 121426.95 121414.8 121412.8 121412.7 121412.7 121412.4
($/hr)

of other test functions, in which the EMFO shows very competitive data of these test systems are customized from Refs. [55e57]. To
compared to other algorithms. demonstrate the effectiveness of the proposed method, five studied
cases with different complexity have been considered as shown in
4.2. Simulation results of test system Table 2.

Three test systems, 6-unit, 40-unit and 80-unit, are introduced 1) Results of 6-unit system
to verify the effect of proposed EMFO for studying ED problem. The
The proposed EMFO and MFO algorithm are developed in cases

Fig. 9. Convergence characteristic of 40-unit system for cases 1 and 2. Fig. 10. Convergence characteristic of 40-unit system for case 3.
1072 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

Table 6
Results of the minimum cost and emission for 40 -unit system for Cases 4 and 5.

Units P max EMFO Case 4 Case 5

MFO MTLA [68] EMFO MFO MODE [56] PDE [56] NSGA-II [56] SPEA 2 [56] GSA [56]

G1 114 108 110 109.69 43.405 114 113.5295 112.1549 113.8685 113.9694 113.9989
G2 114 109.0145 109.3402 108.23 113.95 114 114 113.9431 113.6381 114 113.9896
G3 120 109.0789 109.3041 108.31 105.86 118.9874 120 120 120 119.8719 119.9995
G4 190 181 183 182.49 169.65 178.987 179.8015 180.2647 180.7887 179.9284 179.7857
G5 97 89 88 87.12 96.659 96 96.7716 97 97 97 97
G6 140 135.0871 134.1528 132.28 139.02 138.2079 139.276 140 140 139.2721 139.0128
G7 300 274 274.3094 275.36 273.28 300 300 299.8829 300 300 299.9885
G8 300 288 288 287.54 285.17 300 298.9193 300 299.0084 298.2706 300
G9 300 290 290 290.32 241.96 296.1205 290.7737 289.8915 288.889 290.5228 296.2025
G10 300 130 130 130 131.26 130.2678 130.9025 130.5725 131.6132 131.4832 130.385
G11 375 244.2104 244.0254 243.82 312.13 243.0575 244.7349 244.1003 246.5128 244.6704 245.4775
G12 375 204 204 204.01 362.58 318.3405 317.8218 318.284 318.8748 317.2003 318.2101
G13 500 304 305 302.64 346.24 394.8075 395.3846 394.7833 395.7224 394.7357 394.6257
G14 500 395 394.5182 393.58 306.06 395 394.4692 394.2187 394.1369 394.6223 395.2016
G15 500 388 388 387.58 358.78 305.1014 305.8104 305.9616 305.5781 304.7271 306.0014
G16 500 395.1877 395.9884 395.8 260.68 395.3409 394.8229 394.1321 394.6968 394.7289 395.1005
G17 500 489 490 489.05 415.19 489 487.9872 489.304 489.4234 487.9857 489.2569
G18 500 487.2547 486.7766 486.76 423.94 488.8998 489.1751 489.6419 488.2701 488.5321 488.7598
G19 550 423.987 423.2334 423.89 549.12 500 500.5265 499.9835 500.8 501.1683 499.232
G20 550 514 515 515.93 496.7 455.3408 457.0072 455.416 455.2006 456.4324 455.2821
G21 550 523 522 521.44 539.17 433 434.6068 435.2845 434.6639 434.7887 433.452
G22 550 527 526.2401 526.07 546.46 433.9455 434.531 433.7311 434.15 434.3937 433.8125
G23 550 527 526.1238 525.18 540.06 445.5136 444.6732 446.2496 445.8385 445.0772 445.5136
G24 550 430 433 432.03 514.5 452 452.0332 451.8828 450.7509 451.897 452.0547
G25 550 525 524 524.07 453.46 493.0258 492.7831 493.2259 491.2745 492.3946 492.8864
G26 550 434 433.3334 433.34 517.31 432.9978 436.3347 434.7492 436.3418 436.9926 433.3695
G27 150 30 29 28.52 14.881 11 10 11.8064 11.2457 10.7784 10.0026
G28 150 45.6587 46.2658 46.19 18.79 11.0254 10.3901 10.7536 10 10.2955 10.0246
G29 150 64.2548 65.5211 64.59 26.611 11.0355 12.3149 10.3053 12.0714 13.7018 10.0125
G30 97 88.0987 89.3214 89.64 59.581 95.2125 96.905 97 97 96.2431 96.9125
G31 190 162.087 161.8954 162.26 183.48 188.1205 189.7727 190 189.4826 190 189.9689
G32 190 183.547 183.0213 182.97 183.39 175 174.2324 175.3065 174.7971 174.2163 175
G33 190 183.5 183 184.47 189.02 188 190 190 189.2845 190 189.0181
G34 200 172 171 170.13 198.73 200 199.6506 200 200 200 200
G35 200 162.6012 162.9847 162.28 198.77 200 199.8662 200 199.9138 200 200
G36 200 175 175 175.09 182.23 200 200 200 199.5066 200 199.9978
G37 110 94 93 91.31 39.673 108.2654 110 109.9412 108.3061 110 109.9969
G38 110 112.458 112 110 81.596 108.2654 109.9454 109.8823 110 109.6912 109.0126
G39 110 97 96.3214 95.39 42.96 108.2301 108.1786 108.9686 109.7899 108.556 109.456
G40 550 422 422 420.44 537.17 420.9987 422.0682 421.3778 421.5609 421.8521 421.9987

