Indonesi
an
Journa
l
of
El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
1
4
,
No.
1
,
A
pr
il
201
9
, p
p.
319
~
326
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
4
.i
1
.pp
319
-
326
319
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Predi
ction o
f ove
rc
u
rrent
relay mi
scoordin
atio
n
ti
me using
urtifici
al n
eural n
etwork
S.
K
arupi
ah
1
, M.H
. Huss
ain
2
, I
. Musir
in
3
,
S.R.A.
Rahim
4
1,2,4
School
of
Ele
ct
ri
ca
l
S
y
st
em E
ngine
er
ing, Univ
ersit
i
Malay
si
a
P
erl
is
(UniMA
P), P
erl
is,
Ma
lay
si
a
3
Cent
re
of
E
lectr
ic
a
l
Pow
er Engineeri
ng
Studie
s
&
Facu
lty
of
E
lectr
i
ca
l
Eng
ine
er
i
ng,
Univer
si
ti
T
e
knologi
MA
RA
(UiTM)
Shah
Al
am,
Sel
angor
,
Malay
si
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
un
23
, 2
018
Re
vised
Sep
3
0,
2018
Accepte
d
Dec
12
, 201
8
Overc
urre
n
t
re
lay
p
lay
s
an
important
role
in
the
prote
c
ti
on
of
po
wer
s
y
st
em.
For
prote
ct
ion
,
prope
r
coor
di
nat
ion
of
re
l
a
ys
with
an
appr
opria
t
e
re
l
a
y
sett
ings
n
ee
d
to
be
done
.
Coord
i
nat
ion
ca
n
be
d
one
b
y
se
lecti
ng
an
op
ti
m
al
Ti
m
e
Multi
plier
Sett
ing
(TMS)
and
Plug
Set
ti
n
g
(PS
)
conside
ring
the
fa
ult
cur
re
nt
at
th
e
re
lay
loc
a
ti
on
.
Conti
nuousTime
Inte
rva
ls
(CTI
)
m
ust
be
m
ai
nta
in
ed
betw
ee
n
primar
y
r
el
a
y
and
sec
on
dar
y
re
l
a
y
to
e
nsure
cor
re
c
t
seque
ntial
oper
a
ti
on
of
the
re
lays
.
How
eve
r,
m
iscoordi
na
ti
on
c
an
occ
urs
due
to
sec
o
ndar
y
re
l
a
y
tri
ps
fa
st
er
t
han
primar
y
re
l
a
y
.
Th
is
pape
r
pre
sents
an
appr
oac
h
for
p
re
dicting
over
c
urre
nt
re
l
a
y
m
iscoordi
na
ti
on
t
ime
using
Artifi
c
ia
l
Neura
l
Network
(AN
N)
al
gorit
hm
in
MA
TL
AB
software
.
Th
e
eff
iciency
of
th
e
proposed
appr
oac
h
have
b
ee
n
tes
te
d
succ
essfull
y
on
17
bus
te
st
s
y
s
te
m
.
Th
e
sim
ula
ti
on
re
sults
indi
c
at
ed
tha
t
the
AN
N
Le
venb
er
-
Mae
quar
dt
al
gor
it
hm
is
ca
pable
to
pre
dict
the
m
iscoordi
na
ti
on
tim
e
occ
urs
bet
wee
n
th
e
pri
m
ar
y
and
se
cond
ar
y
re
lay
oper
at
i
ng
ti
m
e.
Ke
yw
or
d
s
:
Ar
ti
fici
al
n
e
ur
a
l netw
or
k
Mi
scoord
i
natio
n
ti
m
e
Ov
e
rc
urren
t
relay
Plug set
ti
ng
T
i
m
e
m
ulti
plier
sett
ing
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
S.K
a
rupia
h,
School
of Elec
tric
al
System
En
gi
neer
i
ng,
Un
i
ver
sit
i
Ma
la
ysi
a Perlis (
U
niMAP
),
Pauh P
utra
ca
m
pu
s,
02600 A
ra
u,
P
erli
s,
Ma
la
ysi
a
.
Em
a
il
:
m
uh
dh
a
tt
a@u
nim
ap.
ed
u.
m
y
1.
INTROD
U
CTION
The
im
po
rtanc
e
of
powe
r
syst
e
m
no
wad
ay
s
is
si
m
pl
y
un
den
ia
ble
wh
e
r
e
alm
os
t
ever
yt
hin
g
in
this
world
need
s
powe
r
to
ope
rate.
Wh
en
it
c
om
es
to
po
we
r
syst
e
m
,
there
are
m
any
challen
ge
s
that
nee
d
to
be
faced
.
T
he
m
ain
rea
son
is
t
ha
t
it
deals
with
ver
y
high
pow
er
le
vels.
Sinc
e
it
is
the
m
ai
n
powe
r
s
ource
in
th
e
world
,
this
sys
tem
need
to
be
desig
ne
d
care
f
ully
by
co
ns
id
erin
g
al
l
the
as
pects.
T
he
m
ain
as
pect
that
ne
ed
to
be
c
onsidere
d
is
the
protect
io
n.
I
n
gen
e
ral,
high
vo
lt
ag
es
and
cu
rr
e
nts
c
an
dam
age
a
s
yst
e
m
if
no
t
prop
e
rly
m
anag
ed
. Wh
e
n
it
co
m
es to t
he powe
r gene
rati
ng syst
em
, i
t i
s i
m
po
rta
nt to prote
ct
it
s equipm
ent and
de
vices.
Re
la
y
is
on
e
of
the
m
ai
n
pr
ot
ect
ion
dev
ic
e
that
is
us
ed
widely
al
l
aro
und
the
world
.
It
is
no
rm
al
l
y
us
e
d
with
a
ci
rcu
it
br
ea
ker
.
T
he
m
a
in
aim
of
powe
r
syst
em
pr
otect
ion
is
to
isolat
e
the
fau
lt
y
sect
ion
s
of
the
powe
r
syst
em
from
the
healt
hy
pa
rt
of
the
syst
e
m
du
rin
g
t
he
occ
urre
nce
of
the
fau
lt
c
urren
t
[1
]
.
