TELKOM
NIKA
, Vol.14, No
.4, Dece
mbe
r
2016, pp. 12
20~122
7
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i4.3473
1220
Re
cei
v
ed Ma
y 10, 201
6; Revi
sed
No
ve
m
ber 15, 201
6; Acce
pted
No
vem
ber 2
4
, 2016
Chaos-Enhanced Cuckoo Search for Economic
Dispatch with Valve Point Effects
M. W. Musta
f
a, Abdirahm
an M. Abdila
hi*, M. Mustapha
F
a
cult
y
of Elec
trical Eng
i
ne
eri
ng, Univ
ersiti T
e
kno
l
og
i Mal
a
ysia, 813
10 Sku
dai, Joh
o
r, Mal
a
y
s
ia
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: abdir
ahma
n
@
fkegra
duate.
utm.m
y
A
b
st
r
a
ct
Econo
mic dis
p
atch deter
mine
s the opt
i
m
al g
ener
ation
outp
u
ts to mi
ni
mi
z
e
the toal fu
el c
o
st w
h
il
e
satisfying
the
l
oad
d
e
m
an
d
a
nd
op
eratio
na
l
constra
i
nt
s. M
oder
n
opti
m
i
z
a
t
ion tec
h
n
i
qu
e
s
fail
to s
o
lve
t
h
e
prob
le
m
in
a r
obust
man
ner
and
fin
d
in
g r
o
bust g
l
o
bal
o
p
timi
z
a
tio
n
tec
h
n
i
qu
es is
n
e
ces
s
ary for
efficie
n
t
system
operation.
In this st
udy, the potentiality of
introducing chaos
in
to the standar
d Cu
ckoo S
earch (
C
S)
in ord
e
r to further en
ha
nce it
s glob
al searc
h
ability is
inv
e
stigate
d
. Deter
m
i
n
istic cha
o
tic ma
ps are ra
nd
om-
base
d
tech
niq
ues that c
an p
r
ovid
e a
bal
an
ced ex
plor
atio
n an
d ex
plo
i
ta
tion se
arch
es
for the al
gor
ith
m
.
F
our differe
nt varia
n
ts are ge
n
e
rated
by caref
u
lly ch
oosi
ng f
our differe
nt lo
cations (w
ithin
the stand
ard C
S
)
w
i
th potential
ado
ptio
n of a cand
idat
e cha
o
tic ma
p.
T
hen
detail
ed studi
es are carrie
d
out on be
nch
m
ar
k
power system
problem
s
with four diffe
rent locations to find
out
the
m
o
st efficient one.
The best of all test
cases
gen
erat
ed
is ch
ose
n
a
nd c
o
mpar
ed
w
i
th alg
o
rith
ms
pres
ented
i
n
t
he
literatur
e. T
he r
e
sults s
h
o
w
that
the pr
op
osed
meth
od
w
i
th th
e pr
opos
ed
ch
aotic
map
outp
e
rforms
stan
da
rd CS. A
dditi
o
nally,
the c
h
a
o
s
-
enh
anc
ed CS
has a very go
o
d
perfor
m
a
n
ce
in co
mp
ariso
n
w
i
th QPSO an
d NSS.
Ke
y
w
ords
:
pow
er p
l
a
n
ts o
perati
on,
econ
omic
disp
atch,
Cuck
oo s
earc
h
, metah
eur
isti
c al
gorith
m
, c
h
aoti
c
ma
ps
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introducti
on
De
spite the
global
agen
d
a
of increa
si
ng the
sha
r
e
of rene
wa
bl
e ene
rgy p
r
o
ductio
n
,
thermal p
o
wer pla
n
ts co
ntribute p
r
ed
ominantly
to
the global
electri
c
ity produ
ction an
d
will
contin
ue to
do so i
n
the
foreseea
ble
future. This warrant
s a
global
con
c
e
r
n to lower t
heir
operating co
sts.
Allocatin
g
power
gen
erated in
the
s
e
plant
s in
the
lea
s
t po
ssibl
e
op
eratin
g
cost
while me
etin
g the system
con
s
trai
nts h
a
s be
en one
of the main concern
s
of the utility opera
t
ors
globally. To address this,
enginee
rs use the econo
mi
c dispatch (ED) fo
rmulat
ions which is a
pra
c
tical p
o
wer sy
stem opt
imization p
r
o
b
lem.
Becau
s
e
of the non
-line
a
rity, non-co
nvexity
and the multimodal chara
c
te
risti
c
s pre
s
en
t
in the cost f
unctio
n
of th
e ED, ad
opti
ng a m
e
tahe
uristi
c o
p
timization te
ch
ni
que
(whi
ch
i
s
the
state-of
-the
-a
rt glob
al opti
m
ization
tech
nique [1
]
)
ha
s two
majo
r
advantag
es
o
v
er the u
s
a
g
e
of
the co
nventi
onal te
chniq
ues. Fi
rstly, metaheu
ri
stic optimizatio
n
techni
que
s
(MOT
s) l
ead
to
better problem modelling that reduce assumptions
related to problem charact
e
rization in term
s
of nonline
a
rit
y
. Secondly, MOTs
have b
e
tter ability to obtain optim
al sol
u
tions
a
s
compa
r
e
d
wit
h
a conventio
n
a
l techniq
ue. Both of these
aspe
cts
lea
d
to optimal generato
r
loadi
n
g
s for lea
s
t cost
operation
wit
h
in the
po
we
r
system. A
s
a
re
sult,
sy
stem op
erato
r
s
can
be
nefit from
sig
n
ificant
co
st saving
s
over the years.