Total cost 125104.7 125208.6 137984.38 123170 125759 125790 125730 125830 125810 125780
($/hr)

Emission 250264.1 270168.6 272986.88 208460 209387.7 211190 211770 210950 211100 210930
(ton/hr)

1e5 on 6-unit system. Table 3 shows the results of cases 1e5 for the cost is 15374.57 $/hr by using EMFO algorithm and 15434.31$/hr by
first test system using MFO. It is found that the proposed EMFO using MFO comparing with the minimum cost in HS which gives
based approach gives the most economical solution compared with 15449 $/hr.
PSO [58], GA [58], NPSO-LRS [59], PSO-LRS [59], NPSO [59], MTS The results of single objective of case 2 are compared with CSA
[60], BSA [57], HS [61] and HHS [61] in case 1. In this table, the total

Fig. 11. Pareto front obtained by MFO for case 4. Fig. 12. Pareto front obtained by MFO for case 4.
 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078 1073

Table 7
Best solution for 80-unit test system using EMFO for case1.

Unit Output power Unit Output power unit Output power Unit Output power

1 114 21 550 41 105.96 61 550

2 113.29 22 550 42 112.18 62 550
3 109.35 23 550 43 82.374 63 550
4 189.18 24 550 44 185.88 64 550
5 90.838 25 549.98 45 55 65 550
6 132.17 26 550 46 97.558 66 549.99
7 299.96 27 10.605 47 272.65 67 10.914
8 295.87 28 10 48 299.58 68 10.425
9 300 29 10.119 49 299.92 69 12.502
10 200.64 30 60.756 50 190.45 70 96.379
11 103.04 31 190 51 146.75 71 190
12 154.52 32 190 52 141.92 72 190
13 199.9 33 190 53 125.04 73 189.93
14 291.17 34 137.17 54 309.09 74 197.87
15 190.43 35 170.06 55 277.57 75 200
16 330 36 199.99 56 325.04 76 200
17 499.96 37 96.049 57 500 77 108.1
18 452.62 38 79.877 58 499.29 78 93.373
19 549.99 39 109.83 59 549.99 79 63.815
20 543.89 40 545.34 60 549.98 80 549.93
Total output power (MW) 21000 Total costs ($/hr) 239783.7994

[55], ICS [62], BBO [63] and SOH-PSO [64] are listed in Table 3. The using proposed EMFO and 15441.15 $/hr by using proposed MFO. In
total generation fuel cost with valve point effect is15423.51$/hr by case 3, the total emission is 1735.914 ton/hr by using proposed
EMFO and 1739.0971 ton/hr by using proposed MFO and also are
listed in Table 3.
The muti-objective studied cases (4,5) are listed in Table 3. In
case 4, the total generation fuel cost and emission are 17220.62 $/hr
and 1780.419 ton/hr by using proposed EMFO and 17235.13092
$/hr, 1782.0322 ton/hr by using proposed MFO. In case 5, the total
generation fuel cost with valve point effect and emission are
17393.28 $/hr and 1777.9141 ton/hr by using proposed EMFO and
17417.1009 $/hr and 1779.974 ton/hr by using proposed MFO. The
convergence characteristic of this test system obtained from MFO
in cases 1 and 2 is shown in Fig. 5 where the proposed algorithm
reaches the best value after 100 and 120 iterations respectively. In
addition, Fig. 6 shows the convergence of the emission function,
where the proposed algorithm reaches the best value after 100 it-
erations. In cases 4 and 5, multi-objective optimization is
employed. Therefore, a set of solutions is obtained which is denoted
Fig. 13. Convergence characteristic of 80-unit system for cases 1. as Pareto optimal solution [65]. The Pareto front is obtained and it

Table 8
Best solution for 80-unit test system using EMFO for case 2.

Unit Output power Unit Output power unit Output power Unit Output power

1 112.404 21 523.2835 41 114 61 523.2037

2 113.547 22 523.2828 42 110.647 62 523.4
3 97.454 23 523.2793 43 97.5 63 523.312
4 179.735 24 523.2968 44 179.645 64 523.2965
5 95.001 25 523.2236 45 88.245 65 523.412
6 140 26 523.2925 46 140 66 523.3497
7 260 27 10 47 260 67 10
8 284.623 28 10 48 285.124 68 10
9 285.175 29 10 49 284.654 69 10
10 130 30 88.12268 50 130 70 90
11 94 31 190 51 168.001 71 190
12 168.501 32 190 52 94 72 190
13 214.845 33 190 53 125 73 190
14 394.148 34 165.221 54 304.5006654 74 165.321
15 393.571 35 200 55 394.325 75 200
16 393.698 36 200 56 394.325 76 200
17 489.4424 37 110 57 488.4 77 110
18 489.28561 38 110 58 489 78 110
19 511.2839 39 110 59 511.457 79 110
20 511.301 40 511.321 60 511.545 80 512
Total output power (MW) 21000 Total costs ($/hr) 242901.8785
1074 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

