Journals of Advancement in Machines

Volume 4 Issue 2

Solution for Different Convex and Non-Convex Economic

Dispatch Problems for Large Scale System by Stochastic Search
Algorithm
K. Sarker1, B. Roy2, J. Sarker3, D. Santra4
1, 3
Assistant Professor, 2Student, 4Associate Professor
1, 2, 3, 4
Department of Electrical Engineering,
1, 2
Saroj Mohan Institute of Technology, Hooghly, West Bengal, India
3
Techno India, Salt Lake, Kolkata, India
4
RCC Institute of Information Technology, Kolkata, India
Email: [email protected]

Abstract
The main objective of the Economic Load Dispatch (ELD) is to determine the power by all
committed generating units so that generating cost is minimized, while satisfying the load
demand and inequality constraints. This paper represents an algorithm motivated by the ‘law
of gravity’ and interaction among the masses to solve ELD problems, called ‘Gravitational
Search Algorithm’. This proposed algorithm has been tested on some standard power systems
including 40-unit, 110-unit, 140-unit, 160-unit, 320-unit systems using different non-linear
effect like valve point loading, multi fuel option etc. This result obtained by using proposed
method compared with other methods shown less computational time over other existing
algorithm.

Keywords: Economic Load Dispatch, Gravitational Search Algorithm, Multi fuels, Valve
point loading effect

INTRODUCTION are complex and non-linear because of

Engineers are very much concerned about valve point loading effect [5, 6], ramp rate
the cost of products and services. In power limit [7], prohibited operating zones [8, 9]
system, minimization of production cost in etc.
any thermal unit is important. Economic
load dispatch is an important and one of Due to the restriction of the traditional
the fundamental optimization techniques techniques for non-convex characteristics
in power system operation. The aim of in solving ELD problems, there are huge
ELD problem is to reduce transmission optimization techniques which have been
loss, minimize generating cost of thermal effectively used to solving this problems
units and optimize the power generation at like Biogeography Based Optimization
particular load demand and subjected to (BBO) [2, 7], Bat Algorithm (BA) [9],
satisfying equality and inequality Improved Genetic Algorithm (IGA) [10],
constraints [1, 2]. A lot of examinations Pattern Search(PS) [11], Improved Particle
have been done on ELD problems till date Swarm Optimization (PSO) [12], Neural
[3]. The cost function of ELD problem is Network (NN) [13], Firefly algorithm
expressed as particular quadratic function (FA) [14], Cuckoo Search Algorithm
and the problem is resolved by numerous (CSA) [15], Real-Coded Genetic
derivative based techniques as Lagrangian Algorithm(RCGA) [16], Imperialist
multiplier method, lambda iteration Competitive Algorithm (ICA) [17] Flower
method, gradient method, Newton’s Pollination Algorithm (FPA) [18], Grew
method, etc., [4]. In reality, characteristics Wolf Optimiser (GWO) [19], Shuffled

23 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

Frog Leaping Algorithm (SFLA) [20], cost curves are nonlinear in nature [43].
Genetic Algorithm(GA) [21] etc. Hence, global optimization techniques,
such as the genetic algorithm (GA),
This paper is organised as following: particle swarm optimization (PSO),
Related work is discussed in section III. In simulated annealing (SA), firefly
section IV the basic ELD problem algorithm (FA) and gravitational search
formulation is discussed along with algorithm (GSA) have been studied in the
various linear and non-linear constraints. last three decades and have been
Section V briefly stresses on the basic successfully used to solve the ELD, though
features of Gravitational Search Algorithm each method has its own limitations and
(GSA) and optimization strategy of solution inaccuracies. However, GSA
proposed GSA algorithm. In section VI, being quite effective (with certain
the discussion of simulation outcomes for limitation) for ELD, till date no
five different test cases followed by experimentation has been done with GSA
comparison and analysis is presented. for the large scale power system
Finally, the study is concluded with optimization problems. Rather different
achievement and limitation of the present hybrid algorithms involving GSA has been
study in Section VII. experimented with ELD.

Motivation The aim of this work is to apply the GSA

The economic load dispatch (ELD) in its pure form to the large scale ELD
optimization problem is one of the problem and compare its effectiveness and
essential issues in power systems towards feasibility with other heuristic methods
acquiring optimal profits with the like the Biogeography-Based Optimization
reliability, stability and security [43]. (BBO) [2], oppositional invasive weed
Basically, the ELD problem is a optimization OIWO [23], cuckoo search
constrained optimization problem in power algorithm CSA [25], Oppositional based
generation systems that have the aim of grey wolf optimization algorithm OGWO
separating the total power demand among [28], Oppositional real coded chemical
different on-line generators which are reaction optimization ORCCRO [31],
usually sharing the generation load while Shuffled differential evolution (SDE) [33],
fulfilling equality and inequality Hybrid differential evolution with
constraints having complex and nonlinear biogeography-based optimization
characteristics. Consequently, good DE/BBO [35], hybrid hierarchical
solutions of the ELD problem would result evolution (HHE) [38] hybrid differential
in immense economical benefits for the evolution and harmony search (DE-HS)
power generation industry as well as the [39] etc. that have been applied to large
end users. scale ELD problem.