Earlier fo
rmul
ations of the
ED probl
em
were ta
ckl
e
d
usin
g cove
ntional math
ematical
techni
que
s such a
s
interio
r
point metho
d
, lambda-ite
r
ation metho
d
and linea
r prog
ram
m
ing
[2].
Ho
wever,
tho
s
e
metho
d
s cannot
solve
the E
D
p
r
obl
e
m
when
form
ulated i
n
a
no
n-line
a
r conte
x
t
and they
suff
er from “cu
r
se of dim
e
n
s
io
nality”
proble
m
s. In
recent
times, the
kn
owle
dge
gro
w
th
of MOTs
ha
s given an o
p
portunity to o
p
timize t
he E
D
problem i
n
a more practi
cal
way than
the
mathemati
c
al
techniq
u
e
s
. Re
sea
r
che
r
s
have im
plem
ented ma
ny MOTs to
sol
v
e the probl
e
m
su
ch a
s
Gen
e
tic Algorith
m
(GA) [3], Evolutionar
y
Algorithm
s [4], Particle Swarm Optimi
za
tion
(PSO) [5], Evolutiona
ry Progra
mming
(EP) [6], T
abu
Search
(TS
)
[7], Simulated Anne
aling
(SA)
[8], Firefly Algorithm
(FA
)
[9] and Gl
owworm Swarm
Optimization
(G
WO
) [10].
Ho
weve
r, m
any
of these tech
nique
s and t
heir varia
n
ts
have sho
w
n
a lack of ability to
obtain con
s
i
s
tent an
d
robu
st optima
l
result
s re
du
cing thei
r e
ffectivene
ss in a
practi
cal op
e
r
ation.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Cha
o
s-Enh
a
n
c
ed
Cu
ckoo
Search for Econom
ic Di
sp
a
t
ch with Valve Point… (M. W. Mustafa
)
1221
One si
gnifica
nt issu
e with these dive
rse
MO
Ts when
impleme
n
ted in the ED pro
b
lem is
their inh
e
ren
t
rando
mne
s
s involved i
n
their
p
e
rf
orma
nce [11
]. Unlike th
e determini
stic
techni
que
s, most of the methodol
ogie
s
rep
o
rted in
t
he diverse literatu
r
e
s
indicate the inabili
ty of
the alg
o
rithm
to a
c
hieve
a
b
sol
u
te
robu
stness to
a
c
hi
eve glo
bal
so
lutions.
To ta
ckl
e thi
s
i
s
su
e,
the developm
ent of a robu
stne
ss-o
rient
ed optimization algo
rithm is
ne
ce
ssary. In this pape
r, an
integrate
d
ap
proa
ch
that combine
s
a
n
e
fficient
Cu
cko
o
Search
(CS) alg
o
rithm
wi
th determi
nist
i
c
cha
o
tic sy
ste
m
s wa
s devel
oped, to solv
e highly multimodal, nonli
n
ear p
r
obl
ems.
CS ha
s
sh
own better characteri
st
ics i
n
optimal a
nd
efficient
gl
ob
al search
abil
i
ty than
many MOT
s
in the literatu
r
e [12]. Th
e
combi
nat
ion
of the Levy flight pro
c
e
ss with st
anda
rd
Gaussi
an probability distri
bution process make the
algorithm gl
obal search technique efficient.
With the ad
option of ch
aotic map
s
,
better bal
a
n
ce b
e
twe
e
n
the diversif
ication a
nd the
intensifi
c
ation
is achieve
d
with suffici
ent
rando
miza
tio
n
[13]. In line
with this, this pap
er aim
s
to
investigate
the effe
ct of
introdu
cin
g
chaotic
map
on the
pe
rfo
r
man
c
e
of t
he
stand
ard
CS
algorith
m
, in the co
ntext of optimal ED
sche
duli
ng. T
h
ough, ma
ny studies h
a
ve i
n
vestigate
d
the
improvem
ent
of different
M
O
Ts u
s
ing
de
termini
s
ti
c ch
aotic sig
nal
s,
ho
wever,
th
ere are
yet
a
n
y
studie
s
on th
e improve
m
e
n
t of the performan
ce
of
CS algo
rithm
using
ch
aoti
c
map
s
. Also,
anothe
r co
ntri
bution on rob
u
st glob
al se
arch tech
ni
qu
es for
solving
ED probl
ems is pre
s
ente
d
in
this
work
.
2. Problem Formulation
The e
c
on
omi
c
di
spat
ch i
s
norm
a
lly form
ulated a
s
a
n
optimizatio
n
probl
em
whe
r
eby the
operating
co
st of all the generatin
g
plant
s are minimi
zed su
bje
c
t to
physi
cal sy
stem co
nstraint
s.
The obje
c
tive
of the proble
m
is mathem
atica
lly define
d
by the following
co
st function:
1
mi
n
D
N
ii
i
F
P
(
1
)
2
ii
i
i
i
i
i
F
Pa
P
b
P
c
(
2
)
Whe
r
e
ܽ
,
ܾ
, an
d
ܿ
are the
co
st coeffici
ents of the generation unit
݅
;
ܲ
is the sched
ule
d
power
of the unit
݅
;
ܰ
is the total nu
mber of onli
n
e gene
rato
rs;
and
ܨ
is
the fuel c
o
s
t.
