TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 8, August 201
4, pp. 5969 ~ 5975
DOI: 10.115
9
1
/telkomni
ka.
v
12i8.608
5
5969
Re
cei
v
ed Ap
ril 9, 2014; Re
vised Ma
y 24
, 2014; Accep
t
ed Jun
e
10, 2014
Fuzzy Neural Network for Classificatio
n Fault in
Protection System
Azriy
e
nni*
1
, Mohd
Wa
zir
Musta
f
a
2
, Naila Zareen
3
1
F
a
cult
y
of Ele
c
trical Eng
i
ne
e
r
ing, Un
iversiti
T
e
knologi Ma
la
ysi
a
,
Skuda
i, 813
10,
Johor Bahr
u, Mala
ysi
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: azri
yen
n
i@
g
m
ail.com
A
b
st
r
a
ct
Novel i
n
tel
lig
e
n
t techniq
ue is
a comb
in
ation
of
fu
zz
y
a
nd
neur
al netw
o
rk
techniq
ues th
at can be
used
to c
l
assif
y
faults
in
el
e
c
tric pow
er sy
stem
protecti
o
n
. T
here
h
a
ve
tw
o pro
b
le
ms
in
the
prot
ectio
n
system, w
h
ic
h
are:
un
desir
e
d
trip
pin
g
and
fail
to
oper
ate. Loss
of
po
w
e
r supp
ly to
rel
a
ys a
n
d
ci
rcuit
break
ers or failure in pr
otective dev
ices
m
a
y cause
failur
e
s in pr
otection system
. Cons
truction of neural
netw
o
rks to explor
e fact to identify f
ault co
mpon
ent is from
control ce
nter
. T
he obj
ective
of this paper is
to
develop
novel concept for
classification f
a
ilur
e
s pr
ot
ection system
ar
e using Fu
z
z
y
Neural Network
techni
qu
e. Method
olo
g
y co
nsi
s
ts of Neur
al N
e
tw
ork
and F
u
zz
y
.
T
he
Ne
ur
al n
e
tw
ork is al
so consc
i
e
n
tio
u
s
for esti
matin
g
degr
ee
of
me
mb
ershi
p
i
n
sy
stem c
o
mpo
n
e
n
ts from corre
spon
din
g
ar
ea
in cl
assific
a
tio
n
o
f
disor
ders. T
h
e
input var
i
a
b
le
s of neur
al n
e
tw
ork form
ed
o
f
binary n
u
m
b
e
rs. Valu
e of 1 ind
i
cates
if faul
t
occurs an
d va
lue of
0 in
dica
t
e
s no-f
ault
oc
curs.
F
u
zz
y r
e
lat
i
o
n
s w
ill rep
r
esent
by f
u
zzy.
T
hese F
u
zzy
relati
ons can b
e
repres
ented
by fu
zz
y
di
agra
m
cons
isti
ng of
three sets of n
ode that w
ould
be consi
der
ed
to
repres
ent co
mpon
ents, rel
a
y
s
and
circ
ui
t bre
a
k
e
r
s. Fu
z
z
y
d
i
ag
ram
i
s
bui
l
t
as c
ausa
l
o
perati
on
of rel
a
ys
and c
i
rcuit br
e
a
kers o
n
ev
en
t of the fault i
n
protec
ti
on s
ystem. T
he c
a
usality
is repr
e
s
ented
in
arro
w
s
.
F
i
nally,
the
con
c
ept of
F
u
zz
y
Neur
al
Netw
ork can
b
e
pr
o
p
o
s
ed as altern
ative
to
solv
e iss
ue of
fai
l
ures
t
hat
occur in the pr
otection system
.
Ke
y
w
ords
: cla
ssificatio
n
fault
,
failures, fu
zz
y
,
me
mbers
h
i
p
degr
ee, ne
ural
netw
o
rks
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Hybrid
intelli
gent te
chni
q
ue i
s
com
p
o
s
ed
of
fu
zzy and
ne
ural
netwo
rk te
ch
nique, i
n
whi
c
h fuzzy tech
niqu
es a
r
e use
d
as fa
ult identif
icati
on and n
eura
l
network te
chniqu
es a
r
e
use
d
as a cla
s
sification fault. There have two
probl
em
s
in the prote
c
tion
system, whi
c
h are: unde
si
red
tripping a
nd fail to operate
.