Over the years, many attempts have been Related Work

made to solve this problem with a variety In the past decades, a global optimization
of constraints and/or multiple objectives, technique GA is extensively used in
through various mathematical optimization problems. But it suffers from
programming and optimization techniques some disadvantages like using of complex
[44]. In the conventional methods, for operator for selection, taking long
example the lambda-iteration method, the computational time etc. Glover [45]
gradient methods etc., have necessary introduces TS algorithm which is a meta-
assumption that the fuel cost curves of the heuristic local search algorithm and it is
units are monotonically increasing piece- used to find improved solution from
wise linear functions, but practically the current solution. Due to the local and

24 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

random interaction between particles in test system, was compared with previous
swarm-based algorithm, PSO is widely methods and the conventional genetic
used in ELD problem. But it suffers from algorithm (CGA) with the MU
slow convergence and it gets trapped while (CGA_MU), revealing that the proposed
solving complex optimization problems. IGA_MU is more effective than previous
Bhattacharya et al., proposed approaches, and applies the realistic ED
biogeography-based optimization (BBO) problem more efficiently than does the
[2] to solve both convex and non-convex CGA_MU. Specially, the proposed
ELD problems with considering different method is extremely capable for the large-
constraints of thermal power plants. scale system of the real ED process.
Biogeography related with the
geographical distribution of biological In 2015, Barisal et al., [23] introduced
species are geographical distributed. Oppositional Invasive Weed Optimization
Mathematically biogeography illustrates (OIWO) by hybridizing IWO,
how a species starts, migrates from one to characterized by the colonizing behaviour
another environment and gets exhausted. of weed plants, and OBL empowered by
This algorithm explore for the global quasi opposite numbers. IWO is anovel
optimum primarily through two steps: population based stochastic, derivative-
migration and mutation. The effectiveness free optimization algorithm developed by
of the proposed algorithm has been Mehrabian and Lucas [46]. The algorithm
verified on 6, 10, 20 and 40-unit test exploits some of the interesting
systems. The proposed technique has been characteristics of weed plants namely fast
established to perform better in a number reproduction, distribution and self
of cases. adaptation to the changes in climatic
conditions. OBL assists to reach global
Ghasemi [6] proposed multi objective optima of IWO which in turn reaches
interactive honey bee mating optimization encouraging results of ELD problem.
(IHBMO) in 2013 for large scale nonlinear
EELD with Valve Point Effect, that is non- Sahoo et al., [25] in 2015 Cuckoo Search
convex problem. In this technique, Pareto algorithm (CSA) was applied to non-
dominance idea is used to produce and sort convex economic load dispatch problems.
the dominated and non-dominated To verify the robustness of the proposed
solutions. Furthermore, fuzzy set theory is Cuckoo Search based algorithm, multiple
engaged to extract the best compromise constraints are also incorporated in the
solution. The propose method has been system. In this algorithm, the levy flights
individually examined and applied to the and the behavior of alien egg discovery is
6, 14 and 40 unit test systems. used to search the optimal solution. In
comparison with the solution, quality and
In 2005, Chiang [10] presented an execution time obtained for 6, 15, 40,140,
improved genetic algorithm with 320-unit test system, the algorithm seems
multiplier updating (IGA_MU) to solve to be a promising method to solve realistic
non-convex ED problems. The IGA dispatch problems.
arranged with an improved evolutionary
direction operator and a migration Mandal et al., in 2014 represented a novel
operation can competently search and and efficient krill herd algorithm (KHA)
actively explore solutions, and the MU is [27] to solve both convex and non-convex
in use to handle the equality and inequality multi-constraint ELD problems. To
constraints of the PED problem. The improve the overall effectiveness and
advantages of the method, which was performance of the proposed algorithm,
applied to 10, 13, 20, 40, 80, and 160-units the crossover and mutation operation of