Thi
s
is
to
ens
ur
e
the
hea
lt
hy
par
t
can
fu
nctio
n
norm
al
ly
witho
ut
an
y
sever
e
dam
a
ge.
syst
em
.
The
sp
ee
d
w
here
the
ci
rcu
it
br
ea
ke
r
disco
nnect
s
de
pends
on
how
fast
the
trip
ping
sig
nal
is
ob
ta
ined
f
r
om
the
protect
ive
rela
y.
It
is
al
ways
i
m
po
rt
ant
f
or
the
sig
na
l
transm
itted
to
the
ci
rc
uit
br
eaker
is
fast
en
ough.
T
her
e
f
ore,
it
dep
e
nds
on
the
relay
sett
ing
s e
sp
eci
al
ly
h
ow
f
ast
they are
set
to react
dur
i
ng
f
a
ult occ
urrenc
e.
The
operati
on
tim
e
of
relay
a
t
healt
hy
sect
ion
can
not
be
too
slo
w
a
nd
to
o
fa
st
beca
us
e
too
sl
ow
will
be
res
ult
in
sy
stem
da
m
age
and
t
oo
fast
ca
n
cause
m
isc
oo
r
din
at
io
n
ti
m
e
of
relay
operat
ion
betwee
n
th
e
two
relay
s
[3
]
.
In
order
to
pr
e
ve
nt
the
m
isc
oo
r
din
at
io
n
tim
e
of
relay
operat
ion
,
a
tim
e
int
erv
al
nee
d
to
be
set
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2
502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
4
, N
o.
1
,
A
pr
il
201
9
:
319
–
326
320
betwee
n
the
two
relay
ope
ra
ti
on
s
ti
m
e
[3
-
4]
.
The
relay
at
fau
lt
y
sect
ion
is
cal
le
d
the
pri
m
ary
relay
whil
e
the
on
e
at the
healt
hy secti
on is ca
ll
ed
the sec
onda
ry r
el
ay
.
Ar
ti
fici
al
ne
ural
netw
orks
(
ANN)
can
be
us
e
d
in
t
he
predict
ion
of
m
i
sco
ordinati
on
t
i
m
e
of
relay
op
e
rati
ons.
A
NN
are
kn
own
as
bio
lo
gical
ly
insp
ire
d
com
pu
te
r
program
s
an
d
a
re
desig
ned
to
sim
ulate
in
t
he
exact
way
ho
w
hum
an
br
ai
n
proc
esses
in
f
or
m
at
ion
.
A
N
N
is
cu
rr
e
ntly
on
e
of
t
he
m
os
t
pr
om
isi
ng
m
od
el
li
ng
te
chn
iq
ue.
T
he
relay
op
erati
on
can
be
trai
ne
d
by
us
i
ng
the
ANN
w
here
th
e
ANN
will
de
te
rm
ine
the
ongo
i
ng
sign
al
s
betwee
n faults a
nd all
o
the
r
c
onditi
ons.
2.
PROBLE
M
FO
RMULAT
ION
The
m
otivati
on
of
t
his
stu
dy
are
to
determ
ine
an
a
ppr
opriat
e
set
ti
ng
s
of
T
MS
an
d
P
S
of
relay
durin
g
fau
lt
occ
urr
e
nc
e,
to
pro
pose
new
relay
operati
on
ti
m
e
a
nd
t
o
predict
m
isc
oo
r
din
at
io
n
tim
e
of
rela
y
us
ing
ANN.
All
of
these
requirem
ents
sh
al
l
be
fu
lfil
le
d
acco
r
ding
to
ty
pe
of
relay
,
li
nea
r
or
no
n
-
li
nea
r
relay
char
act
e
risti
cs, prim
ary and
bac
kup rela
y co
ns
trai
nts
as
w
e
ll
as coordinati
on constrai
nts
[5
-
6].
2.1
Li
near or
Non
-
Li
near Rel
ay
C
h
ar
act
eri
s
tic
The
non
-
li
nea
r
relay
char
ac
te
risti
cs
fu
nction
is
base
d
on
the
sta
nda
rd
IEC
60255
-
4[7].
T
he
char
act
e
risti
cs
of
t
he
relay
is
Norm
al
l
y
In
ve
rse
De
finite
Mi
nim
u
m
Ti
m
e
(I
DMT)
ty
pe
.
T
he
ope
rati
ng
ti
m
e
fo
r
this ty
pe of rel
ay
can be e
xpr
essed
a
s foll
ow
s:
-
i
i
i
T
M
S
L
PS
K
t
1
r
a
t
i
o
p
r
i
m
a
r
y
r
a
t
i
o
o
n
d
a
r
y
sc
CT
c
u
r
r
e
n
t
f
a
u
l
t
x
CT
I
i
s
e
c
1
0
0
s
e
c
r
at
i
o
ond
ar
y
p
CT
x
s
e
t
t
i
n
g
c
u
r
r
e
n
t
I
i
i
i
p
sc
i
I
I
PS
wh
e
re
I
sc
i
is
t
he
fa
ult
cu
rr
e
nt in
CT
sec
onda
r
y
and
I
pi
is
t
he
pick
up
c
urren
t set
ti
ng
o
f
the
i
-
th
relay
.
PS
val
ues
is
the
rati
o
of
fa
ult
current
in
CT
seconda
ry
to
pick
up
cu
rrent
set
ti
ng
.
T
he
TMS
values
is
set
at
0.
5
a
nd
t
he
curr
ent
set
ti
ng
is set
at 1
25%
. T
he
c
onsta
nt
f
act
or
,
K =
0.14
, L
=
0
an
d
α
=
0.0
2
.
2.2
Coordin
ati
on
Constr
aints
In
orde
r
to
pr
e
dict
the
m
isc
oo
r
din
at
io
n
ti
m
e
for
al
l
the
rel
ay
s,
the
val
ues
of
t
he
relay
operati
ng
ti
m
e
of
both
pri
m
ar
y
and
seco
ndar
y
need
to
be
de
te
rm
ined
fi
rst.