has been depicted in Figs. 7 and 8. The graphical results signifi- Case 2 is proposed to minimize the generation fuel cost with
cantly reveal that the obtained solutions are well-distributed and valve point effect and it is shown in Table 5. It is compared with PSO
covered the entire Pareto front of Cases 4 and 5. [66], PPSO [66], CSA [55], BBO [67], IABC [41], ABCDP [41], IABC-LS
[41], ABCDP-LS [41] and ABC [41]. Using EMFO leads to most eco-
2) Results of 40-unit system nomic emission compared with different methods. The conver-
gence characteristic of this test system obtained from MFO in cases
The second test system has 40 generating units. The simulation 1 and 2 are shown in Fig. 9. In addition, Fig. 10 shows the conver-
results are obtained by using proposed EMFO and MFO for cases 1 gence of the emission function.
and 3 in Table 4. The total cost obtained by EMFO is 120389.4 $/hr Table 6 shows the results of multi-objective functions for case 4
and 121308.6284 $/hr by using MFO. It is compared with MBFA [64] and case 5. Case 4 is proposed to minimize the generation fuel cost
and OGSA [64]. The proposed EMFO provides the lowest minimum and emission by using EMFO and MFO algorithms. The total cost is
cost among all other competitive methods in the literature. For 125104.7 $/hr and emission is 250264.1ton/hr compared to MTLA
economic emission dispatch, case3 are listed in Table 4. To mini- [67]. Case 5 is proposed to minimize the generation fuel cost with
mize the emission, EMFO and MFO are used and compared with valve point effect and emission by using EMFO and MFO algorithms.
other methods such as IABC [41], ABCDP [41], IABC-LS [41], ABCDP- The total cost is 123170 $/hr and emission is 208460 ton/hr
LS [41], ABC [41] and MBFA [64]which the total emission is compared to MODE [55], PDE [55], NSGA-II [55], SPEA 2 [55] and
176482.9 ton/hr by using EMFO and 176629.1 ton/hr by using MFO GSA [55]. Results in Table 6 show that: EMFO has the least total cost
respectively. and emission for the studied cases. In cases 4 and 5, multi-objective

Table 9
Best solution for 80-unit test system using EMFO for minimizing the total emission (case 3).

Unit Output power Unit Output power unit Output power Unit Output power

1 113.9997243 21 418.8459586 41 113.9784 61 418.0022

2 113.7379415 22 415.5025768 42 113.7 62 417.7792
3 81.89471823 23 414.3628902 43 108.0896 63 414.596
4 158.8768486 24 417.0096788 44 159.3282 64 413.3588
5 71.60544837 25 415.9925598 45 73.30211 65 418.5347
6 120.5910362 26 418.4022096 46 114.3446 66 420.4356
7 280.9490319 27 149.9999171 47 286.6427 67 150
8 276.368889 28 149.9999397 48 272.279 68 150
9 284.1128145 29 149.9999853 49 287.771 69 150
10 279.9948641 30 66.9288014 50 290.4878 70 93.38114
11 282.6289825 31 158.3164724 51 285.6921 71 153.7112
12 285.6581956 32 160.0855177 52 287.0221 72 161.6624
13 415.0397192 33 157.9581089 53 419.6174 73 158.2236
14 417.7272226 34 199.9999743 54 415.6176 74 200
15 414.7869754 35 199.9999991 55 415.6001 75 199.9988
16 418.9747865 36 200 56 417.7571 76 199.9999
17 412.83754 37 96.94025834 57 418.5315 77 82.16123
18 414.8997045 38 91.18564104 58 415.5713 78 97.23573
19 413.9892496 39 90.33849044 59 419.5352 79 93.21022
20 412.0937308 40 415.4987849 60 415.9485 80 418.7578
Total output power (MW) 21000 Total Emission (ton/hr) 139769.7

Table 10
Best solution for 80-unit test system using EMFO for minimizing the total cost and emission (case 4).