25 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

differential evolution (DE) are In 2013, Reddy et al., proposed a hybrid

incorporated with the proposed method. shuffled differential evolution (SDE) [32]
The different versions of KHA are algorithm which combined the benefits of
successfully applied to different scale like shuffled frog leaping algorithm (SFLA)
6, 15, 40, 80-unit power systems for and DE to solve non-convex ELD problem
solving different ELD problems. with transmission loss. SDE, characterized
by a fresh mutation operator, mixes the
Pradhan et al., presented [28] oppositional exploration and exploitation mutation
based grey wolf optimization (OGWO) capabilities of DE and therefore helps
algorithm in 2017 for ELD problem. The conquer the inherent limitation of DE and
proposed algorithm merges two basic SFLA in solving large scale non-convex
ideas. Initially, the hunting behaviour and ELD.
social hierarchy of grey wolves are used to
search optimal solutions and then, In 2010, Bhattacharya et al., suggested
oppositional concept is integrated with the combination of differential evolution (DE)
grey wolf optimization (GWO) algorithm
and with biogeography-based optimization
to accelerate the convergence speed of the
(BBO) [34] used to solve non-convex
conventional GWO algorithm. The
ELD. This hybrid technique improves the
performance of the algorithm has been
searching capability of DE employing
measured on being applied on small,
BBO algorithm successfully and can
medium and large scale test systems for
solving of 13, 40 and 160-unit ELD produce the promising candidate solutions.
problems. In this paper, the migration operator of
BBO together with mutation, crossover
In 2013, Roy et al., presented [29] and selection operators of DE have been
teaching learning based optimization combined together to effectively develop
(TLBO) for non-convex ELD problem the decency of both DE and BBO to
considering VPL effect. TLBO is based on improve the convergence speed and to
two concepts of education that is teaching improve quality of solution. This
phase and learning phase. In the technique has been applied on customary
beginning, learners get knowledge during 40-unit test case system with valve point
the teaching methodology of teacher and effect, i.e., non convex solution.
lastly learners enhance their knowledge by
communications among themselves. Vishwakarma et al., in 2012 presented a
TLBO technique is a promising method simulated annealing (SA) approach [35]
for solving ELD. for solving ELD problems. Global
optimization approach is motivated by
In 2014, Bhattacharjee et al., [31] have annealing process of thermodynamics. The
represented the oppositional real coded proposed technique works very fast, this
chemical reaction optimization (ORCCRO) feature of algorithm is striking when
by fitting in the concept of quasi-opposition applied for a large ELD system.
based learning (QOBL) in RCCRO inspired Simulation has been presented over two
by the computational efficiency of QOBL. different cases as 38, 110-units, both cases
Effectively, the convergence speed of having convex fuel cost characteristics.
RCCRO gets accelerated by comparing the The results produced by the method have
fitness of a solution approximate to its been compared with other approaches and
opposite and retaining the fitter one in the finally SA proves glowing feasibility,
arbitrarily chosen population set. ORCCRO robustness and fast convergence for
has been successfully tested on non-convex optimization of ELD problems.
large scale ELD problems.

26 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

In 2013, Kim et al.[36] implemented units and compared with similar works
Mean-Variance Optimization (MVO) reported in recent literature in terms of
algorithm with Kuhn Tucker condition and solution quality and computational
swap process to improve a global efficiency which showed encouraging
minimum searching capability for solving results, suggesting that the proposed
economic dispatch (ED) problems with approach was capable of efficiently
non-convex cost functions. Kuhn-Tucker determining higher quality solutions of ED
condition is a mathematical judgment for problems.
heuristic method, i.e., an experience-based
technique for solving the problem. Swap Wang et al., proposed [42] estimation of
process is a technique that changes the distribution and differential evolution
generator outputs in the direction of cooperation (ED-DE) in 2010, which is a
reducing total generation cost for consecutive hybrid of two efficient
searching local minimum. These two evolutionary computation (EC) techniques:
methods are applied to MVO algorithm so estimation of distribution and differential
that optimal solution can be obtained. evolution. The advantages of ED-DE over
the previous ELD optimization algorithms
Khoa et al., presented swarm based mean- are experimentally confirmed on non-
variance mapping optimization (MVMO) convex ELD problems of 10, 20, 40, 80
[37] in 2015 for solving ED problem. The and 160-unit ELD problems. To further
proposed technique is the extension of the evaluate the efficiency and effectiveness of
original single particle mean-variance ED-DE, this algorithm is compared with
mapping optimization (MVMO). The other recent published evolutionary
original feature is the special mapping algorithms on typical function
function applied for the mutation based on optimization tasks showing better result
the mean and variance of n-best than others.
population. The MVMOS do better than
the classical MVMO in global search GSA [9] is a meta-heuristic algorithm
ability due to the development of the based on Newton’s law related to gravity
mapping. The proposed MVMOS is study and motion. According to Newton, every
on 3, 13, 20-units and large-scale 140 units object in the universe attracts each other
system and the achieved results are by the gravitational force. The magnitude
compared with many other known of the force is directly proportional to the
methods given in the paper. Test results product of masses and inversely
prove that the proposed technique can proportional to the square of the distance
competently apply for solving ED between them. The direction of movement
problem. of the particle occurs towards the particle
of higher masses. In GSA every object in
In 2008, Kuo proposed [41] a new the universe is treated as an agent. The
approach and coding scheme for solving heavier masses move more slowly than the
ELD problem through simulated annealing lighter one. This ensures the exploitation
like particle swarm optimization (SA- steps of this algorithm.
PSO). This novel coding scheme could
effectively prevent obtaining feasible This paper presents a heuristic approach
solutions through the application of based on GSA for the solution of
stochastic search methods, thereby economic power dispatch with non-linear
dramatically improving search efficiency constraints. GSA is applied to standard test
and solution quality. The effectiveness and system of 40-unit, 110-unit, 140-unit, 160-
feasibility of the proposed method were unit, and 320-unit. For more realistic
verified by applying it to 6, 13, 15, 40- solution, loss co-efficient and different