Ho
wever, thi
s
formulation
of the co
st functi
on
simplifi
e
s the g
ene
rator characte
ristics to
a smooth q
u
adrati
c
functi
on leadin
g
to poor proble
m
modelling.
To model a
more p
r
a
c
tica
l cost
function, on
e must con
s
ide
r
the valve point e
ffect and the equatio
n is modified a
s
sho
w
n bel
ow:
2m
i
n
sin
ii
i
i
i
i
i
i
ii
i
FP
a
P
b
P
c
e
f
P
P
(
3
)
Whe
r
e
݁
and
݂
are the valve point effect co
efficients;
ܲ
is the minimum
generation li
mit;.
In additio
n
t
o
the
above
problem
formulation
s
, th
e sy
stem i
s
subj
ecte
d to
physi
ca
l
con
s
trai
nts.
The
system
po
we
r b
a
la
nce
con
s
tr
ai
nt is an
eq
uality co
nst
r
aint an
d
ca
n b
e
rep
r
e
s
ente
d
in mathemati
c
al form as foll
ows:
1
0
D
N
iD
L
i
PP
P
(
4
)
Whe
r
e
ܲ
is the
total system deman
d; and
ܲ
is the syste
m
loss. The li
mits of each
gene
rating
unit co
nstitut
e
as the in
eq
uality con
s
tra
i
nt of
the p
r
o
b
lem
whi
c
h
can be
written
mathemati
c
a
l
ly
as:
mi
n
m
ax
ii
i
PP
P
(
5
)
Whe
r
e
ܲ
and
ܲ
௫
are the minim
u
m and the m
a
ximum limits of the power
gene
ration u
n
i
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1220 – 122
7
1222
3. Cucko
o Search
Algori
t
hm
Cu
ckoo search (CS) al
go
rithm wa
s devel
oped
by Xin-She Yang an
d Suash
Deb
in 200
9
[14]. The algorithm is ba
sed o
n
the theory of
Cu
ckoo
s, particularly their b
r
ood p
a
ra
siti
sm
cha
r
a
c
teri
stics in combi
nat
ion with the l
e
vy flight
con
c
ept. Sectio
n
s
3.1 to 3.4 o
u
lined the
sta
ges
for impleme
n
ting CS algo
rit
h
m.
3.1. Representation and Initializ
ation
The de
cisi
on
variable
s
a
r
e
rep
r
e
s
ente
d
within the opti
m
ization a
s
follows:
mi
n
m
ax
mi
n
*
ii
i
i
P
P
rand
P
P
(
6
)
W
h
er
e
ma
x
i
P
is the
maximum
po
wer o
u
tput of
ea
ch
unit i;
ra
nd
is a
unifo
rm
d
i
stribute
d
ran
dom
gene
rato
r.
3.2. Fitness
Function
The fitness e
v
aluation for
all populatio
n
indi
viduals i
s
perform
ed b
a
se
d on the followin
g
fitness e
quati
on:
1
D
N
ii
i
i
D
L
i
Fitne
s
s
P
F
P
P
P
P
(
7
)
The paramet
er
is the
pe
nalty facto
r
multiplier to
amplify the
e
rro
r val
u
e
s
so that it
wea
k
e
n
s the
good
ne
ss of the fitness fun
c
tion when th
ere a
r
e eq
ual
ity constraint violations [15
].
An additional
con
s
traint h
andlin
g mod
u
le is imple
m
ented to cater effective
l
y the const
r
aint
violations. In
this pap
er, th
e method p
r
e
s
ente
d
in [16
]
was u
s
e
d
to
effectively obtain the glo
bal
optimal value
s
.
3.3. Levy
F
l
ight
Within CS,
n
e
w sol
u
tion
s are
gen
erate
d
u
s
ing
the
L
e
vy flight pro
c
e
ss. In
this
pro
c
e
ss,
the glob
al b
e
s
t
best
P
index is u
t
ilized a
nd th
e optimal
pat
h for th
e L
e
vy flights. The
upd
ating
formula for th
e Levy flight
pro
c
e
ss i
s
given as follo
ws:
ne
w
ii
P
P
randn
Scal
e
L
ev
y
(
8
)
Whe
r
e
max
m
in
10
0
ii
PP
Sc
ale
and
ra
nd
is a norm
a
lly distribute
d
stocha
stic nu
mber. Additio
nally,
the
()
Le
v
y
function for every iterat
ion is dete
r
mi
ned a
s
follows:
1/
ib
e
s
t
u
Le
v
y
P
P
v
(
9
)
Whe
r
e
u
and
v
are drawn fro
m
normal di
st
ribution. Th
at is:
22
~(
0
,
)
,
~(
0
,
)
uv
Nv
N
(
1
0
)
With,
1/
1/
2
1s
i
n
/
2
,
1
1/
2
2
uv
(
1
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Cha
o
s-Enh
a
n
c
ed
Cu
ckoo
Search for Econom
ic Di
sp
a
t
ch with Valve Point… (M. W. Mustafa
)
1223
Whe
r
e
(.
)
is the gamma di
stri
bution fun
c
tio
n
.
The main ad
vantage of this Levy flight is to
perfo
rm a global
exploratio
n search by
perfo
rming
o
c
casi
onal l
o
n
g
-di
s
tan
c
e ju
mps. It is the
s
e
sud
den j
u
mps th
at mig
h
t increa
se t
he
sea
r
ch efficie
n
cy of the CS
significantly
particula
rly for multimodal,
nonlin
ear p
r
o
b
lems.