Missin
g
po
wer
sup
p
ly to relay or failu
res in
prote
c
t
i
ve device
s
can
cau
s
e
t
he f
a
ilure
s.
W
hen
sy
st
em
di
sruption
occu
rs, the
syst
em o
f
pr
o
t
ec
tio
n
fa
ilur
e
s
ca
n
gene
rate d
a
m
age to eq
u
i
pment re
sult
ing in several importa
nt element
s out
of the syst
em.
Disoperation of protection syst
em can reduce lev
e
l of sy
stem
reliability. In some
cases,
unde
sired t
r
i
pping
can
occur for fault
outsid
e
p
r
ote
c
tion
zo
ne. If prote
c
tive d
e
vice
doe
s
n
o
t
operate pro
p
e
rly whe
n
fau
l
t occu
rs, the
n
back-up prot
ection devi
c
e
s
will isol
ate faults in are
a
of
transmissio
n
system. After fault o
c
curs to p
o
we
r
st
ream
in th
e
system
will
chang
e d
ue t
o
topology
cha
nge. It will l
e
ad to ove
r
lo
ading
and
tri
pping
due
to
overlo
ad. Protection
sy
stem
comp
one
nts su
ch
a
s
Cu
rrent
Tra
n
sfo
r
mer (CT
)
, Vo
ltage Tran
sfo
r
mer (VT
)
, Relay, and
Circuit
Breaker
will fail unknown. For this m
o
del, the ci
rcuit
breaker can not obvious fault current i
s
tripped
d
ue
t
o
a
failure
in
the mechan
ism. In the case of circuit
brea
ker fail
ure, und
esi
r
e
d
tripping
can
occur,
and
re
lays o
n
a
d
ja
cent bu
s fa
ult may o
c
cur
outsid
e
of
pri
m
ary p
r
ote
c
ti
on
zon
e
.
In this
re
se
a
r
ch,
explain
e
d
that m
e
ssage
s fr
om control ce
nter
as
so
ciate
d
with
fault
prote
c
tion
de
vice di
sop
e
ra
tion. In many
ca
se
s, it
is
difficult to ma
ke
co
nclu
sio
n
s
about
what
h
ad
happ
ened. E
s
pe
cially, wh
en p
r
ote
c
tion
schem
e d
o
e
s
n
o
t wo
rk
well and
comm
unication fail
ure
s
may o
ccu
r. T
o
ide
n
tify fault prote
c
tion
device, it
i
s
very impo
rta
n
t to be
ha
n
d
led i
n
real ti
me.
Whe
n
a fault occurs on p
o
w
er
system
compon
ent
s, then statu
s
of adjace
n
t co
mpone
nts will
be
affected whe
n
the prote
c
tion syste
m
op
erated.
Many intellige
n
t techniq
ues applicatio
n for f
ault has b
een propo
se
d
in literature [1-3], [5-
6], [9-10]. So
uz
a
et.al
(20
0
4
) pre
s
e
n
ted
a combi
natio
n method use
s
Neu
r
al Networks and Fu
zzy
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
0
46
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 596
9 –
5975
5970
Logi
c th
roug
h
ala
r
m p
r
o
c
e
ssi
ng
and
ide
n
tification
of f
ault [1]. In th
e same
time,
Yu
et.al
(2
0
04)
prop
osed fuzzy identificati
on usi
ng fuzzy
neural net
work [2]. And also, Flau
zino
et.al
(20
12)
recomme
nde
d Hybri
d
Intelligent a
r
chitecture fo
r th
e fault identi
fication in p
o
w
er
distri
buti
on
system [3]. It detect
s
fault
in the tran
smissi
on lin
e
and
cla
ssifi
ca
tion schem
e
based o
n
si
n
g
le
measurement
using
sinu
soi
dal wavefo
rm
[4]. Kemal
et.al
(2009
) al
so expre
s
sed i
m
pleme
n
tatio
n
of the
Neu
r
o-Fu
zzy infe
rence
system
in el
ect
r
ical
tran
smi
ssi
o
n
line
protection to a
d
d
r
ess
different i
s
su
es i
n
power
system
[5].