27 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

non-linear factors like valve point loading maximum value power of the ith
effects and with multi-fuel source are transmission line.
considered. This MATLAB simulation
result has been compared by many well- Real Power Balance Constraint
known heuristic search algorithms, among In this world, power generated PGi by the
which the result of our proposed technique generators must be equal to the sum of
is better. power demand PD by the consumers and
total power loss PL in the transmission line.
ECONOMIC LOAD DISPATCH That is expressed by the following
PROBLEM equation:
The ELD problem is a nonlinear n
 PGi  PD  PL  0
programming optimization technique [3]. i 1 (3)
The main objective of ELD is to minimize
the fuel cost by generating real power As the total power loss function of power
output for a specific period of operation generation, so it can be calculated by
while satisfying several equality and solving the power equation as follows:
inequality constraints. Two models of m m m (4)
PL    PGi Bij PGj   B0i PGi  B00
ELD are considered, viz. convex ELD i 1 j 1 i 1
problem which assumes the quadratic cost
function along with the system load Where, Bij is the ijth element of loss co-
demand and non-convex ELD problem efficient square matrix, B0i is mathematical
which contains generator nonlinearities model can be mathematically described as
such as valve point loading effects, ramp follows.
rate limits, prohibited zones and multi-fuel
options. Both convex and non-convex ELD with valve point loadings
ELD problems are discussed in this paper. The total cost FTotal of power generation in
Generally, ELD mathematical model can any thermal unit is expressed by equation
be mathematically described as follows. (1). As ELD problem with valve point
loading introduces ripple in the heat-rate
IELD with Quadratic Cost Function curve, so it becomes complex. The model
The total cost of ELD problem may be of valve point loading has been discussed
written as: by introducing a sinusoidal function with
(1) the quadratic equation [21, 22]. The
n
 i 1
   min i1 ai  bi PGi  ci PGi2 
n
FTotal  min  Fi PGi variation of fuel cost Fi (PGi) due to effect
of valve point loading with the change of
Where Fi(PGi) is the ith generator cost generated output power PGi is shown in
function and generally expressed as Fig. 2.The actual cost function with valve
quadratic equation, ai, bi and ci are the cost point is given by equation (5).
coefficient of ith generator, n is the
generator connected to the system, PGi is

n
i 1
2 min

FT  min  ai  bi PGi  ci PGi  ei  sin  fi   PGi  PGi  (5)

the power output of the ith generator.

ELD with VPL effects and multi-fuel
Generator Capacity Constraints options
The power generated by each generator For m number of generator and NF fuel
must be in between the maximum and option for each generator the cost function
minimum value as follows: with valve point and multi-fuel options is
i  1, 2, 3, ...., n
min max
PGi  PGi  PGi (2) expressed by equations (6) and (7), where
min PGikmin and PGikmax are the minimum and
PGi is the minimum value below which maximum power generation limit of the ith
it becomes uneconomical and PGimax is the generator with fuel option k respectively.

28 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

min max
if PGik  PGi  PGik forfuel option k; k=1,2,...N F
(6) passive gravitational mass. The solution of
the problem is represented by the position
F
T
2
  min
 aik  bik PGi  cik PGi  eik  sin fik  PGik  PGik  (7) of the respective mass. Good solution and
worst solution are represented by heavier
and lighter mass respectively. Two well-
For more realistic representation of
known equations used in GSA are:
generating unit, each unit is supplied from M 1M 2
different fuel sources. The main aim is to F=G 2
(9)
R
get accurate and practical economic
And equation of acceleration of a particle
dispatch solution by combining sinusoidal
when a force applied to it, written as:
and quadratic function. The generator with
F
multi-fuel source becomes more a= (10)
M
appropriate when cost function with
Gravitational constant value G(t) is
piecewise quadratic function is represented
represented as
by (8) below. 
t 
G  t   G  t0    0  (11)
 a  b P  c P 2  e  sin
 i1 i1 Gi i1 Gi i1
  min
fi1 PGi1  PGi1   t 
 if P min  P  P , fuel-1 In the above equations, M1 and M2 are two
 Gi1
  