3.4. Discov
e
r
y
and Rand
omization
The second
su
ccessive u
pdating e
qua
tion use
d
for the CS algo
rithm is ba
se
d on the
con
c
e
p
t of d
i
scovery of
the
Cu
ckoo
egg
within
the h
o
st
ne
st. This con
c
e
p
t brin
gs in
a
rand
omi
z
atio
n feature an
d
a lo
cal
sea
r
ch b
a
sed
o
n
a ra
ndom
wa
lk for th
e alg
o
rithm. Th
e n
e
w
solutio
n
s g
e
n
e
rated a
s
a result of this concept is obt
ained a
s
follo
ws:
ne
w
ii
a
m
n
P
P
ra
nd
H
P
r
and
P
P
(
1
2
)
Whe
r
e
and
mn
PP
are two differe
nt solution
s sel
e
cted rand
omly by permutatio
n
;
.
H
is a
Heavi
s
ide fun
c
tion controll
ed by a swit
ching pa
ram
e
ter
a
P
.
4. Chao
tic Maps (CM)
In this sub-section, a Cheb
yshev chaotic map is introduce
d. The CM whi
c
h will
be used
in the rest o
f
the experi
m
entation
s
i
s
de
scri
bed
and fo
rmulat
ed. Equation
(2.3) gives
the
mathemati
c
al
expressio
n
of the implemented Ch
e
vyshev chaot
i
c map with
the initial value
0
0.1
x
, and Figure 1 illustrate
s the visual
cha
r
acte
ri
stics of the impleme
n
ted CM [13].
1
1
co
s
(
co
s
)
tt
xt
x
(
1
3
)
Figure 1. Illustration of cha
o
tic map
s
used in the stud
y
5. Implementation
In this
study, four
different
schem
es of
in
tegratin
g a
Che
b
ysh
e
v chaotic
map i
n
to th
e
stand
ard
CS
have b
een i
n
vestigate
d
, each of t
hem
pro
d
u
c
ing
a
uniqu
e vari
a
n
t of CS. Th
en
each variant i
s
expo
sed to
the prop
osed
Che
b
ys
h
e
v map explaine
d
in the earlie
r
se
ction.
The
choi
ce
o
f
the favourit
e lo
cation
s i
s
di
re
cted
to t
he u
pdating
equatio
ns
of
the CS
algorith
m
. Th
e stan
da
rd
CS algorith
m
h
a
s two
su
cce
ssive
upd
atin
g equ
ation
s
a
n
d the
r
efore t
h
e
modificatio
n
s are di
re
cted
to these two
equatio
ns
. Th
e pse
u
do
ran
dom ge
nerators
(pa
r
ticula
rly
the uniform
and norma
l distributio
n
rando
m nu
mber g
ene
ra
tors) and
CS’s cont
rolla
ble
para
m
eter a
r
e subje
c
ted u
nder thi
s
investigation
whet
her the CM can altern
atively replace them.
Table 1 bel
o
w
sum
m
ari
s
e
s
the implem
ented modifi
cations.
The eq
uation
s
in Ta
ble 1
are p
r
eviou
s
l
y
formulated
in se
ction 3.
Last
colum
n
i
ndicate
s
the
co
rrespo
nding stage name,
whic
h
are d
e
scri
b
ed in sectio
n
s
3.3 a
n
d 3.
4. The eq
uat
ions
involved in the modificatio
n
s a
r
e eq
uati
ons
(2.2
)
(for variants A a
nd B) an
d eq
uation (1.9) f
o
r
variant C. Therefo
r
e, the
symbolic d
e
s
cription
s of the equ
ation
s
presented in
Table 1 rem
a
in
the same a
s
descri
bed in
se
ction 3.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1220 – 122
7
1224
Table 1. Illust
ration of Modi
fication
s Implemented fo
r the Differe
nt Variant
s
Variant
Equations
Involved
Stage
A
O
r
iginal
ne
w
ii
a
m
n
P
P
rand
H
P
P
P
Discover
y
Modified
ne
w
ii
t
a
m
n
PP
x
H
P
P
P
B
O
r
iginal
ne
w
ii
a
m
n
P
P
rand
H
P
P
P
Discover
y
Modified
new
ii
t
m
n
P
P
rand
H
x
P
P
C
O
r
iginal
new
ii
PP
r
a
n
d
n
S
c
a
l
e
L
e
v
y
Lev
y
Flight
Modified
21
ne
w
ii
t
P
P
x
S
cale
Le
vy
D
Modified Combination
of
modificati
ons in
Variants A and B
Discovery
5.1. Experiment Settings
All the exp
e
r
iment
s im
plemented
in
this p
ape
r
were
ca
rrie
d
out u
s
in
g
MATLAB
softwa
r
e. A
st
anda
rd
po
we
r sy
stem te
st
ca
se
with thi
r
teen unit
s
al
o
ng
with valve
loadin
g
effe
cts
is used to test the algorithm’s
c
apability to solve the
ED problem
s.
The data of the test
system i
s
obtaine
d from
[6].