Unit Output power Unit Output power unit Output power Unit Output power

1 113.9999901 21 433.7768021 41 92.61237083 61 431.6077583

2 113.6499609 22 434.7358723 42 73.56916769 62 425.7248365
3 112.3172297 23 436.9140168 43 89.03423221 63 433.9700089
4 150.6438562 24 435.2507888 44 170.374087 64 431.9074251
5 76.3727202 25 429.6569678 45 74.91921894 65 435.7249398
6 122.4982883 26 435.1647635 46 120.8659756 66 436.190046
7 292.195806 27 75.84471392 47 279.0513343 67 112.9281203
8 279.3375978 28 101.9397203 48 279.5596165 68 86.4045005
9 293.1292868 29 67.49712588 49 292.2765141 69 68.22949278
10 290.9995441 30 95.90819804 50 277.0593706 70 75.25330419
11 289.8689037 31 166.685163 51 287.2516483 71 153.4069455
12 291.0160394 32 167.9953735 52 286.3244034 72 171.7399109
13 428.028883 33 172.2518036 53 426.3931601 73 167.3577935
14 422.0941444 34 143.375968 54 424.7794849 74 200
15 428.8010156 35 199.9999245 55 429.9879239 75 199.9999719
16 427.8846201 36 199.9988759 56 424.3594949 76 199.9999521
17 430.5259955 37 81.03138605 57 433.7128997 77 102.8862173
18 421.2731809 38 101.4899497 58 424.5210638 78 57.44468002
19 426.7053518 39 98.57050959 59 429.1863283 79 81.70532722
20 430.6994196 40 429.5574786 60 430.525495 80 431.4677432
Total output power (MW) 21000 Total costs ($/hr) 270366.1139
Total Emission (ton/hr) 169974.8431
 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078 1075

optimization is employed. Therefore, a set of solutions is obtained

which is denoted as Pareto optimal solution. The Pareto front is
obtained and it has been depicted in Figs. 11 and 12.

3) Results of large scale 80-unit system

The proposed EMFO is integrated to obtain the solution of ED

problem in case of large scale system such as 80-units. Results
obtained from proposed EMFO for single objective of case 1 have
been showed in Table 7 which the total cost is 239783.7994 $/hr.
Fig. 13 displays the convergence characteristics for this case study.
Best solution obtained from proposed EMFO for single objective of
minimizing the total cost with valve point effect (case 2) have been
showed in Table 8. It is compared with THS (t ¼ 8) [69], FAPSO [70],
CSO [71], PSO [71], SCA [71], SSA-I [72], SSA-II [72] and MSSA [72] in
Fig. 14. Convergence characteristic of 80-unit system for cases 2.

Table 11
Best solution for 80-unit test system using EMFO for minimizing the total cost with valve point effect and emission (case 5).

Unit Output power Unit Output power unit Output power Unit Output power

1 113.9998265 21 433.519143 41 111.6183569 61 432.498653

2 112.9842716 22 431.0020765 42 113.9911717 62 433.4055515
3 119.887989 23 432.6235191 43 111.4607192 63 432.9913096
4 166.4089869 24 433.3823754 44 171.8888473 64 433.351739
5 92.02972364 25 433.5186863 45 92.39950267 65 433.4732216
6 116.39442 26 431.5517683 46 122.1456179 66 431.6136039
7 285.1152312 27 52.50799904 47 286.2973615 67 105.568981
8 293.4156898 28 94.81926926 48 288.7699688 68 77.12133036
9 285.6266712 29 79.2246787 49 295.0196336 69 96.06526708
10 286.4371167 30 96.98945326 50 281.6644625 70 68.54109004
11 293.1004786 31 164.6411792 51 284.1616346 71 162.3134003
12 289.535277 32 160.9020518 52 287.279288 72 160.3916968
13 420.9093974 33 167.443694 53 418.8579939 73 167.1915322
14 421.3507449 34 199.9999712 54 426.7211614 74 198.338859
15 417.9763539 35 199.9998115 55 418.5805059 75 199.9999996
16 417.9396216 36 199.9999661 56 421.8241461 76 172.2766316
17 420.9704091 37 55.49225244 57 418.7913012 77 99.74313785
18 427.163854 38 98.12864754 58 420.9101058 78 69.25467443
19 422.3245673 39 69.86984591 59 424.1025905 79 101.5341278
20 421.5259397 40 421.5193918 60 422.9145018 80 422.6939712
Total output power (MW) 21000 Total costs ($/hr) 276654.0157 Total Emission (ton/hr) 162932.9399

Table 12
Statistical indices of test system for Case 1.

test system Method Total generation cost ($/hr)

Minimum Average Maximum Standard deviation

6-unit GA [58] 15459 15469 15524 NA

MTS [60] 15450.06 15451.17 15453.64 0.9287
NPSO-LRS [59] 15450 15450.5 15452 NA
PSO-LRS [59] 15450 15454 15455 NA
NPSO [59] 15450 15452 15454 NA
PSO [58] 15450 15454 15492 NA
BSA [57] 15449.9 15449.9 15449.91 0.001
MFO 15434.31 15434.27 15434.42 0.0008
EMFO 15374.57 15376.42 15378.15 0.0006
40-unit ICA-PSO [62] 121413.2 121428.14 121453.56 NA
CPSO-SQP [62] 121458.54 122028.16 NA NA
ICS [62] 121399.52 121428.352 121469.865 5.27
MFO 121308.63 121365.425 121412.014 4.98
EMFO 120389.4 120449.35 120489.43 4.02
80-unit THS (t ¼ 8) [69] 243192.69 243,457.36 NA 120.99
FAPSO [70] 244273.54 NA NA NA
CSO [71] 243195.38 243546.63 244038.74 NA
SCA [71] 250864.05 254579.79 261099.64 NA
SSA-I [72] 243173.59 243306.11 243490.81 72.663
SSA-II [72] 243398.17 243575.35 243760.89 68.983
MSSA [72] 242909.25 243037.25 243229.95 53.762
MFO 242908.35 243036.36 243230.14 53.641
EMFO 242900.01 243025.84 243250.87 51.651

NA: Data not available.