Gi Gi1
2 min different masses, F represents force, a is
ai 2  bi 2 PGi  ci 2 PGi  ei 2  sin fi 2 PGi 2  PGi 2
Fi  PGi  = 
acceleration, R represents the distance
if P  P  P , fuel-2
 Gi1 Gi Gi 2 (8) between two masses, t is the actual time
 2
 ain  bin PGi  cin PGi  ein  sin   min
fin PGin  PGin 
and G(t0) is the value of the gravitational
constant at the initial time,t0 respectively.
 max
 if PGin 1  PGi  PGi , fuel-n β <0.
Active gravitational mass (Ma), passive
METHODOLOGY gravitational mass (Mp) and inertial mass
Gravitational Search Algorithm (Mi) are defined in physics.
Till date a number of evolutionary The equation representing the decrease in
algorithms have been studied in power gravitational constant can be represented
system to obtain the optimal solution. as
t
Among them, GSA is a newer technique, 
having capability to handle the multi- G  t   G0e T (12)
dimensional problem. GSA has been Where, α is a user specified constant, T is
implemented up till now on limited the total number of iteration, and t is the
number of power system problems, such as current iteration. If ith active and passive
for post-outage bus voltage magnitude gravitational masses are equal, then
calculations, combined economic and        and for i  1, 2, , 
ai pi ii i
emission dispatch problems of power
systems, optimal power flow and number of masses, these gravitational
parameters identification of hydraulic masses can be represented in terms of their
turbine governing system, multi–objective respective fitness values and the equations
economic emission load dispatch, and can be represented as:

fitnessi t  worst t  ,
solution of unit commitment problem. 
mi t 
best  t   worst  t 
(13)
mi  t 
GSA is based on Newtonian law of
gravity. In this algorithm, the solutions are i t  
 
j 1 m j t
analyzed in terms of masses of respective
agents. Each mass has their own position, The total force acting on mass i in d
inertial mass, active gravitational mass and dimensions may be represented as,

29 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

Fi
d
t   j Kbest , j 1 rand j Fijd t 
(14)
th
The acceleration in d dimension, velocity
Where, r and j is the random number ( v ) and position (x) at time (t+1) of object
between 0 and 1, Kbest is the set of first K i may be expressed as:
objects with the best fitness value and vid  t  1  randi  vid  t   aid , xid t  1  xid  vid t  1
(15)
biggest mass, Fijd is the force on mass i Where, r and 1 is the random number
from mass j in d dimensions. between 0 and 1.
The procedural steps of ELD problem
solution with GSA is shown in Fig. 1.

Figure 1: Flow chart of GSA algorithm.

Results and Discussion Case Study 1: 40 generator test system

The proposed GSA has been applied to with non-convex constraints
solve ELD problems of different test For 40-unit test, system transmission loss
systems to demonstrate its performance in with VPL has been considered. The system
comparison to several established load demand is 10500 MW. The input data
optimization techniques reported in and B- loss coefficients are taken from [6,
literature. The GSA has been implemented 11, 23]. We have calculated cost and loss
using Matlab-8.1 environment on a core for different operating conditions. The best
i5/4GB/500 GB/Win 8.1 personal power output for 40 generator system
computers. Details simulation results could considering transmission loss with VPL
not be presented due to page limitation. compare with other recently published
methods shown in Table 1.

30 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

Table 1: Comparison between different methods taken after 50 trials (40-generators system)
Load Demand-10,500 MW.
Methods Power (MW) Ploss (MW) Cost ($/h)
Proposed GSA 11425.5429 925.5429 136445.452
OIWO [27] 11457.2965 957.2965 136452.68
KHA[27] 11478.9251 978.9251 136670.370
OGWO[28] 11474.43 974.43 136440.62
TLBO [29] 11502.63 1002.63 137814.17
QOTLBO[29] 11508.96 1008.96 137329.86
GAAPI[30] 11545.06 1045.06 139864.96
ORCCRO[31] 11458.75 958.75 136855.19
SDE[32] 11474.43 974.43 138157.46
OCRO[33] 11468.9607 968.9607 136563.48
DE/BB0[34] 11457.83 957 136950.77

From this Table 1, it can be seen that shown is better. The convergence
proposed method in terms of total cost and characteristics loss withVPL and is
loss are calculated for 40 generating units depicted in Fig. 2, which is only 7 nos.
compared with different algorithms as iterations have been made.

5
x 10 Total Fuel Cost=136445.452 ($/hour) with 40 Generators
Total Generation Cost ($/hour)

2.4

2.2

1.8

1.6

1.4
0 10 20 30 40 50
Number of Iterations

Figure 2: Convergence characteristics of 40 generatorssystem with transmission loss VPL.

Case Study 2: 110 generator test system cost function minimum fuel cost is
The input system data of 110 generating units 197983.453 $/h obtained by GSA algorithm
for the load demand of 15000 MW is taken and compared to other method for ELD gives
from [23].The power output with quadratic better result is shown in Table 2.