In orde
r to obtain the righ
t paramete
r
s
of
the algorit
hm, we have
carried o
u
t a detaile
d
para
m
etri
c st
udy by varying one p
a
ra
meter at
a time. The adv
antage
s of the CS alg
o
ri
thm
inclu
de its n
a
t
ure of havin
g small
num
ber of
cont
rol
l
able pa
ram
e
ters u
n
like th
e PSO. The f
i
rst
para
m
eter to
be tu
ned
i
s
the
fixed
numbe
r th
at re
pre
s
e
n
ts
the p
r
oba
bility of ra
ndom
ly
discoveri
ng a
Cu
ckoo’
s egg
in
th
e host
n
e
st (d
onated as
P
ₐ
). Besid
e
s,
other i
nhe
rent
para
m
eters
such
as th
e p
opulatio
n si
ze (d
onated
a
s
N) a
nd the
Levy flight expone
nt - d
o
n
a
ted
as
β
(Beta
)
- are tune
d. Table 2
sho
w
s the li
st of param
eters
being te
sted
and the
cho
s
en
values for fin
a
l alg
o
rithm
experim
entati
on. In al
l
ex
ecute
d
exp
e
riments, th
e
stopping
criterion
wa
s the maximum iteration
.
Table 2. Para
meter Setting
in the Algorithm De
sign
Parameter
Descrip
tion
Teste
d
Value
s
13 Uni
t
β
Lev
y
Flight Exp
o
nent
0.3
,
2
.
0
in steps of 0.1
0.55
P
ₐ
Probabilit
y
of Discovery
0,
1
in steps of 0.1
0.90
N Population
Size
10
,
1
00
in steps of 10
50
T
ᵩ
Max
i
mum
Iteration
5x10
³, 10
x10³,
2
0
x10
³
,25
x
10³
20x10
³
PF Penalt
y
Facto
r
50, 100, 500,
10
00
100
6. Results a
nd Discu
ssi
on
This sectio
n
pre
s
ent
s th
e
re
sult
s of t
he inve
s
t
ig
a
t
io
n
c
a
r
r
i
ed
ou
t in
th
is
s
t
ud
y. T
h
e
perfo
rman
ce
of ea
ch va
riant in
respe
c
t wi
th
th
e standard CS (SCS)
are
in
vestigated a
nd
discu
s
sed in the followi
ng sub-se
ction
s
.
6.1. Perform
a
nce o
f
Cha
o
tic Maps
w
i
thin Each Va
riant
In this
su
b-section, th
e p
e
rform
a
n
c
e
impac
t
s
of
a
Ch
ebyshev CM on
fo
ur differen
t
locatio
n
s a
r
e
looked
at, in
comp
ari
s
o
n
with the
SCS.
Each va
riant
is loo
k
ed
at
sep
a
rately. T
abl
e
3 summa
rize
s the
de
script
ive statisti
cal
para
m
eter
s
o
f
each exp
e
ri
ment for all v
a
riant
s. O
b
se
rve
the numbe
rin
g
informatio
n
of CS
-A, CS-B, CS-C a
n
d
CS-D is in
line with the
variant letteri
ng
given in Tabl
e 1. SCS stands for Stan
d
a
rd CS.
It can b
e
ob
served i
n
Ta
bl
e 3 that
CS-A
perfo
rm
ed better than the res
t
of the tes
t
cases
whe
n
loo
k
ed
at the mean, median, sta
n
dard d
e
vi
atio
n (abb
reviate
d
as Std here after) an
d the
range. However, in terms
of the
ability
of the algorit
h
m to random
ly locate the optimal
cost, all
the different test ca
se
s a
c
hiev
ed the d
e
sired poi
nt. Con
c
lu
sively, the best ca
se wa
s test case
CS-A in
which the
Che
b
yshev map
re
pl
ace
s
th
e rand
om nu
mbe
r
g
enerated
ba
sed o
n
Ga
ussain
prob
ability di
stributio
n. Thi
s
sho
w
s th
at
a dete
r
mi
ni
stic chaoti
c
si
gn
al (su
c
h a
s
Chebyshev ma
p)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Cha
o
s-Enh
a
n
c
ed
Cu
ckoo
Search for Econom
ic Di
sp
a
t
ch with Valve Point… (M. W. Mustafa
)
1225
might have better ability to stabili
ze the performan
ce of the algorithm while achieving global
optimal sol
u
tions.
Table 3. De
scriptive stati
s
t
i
cal re
sult
s of the variou
s variant
s (1,8
00
MW)
Variants
Min Cos
t
Mean C
o
st
Median
Cos
t
Max Cos
t
Std C
o
st
Range
Standard CS
SCS
17,963.84
17,969.95
17,969.80
17,990.77
5.50
26.93
Cheb
y
s
he
v
CS-A
17,963.83
17,965.05
17,963.86
17,968.99
2.15
5.16
CS-B
17,963.83
17,966.34
17,963.91
17,975.91
3.99
12.08
CS-C
17,963.84
17,970.01
17,969.01
17,985.09
5.55
21.25
CS-D
17,963.83
17,969.15
17,968.95
18,012.39
6.99
48.56
6.2. Algorith
m
ic Robus
tn
ess of Eac
h
Test
Cas
e
With statistical
pe
rforman
c
e pa
ramete
rs
o
n
ly,
it is diffic
u
lt to observe the optimal
c
o
s
t
distrib
u
tion of
the trial
set result
s. For
example,
loo
k
in
g at Table
3
one
can
ob
se
rve that for te
st
variant B the mean
cost i
s
actually hig
h
e
r than t
he m
edian
co
st. This indi
cate
s
that despite t
he
algorithm’s ability with half of the total
trial set
to achieve mini
m
u
m cost that is close to the
desi
r
ed
nu
m
ber, the
few
high valu
es o
f
optimal
co
st occu
rren
ce
s of some t
r
ial
s
coul
d a
c
tua
lly
lead to
high
mean
s. Th
erefore, the
nu
mber-o
rient
e
d
table
form
a
t
is n
o
t en
ou
gh to
de
scrib
e
the
variability an
d di
stributio
n
of re
sult
s in
cludi
ng o
u
tlie
rs,
q
u
a
r
tiles and other
s. To
a
c
tually show
that, an exten
ded a
nalysi
s
of this exp
e
ri
ment re
sults
are
plotted
using box
-whisker pl
ot. Figu
re 2
indicates the
box plot of a 13 unit test system for ea
ch variant impl
emented.