The
ex
pressi
on
bel
ow
can
be
us
ed
to
pre
dict
the
m
isc
oo
r
din
at
io
n
ti
m
e.
c
p
s
T
T
T
(5)
Wh
e
re
T
s
is
the
op
er
at
ing
ti
m
e
of
the
seco
nd
a
ry
relay
,
T
p
is
the
op
erati
ng
tim
e
of
the
pr
im
ary
relay
a
nd
T
c
is
the coo
rd
i
natio
n
ti
m
e int
erv
al
w
hic
h varies
from
0
.2
s
–
0.5s
dep
e
nds
on
dif
fer
e
nt circ
um
st
ances.
The
val
ue
of
t
he
(CTI
)
betw
een
pri
m
ary
and
seco
ndary
re
la
y
need
to
be
in
the
range
of
0.
2s
to
0.5s
[8
-
9].
If the C
TI
fall
s i
n
this
range,
t
he rel
ay
p
ai
rs
ca
n be
sai
d
ha
vi
ng
go
od coor
d
i
natio
n.
If the CT
I v
al
ue
fall
s
ou
tsi
de
this
ra
ng
e
an
d
ha
ving
po
sit
ive
valu
e,
the
c
oor
dina
ti
on
ca
n
be
s
ai
d
not
ideal
.
I
f
the
CT
I
valu
e
fall
s
ou
tsi
de
the
ra
ng
e
an
d
ha
ving
neg
at
i
ve
value,
it
ca
n
be
sai
d
that
t
he
r
el
ay
pairs
ha
vi
ng
m
isc
oord
i
natio
n
pro
blem
.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Tit
le
o
f m
anusc
ript is s
ho
rt
and cl
ear
, impli
es rese
ar
ch
res
ults (
First
Au
th
or
)
321
3.
METHO
DOL
OGY
In
t
his
pa
pe
r,
r
adial
netw
ork
i
s
us
e
d
to
analy
ze
the
c
oord
i
na
ti
on
.
T
he
ra
dial
netw
ork
c
hosen
for
this
researc
h
is
a
po
we
r
distri
bu
ti
on
ra
dial
netw
ork
ta
ke
n
from
[1
0].
T
he
sin
gl
e
li
ne
diagr
am
is
dep
ic
te
d
in
Figure
1
.
The
te
st
sy
stem
com
pr
ise
s
of
17
buses,
28
li
ne
s
a
nd
trans
form
ers,
s
ing
le
ge
ner
at
or
a
nd
28overc
urren
t
relay
s.
T
he
c
om
ple
te
n
et
wor
k
is a
naly
zed i
n
te
rm
s o
f
t
he powe
r
fl
ow and the
total
loa
d use
d.
It is im
po
rta
nt to
know the
im
po
rtant
par
am
et
ers
of the
n
et
wor
k
in
or
der
t
o de
te
rm
ine the
f
au
lt
cu
rr
e
nt.
Figure
1
.
17
-
bus test
syst
em
ANN
will
be
us
e
d
to
determ
ine
the
op
ti
m
a
l
op
erati
ng
tim
e
of
the
relay
s
in
the
rad
ia
l
netw
ork.
I
n
oth
e
r
w
ords,
t
he
best
relay
op
erati
ng
ti
m
e
of
the
relay
pair
wh
ic
h
are
the
pr
im
ary
relay
and
t
he
bac
ku
p
rela
y
will
be
determ
ined
by
us
in
g
ANN.
T
he
radi
al
network
wi
ll
be
desig
ne
d
by
us
in
g
the
E
asy
Po
we
r
s
of
t
war
e
.
Ma
nu
al
cal
c
ulati
on
will
be
done
in
orde
r
to
determ
ine
cer
ta
in
pa
ram
et
er
s
su
c
h
as
the
operati
ng
ti
m
e
of
t
he
relay
,
pr
im
ary
cur
r
ent
val
ue,
TMS,
PS
an
d
so
on.
T
he
c
al
culat
ion
will
be
ver
i
fied
by
us
ing
Easy
Powe
r
si
m
ulati
on
r
es
ul
ts.
The
obta
ine
d
va
lues
from
calcu
la
ti
on
,
wh
ic
h
is
ver
ifie
d
by
si
m
ulati
on
is
f
eed
in
into
f
unct
ion
fitt
in
g
trai
ning of AN
N
in MA
TLA
B i
n
orde
r
to
de
te
rm
ine an
im
pro
ved
optim
al
op
e
rati
on tim
e
o
f
t
he
relay
s. Hence
,
the
obta
ine
d
va
lues
f
r
om
the
ANN
ca
n
be
use
d
t
o
produce n
ew
tim
e
set
ti
ng
f
or
the
relay
o
pe
rati
on
a
nd also
to
pr
e
dict m
isc
oo
rd
i
nation t
i
m
e o
f
the
relay
op
erati
on. F
i
gure
2
sim
plifie
s all
the e
xecu
ti
on
process
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2
502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
4
, N
o.
1
,
A
pr
il
201
9
:
319
–
326
322
Figure
2. Re
s
e
arch Flo
wc
har
t
4.
RESU
LT
S
A
ND AN
ALYSIS
The
pro
posed
ANN
was
e
xe
cuted
on
In
te
l
Core
i5
2.5
3
GH
z
with
4
G
B
RAM
an
d
s
i
m
ulate
d
in
MATLAB
.
Th
e
stud
y
rev
eal
e
d
the
feasibil
it
y
of
the
pro
posed
A
NN
to
predict
m
isc
oo
r
di
nation
pr
ob
le
m
.
The
syst
e
m
par
am
e
te
rs
with
fa
ul
t
current
isc
al
cu
la
te
d
and
ta
bu
l
at
es
in
Table
1.
It
can
be
obs
erv
e
d
that
the
fau
lt
s
occurre
d
in
a
transm
issi
on
lin
e
is
the
sy
m
m
et
rical
balanced
three
ph
as
e
fau
lt
.
Acc
ord
ing
to
this
ta
bl
e,
the
si
m
ulati
on
res
ults
an
d
the
ca
lc
ulati
on
re
su
lt
s
wer
e
al
m
os
t
the
sa
m
e
and
the
di
ff
e
ren
ce
wer
e
sm
al
l
and
fall
s
inside t
he
acce
ptable
reg
i
on
wh
ic
h
is
1% o
f
the sim
ulate
d
resu
lt
s.