Table 2: Comparison between different methods taken after 50 trials (110-generators system)
Load Demand-15,000 MW.
Methods Best costs($/hr) Time(s) No. of hits
IHBMO [6] 197,989.1358 - -
OIWO[23] 197,989.1358 31 -
ORCCRO [31] 198,016.29 0.15 48
DE/BBO [34] 198,231.06 0.46 43
SA [35] 198,352.6413 - -
Proposed GSA 197983.4538 0.33 20

Case Study 3: 140 generator test system algorithm gives better result than the result
The input data are taken from [24].The obtained by other methods. This system
proposed algorithm provides total cost of consisted of 140 thermal units in
1557461.7934 $/hr for load demand of comparison of generation cost as given in
49342 MW. In terms of results, GSA Table 3.

31 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

Table 3: Comparison between different methods taken after 50 trials (140-generators

system) Load Demand-49342MW.
Methods Costs($/h) Time (s) No. of hits
KMVO[36] 1568,450.94 - -
MVMOS[37] 1557461.803 - -
HHE[38] 1 655 679.4116 - -
OGWO[28] 1559709.97 - -
DEHS[39] 1655679 - -
SDE [32] 1,560,236.85 - -
OIWO[23] 1559405.4669 46.8 46
CTPSO[40] 1655685 50.1 -
DEL[24] 1657962.7166 - 30
CSA[25] 1559547.4708 26.37 -
CCPSO [41] 1655685 42.9 -
KMVO [37] 1577607 - -
Proposed GSA 1557461.7934 22 50

Case Study 4: 160 generator test system generating level of each generator is not
In Table 4, system consists of 160 given in Table due to page limitation.
generator systems with valve point loading Transmission loss is not considered in this
effect and multi-fuel options. The load case study. The best generating cost using
demand is considered as 43200 MW. The proposed algorithm is 9978.8593 $/hour
input data of 10 units are replicated up to and this result is quite better than the result
160 units and are taken from [23]. The obtained other algorithm shown in Table 4.

Table 4: Comparison between different methods taken after 50 trials (160-generators

system).
Methods Best Cost ($/h) Time (s) No. of hits
OIWO[23] 9981.9834 17.3 46
ORCCRO [31] 10,004.20 19 48
BBO [2] 10,008.71 44 40
DE/BBO [34] 10,007.05 35 42
ED-DE [42] 10,012.68 - -
CGA-MU [10] 10,143.73 - -
IGA-MU [10] 10,042.47 - -
CSA [25] 9982.085 29.97 -
Proposed GSA 9978.8593 15.3 50

Case Study 5: 320 generator test system transmission loss is not considered in
320 thermal units are considered with the cost function. The cheapest
valve point loading effect and the load generating cost using proposed method
demand is set up to 86400 MW. The is 19925.4291$/hr. The result is quite
input data of 10 units are replicated up better than result obtained by using CSA
to 160 units and 320 units [23].The [25] shown in Table 5.

Table 5: Comparison between different methods taken after 50 trials (320-generators

system).
Methods Best Cost ($/h) Time (s) No. of hits
CSA[25] 19964.171 59.82 -
Proposed GSA 19925.4291 40 15

32 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

CONCLUSION 6. Ghasemi A, (2013), “A Fuzzified

GSA algorithm has been proposed for Multi Objective Interactive Honey Bee
solving ELD problems. In the proposed Mating Optimization for
algorithm, GSA avoids the divergence of Environment/Economic Power
the agents in the search space. The GSA Dispatch with Valve Point Effect”,
algorithm is used to explore the search and Electrical Power and Energy System,
tunes the control variables in order to find Volume 49, Issue 1, pp. 308321.
global optimal solution. The results 7. Agrawal N., Agrawal S., Swarnkar
obtained show the effectiveness of the
K.K., Wadhwani S., Wadhwani A.K.,
proposed algorithm over other well
(2012), “Economic Load Dispatch
established optimization techniques
applied in modern power system. This Problem with Ramp Rate Limit using
comparison results reveal the BBO”, International Journal of
effectiveness, robustness, high quality Information and Education
solution, feasibility, stable convergence Technology, Volume 2, Issue 5, pp.
characteristics and good computation 419424.
efficiency of the proposed GSA technique. 8. Gherbi Y. A., Bouzeboudja H., Lakdja
F. (2014), “Economic dispatch
REFERENCES problem using bat algorithm”,
1. Gaing, Z L (2003), “Particle swarm Leonardo Journal of Sciences, Volume
optimization to solve the economic 13, Issue 24, pp. 7584.
dispatch considering the generator 9. Chiang C. L., (2005), “Improved
constraints”, IEEE Transactions on genetic algorithm for power economic
Power System, Volume 18, Issue 3, pp. dispatch of units with valve-point
11871195. effects and multiple fuels”, IEEE
2. Bhattacharya, A, Chattopadhyay, P K, Transactions on Power Systems,
(2010), “Biogeography Based Volume 20, Issue 4, pp. 16901699.
Optimization for Different Economic 10. Al-Sumait, J.S., AL-Othman, A.K.,
Load Dispatch Problems”, IEEE Sykulski, J.K. (2007), “Application of
Transactions on Power System, pattern search method to power system
Volume 25, Issue 2, pp. 1064–1077. valve-point economic load dispatch”,
3. Sinha N., Chakrabarti R., International Journal of Electrical
Chattopadhyay P K (2003), Power & Energy Systems, Volume 29,
“Evolutionary Programming Issue 10, pp. 720–730.
Techniques for Economic Load 11. Hota, P. K., Barisal, A. K.,
Dispatch”, IEEE Transactions on Chakrabarti, R. (2009), “An improved
Evolutionary Computation, Volume 7, PSO
Issue 1, pp. 8394. technique for short-term optimal
4. Wood A J, Wollenberg B F (1996), hydrothermal scheduling”, Electric
“Power Generation, Operation and Power
Control”, John Wiley & Sons, System Research, Volume 79, Issue 7,
5. Duman S, Yorukeren N, Atlas I H, pp. 1047–1053.
(2015), “A Novel Modified Hybrid 12. Mohammadi, A., Varahram, M. H.
PSOGSA Based on Fuzzy Logic for (2006), “Using neural network for
Non-Convex Economic Dispatch solving of on-line economic dispatch
Problem with Valve-Point Effect”, problem”, Proceedings of
Electrical Power and Energy System, International Conference on Intelligent
Volume 64, Issue 1, pp. 121135. Agents for Modelling, Control and