1076 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

improve the solution quality. The proposed EMFO is validated for

single and multi-objectives functions. The capability of the EMFO is
proved through minimizing the total generation cost and emission
for three 6, 40, and 80-unit test systems. The obtained results were
compared with recent optimization methods that reported in the
literature. It was concluded that, the proposed EMFO leads to the
most economical solution and minimum emission levels. According
to the simulation results, the effectiveness, robustness and perfor-
mance of EMFO achieve the best of all objectives for all studied
cases. The salient findings of this work are summarized as follows:

 This paper has been developed an enhanced version of MFO

algorithm that merges the merits of Le
vy-flight.
Fig. 15. Convergence characteristic of 80-unit system for cases 3.
 The proposed EMFO enhances the solution quality by enhancing
exploration and exploitation processes.
 Application of the EMFO is employed for 10 benchmarking test
functions as well as practical power systems.
Table 12. The total cost with valve point effect is 242901.8785 $/hr.
 In the economical regard, the EMFO algorithm is applied to non-
Using EMFO leads to most economic dispatch compared with
smooth cost function ED framework with considering valve-
different methods. The convergence characteristics for case 2 is
point concerns with equality and inequality operational
shown in Fig. 14. Best solution for 80-units test system using EMFO
constraints.
for minimizing the total emission in case 3 is 139769.7 ton/hr and
 Single and multi-objective functions formulation for practical
shown in Table 9. The convergence characteristic of cases 3 is
small and large-scale power systems are successively solved for
shown in Fig. 15. Tables 10 and 11 show the results of multi-
three standard systems up to 80-unit tests.
objectives functions for case 4 and case 5. In case 4, the total cost
 Significant economic benefit is achieved using the proposed
and emission are 270366.1139 $/hr and 169974.8431ton/hr
EMFO.
respectively. In case 5, the best solution of total cost with valve
 The proposed EMFO is efficient for solving nonconvex multi-
point effect and emission obtained from proposed EMFO are
objective optimization problems where the optimal solutions
276654.0157 $/hr and 162932.9399ton/hr respectively.
can be obtained and it reflected the closeness to the best solu-
tion and the extreme from the worst solution.
4.3. Statistical indices  Statistical evaluation of the proposed EMFO compared with
previous optimization algorithms reflects the EMFO robustness
For 6-unit,40- unit and 80-unit test systems, to validate the and capability for solving non-smooth ED problem.
robustness of the proposed EMFO method the number of 50,100  The EMFO algorithm has good convergence characteristics.
and 100- trial are respectively considered. The population size and  It can be considered as a promising alternative algorithm for
maximum iterations are set 10, 500 and 1000 for the three test solving problems in practical large-scale power systems.
systems, respectively. Table 12 shows the statistical indices of the
results for the studied systems compared with other methods. To
validate its robustness for solving the ED problem in each test
system, the low value of standard deviation of total generation cost
achieved. The optimal results of EMFO vary from 15434.31 ($/hr) to Appendix A
15434.42 ($/hr) for 50 trials, which also shows high robustness for
the proposed method. The Statistical results for 40-unit test system Table A1 shows the main data of benchmarking test functions.
by EMFO and other methods are presented in Table 12. The EMFO The dimension of all test functions equal 100.
method can obtain lower minimum, average, and maximum cost
than the other recent methods. The results show that highly similar
optimal results with negligible standard deviation (4.98 $/hr). The Table A1
range of total generation costs obtained by EMFO for 100 runs is Data of benchmark functions

between 121308.6284 ($/hr) to 121412.0145 ($/hr), which also Function Range f min
shows that EMFO converges to almost the same optimal in large Pn
F1 ðxÞ ¼ 2
i¼1 xi
[-100,100] 0
size system. The Statistical results for large-scale 80-unit test sys- Y
n
Pn [-10,10] 0
tem by EMFO and other methods. The EMFO method can obtain F2 ðxÞ ¼ i¼1 jxi j þ jxi j
i¼1
lower minimum, average, and maximum cost than the other recent Pn Pi 2 [-100,100] 0
F3 ðxÞ ¼ i¼1 ð j1 xj Þ
methods. The range of total generation costs obtained by MFO for
F4 ðxÞ ¼ maxfjxi j ; 1  i  ng [-100,100] 0
100 runs is between 242900.01($/hr) to 243250.87 ($/hr) and the P 2 2 [-30,30] 0
F5 ðxÞ ¼ n1
i¼1 ½100ðxiþ1  xi Þ þ ðxi  1Þ 
2
best total generation costs is 242901.8785 ($/hr). EMFO show very P
F6 ðxÞ ¼ ni¼1 ð½xi þ 0:5Þ2 [-100,100] 0
competitive results and effeteness compared to other algorithms in P
F7 ðxÞ ¼ ni¼1 ix4i þ randomð0; 1Þ [-1.28,1.28] 0
small and large scale systems. P pffiffiffiffiffiffiffi
F8 ðxÞ ¼ ni¼1  xi sinð jxi jÞ [-500,500] 418.9829  5
Pn
F9 ðxÞ ¼ i¼1 ½xi  10 cosð2pxi Þ þ 10
2 [-5.12,5.12] 0
5. Conclusion 0 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 [-32,32] 0
1 Xn 2 A
F10 ðxÞ ¼  20 exp@  0:2 i¼1 i
x 
n
In this paper, an enhanced moth-flame optimization algorithm !
has been implemented for solving the non-smooth economic 1 Xn
exp i¼1
cosð2pxi Þ þ 20 þ e
n
dispatch problem with valve point effects and emissions. The EMFO
combines the merits of MFO algorithm and levy-flight method to
 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078 1077