Table
1.
C
om
par
iso
n betwee
n ca
lc
ulate
d
a
nd
sim
ulatio
n res
ults f
or f
a
ulted
values
Bus
Ca
lcula
ted
Fault
Va
lue
(
k
A)
,
a
Si
m
ula
tion Fault
Va
lue
(
k
A)
, b
Diff
er
ence
(b
a
n
d a)
(
k
A)
Accept
a
ble
rang
e
(
1
%
of
b),
(
k
A)
1
2
6
.24
3
2
6
.24
3
0
.00
0
0
–
0
.26
2
2
1
2
.14
5
1
2
.14
3
0
.00
2
0
–
0
.12
1
3
8
.84
4
8
.84
6
0
.00
2
0
–
0
.08
8
4
8
.26
1
8
.26
2
0
.00
1
0
–
0
.08
3
5
7
.84
7
7
.84
7
0
.00
0
0
–
0
.07
8
6
7
.07
0
7
.07
0
0
.00
0
0
–
0
.07
1
7
1
0
.24
0
1
0
.23
8
0
.00
2
0
–
0
.10
2
8
9
.67
8
9
.68
0
0
.00
2
0
–
0
.09
7
9
7
.70
5
7
.70
5
0
.00
0
0
–
0
.07
7
10
6
.70
3
6
.70
5
0
.00
2
0
–
0
.06
7
11
6
.20
9
6
.20
8
0
.00
1
0
–
0
.06
2
12
8
.74
9
8
.75
1
0
.00
2
0
–
0
.08
8
13
7
.80
4
7
.80
5
0
.00
1
0
–
0
.07
8
14
6
.20
4
6
.20
3
0
.00
1
0
–
0
.06
2
15
5
.97
3
5
.97
3
0
.00
0
0
–
0
.06
0
16
5
.44
3
5
.44
1
0
.00
2
0
–
0
.05
4
17
5
.26
4
5
.26
3
0
.00
1
0
–
0
.
0
5
3
4.1.
Fau
lt
curren
t e
ach
rel
ay
an
d rela
y oper
ati
ng
time
Fr
om
Table
2.
it
can
be
see
n
t
hat
the
cu
rr
e
nt
set
ti
ng
an
d
t
he
pic
kup
c
urre
nt
f
or
al
l
the
re
la
ys
are
the
sam
e.
All
the
relay
s
at
set
to
op
e
rate
at
12
5%
cur
re
nt
set
ti
ng.
It
can
be
se
en
that
the
value
of
p
ic
kup
c
urren
t
is
so
le
ly
dep
e
nd
s
on
cu
rr
e
nt
tra
ns
f
or
m
er
seco
nd
a
ry
rati
o.
T
he
ob
ta
ine
d
PS
value
will
be
use
d
f
or
the
cal
c
ulati
on
of
the
relay
operati
ng
ti
m
e.T
a
ble
2als
o
ta
bula
te
s
the
valu
e
of
relay
ope
r
at
ing
tim
e
fo
r
al
l
relay
s.
In
orde
r
t
o
determ
ine
the
relay
op
e
rati
ng
ti
m
e,
the
value
of
TM
S
is
require
d.
Af
te
r
Ar
ti
fici
al
Ne
ural
Netw
ork
(
ANN)
util
iz
ed,
the
ne
w
relay
operat
ing
tim
e
will
be
determ
ined.
Th
us
,
the
n
e
w
TMS
values
will
be
ob
ta
in
ed
a
nd
will
b
e
pro
po
s
ed fo
r
the
r
el
ay
sett
ing
.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Tit
le
o
f m
anusc
ript is s
ho
rt
and cl
ear
, impli
es rese
ar
ch
res
ults (
First
Au
th
or
)
323
Table
2.
Fault
current
f
or
ea
c
h rela
y an
d
rel
ay
o
pe
rati
ng ti
m
e
Rela
y
Curren
t
Se
tt
ing
(
%
)
Pic
k
up
Curre
nt
(A)
Fault Cu
rren
t
in
CT Seco
nda
ry
(A)
PS
TMS
Rela
y
O
pera
ti
ng
Ti
m
e
(s)
So
u
rce
125
6
.25
8
7
.47
7
1
3
.99
6
0
.10
0
.25
8
1
125
6
.25
4
0
.47
7
6
.47
6
0
.10
0
.36
8
2
125
6
.25
2
9
.48
7
4
.71
8
0
.10
0
.44
4
3
125
6
.25
1
8
.77
7
3
.00
4
0
.10
0
.62
9
4
125
6
.25
1
8
.77
7
3
.00
4
0
.10
0
.62
9
5
125
6
.25
1
6
.52
2
2
.64
4
0
.10
0
.71
3
6
125
6
.25
1
7
.11
3
2
.73
8
0
.10
0
.68
8
7
125
6
.25
1
8
.20
0
2
.91
2
0
.10
0
.64
8
8
125
6
.25
8
8
.37
5
1
4
.14
0
.10
0
.25
7
9
125
6
.25
3
4
.12
7
5
.46
0
.10
0
.40
5
10
125
6
.25
1
6
.49
5
2
.63
9
0
.10
0
.71
4
11
125
6
.25
1
7
.15
6
2
.74
5
0
.10
0
.68
6
12
125
6
.25
1
6
.49
5
2
.63
9
0
.10
0
.71
4
13
125
6
.25
3
8
.52
5
6
.16
4
0
.10
0
.37
8
1
4
125
6
.25
2
1
.96
3
3
.51
4
0
.10
0
.55
0
15
125
6
.25
1
9
.98
8
3
.19
8
0
.10
0
.59
5
16
125
6
.25
6
2
.08
0
9
.93
3
0
.10
0
.29
8
17
125
6
.25
3
9
.77
7
6
.36
4
0
.10
0
.37
1
18
125
6
.25
2
2
.05
0
3
.52
8
0
.10
0
.54
8
19
125
6
.25
2
1
.61
3
3
.45
8
0
.10
0
.55
7
20
125
6
.25
2
8
.19
5
4
.51
1
0
.10
0
.45
8
21
125
6
.25
1
1
.35
0
1
.81
6
0
.10
1
.16
6
22
125
6
.25
1
1
.12
5
1
.78
0
0
.10
1
.20
7
23
125
6
.25
1
2
.23
1
1
.95
7
0
.10
1
.03
6
24
125
6
.25
2
2
.66
7
3
.62
7
0
.10
0
.53
6
25
125
6
.25
2
2
.66
7
3
.62
7
0
.10
0
.53
6
26
125
6
.25
2
0
.35
0
3
.25
6
0
.10
0
.58
6
27
125
6
.25
1
9
.57
8
3
.13
2
0
.10
0.