Figure 2. The
box plot of five different test ca
se
s (1,8
00 MW)
In Figure
2, the result dis
t
ribution and
the
cos
t
range c
o
ve
red for v
a
rious
tes
t
c
a
s
e
s
and
variants i
s
p
r
ese
n
ted. The
actual p
e
rfo
r
mance in
terms of ro
bu
stness i
s
truly evident within
thi
s
box plot tha
t
indicate
s t
he optimal
result
s di
strib
u
tion an
d th
e co
nsi
s
ten
cy level of the
impleme
n
ted
variants. Th
e
lowe
r the box
and the sm
al
ler the box
si
ze the b
e
tter
the algo
rithm
in
achi
eving
o
p
timal a
n
d
co
nsi
s
tent
re
sult
s
re
spe
c
tively. Proceedi
ng
from the
p
r
evious
observation
s in sub-se
ction 6.1 to th
e obse
r
vati
o
n
s from Figu
re 2, it can be observe
d
that
variant A shows very
small
optimal resul
t
variab
ility in
respect to other experiment
ed locations.
6.3. Optimal Solutions
After the op
timal tuning
experim
ents, anot
her fin
a
l experi
m
en
t was
run
with th
e
recomme
nde
d optimal values for the p
a
ram
e
ters
as describ
ed in
the previou
s
sectio
n with
a
maximum ite
r
ation of 2
0
,00
0
for th
e thirt
een u
n
it
sy
stems. T
he o
u
tput se
ttings
o
f
the gen
erators
are sho
w
n in Table 4. The results of so
me other
met
hod
s pre
s
e
n
ted in recent literature are a
l
so
inclu
ded for
compa
r
ison.
1.
7965
1.
797
1.
7975
1.
798
1.
7985
1.
799
1.
7995
1.
8
1.
8005
1.
801
1.
8015
x 1
0
4
SC
S
A
B
C
D
V
a
r
i
ant
s
O
p
ti
m
a
l
C
o
s
t
(
$
/h
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1220 – 122
7
1226
Table 4. Best
solution o
u
tp
ut powe
r
sol
u
tion setting
s for the gen
era
t
ors in
comp
a
r
iso
n
with oth
e
r
method
s in the literature (1
3-Unit System)
Unit
Q
PSO
[17]
NSS [18]
CCS-
A
1
1
538.56
448.80
628.32
2
224.70
300.50
222.76
3
150.09
299.20
149.59
4
109.87
60.00
109.87
5
109.87
109.90
109.87
6
109.87
109.90
109.87
7
109.87
61.90
60
8
109.87
109.90
109.87
9
109.87
109.90
109.87
10
77.41
40.00
40
11
40.00
40.00
40
12
55.01
55.00
55
13
55.01
55.00
55
Total
cos
t
($/
h
)
17,969.01
17,976.95
17,963.83
6.4. Compari
s
on
w
i
th the
Existing Methods
In order to show
CS’s
effect
iveness and suitability in ED pr
oblems, result
s of different
method
s for
both sy
stem
s with valve-p
o
int loadi
ng
effects a
r
e
shown in T
abl
es 5. T
he ta
ble
summ
ari
s
e
s
the optimal co
st result achievem
ent of eleven different metho
d
s pu
blish
e
d
i
n
journ
a
ls that
are
found
in
the maj
o
r en
e
r
gy a
nd
engin
eerin
g d
a
tab
a
se
s. In te
rm
s of th
e a
b
ility of
the method t
o
achi
eve minimum op
era
t
ing co
st out
of the set tria
ls, the re
sults of the propo
se
d
method are b
e
tter than tho
s
e of the method
s list
ed in
Table 5. Moreover, the propo
sed meth
od
is the b
e
st p
r
ese
n
ted in te
rms
of havin
g bot
h lo
w st
anda
rd d
e
via
t
ion and th
e global o
p
timu
m
point at the same time.
Table 5. Co
m
pari
s
on of results for existin
g
met
hod
s for a thirteen uni
t system with
a deman
d of
1800 M
W
Metho
d
Min Cos
t
($
/h)
Mean C
o
st
($/h
)
Max Cos
t
($/
h
)
SD ($/h
)
MSL [19]
18,158.68
IFEP [6]
17,994.07
18,127.06
18,267.42
-
AIS [20]
17,977.09
NDS [21]
17,976.95
17,976.95
17,976.95
0
HMAPSO [22]
17,969.31
17,969.31
17,969.31
0
FCASO-SQP [2
3
]
17,964.08
18,001.96
-
-
CS-
A
17,963.83
17,965.05
17,968.99
2.15
7. Conclusio
n
Econo
mic di
spatch i
s
an i
m
porta
nt power sy
stem opt
imization p
r
o
b
lem. Findin
g
elegant
techni
que
s th
at can
efficie
n
tly tackle thi
s
proble
m
bri
ngs
sig
n
ifican
t system op
eration savings.