Then, A
z
iz
et.al
(20
11) p
r
opo
sed
ap
pli
c
ation
the
hi
gh
impeda
nce d
e
tection
of fa
ult and
classification i
n
di
stributio
n
systems [6]. In
the p
r
o
b
lem
o
f
cla
ssifi
cation
fault in po
we
r sy
stems u
s
i
ng intellig
ent
techni
que
h
a
s b
e
en
re
po
rted by
seve
ral
resea
r
chers [
1
, 2, 5, 9, 1
0
]. There
are
also, fa
ult identificatio
n
i
n
po
we
r sy
stem di
stributi
o
n
system
usi
n
g
hybrid
intelli
gent h
a
ve be
en d
e
cla
r
e
d
by re
sea
r
che
r
s [3,
6]. Ho
wever, to the
b
e
st
our
kn
owle
d
ge, they hav
e not b
een
discu
ssi
ng a
b
ou
t
fault cla
ssifi
cation
i
n
failure prote
c
tion
system.
The
aim of thi
s
st
udy was to
propo
se th
e n
o
v
el co
ncept f
o
r fa
ult cl
assif
i
cation
a
s
wel
l
as
hybrid intellig
ent techni
que
and to devel
op ne
w stru
ct
ure in Fu
zzy Neu
r
al Netwo
r
k.
2. Res
earc
h
Method
For a
clearer illustration of failures protection
system in Figure
1 is
simpl
e
electri
c
al
pow
er
sy
st
e
m
.
Two
sy
st
e
m
s
G1
and
G
2
s
our
ce
e
n
e
r
gized
co
nne
cted i
n
seri
es and
ea
ch
on
e i
s
limited by the circuit bre
a
ke
r. G1 is
a prot
e
c
tion
system
com
pone
nt for b
a
ckup p
r
ote
c
tion
system
comp
onent G2. Th
e system
of p
r
otectio
n
G2
acts a
s
a b
a
ckup
system f
o
r the p
r
ote
c
tio
n
of G1.
There are two kind
s of failure
s in
the protection
syste
m
, namely:
a)
Failed to ope
rate
b) Un-t
rippi
ng
Figure 1. Model of Powe
r System
Protectio
n
sy
stem is d
e
si
gned to i
s
ol
ate
system
compon
ents i
n
power
syst
ems for
every time a fault occurs. It should b
e
d
one very
qui
ckly to lowe
r ri
sk of d
a
mag
e
to the syste
m
prote
c
tion d
e
v
ices. Spe
e
d
,
selectivity and coo
r
din
a
t
ion are
so
me of the most ho
ped
for
cha
r
a
c
teri
stic in protectio
n
system. The
n
, prot
e
c
tion device
mu
st operate
and coo
r
din
a
ted
i
n
protection
scheme to ensure that
only
component of fault will be
disconnected. The protecti
on
device shoul
d also have b
a
ckup p
r
ote
c
tion. If protec
tive device is resp
on
sible fo
r isolatin
g fau
l
t
comp
one
nt that is not workin
g pro
perly,
then other protection device
s should
wo
rk.
Cla
ssifi
cation
fault defined indepe
nde
ntly of faul
t
detectio
n
[4]. This traini
n
g
pro
c
ed
ure
is
sep
a
rate
d according to di
stance p
r
ote
c
tion sc
hem
es i
n
a variety of system
condit
i
ons.