606
28
125
6
.25
1
8
.48
8
2
.95
8
0
.10
0
.63
8
4.2.
Art
ific
ial
Neu
ral N
e
twor
k (ANN)
In
order
to
unde
rgo
trai
ni
ng
i
n
A
NN,
the
in
pu
ts
f
or
the
net
work
nee
d
to
be
determ
ined.
The
net
work
require
d
tw
o
in
pu
ts
w
hich
are
input data a
nd
ta
rg
et
d
at
a.
Th
e input data is t
he
relay
ope
rati
ng
ti
m
e as in
Tab
l
e
2 wh
il
e tar
get
data is s
how
n
i
n
Ta
ble
3.
Table
3.
A
NN
ta
rg
et
data
Rela
y
Rela
y
O
pera
ti
ng
Ti
m
e
(
s)
Rela
y
Rela
y
O
pera
ti
ng
Ti
m
e
(
s)
So
u
rce
0
.90
0
1
0
.80
0
15
0
.50
0
2
0
.70
0
16
0
.40
0
3
0
.60
0
17
0
.90
0
4
0
.60
0
18
0
.80
0
5
0
.50
0
19
0
.80
0
6
0
.50
0
20
0
.70
0
7
0
.50
0
21
0
.60
0
8
0
.40
0
22
0
.60
0
9
0
.80
0
23
0
.60
0
10
0
.70
0
24
0
.50
0
11
0
.70
0
25
0
.50
0
12
0
.70
0
26
0
.40
0
13
0
.60
0
27
0
.40
0
14
0
.50
0
28
0
.40
0
As
in
Table
3
.
the
relay
op
era
ti
ng
tim
e
is
the
ta
rg
et
data
for
ANN
trai
ning
[1
1].
T
he
m
ain
reas
on
is
the
relay
w
hich
is
nea
rer
to
the
power
s
ource
need
to
op
erate
slo
wer
ti
m
e
com
par
ed
to
the
relay
w
hich
is
furthe
r
to
the
powe
r
s
ource
. T
his is to
ensu
re
that the
re is
no m
isc
oo
r
din
at
ion
betwee
n re
la
y op
e
rati
ons.
Table
4
.
ta
bu
l
at
es
the
ou
t
pu
t
of
A
N
N
w
hic
h
dif
fers
with
the
both
in
pu
t
data
and
ta
rget
data.
The
ou
t
pu
t
of
A
N
N
is
base
d
on
be
st
fit
or
cu
r
ve
f
it
ti
ng
te
chn
i
qu
e
by
us
in
g
in
put
an
d
ta
rg
et
da
ta
wh
ic
h
was
giv
e
n
to
the
A
N
N.As
ind
ic
at
ed
i
n
T
able
4,
t
he
er
r
ors
val
ues
are
th
e
di
ff
e
re
nce
be
tween
the
ta
rg
e
t
data
an
d
the
ANN
ou
t
pu
t.
The
po
sit
ive
values
of
errors
in
dicat
es
that
the
ANN
outp
ut
is
faste
r
tha
n
the
ta
rg
e
t
data.
T
he
negat
ive
values
in
dicat
e
s
that
the
ANN
ou
t
pu
t
is
slo
w
er
than
t
he
ta
rget
data.
It
can
be
note
d
t
hat
m
or
e
tha
n
50
%
of
al
l
relay
s
hav
e
ti
m
e
i
m
pr
ov
em
ent
based
on
the
ANN
outp
ut.
Hen
ce
,
the
A
N
N
outp
ut
can
be
acce
pted
as
the
ne
w
relay
operati
ng
tim
e. Th
e
ANN ou
t
pu
t al
s
o wil
l be
us
e
d
f
or m
isc
oo
r
din
at
ion p
re
dicti
on
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2
502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
4
, N
o.
1
,
A
pr
il
201
9
:
319
–
326
324
Table
4.
A
NN
ou
t
pu
t t
im
e fo
r
all
r
el
ay
s
Rela
y
Inpu
t
t
i
m
e (
s
)
Ta
rg
et
ti
m
e (
s
)
ANN
O
utp
ut
Ti
m
e
(
s)
Erro
r
(s)
Rela
y
Inpu
t
ti
m
e (
s
)
Ta
rg
et
ti
m
e (
s
)
ANN
O
utp
ut
Ti
m
e
(
s)
Erro
r
(s)
So
u
rce
0
.25
8
0
.90
0
0
.42
0
0
.48
0
15
0
.59
5
0
.50
0
0
.43
5
0
.06
5
1
0
.36
8
0
.80
0
0
.49
5
0
.30
6
16
0
.29
8
0
.40
0
0
.45
6
-
0
.05
6
2
0
.44
4
0
.70
0
0
.62
8
0
.07
2
17
0
.37
1
0
.90
0
0
.50
0
0
.40
0
3
0
.62
9
0
.60
0
0
.41
5
0
.18
5
18
0
.54
8
0
.80
0
0
.56
4
0
.23
6
4
0
.62
9
0
.60
0
0
.41
5
0
.18
5
19
0
.55
7
0
.80
0
0
.54
3
0
.25
7
5
0
.71
3
0
.50
0
0
.59
0
-
0
.09
0
20
0
.45
8
0
.70
0
0
.62
4
0
.07
6
6
0
.68
8
0
.50
0
0
.55
7
-
0
.05
7
21
1
.16
6
0
.60
0
0
.57
4
0
.02
7
7
0.