In
line
with thi
s
, this stu
d
y ai
ms to
imp
r
ov
e the
st
an
da
rd CS
by inte
grating
with
chaotic ma
ps
to
develop
the
new chaoti
c
-enha
nced
CS
. Four different lo
cation
s
of the
stand
a
r
d
CS h
a
ve b
een
introdu
ce
d to Cheby
shev CM. The pe
rforman
c
e
s
of
these varia
n
ts have be
en investigat
ed.
Based
on
sta
t
istical and
ro
bustn
ess co
mpari
s
o
n
s,
th
e first lo
catio
n
(va
r
iant A
)
prod
uces the
best
perfo
rming
al
gorithm. T
h
e
re
sults
reve
aled that va
riant A is the
best g
ene
ra
ted algo
rithm
s
among
all th
e four test
case
s i
n
vestig
ated in thi
s
study. By cho
o
sin
g
a
s
CS-A (varia
nt A
with
Che
b
ysh
e
v) the be
st ge
ne
rated al
go
rith
m in this
st
ud
y
,
it
was f
u
rt
h
e
r inv
e
st
igat
e
d
in t
e
rm
s of
i
t
s
perfo
rman
ce
with e
s
tabli
s
h
ed meth
od
s i
n
the lit
erature. The p
r
op
o
s
ed
me
tho
dol
ogy proves th
at
it outperform
s established
method
s in terms of ro
b
u
st
ness and a
c
h
i
eving con
s
i
s
tent results. Due
to high
robu
st results, t
he m
e
thod
ology is
a
b
le
to elimi
nat
e the
inhe
re
nt ra
ndomi
z
ation
cha
r
a
c
teri
stics relate
d to the heuri
s
tic m
e
thod
s
wh
en
applie
d in the
econ
omic di
spatch field.
Ackn
o
w
l
e
dg
ements
The a
u
tho
r
s ackn
owle
dg
e Malay
s
ian
Minist
ry of highe
r Ed
u
c
ation
and
Universiti
Tekn
ologi M
a
laysia for sup
porting thi
s
work.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Cha
o
s-Enh
a
n
c
ed
Cu
ckoo
Search for Econom
ic Di
sp
a
t
ch with Valve Point… (M. W. Mustafa
)
1227
Referen
ces
[1]
XS Ya
ng. Natu
re-insp
i
re
d metahe
ur
istic al
gor
ithms. Luniv
e
r press. 201
0.
[2]
BH Cho
w
d
h
u
r
y
, S Ra
hman.
A revie
w
of
re
cent adv
ances
in eco
nomic
d
i
spatch.
IEEE Transactions
on Power System
s.
1
990; 5: 124
8-12
59.
[3]
CT
Su, Nonco
n
ve
x po
w
e
r ec
onom
ic disp
atch b
y
im
prove
d
gen
etic alg
o
rit
h
m
w
i
t
h
multip
l
i
er up
datin
g
method.
Electri
c
pow
er compo
nents an
d systems.
2
004; 3
2
: 257-2
73.
[4]
A Afand
i, I F
a
d
lika, A A
ndok
o.
Comp
arin
g Pe
rf
ormances of Evoluti
onar
y
Al
gorithms
on th
e Emissi
on
Dispatc
h
a
n
d
Econom
ic
Dispatc
h
Pro
b
lem.
T
E
LKO
M
NIKA T
e
lec
o
mmunic
a
tio
n
Co
mp
utin
g
Electron
ics an
d Contro
l.
201
5; 13: 118
7-11
93.
[5]
H Sh
ahi
nza
d
e
h
, SM Nasr-
A
zada
ni, N
J
ann
esari.
A
p
p
licatio
ns
of p
a
rt
icle s
w
a
r
m
optimiz
atio
n
alg
o
rithm to so
lving t
he ec
on
omic lo
ad
disp
atch of
units in po
w
e
r s
y
st
ems
w
i
t
h
valv
e-point effects
.
Internatio
na
l Journ
a
l of Electr
ic
al a
nd Co
mp
uter Engi
ne
erin
g.
2014; 4: 85
8
.
[6]
N Sinh
a, R Chakrab
a
rti, P Chattopa
dh
ya
y
.
Evoluti
onar
y pr
ogrammi
ng tec
hni
ques for ec
onom
ic loa
d
disp
atch.
IEEE Transactions
on Evol
utionar
y Com
putation.
2003; 7: 83-
94
.
[7]
W
M
Lin, F
S
C
hen
g, MT
T
s
a
y
. An im
prove
d
tabu
s
earch
for
eco
nomic
dis
p
atch
w
i
t
h
mu
lti
p
le
min
i
ma.
IEEE Transactions on Power
System
s.
20
02
; 17: 108-1
12.
[8]
H Hardi
ans
ya
h
,
J Junaidi, Y MS. An Efficient
Simulate
d Anne
ali
ng Al
g
o
rithm for Economic L
o
a
d
Dispatc
h
Probl
ems.
TELKOMNIKA Teleco
mmu
n
ic
ation C
o
mp
utin
g Electr
onics a
nd C
o
ntrol.
201
2;
11(1): 37-
46.
[9]
XS Y
a
n
g
, SSS
Hosse
ini, A
H
Gandom
i. F
i
ref
l
y al
gorit
hm for
solvi
ng
no
n-c
onve
x
eco
nom
ic dis
patch
prob
lems
w
i
th
valve l
oad
in
g e
ffect.
Applied S
o
ft Computin
g.
201
2; 12: 118
0
-
118
6.
