18-01-2013, 02:59 PM
A Genetic Algorithm Based Economic Load Dispatch Solution for Eastern Region of EGAT System having Combined Cycle and Cogeneration Plants
[attachment=47541]
Abstract:
This paper has used Genetic Algorithms (GAS)
to solve economic load dispatch (ELD) problem for eastem
region of Electricitv Generating Authority of Thailand
(EGAT) svstem. A piecewise model of cost characteristics
has been derived for the combined cycle and the
cogeneration power plants in the svstem and utilized for
the ELD formulation. Results of ELD have been obtained
bv using a conventional GA and a Micro GA models. Both
the models effectively solve the ELD problem and reduce
the operating cost. However, the Micro GA model ot'fers
faster convergence.
Keywords: Economic load dispatch, Genetic algorithms,
Cogeneration plants, Combined cycle plants
INTRODUCTION
Economic load dispatch (ELD) is one of the
main functions of modem Energy Management
Svstem (EMS) which determines the optimal real
power settings of generating units with an objective
to minimize total fuel cost of thermal plants. Several
conventional algorithmic methods have been used
11.21 to solve the ELD problem. However these
methods require the objective function in continuous
differentiable form and sometimes fail to provide
global minima. The generator fuel costs are
generally approximated in the continuously
Merentiable quadratic form.
Some power plants, such as combined cycle
and cogeneration plants, have more than one type of
turbine operation. The open cycle operation of the
gas turbine units causes the cost characteristics of
such plants to be, generally. non smooth and
continuously non-differentiable. Genetic Algorithms
(GAS) have been found suitable to effectively handle
the non-continuous or nondlfferentiable objective
functions [3.5].
GAS are optimization methods employing
search process imitated from the mechanism of
biological selection and biological genetics [3]. They
combine survival of the fittest among those feasible
solutions in the form of string structures (or genes :
mostly in binary form) with a randomized
information exchange to form a search algorithm. In
every generation, a new sct of string solutions is
created froin thc fittest of the old string solutions set.
+ 011 leave frnm Deptt. oFElect. En%.. 117' Kanprrr. India
IEEE Catalogue No: 98EX137
0-7803-4495-2/98/$10.00 1998 IEEE 165
Whle randomized, GAS are no simple
random walk. They efficiently use historical
information to speculate on new search points with
e'xpected improved performance
The present work has considered the ELD
studies on the eastern regon of EGAT system which
has a large amount of thermal plants including
cogeneration and combined cycle plants. Genetic
Algorithms have been selected for the ELD solution.
Three loading cases in this region viz light load. day
load and peak load have been considered. Two
models of GAS, a conventional GA and a Micro GA
[9] have been utilizedl. An experimentation to select
optimum values of some of the GA parameters have
also been conducted.
GENETIC ALGORITHMS
. General
GAS are derived from a simple model of
population genetics. They have followmg five
components [6].
i) Stmg represeintation
ii) An initial popiulation of strings.
iii) Evaluation funiction or fitness function
iv) Three GA Operators : Reproduction,
Crossover and Mutation
v) Value of the GA operator and other
parameters
Instead of explloiting the variables as actual
parameters, the coded binary strings are often used
for the variables' representation. The length of stnng
depends on the precision required. Before starting
the iterative routine of the GA, a population of
strings must be initially randomized. GA considers
only a single evaluation function or fitness function.
The performance of each string is evaluated
according to its fitness. calculated by the evaluation
function.
Three basic operators used in a conventional
GA are Reproduction. Crossover and Mutation. The
more fitness string should have more chance to be
copied or selected to the next generation. The biased
Roulette wheel IS usedl to achieve the "survival of the
tittest"-aspat of GAS For the population size of .V.
probability of reproduction of each ith string with
fitnessfi i.e. Preprod.; , is
After the new generation population is
copied or selected by Reproduction routine, the
randomized couple of strings are picked up to offer
'crossover'.
Although Reproduction and Crossover
effectively search and recombine existing
chromosomes. they do not create any new genetic
matenal in the population. Mutation is capable of
overcoming this shortcoming. The Mutation is to
change any string bit from 1 to 0 or 0 to 1.