688
0
.50
0
0
.44
4
0
.05
6
22
1
.20
7
0
.60
0
0
.59
2
0
.00
8
8
0
.25
7
0
.40
0
0
.41
9
-
0
.01
9
23
1
.03
6
0
.60
0
0
.66
5
-
0
.06
5
9
0
.40
5
0
.80
0
0
.58
5
0
.21
5
24
0
.53
6
0
.50
0
0
.58
3
-
0
.08
3
10
0
.71
4
0
.70
0
0
.59
1
0
.10
9
25
0
.53
6
0
.50
0
0
.58
3
-
0
.08
3
11
0
.68
6
0
.70
0
0
.55
2
0
.14
8
26
0
.58
6
0
.4
00
0
.45
6
-
0
.05
6
12
0
.71
4
0
.70
0
0
.59
1
0
.10
9
27
0
.60
6
0
.40
0
0
.41
9
-
0
.01
9
13
0
.37
8
0
.60
0
0
.51
4
0
.08
6
28
0
.63
8
0
.40
0
0
.42
5
-
0
.02
5
14
0
.55
0
0
.50
0
0
.56
0
-
0
.06
0
Figure
3. Per
f
orm
ance p
lot
of
the traine
d A
N
N
Fi
gure
4. S
ta
te
p
lot
of
t
he t
r
ai
ned AN
N
Figure
3
a
nd
4
disp
la
ys
the
pe
rfor
m
ance
plo
t
of
the
trai
ne
d
A
NN
wh
ic
h
giv
e
s
the
be
st
validat
ion
perform
ance
of
0.0
26439
at
e
po
c
h
1
an
d
the
sta
te
plo
t
of
t
he
trai
ne
d
ANN
w
her
e
t
he
A
NN
trai
ning
ga
ve
th
e
best
netw
ork
perform
ance
fu
nctio
n
at
gr
a
die
nt
of
0.000
2573
at
epo
c
h
7.
Figure
5
s
hows
the
re
ge
ressio
n
perform
ance p
l
ot whe
re th
e
r
e
gr
essi
on c
orrelat
ion
is
0.2
9098.
Figure
5. Re
gr
ession pe
rfo
rm
ance
plo
t
4.3.
N
ew
R
el
ay O
pera
ting T
im
e
The
ne
w
r
el
ay
op
e
rati
ng
tim
e
is
ob
ta
ine
d
f
ro
m
the
ANN
ou
t
pu
t
as
in
T
able
4.
T
hus,
t
he
ne
w
TM
S
values
w
il
l be
determ
ined
by
us
in
g
e
q. 1. T
he
n
e
w
TMS
v
al
ues
a
re tab
ulat
es in Ta
ble
5.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Tit
le
o
f m
anusc
ript is s
ho
rt
and cl
ear
, impli
es rese
ar
ch
res
ults (
First
Au
th
or
)
325
Table
5.
T
he
new TMS
v
al
ue
s
Rela
y
New
Relay
O
pera
ting
Ti
m
e
(
s)
TMS
Rela
y
New
Relay
O
pera
ting
Ti
m
e
(
s)
TMS
Rela
y
New
Relay
O
pera
ting
Ti
m
e
(
s)
TMS
So
u
rce
0
.42
0
0
.16
10
0
.59
1
0
.08
20
0
.62
4
0
.14
1
0
.49
5
0
.13
11
0
.55
2
0
.08
21
0
.57
4
0
.05
2
0
.62
8
0
.14
12
0
.59
1
0
.08
22
0
.59
2
0
.05
3
0
.41
5
0
.07
13
0
.51
4
0
.14
23
0
.66
5
0
.06
4
0
.41
5
0
.07
14
0
.56
0
0
.10
24
0
.58
3
0
.11
5
0
.59
0
0
.08
15
0
.43
5
0
.07
25
0
.58
3
0
.11
6
0
.55
7
0
.08
16
0
.45
6
0
.15
26
0
.45
6
0
.08
7
0
.44
4
0
.07
17
0
.50
0
0
.13
27
0
.41
9
0
.07
8
0
.41
9
0
.16
18
0
.56
4
0
.10
28
0
.42
5
0
.07
9
0
.58
5
0
.14
19
0
.54
3
0
.10
4.4.
Predi
ction
of
mi
scoo
r
dina
ti
on
time be
twe
en rela
ys
Pr
e
dicti
on
of
m
isc
oo
r
din
at
io
n
tim
e
can
be
done
by
look
ing
at
the
dif
f
eren
ce
betwee
n
the
relay
op
e
rati
ng
tim
e
of
pri
m
ary
re
la
ys
and
seco
ndary
relay
s.
T
he
dif
fer
e
nce
is
cal
le
d
coord
i
nation
ti
m
e
interval
(CTI
).
T
he
CT
I
can
be
deter
m
ined
by
usi
ng
the
e
q.
5.
T
he
relay
op
e
rati
ng
ti
m
e
cal
cul
at
ion
are
no
t
ba
sed
on
the
A
N
N
ou
t
pu
t
but
the
c
al
culat
ed
ou
t
put.
T
he
c
om
par
iso
n
betwee
n
the
m
isc
oo
r
din
at
io
n
ti
m
e
of
t
he
cal
culat
ed ou
t
put an
d
t
he AN
N ou
t
pu
t i
s
sho
wn in T
able
6
.
belo
w.
Table
6.