References 2016. p. 734e8. Region 10 Conference (TENCON).

[31] Zawbaa HM, Emary E, Parv B, Sharawi M. Feature selection approach based on
moth-flame optimization algorithm. In: Evolutionary computation (CEC).
[1] Chen G, Ren J, Feng EN. Distributed finite-time economic dispatch of a
IEEE; 2016. p. 4612e7. IEEE Congress on.
network of energy resources. IEEE Trans Smart Grid 2017;8(2):822e32.
[32] Parmar SA, Pandya MH, Bhoye M, Trivedi IN, Jangir P, Ladumor D. Optimal
[2] Yang Z, Xiang J, Li Y. Distributed consensus based supplyedemand balance
active and Reactive Power dispatch problem solution using Moth-Flame
algorithm for economic dispatch problem in a smart grid with switching
Optimizer algorithm. April. In: Energy efficient technologies for sustainabil-
graph. IEEE Trans Ind Electron 2017;64(2):1600e10.
ity (ICEETS), 2016 international conference on. IEEE; 2016. p. 491e6.
[3] Zhan J, Wu Q, Guo C, Zhou X. Economic dispatch with non-smooth objecti-
[33] Bentouati B, Chaib L, Chettih S. Optimal power flow using the moth flam
vesdPart I: local minimum analysis. IEEE Trans Power Syst 2015;30(2):
optimizer: a case study of the Algerian power system. Indonesian J Electrical
710e21.
Eng Comput Sci 2016;1(3):431e45.
[4] Elsayed W, Hegazy Y, Bendary F, El-Bages M. A review on accuracy issues
[34] Buch H, Trivedi IN, Jangir P. Moth flame optimization to solve optimal power
related to solving the non-convex economic dispatch problem. Elec Power
flow with non-parametric statistical evaluation validation. Cogent Engineer-
Syst Res 2016;141:325e32.
ing 2017;4(1):1286731.
[5] Elsayed WT, El-Saadany EF. A fully decentralized approach for solving the
[35] Trivedi IN, Jangir P, Parmar SA, Jangir N. Optimal power flow with voltage
economic dispatch problem. IEEE Trans Power Syst 2015;30(4):2179e89.
stability improvement and loss reduction in power system using Moth-Flame
[6] Chowdhury BH, Rahman S. A review of recent advances in economic dispatch.
Optimizer. Neural Comput Appl 2016:1e16.
IEEE Trans Power Syst 1990;5(4):1248e59.
[36] Mei RNS, Sulaiman MH, Mustaffa Z, Daniyal H. Optimal reactive power
[7] Liang Z-X, Glover JD. A zoom feature for a dynamic programming solution to
dispatch solution by loss minimization using moth-flame optimization tech-
economic dispatch including transmission losses. IEEE Trans Power Syst
nique. Appl Soft Comput 2017;59:210e22.
1992;7(2):544e50.
[37] Kamaruzaman AF, Zain AM, Yusuf SM, Udin A. Levy flight algorithm for
[8] Granelli G, Marannino P, Montagna M, Silvestri A. Fast and efficient gradient
optimization problems-a literature review. Conference Levy flight algorithm
projection algorithm for dynamic generation dispatching. Conference Fast and
for optimization problems-a literature review, vol. vol. 421. Trans Tech Publ,
efficient gradient projection algorithm for dynamic generation dispatching,
p. 496e501.
vol. vol. 136. IET, p. 295e302.
[38] Brown CT, Liebovitch LS, Glendon R. Le
vy flights in Dobe Ju/’hoansi foraging
[9] Wood AJ, Wollenberg BF. Power generation, operation, and control. John
patterns. Hum Ecol 2007;35(1):129e38.
Wiley & Sons; 2012.
[39] Pavlyukevich I. Le
vy flights, non-local search and simulated annealing.
[10] Zhang H, Yue D, Xie X, Hu S, Weng S. Multi-elite guide hybrid differential
J Comput Phys 2007;226(2):1830e44.
evolution with simulated annealing technique for dynamic economic emis-
[40] Pavlyukevich I. Cooling down Le
vy flights. J Phys Math Theor 2007;40(41):
sion dispatch. Appl Soft Comput 2015;34:312e23.
12299.
[11] Lin W-M, Cheng F-S, Tsay M-T. An improved tabu search for economic
[41] Shayeghi H, Ghasemi A. A modified artificial bee colony based on chaos theory
dispatch with multiple minima. IEEE Trans Power Syst 2002;17(1):108e12.
for solving non-convex emission/economic dispatch. Energy Convers Manag
[12] Amjady N, Nasiri-Rad H. Solution of nonconvex and nonsmooth economic
2014;79:344e54.
dispatch by a new adaptive real coded genetic algorithm. Expert Syst Appl
[42] Guo C, Zhan J, Wu Q. Dynamic economic emission dispatch based on group
2010;37(7):5239e45.
search optimizer with multiple producers. Elec Power Syst Res 2012;86:8e16.
[13] El-Sehiemy RA, El-Hosseini MA, Hassanien AE. Multiobjective real-coded ge-
[43] Attia AF, El Sehiemy RA, Hasanien HM. Optimal power flow solution in power
netic algorithm for economic/environmental dispatch problem. Stud Inf Contr
systems using a novel Sine-Cosine algorithm. Int J Electr Power Energy Syst
2013;22(2):113e22.
2018;99:331e43.
[14] Dash S, Mohanty S, Padhi A. Multi-objective economic emission load dispatch
[44] Xie J, Zhou Y, Chen H. A novel bat algorithm based on differential operator and
with nonlinear fuel cost and non-inferior emission level functions for a 30-bus
Le
vy flights trajectory. Comput Intell Neurosci 2013:2013.
IEEE test case system. 2014.
[45] Yang X-S, Deb S. Cuckoo search via Le
vy flights. Conference Cuckoo search via
[15] Venkatesh P, Gnanadass R, Padhy NP. Comparison and application of evolu-
Le
vy flights. IEEE, p. 210e214.
tionary programming techniques to combined economic emission dispatch
[46] Yang XS. Appendix A: test problems in optimization. Eng Optim 2010:261e6.
with line flow constraints. IEEE Trans Power Syst 2003;18(2):688e97.