Clas
s
i
fic
a
tion error is
as
follows
:
a)
Cla
ss 1 for Si
ngle Lin
e
to Grou
nd (SL
G
)
b)
Cla
ss 2 for T
r
iple Line to G
r
oun
d (TL
G
)
In this meth
odolo
g
y, Fuzzy Sets are
capa
ble of
handlin
g problem
s and
certai
nly
qualitative inf
o
rmatio
n p
r
o
v
ided by hu
man exp
e
rts
based o
n
the
i
r kno
w
ledg
e
and
experi
e
n
ce
that given
so
lution in
som
e
p
r
oble
m
. B
e
sid
e
tha
t,
Neural
Netwo
r
ks a
r
e
difficu
lt to re
prese
n
t
some
of quali
t
ative data, generaliz
ation
ability and fa
ult toleran
c
e.
Fuzzy Sets repre
s
e
n
t so
me
stru
ctural rela
tionshi
p bet
ween
several d
i
fferent pa
tterns of fault i
n
system
com
p
onent
s. He
re
is
a Fuzzy de
scribe
stru
cture to establi
s
h p
a
tterns
of system comp
one
nt fault. While
the tempora
r
y
output of Fuzzy Relatio
n
s t
o
be achieve
d
are:
a)
0 for not fault con
d
ition
s
.
b)
1 for fault con
d
itions
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fuzzy
Neural Network
for Cl
as
s
i
fic
a
tion Fault in Prot
ection Syste
m
(Azriye
n
ni)
5971
For e
a
ch fuzzy diag
ram
con
s
i
s
ts of i
nput varia
b
le
s rel
a
ys a
n
d
circuit brea
kers
can
operate in fa
ult con
d
ition
involving co
mpone
nts in
the are
a
mo
nitored. T
h
e
input varia
b
le
in
binary num
be
rs; be eq
ual to 1 if a signal
has bee
n re
ceived o
r
equ
al to 0 if signal has not be
en
received.
Figure 2. Fuzzy Diag
ram f
o
r Tran
smissi
on Line
Fault cl
assifi
cation i
s
ta
ken fro
m
vari
ous
fault ci
rcumstan
ce
s,
carri
ed o
u
t for variou
s
types of fault, such a
s
sin
g
le line to ground an
d
triple line to ground. As value of stru
ctu
r
e
Neu
r
al Network in
put voltage and
current is to get
fault classification. Fi
gure
3 is stru
cture
of
neural net
wo
rk
as
cla
s
sifiers fault. Th
e structu
r
e
di
agra
m
of Ne
ural
Networks cl
assifie
s
fault
based o
n
inf
o
rmatio
n fro
m
incoming
messag
es or
i
g
inating f
r
om
re
stri
cted a
r
ea
system.
Using
sev
e
r
a
l
Neu
r
al N
e
t
w
o
r
k
s
doe
s t
h
i
s
;
ea
ch
cla
s
sif
i
cati
on fault i
s
re
spo
n
si
ble for determining
the
area of fault in comp
one
nt system.
Figure 3. Model of Neu
r
al
Networks
∑
(
1
)
The Sigmoid;
Sgn (x
) =
1
0
1
1
0
2
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
0
46
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 596
9 –
5975
5972
The output;
(
3
)
For line
a
r fun
c
tion;
1,
,
0,
(
4
)
3. Resul
t
s
In this sectio
n is
sho
w
n
al
gorithm
of fault
identificati
on that the
r
e
five module
s
comp
ose
system: d
a
ta
acq
u
isitio
n, tran
sient id
entificati
on, occurre
n
ce
o
r
condition n
o
rmal co
nditi
on,
sign
al pro
c
e
s
sing, de
ci
sio
n
fault, and fault cla
ssi
fi
ca
tion as sho
w
n in Figure 4
.
The prop
osed
system
op
erates
usi
n
g
o
n
ly data
obta
i
ned f
r
om
su
bstation acquisition,
w
h
ic
h is d
e
r
i
ve
d fro
m
three ph
ase
voltages an
d curre
n
ts. Voltage and
current a
r
e
the input of neural net
w
o
r
k
algorith
m
, an
d it will be
done
after cl
ass fault. Pri
o
r to cl
assifi
cation
of fau
l
ts that will
be
perfo
rmed
by
neu
ral n
e
twork,
also kno
w
n to
determ
i
ne ne
ce
ssa
r
y degree of
membe
r
ship
of
f
u
zzy
set
s
.