Mi
sc
oor
din
at
io
n
ti
m
e
b
et
wee
n rela
y
pairs
Rela
y
P
a
ir
Ca
lcula
ted
O
utp
u
t
Ti
m
e(s
)
ANN
O
ut
put
Ti
me (s)
CTI
(
s)
M
isco
o
rdina
tion
CTI
(
s)
M
isco
o
rdina
tion
R
s
o
u
r
ce
an
d
R1
-
0
.11
0
Yes
-
0
.07
5
Yes
R1
and
R2
-
0
.07
6
Yes
-
0
.13
3
Yes
R2
and
R3
-
0
.18
5
Yes
0
.21
3
NO
R2
and
R4
-
0
.18
5
Yes
0
.21
3
NO
R3
and
R5
-
0
.08
4
Yes
-
0
.17
5
Yes
R3
and
R6
-
0
.05
9
Yes
-
0
.14
2
Yes
R3
and
R7
-
0
.01
9
Yes
-
0
.02
9
Yes
R4
and
R5
-
0
.08
4
Yes
-
0
.17
5
Yes
R4
and
R6
-
0
.05
9
Yes
-
0
.14
2
Yes
R4
and
R7
-
0
.01
9
Yes
-
0
.02
9
Yes
R5
and
R8
0
.45
6
NO
0
.17
1
NO
R6
and
R8
0
.43
1
NO
0
.13
8
NO
R7
and
R8
0
.39
1
NO
0
.02
5
NO
R1
and
R9
-
0
.03
7
Yes
-
0
.09
0
Yes
R9
and
R1
0
-
0
.30
9
Yes
-
0
.00
6
Yes
R9
and
R1
1
-
0
.28
1
Yes
0
.03
3
NO
R9
and
R1
2
-
0
.30
9
Yes
-
0
.00
6
Yes
R2
and
R1
3
0
.06
6
NO
0
.11
4
NO
R1
3
and
R1
4
-
0
.17
2
Yes
-
0
.04
6
Yes
R1
3
and
R1
5
-
0
.21
7
Yes
0
.07
9
NO
R1
4
and
R1
6
0
.25
2
NO
0
.10
4
NO
R1
5
and
R1
6
0
.29
7
NO
-
0
.02
1
Yes
R1
and
R1
7
-
0
.00
3
Yes
-
0
.00
5
Yes
R1
7
and
R1
8
-
0
.17
7
Yes
-
0
.06
4
Yes
R1
7
and
R1
9
-
0
.18
6
Yes
-
0
.04
3
Yes
R2
and
R2
0
-
0
.01
4
Yes
0
.00
4
NO
R2
0
and
R2
1
-
0
.70
8
Yes
0
.05
0
NO
R2
0
and
R2
2
-
0
.74
9
Yes
0
.03
2
NO
R2
0
and
R2
3
-
0
.57
8
Yes
-
0
.04
1
Y
es
R2
1
and
R2
4
0
.63
0
NO
-
0
.00
9
Yes
R2
1
and
R2
5
0
.63
0
NO
-
0
.00
9
Yes
R2
2
and
R2
4
0
.67
1
NO
0
.00
9
NO
R2
2
and
R2
5
0
.67
1
NO
0
.00
9
NO
R2
3
and
R2
4
0
.50
0
NO
0
.08
2
NO
R2
3
and
R2
5
0
.50
0
NO
0
.08
2
NO
R2
4
and
R2
6
-
0
.05
0
Yes
0
.12
7
NO
Rela
y
P
a
ir
Ca
lcula
ted
O
utp
u
t
Ti
m
e(s
)
ANN
O
ut
put
Ti
me (s)
CTI
(
s)
M
isco
o
rdina
tion
CTI
(
s)
M
isco
o
rdina
tion
R2
4
and
R2
7
-
0
.07
0
Yes
0
.16
4
NO
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2
502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
1
4
, N
o.
1
,
A
pr
il
201
9
:
319
–
326
326
Rela
y
P
a
ir
Ca
lcula
ted
O
utp
u
t
Ti
m
e(s
)
ANN
O
ut
put
Ti
me (s)
CTI
(
s)
M
isco
o
rdina
tion
CTI
(
s)
M
isco
o
rdina
tion
R2
4
and
R2
8
-
0
.10
2
Yes
0
.15
8
NO
R2
5
and
R2
6
-
0
.05
0
Yes
0
.12
7
NO
R2
5
and
R2
7
-
0
.07
0
Yes
0
.16
4
NO
R2
5
and
R2
8
-
0
.10
2
Yes
0
.15
8
NO
Table
6
.
T
ab
ul
at
es
the
com
par
iso
n
of
m
isco
or
din
at
io
n
ti
m
e
between
th
e
cal
culat
ed
va
lues
an
d
th
e
ANN
outp
ut.
F
ro
m
the
ta
ble,
it
can
be
see
n
that
the
A
N
N
outp
ut
produces
le
ss
m
isc
oo
r
din
at
ion
c
om
par
ed
t
o
the
cal
culat
ed
ou
t
pu
t.
T
he
ca
lc
ulate
d
ou
t
put
pro
du
ce
s
12
pair
s
of
relay
s
w
hich
s
how
s
go
od
co
ordi
natio
n
betwee
n
relay
pairs.
Howe
ve
r,
the
A
NN
outpu
t
pro
duces
22
pairs
of
rel
ay
s
wh
ic
h
s
hows
good
co
ord
inati
on,
wh
ic
h
m
eans
m
or
e
than
50
%
of
t
otal
rela
y
pairs.
T
he
t
ot
al
relay
pairs
are
41
pai
rs.
Mi
scoord
i
natio
n
ti
m
e
i
s
the
CTI
val
ues
w
hich
ha
ve
ne
gative
values
.
For
e
xam
ple,
if
a
pair
of
rela
ys
have
the
CTI
values
of
-
0.1
29
s,
then
the
m
isc
oo
r
din
at
io
n
ti
m
e
will
be
0.129
s.
T
he
res
ults
in
Table
6
.
al
so
pro
ves
that
the
ANN
out
pu
t
is
m
or
e
reli
able
since
it
is
no
t
only
produces
l
ess
m
isc
oo
r
dina
ti
on
tim
e,
bu
t
al
so
pro
duces
f
ast
er
relay
ope
rati
ng
tim
e.
5.
CONCL
US
I
O
N
ANN
Le
venbe
r
-
Ma
e
qu
a
rd
t
al
gorithm
has
be
en
prese
nted
in
this
pa
per
t
o
predict
m
isco
or
din
at
io
n
tim
e
that
occu
rr
e
d
bet
wee
n
relay
pairs.
From
the
resu
lt
s
ob
ta
i
ned,
it
can
be
re
veale
d
that
the
pro
po
s
e
d
te
chn
iq
ue
dem
on
st
rates
the
s
ign
ific
a
nt
res
ul
ts
betweem
A
NN
outp
ut
a
nd
cal
c
ulate
d
outp
ut.