[47] Kennedy J. Particle swarm optimization. Encyclopedia of machine learning:
[16] Basu M. Dynamic economic emission dispatch using nondominated sorting
Springer. 2011. p. 760e6.
genetic algorithm-II. Int J Electr Power Energy Syst 2008;30(2):140e9.
[48] Rashedi E, Nezamabadi-Pour H, Saryazdi SGSA. A gravitational search algo-
[17] Hetzer J, David CY, Bhattarai K. An economic dispatch model incorporating
rithm. Inf Sci 2009;179(13):2232e48.
wind power. IEEE Trans Energy Convers 2008;23(2):603e11.
[49] Yang X-S. A new metaheuristic bat-inspired algorithm. Nature inspired
[18] Lee J-C, Lin W-M, Liao G-C, Tsao T-P. Quantum genetic algorithm for dynamic
cooperative strategies for optimization (NICSO 2010): Springer. 2010.
economic dispatch with valve-point effects and including wind power system.
p. 65e74.
Int J Electr Power Energy Syst 2011;33(2):189e97.
[50] Yang X-S. Flower pollination algorithm for global optimization. Conference
[19] Rizk-Allah RM, El-Sehiemy RA, Wang G-G. A novel parallel hurricane opti-
Flower pollination algorithm for global optimization. Springer, p. 240e249.
mization algorithm for secure emission/economic load dispatch solution. Appl
[51] Cuevas E, Echavarría A, Ramírez-Ortego
n MA. An optimization algorithm
Soft Comput 2018;63:206e22.
inspired by the States of Matter that improves the balance between explo-
[20] Bentouati B, Chettih S, El-Sehiemy RA-A. A chaotic Krill Herd algorithm for
ration and exploitation. Appl Intell 2014;40(2):256e72.
optimal solution of the economic dispatch problem. Conference a chaotic Krill
[52] Younes M, Khodja F, Kherfane RL. Multi-objective economic emission dispatch
Herd algorithm for optimal solution of the economic dispatch problem, vol.
solution using hybrid FFA (firefly algorithm) and considering wind power
vol. 31. Trans Tech Publ, p. 155e168.
penetration. Energy 2014;67:595e606.
[21] Chiang C-L. Improved genetic algorithm for power economic dispatch of units
[53] Holland JH. Adaptation in natural and artificial systems: an introductory
with valve-point effects and multiple fuels. IEEE Trans Power Syst 2005;20(4):
analysis with applications to biology, control, and artificial intelligence. MIT
1690e9.
Press; 1992.
[22] Liu D, Cai Y. Taguchi method for solving the economic dispatch problem with
[54] Mirjalili S, Gandomi AH, Mirjalili SZ, Saremi S, Faris H, Mirjalili SM. Salp
nonsmooth cost functions. IEEE Trans Power Syst 2005;20(4):2006e14.
swarm algorithm: a bio-inspired optimizer for engineering design problems.
[23] Basu M. Modified particle swarm optimization for nonconvex economic
Adv Eng Software 2017;114:163e91.
dispatch problems. Int J Electr Power Energy Syst 2015;69:304e12.
[55] Basu M, Chowdhury A. Cuckoo search algorithm for economic dispatch. En-
[24] El Ela AA, El-Sehiemy RA. Different transmission loss allocation schemes in
ergy 2013;60:99e108.
deregulated power markets. In: Universities power engineering conference,
[56] Güvenç U, So €nmez Y, Duman S, Yo €rükeren N. Combined economic and
2007. UPEC 2007. 42nd international. IEEE; 2007. p. 634e40.
emission dispatch solution using gravitational search algorithm. Sci Iran
[25] Zare K, Haque MT, Davoodi E. Solving non-convex economic dispatch problem
2012;19(6):1754e62.
with valve point effects using modified group search optimizer method. Elec
[57] Modiri-Delshad M, Kaboli SHA, Taslimi-Renani E, Rahim NA. Backtracking
Power Syst Res 2012;84(1):83e9.
search algorithm for solving economic dispatch problems with valve-point
[26] Hemamalini S, Simon SP. Artificial bee colony algorithm for economic load
effects and multiple fuel options. Energy 2016;116:637e49.
dispatch problem with non-smooth cost functions. Elec Power Compon Syst
[58] Gaing Z-L. Particle swarm optimization to solving the economic dispatch
2010;38(7):786e803.
considering the generator constraints. IEEE Trans Power Syst 2003;18(3):
[27] Mirjalili S. Moth-flame optimization algorithm: a novel nature-inspired
1187e95.
heuristic paradigm. Knowl Base Syst 2015;89:228e49.
[59] Selvakumar AI, Thanushkodi K. A new particle swarm optimization solution to
[28] Trivedi IN, Kumar A, Ranpariya AH, Jangir P. Economic Load Dispatch problem
nonconvex economic dispatch problems. IEEE Trans Power Syst 2007;22(1):
with ramp rate limits and prohibited operating zones solve using Levy Flight
42e51.
Moth-Flame optimizer. In: Energy efficient technologies for sustainability
[60] Pothiya S, Ngamroo I, Kongprawechnon W. Application of multiple tabu
(ICEETS), 2016 international conference on. IEEE; 2016. p. 442e7.
search algorithm to solve dynamic economic dispatch considering generator
[29] Li C, Li S, Liu Y. A least squares support vector machine model optimized by
constraints. Energy Convers Manag 2008;49(4):506e16.
moth-flame optimization algorithm for annual power load forecasting. Appl
[61] Fesanghary M, Ardehali M. A novel meta-heuristic optimization methodology
Intell 2016;45(4):1166e78.
for solving various types of economic dispatch problem. Energy 2009;34(6):
[30] Raju M, Saikia LC, Saha D. Automatic generation control in competitive market
757e66.
conditions with moth-flame optimization based cascade controller. IEEE;
[62] Afzalan E, Joorabian M. An improved cuckoo search algorithm for power
1078 A.A. Elsakaan et al. / Energy 157 (2018) 1063e1078