33 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

Automation, pp. 8187. Society General Meeting, San Diego,

10.1109/CIMCA.2006.228 CA, pp. 18.
13. Yang, X. S., Hosseini, S. S. S., 20. Walter D.C, Shebla G.B., (1993),
Gandomi, A. H. (2012), “Firefly “Genetic Algorithm Solution of
algorithm for solving non-convex Economic Load Dispatch with Valve
economic dispatch problems with Point Loading”, IEEE transactions on
valve loading effect”, Applied Soft power systems, Volume 8, Issue 3, pp.
Computing, Volume 12, Issue 3, pp. 13251332.
11801186. 21. Deslshad M. M., Rahim N. A., (2016),
doi:10.1016/j.asoc.2011.09.017 “Multi-Objective Backtracking Search
14. Hima Bindu A., Reddy M. D, (2013), Algorithm for Economic Emission
“Economic Load Dispatch using Dispatch Problem”, Applied Soft
Cuckoo Search Algorithm”, Int. Computing, Volume 40, Issue 1, pp.
Journal of Engineering Research and 479494.
Apllications, Volume 3, Issue 4, pp. 22. Barisal A.K., Prusty R.C., (2015),
498502. “Large scale economic dispatch of
15. Prathiba R., Balasingh M M., Sakthivel power systems using oppositional
S., (2014), “Flower pollination invasive weed optimization”, Applied
algorithm applied for different Soft Computing, Volume 29, Issue 1,
economic load dispatch problems”, pp. 122137.
International Journal of Engineering 23. Coelho L. D. S., Bora T. C., Mariani
and Technology, Volume 6, Issue 2,pp. V. C., (2014), “Differential Evolution
10091016. Based On Truncated Levy-Type
16. Abedinia O., Amjady N., Kiani K., Flights and Population Diversity
Shayanfar H. A., Ghasemi A., (2012), Measure to Solve EconomicLoad
“Multiobjective environmental and Dispatch Problems”, Electrical Power
economic dispatch using imperialist and Energy System, Volume 57, Issue
competitive algorithm”, International 1, pp.178188.
Journal on Technical and Physical 24. Sahoo S., Dash K. M., Prusty R. C.,
Problems of Engineering (IJTPE), Barisal A. K. (2015), “Comparative
Volume 11, pp. 6370. Analysis of Optimal Load Dispatch
17. Kaveh A., Talatahari S., (2010), “A through Evolutionary Algorithms”, Ain
Novel Heuristic Optimization Method: Shams Engineering Journal, Volume
Charged System Search, Acta Mech, 6, Issue 1, pp.107120.
Volume 213, Issue 34, pp. 267289. 25. Hota P.K., Barisal A.K., Chakrabarti
18. Wong L. I., Sulaiman M. H., R., (2010), “Economic emission load
Mohamed M. R., Hong M. S. (2014), dispatch through fuzzy based bacterial
“Grey Wolf Optimizer for solving foraging algorithm”, Electr Power
economic dispatch problems”, IEEE Energy Syst, Volume 32, Issue 7, pp.
International Conference on Power 794–803.
and Energy (PECon), Kuching, pp. 26. Mandal B., Roy P.K., Mandal S.,
150154. (2014), “Economic load dispatch using
19. E. Sayedi, M. M. Farsangi, M. Barati, krill herd algorithm”, Int J Elect Power
K. Y. Lee (2012), “A modified Energy Syst, Volume 57, pp. 1–10.
Shuffled Frog Leaping algorithm for 27. Pradhan M, Roy P. K., Pal T., (2017),
nonconvex Economic Dispatch “Oppositional based grey wolf
problem”, IEEE Power and Energy optimization algorithm for economic