Like the other stochastic methods , GAS
require a number of parameters which are population
size. probability of crossover. probability of mutation
that must be selected. Usually relatively small
population size, high Crossover probability (Pc) and
low mutation probability (P,,,i)n versly proportional
to the population size, are recommended [7]. The
selection of optimal values of these parameters is
generally done through h t and trail. However. value
of P, generally lies between 0.6 to 0.8 and P,,,
behveen 0.000 1 to 0.1
[attachment=47541]
Abstract:
This paper has used Genetic Algorithms (GAS)
to solve economic load dispatch (ELD) problem for eastem
region of Electricitv Generating Authority of Thailand
(EGAT) svstem. A piecewise model of cost characteristics
has been derived for the combined cycle and the
cogeneration power plants in the svstem and utilized for
the ELD formulation. Results of ELD have been obtained
bv using a conventional GA and a Micro GA models. Both
the models effectively solve the ELD problem and reduce
the operating cost. However, the Micro GA model ot'fers
faster convergence.
Keywords: Economic load dispatch, Genetic algorithms,
Cogeneration plants, Combined cycle plants
INTRODUCTION
Economic load dispatch (ELD) is one of the
main functions of modem Energy Management
Svstem (EMS) which determines the optimal real
power settings of generating units with an objective
to minimize total fuel cost of thermal plants. Several
conventional algorithmic methods have been used
11.21 to solve the ELD problem. However these
methods require the objective function in continuous
differentiable form and sometimes fail to provide
global minima. The generator fuel costs are
generally approximated in the continuously
Merentiable quadratic form.
Some power plants, such as combined cycle
and cogeneration plants, have more than one type of
turbine operation. The open cycle operation of the
gas turbine units causes the cost characteristics of
such plants to be, generally. non smooth and
continuously non-differentiable. Genetic Algorithms
(GAS) have been found suitable to effectively handle
the non-continuous or nondlfferentiable objective
functions [3.5].
GAS are optimization methods employing
search process imitated from the mechanism of
biological selection and biological genetics [3]. They
combine survival of the fittest among those feasible
solutions in the form of string structures (or genes :
mostly in binary form) with a randomized
information exchange to form a search algorithm. In
every generation, a new sct of string solutions is
created froin thc fittest of the old string solutions set.
+ 011 leave frnm Deptt. oFElect. En%.. 117' Kanprrr. India
IEEE Catalogue No: 98EX137
0-7803-4495-2/98/$10.00 1998 IEEE 165
Whle randomized, GAS are no simple
random walk. They efficiently use historical
information to speculate on new search points with
e'xpected improved performance
The present work has considered the ELD
studies on the eastern regon of EGAT system which
has a large amount of thermal plants including
cogeneration and combined cycle plants. Genetic
Algorithms have been selected for the ELD solution.
Three loading cases in this region viz light load. day
load and peak load have been considered. Two
models of GAS, a conventional GA and a Micro GA
[9] have been utilizedl. An experimentation to select
optimum values of some of the GA parameters have
also been conducted.
GENETIC ALGORITHMS
. General
GAS are derived from a simple model of
population genetics. They have followmg five
components [6].
i) Stmg represeintation
ii) An initial popiulation of strings.
iii) Evaluation funiction or fitness function
iv) Three GA Operators : Reproduction,
Crossover and Mutation
v) Value of the GA operator and other
parameters
Instead of explloiting the variables as actual
parameters, the coded binary strings are often used
for the variables' representation. The length of stnng
depends on the precision required. Before starting
the iterative routine of the GA, a population of
strings must be initially randomized. GA considers
only a single evaluation function or fitness function.
The performance of each string is evaluated
according to its fitness. calculated by the evaluation
function.
Three basic operators used in a conventional
GA are Reproduction. Crossover and Mutation. The
more fitness string should have more chance to be
copied or selected to the next generation. The biased
Roulette wheel IS usedl to achieve the "survival of the
tittest"-aspat of GAS For the population size of .V.
probability of reproduction of each ith string with
fitnessfi i.e. Preprod.; , is
After the new generation population is
copied or selected by Reproduction routine, the
randomized couple of strings are picked up to offer
'crossover'.
Although Reproduction and Crossover
effectively search and recombine existing
chromosomes. they do not create any new genetic
matenal in the population. Mutation is capable of
overcoming this shortcoming. The Mutation is to
change any string bit from 1 to 0 or 0 to 1.
Like the other stochastic methods , GAS
require a number of parameters which are population
size. probability of crossover. probability of mutation
that must be selected. Usually relatively small
population size, high Crossover probability (Pc) and
low mutation probability (P,,,i)n versly proportional
to the population size, are recommended [7]. The
selection of optimal values of these parameters is
generally done through h t and trail. However. value
of P, generally lies between 0.6 to 0.8 and P,,,
behveen 0.000 1 to 0.1