economic load dispatch. Inter Trans Electrical Energy Syst 2015;25(6): [68] Bayo
n L, Grau JM, Ruiz MM, Sua
rez PM. The exact solution of the environ-
958e75. mental/economic dispatch problem. IEEE Trans Power Syst 2012;27(2):
[63] Bhattacharya A, Chattopadhyay PK. Biogeography-based optimization for 723e31.
different economic load dispatch problems. IEEE Trans Power Syst [69] Al-Betar MA, Awadallah MA, Khader AT. Bolaji ALa. Tournament-based har-
2010;25(2):1064e77. mony search algorithm for non-convex economic load dispatch problem. Appl
[64] Chaturvedi KT, Pandit M, Srivastava L. Self-organizing hierarchical particle Soft Comput 2016;47:449e59.
swarm optimization for nonconvex economic dispatch. IEEE Trans Power Syst [70] Niknam T, Mojarrad HD, Meymand HZ. Non-smooth economic dispatch
2008;23(3):1079e87. computation by fuzzy and self adaptive particle swarm optimization. Appl
[65] Shaheen AM, Farrag SM, El-Sehiemy RA. MOPF solution methodology. IET Soft Comput 2011;11(2):2805e17.
Generation. Transm Distrib 2017;11(2):570e81. [71] Selvakumar AI, Thanushkodi K. Optimization using civilized swarm: solution
[66] Abdelaziz A, Ali E, Elazim SA. Flower pollination algorithm to solve combined to economic dispatch with multiple minima. Elec Power Syst Res 2009;79(1):
economic and emission dispatch problems. Eng Sci Technol Inter J 2016;19(2): 8e16.
980e90. [72] Elsayed W, Hegazy Y, Bendary F, El-bages M. Modified social spider algorithm
[67] Shaw B, Mukherjee V, Ghoshal S. A novel opposition-based gravitational for solving the economic dispatch problem. Eng Sci Technol Inter J
search algorithm for combined economic and emission dispatch problems of 2016;19(4):1672e81.
power systems. Int J Electr Power Energy Syst 2012;35(1):21e33.