34 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

dispatch problem of power system”, Ain J.H., Lee S. U., Son S.Y., (2013), “An
Shams Engineering Journal, pp. 1–11. Improved Mean-Variance
https://doi.org/10.1016/j.asej.2016.08.0 Optimization for Non-convex
23,(in press) Economic Dispatch Problems”, J
28. Roy P.K., Bhui S., (2013), “Multi- ElectrEng Technol., Volume 8, Issue 1,
objective quasi-oppositional teaching pp. 8089.
learning based optimization for 36. Khoa H. T., Vasant P., Singh B., Dieu
economic emission load dispatch N. V.,(2015), “Solving Economic
problem”, Int J Elect Power Energy Dispatch By Using Swarm Based
Syst, Volume 53, Issue 1, pp. 937–948. Mean-Variance Mapping
29. Ciornei I., Kyriakides E., (2012), “A Optimization”, Global J Technol
GA-API solution for the economic Optim, Volume 6, Issue 3, pp. 18.
dispatch of generation in power system 37. Pandit M., Srivastava L., Sharma M.,
operation”, IEEE Trans. on Power Dubey H. M., Panigrahi B. K., (2014),
System, Volume 27, Issue 1, pp. “Large-scale multi-zone optimal power
233242. dispatch using hybrid hierarchical
30. Bhattacharjee K., Bhattacharya A., evolution Technique”, The Journal of
Dey S.H., (2014), “Oppositional real Engineering, Volume 2014, Issue 3,
coded chemical reaction optimization pp. 71–80.
for different economic dispatch 38. Dinh L. L., Ngoc D. V., Nguyen T.H.,
problems”, Int. J. Electr. Power Anh D.L., (2013), “A hybrid
Energy Syst., Volume 55, Issue 1, pp. differential evolution and harmony
378–391. search for non-convex economic
31. Reddy A. S., Vaisakh K., (2013), dispatch problems”, In the Proceedings
“Shuffled differential evolution for of the PEOCO, pp. 238243.
large scale economic dispatch”, Electr. 39. Park J. B., Jeong Y.W., Shin J.R., Lee
Power Syst. Res, Volume 96, Issue 1, K. Y., (2010), “An Improved Particle
pp. 237–245. Swarm Optimization for Non-convex
32. Hazra S., Roy P. K., Sinha A., (2015), Economic Dispatch Problems”, IEEE
“An efficient evolutionary algorithm transactions on power systems,
applied to economic load dispatch Volume 25, Issue 1, pp. 156166.
problem”, In the Proceedings of the 40. Kuo C.C., (2008), “A novel coding
CCC and IT, Hooghly, India, pp. 16. scheme for practical economic
33. Bhattacharya A., Chattopadhyay P. K., dispatch by modified particle swarm
(2010), “Hybrid differential evolution approach”, IEEE Transactions on
with biogeography-based optimization power systems, Volume 23, pp.
for solution of economic load 18251835.
dispatch”, IEEE transactions on power 41. Wang Y., Bin L., Weise T., (2010),
systems, Volume 25, Issue 4, pp. “Estimation of distribution and
19551964. differential evolution cooperation for
34. Vishwakarma K.K., Dubey H.M., large scale economic load dispatch
(2012), “Simulated annealing based optimization of power systems”,
optimization for solving large scale Inform. Sci., Volume 180, Issue 12, pp.
economic load dispatch problems”, Int. 2405–2420.
J. Eng. Res. Technol.(IJERT), Volume 42. Selvakumar, A.I., K. Thanushkodi.
1, Issue 3, pp.1–8. (2008), “Anti-predatory particle swarm
35. Kim M.J., Song H.Y., Park J.B., Roh optimization: Solution to non-convex

35 Page 23-36 © MAT Journals 2019. All Rights Reserved

 Journals of Advancement in Machines
Volume 4 Issue 2

economic dispatch problems”, Electric 44. Glover F.(1989), “Tabu search - Part
Power System Research, Volume 78, I”, ORSA Journal of Computing,
Issue 1, pp. 2–10. Volume 1, Issue 1, pp. 190206.
43. Panigrahi, B.K., V.R. Pandi, S. Das. 45. Mehrabian A.R. and Lucas C. (2006),
(2008), “Adaptive particle swarm “A novel numerical optimization
optimization approach for static and algorithm inspired from weed
dynamic economic load dispatch”, colonization”, Ecol. Inform., Volume
Energy Conversion and Management, 1, Issue 4, pp. 355366.
Volume 49, Issue 6, pp. 1407–1415.

36 Page 23-36 © MAT Journals 2019. All Rights Reserved