Ba
se
d
on
t
he
stud
y,
t
he
A
N
N
outp
ut
are
s
uitable
tha
n
ca
lc
ulate
d
outp
ut
since
it
sh
ows
good
co
ordi
na
ti
on
.
M
or
e
ov
e
r
,
A
NN
ou
t
pu
t
pro
duce
s less
m
isc
oo
r
di
nation
ti
m
e and
faster
relay
operati
ng ti
m
e.
ACKN
OWLE
DGE
MENT
The
aut
hors
w
ou
l
d
li
ke
to
ac
knowle
dge
the
su
pp
or
t
f
ro
m
the
Fundam
ent
al
Re
search
G
r
ant
Schem
e
(F
RG
S)
un
der
a
grant
num
ber
of
FR
GS
/
1/20
16
/T
K
04
/
UNI
MAP/0
2/9
f
rom
the
Mi
nistry
of
E
ducat
ion
(
MOE)
Ma
la
ysi
a
.
REFERE
NCE
S
[1]
Y.G.
Paint
h
aka
r
,
Bhide,
S.R
.
Fundam
ent
al
s
of
Pow
er
S
y
st
em
Protection
.
5
th
Ed
i
tion
.
New
Delh
i
:
P
re
nti
c
e
-
Hall
of
India
Priv
at
e
L
i
m
it
ed
.
2007
:
1
-
2
.
[2]
[2]
L.
Hewitson.
Prac
tica
l
Pow
er
S
y
stem
Prote
ct
i
on
for
Engi
n
ee
r
s
and
Techni
ci
a
ns.
R
ev.
8
.
1.
IDC
Te
chno
logies,
2011:
3
[3]
[3]
M.H.
Hus
sain,
I
.
Mus
iri
n,
A.
F.
Abidin,
S.R.
A.
Rahi
m
.
Appli
ca
t
ion
of
Modifi
ed
Firef
l
y
Algor
it
hm
for
Solvin
g
Dire
ctional Over
cur
re
nt
Rela
y
.
J
ournal
of El
e
ct
ri
cal
S
yste
ms
“Speci
a
l
Iss
ue AMPE
2015”
.
2015
.
120
-
127
[4]
[4]
M.H.
Hus
sa
in,
I
.
Mus
iri
n,
A.F.
Abidin,
S.
R.
A.
Rah
im.
Multi
-
Obje
ctive
A
pproa
ch
for
Solving
Dire
c
ti
on
al
Overc
urre
nt
R
el
a
y
Problem
using
Modifie
d
Firefly
Al
gorit
hm
.
Inte
rn
ati
onal
Journa
l
of
Computing,
Comm
unic
ati
ons
&
Instrum
ent
ati
on
Engi
n
ee
ring
(
IJ
CCIE)
.
2016.
V
ol.
3
(Iss
ue
1):
21
-
26
[5]
[5]
M.
Ez
ze
din
,
R.
Ka
cz
m
are
k
,
A
novel
m
ethod
for
opti
m
al
coor
dination
o
f
dire
c
ti
ona
l
ov
erc
urre
n
t
re
l
a
y
s
conside
ring
their
availa
b
le
discrete
se
tt
ings
and
s
eve
ra
l
op
era
t
ion
cha
r
acte
rist
ic
s,
El
e
ct
ric
Powe
r
Syste
m
R
ese
arch
,
2011.
vol
.
8
1
,
14
75
-
1481.
[6]
[6]
A.
Li
u,
M.T
,
Yang.
A
New
H
y
brid
Neld
er
-
Me
ad
Parti
cle
Sw
arm
Optimiza
ti
on
for
Coordina
ti
on
Optimiza
ti
on
of
Dire
ctional Over
cur
re
nt
Rela
y
s,
J
ournal
Mathe
ma
ti
cal P
robl
ems i
n
Engi
n
ee
ring
,
2
011.
vol. 2012, 1
-
18.
[7]
[7]
N.Z
.
Jam
al,
O.
Aliman,
M.H.
S
u
la
iman.
Inve
rse
Defi
nite
Over
c
urre
nt
Re
lay
Dis
cri
m
ina
t
ion
Alg
orit
hm
and
Its
Applic
a
ti
on
In
I
ndustria
l
Pow
er
S
y
stems
.
ARPN
Journal
of
Eng
i
nee
ring
and
App
li
ed
S
ci
en
ce
s
.
20
15.
Vol.
10
(No.
23),
17544
-
1754
8.
[8]
[8]
A.J.
Onah.
Rel
a
y
Coordin
ation
in
the
Prote
ct
ion
of
Radially
-
Conne
cted
Pow
er
S
y
stem
Network.
Nige
ria
n
Journal
of
Tech
nology
(
NIJOTECH)
.
2012.
Vol.
31
(No.
1)
.
58
-
62
[9]
[9]
M.H.
Hussai
n,
I.
Mus
iri
n
,
A.
F.
Abidin,
S.R.
A.
Rahi
m
.
Modi
fie
d
Sw
arm
Firefly
Algor
it
hm
for
opti
m
al
setting
s
of
Dire
c
ti
on
al
O
ver
cur
re
n
t
R
el
a
ys.
Applied
Me
ch
anic
s and
Ma
te
r
ial
s.
2015
.
Vol
.
793.
473
-
477
.
[10]
[10]
Network
Protection &
Auto
m
at
ion
Guide
.
A
lstom
Grid.
2011
.
[11]
[11]
O.
Em
m
anue
l,
M.
L.
Othm
a
n,
H.
Hiza
m
,
M
.
M.
Othm
an.
Ar
ti
ficia
l
Neura
l
Network
Applicati
on
in
Opt
imal
Coordina
ti
on
o
f
Dire
ctional
Ov
erc
urre
n
t
Prote
c
ti
ve
R
el
a
y
s
in
El
e
ct
ri
ca
l
M
esh
Distribut
ion
N
et
work.
App
li
ed
Me
chanics and Mate
rials
.
2015.
Vol.
785.
48
-
52
.
Evaluation Warning : The document was created with Spire.PDF for Python.