[10]
DN Ja
ya
kum
a
r, P Venk
atesh.
Glo
w
w
o
rm
s
w
a
rm optim
iz
ation
alg
o
rithm
w
i
th
topsis for s
o
lvi
ng mu
ltip
l
e
obj
ective e
n
vir
onme
n
tal ec
on
omic dis
patch
prob
lem.
Appl
i
ed Soft Co
mp
u
t
ing.
201
4; 23: 375-
386.
[11]
KT
Chaturved
i
, M Pandit, L Srivastava. Sel
f
-organ
izin
g hi
erarchic
al p
a
rti
c
le s
w
arm o
p
ti
mizatio
n
for
nonc
onv
e
x
eco
nomic d
i
spatch
.
IEEE
Transac
tions on Power
System
s.
200
8; 23: 107
9-10
87.
[12]
XS Y
ang, S
Deb. C
u
ckoo
search
: rec
ent
adva
n
ces
an
d ap
plic
atio
ns.
Neur
al C
o
mputin
g a
n
d
Appl
icatio
ns.
2
014; 24: 1
69-1
74.
[13]
AH Gand
omi, GJ Yun, XS
Yang, S T
a
latahar
i. Cha
o
s
-enh
anc
e
d
a
cceler
a
ted p
a
r
t
icle s
w
arm
optimiz
ation.
C
o
mmunic
a
tio
n
s
in Non
lin
ear S
c
ienc
e an
d Nu
mer
i
cal Si
mul
a
tion.
201
3; 18: 327-
340.
[14]
XS Y
a
n
g
, S D
eb.
Cuck
oo s
e
arch vi
a L
é
vy fl
ights
. In N
a
tur
e
& Bi
olo
g
ica
l
l
y
Ins
p
ire
d
C
o
m
putin
g, 20
09.
NaBIC 20
09. 2
009: 21
0-2
14.
[15]
A Abdi
la
hi, M
W
Mustafa, J Jamia
n
, J Us
man.
E
m
pir
i
cal
study of p
a
rti
c
le sw
ar
m opti
m
i
z
at
io
n for
econ
o
m
ic
lo
ad
dis
patch
w
i
th
valve-
poi
nt eff
e
cts
. In 2
014
I
EEE 8th
Intern
ation
a
l
Po
w
e
r
Engi
neer
in
g
and Optimiz
a
ti
on Co
nfere
n
ce
(PEOCO). 2014: 467-4
72.
[16]
JB Park, YW
Jeon
g, JR Shi
n
, KY Lee. An
improv
ed pa
rticle s
w
arm o
p
timizati
on for
nonco
n
ve
x
econ
omic dis
p
atch prob
lems.
Power Systems, IEEE
Transactions on.
2
010
; 25: 156-1
66.
[17]
K Men
g
, HG
W
ang, Z
Do
ng,
KP W
o
n
g
. Qu
antum-i
nspir
e
d
particl
e s
w
a
r
m
optimiz
atio
n fo
r valve-
po
int
econ
omic lo
ad
disp
atch.
Power System
s, IEEE Transactions
on.
201
0; 25: 215-
222.
[18]
MT
T
s
ai, HJ Go
w
,
W
M
L
i
n. A
nove
l
stoch
a
stic
searc
h
meth
od for the s
o
l
u
tion of ec
on
om
ic disp
atch
prob
lems
w
i
t
h
non-c
onve
x
fuel c
o
st functi
ons.
Intern
atio
nal J
ourn
a
l
of Electrica
l
Po
w
e
r & Energy
System
s.
20
11
; 33: 1070-
107
6.
[19]
S Hem
a
mal
i
ni,
SP Sim
on. M
a
cla
u
rin
seri
es
-bas
e
d
Lagr
an
gia
n
met
hod
for ec
on
omic
d
i
spatch
w
i
t
h
valve-
poi
nt effect.
IET
,
Genera
t
ion, T
r
ans
miss
ion & Distri
buti
on.
200
9; 3: 85
9-87
1.
[20]
LdS C
oel
ho, V
C
Maria
n
i. Ch
a
o
tic artificia
l
im
mune
appr
oac
h ap
pli
ed to ec
onom
ic dis
patc
h
of electric
ener
g
y
usin
g
thermal units.
C
haos, Sol
i
tons
& F
r
actals.
200
9; 40: 237
6-23
83.
[21]
W
M
Lin, HJ Go
w
,
MT
T
s
ai
. Combini
ng o
f
di
rect searc
h
and si
gn
al-to-no
ise ratio f
o
r econ
omic
disp
atch optimi
z
ation.
Ener
gy
Conv
ersio
n
an
d Mana
ge
ment
.
2011; 52: 48
7
-
493.
[22]
R Kumar, D S
harma, A Sadu. A hy
brid multi-
ag
ent bas
ed
particl
e s
w
ar
m optimizati
o
n
alg
o
rithm fo
r
econ
omic p
o
w
e
r disp
atch.
Internati
ona
l jo
ur
nal of e
l
ec
trica
l
pow
er & ener
gy systems.
2
0
11; 33: 11
5-
123.
[23]
J Cai, Q
Li, L
Li, H P
e
n
g
, Y
Yang. A
h
y
br
id
CPSO–SQP
method
for ec
onom
ic d
i
spatc
h
co
nsid
erin
g
the valve-
poi
nt effects.
Energy Convers
i
on a
n
d
Mana
ge
ment
.
2012; 53: 17
5
-
181.
Evaluation Warning : The document was created with Spire.PDF for Python.