F
u
z
z
y
sy
st
ems
are
ac
cu
ra
cy
v
a
lues
in Fu
zzy
Logi
c
a
n
d
memb
er
shi
p
v
a
lue i
ndi
c
a
t
e
d
by value in
ra
nge of
0 a
nd
1. A value of
0 ab
solu
te
fal
s
ity and
de
cl
ared
are d
e
cl
ared
value
of 1
for the
ab
sol
u
te a
c
cura
cy
. Fuzzy Sets often mi
stakenly affecte
d
to indi
cate
some
form
at of
probability.
Figure 4. Sch
e
matic Di
ag
ram of Faul
t Identification a
nd Cla
s
sificat
i
on
3.1. Member
ship Func
tions
There a
r
e m
a
ny po
ssibl
e fo
rmats of me
m
bersh
ip
fun
c
tion, fuzzy op
e
r
ation l
a
rg
ely
dra
w
n
whi
c
h set of curve
s
, in thi
s
pap
er u
s
in
g a triangul
ar function. Th
e methodol
o
g
ies to e
s
tab
lish
membe
r
ship
function
s are
broadly cl
assified in t
he followin
g
cate
gorie
s
: Subje
c
tive Evaluation,
elicitation, probability, physical me
a
s
u
r
ement, learni
ng and a
dapt
ation.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fuzzy
Neural Network
for Cl
as
s
i
fic
a
tion Fault in Prot
ection Syste
m
(Azriye
n
ni)
5973
Figure 5. Membershi
p
Tria
ngula
r
Fun
c
ti
on
Min and
max
operators h
a
v
e been
com
m
only ado
pt
e
d
in dete
r
min
i
ng the me
m
bership
function
s that
to reflect Fu
zzy inte
rsecti
on
and
unio
n
set
s
. Adap
tation can
be
accompli
sh
e
d
through
whi
c
h utilize of
pa
rameter famil
y
of operators F
u
zzy Se
t
s
, whi
c
h might
be very useful to
obtain the me
mbershi
p
fun
c
tion
s in seve
ral of issue
s
.
Ham
a
ch
er'
s
model in [1] a
n
d [9], found
to
be mo
st app
ropriate i
n
the
probl
em to d
e
termin
e faul
t in electri
c
p
o
we
r sy
stem.
In this mode
l,
the intersecti
on and u
n
ion
of two fuzzy
sets A and B are define
d
as
follows:
∩
(5)
∪
′
′
(6)
The
distan
ce
error is calculated from the
num
ber
of e
r
rors th
at a
r
e
simila
r to th
e
circle
o
f
sea
r
ch errors. Simulated powe
r
system
usin
g
ETAP
softwa
r
e to provide data e
rro
rs, the fuzzy
rule
s will be d
e
sig
ned a
nd i
m
pleme
n
ted
usin
g MATLA
B
software.
3.2. Neural Net
w
o
r
k
Clas
sifier
Traini
ng n
e
u
r
al net
work i
s
built to d
e
te
rmine
differe
nt fault gradi
ng involving
variou
s
comp
one
nts
of system. Fo
r ea
ch ne
ural
netwo
rk
co
nsi
s
ts of inp
u
t variabl
es
con
s
i
s
ting of volta
g
e
and
curre
n
t a
r
e obtai
ned f
r
om data
re
co
rds erro
rs
th
at have o
c
cu
rre
d at the
substatio
n
. Each
training p
r
o
d
u
ce
s fault grading (cla
ss 1 or cla
s
s
2).
In Figure 6
sho
w
s the st
ructu
r
e of hy
brid
intelligent techniqu
es a
r
e u
s
ed to dete
r
m
i
ne cla
s
si
ficat
i
on fault in po
wer
system p
r
otectio
n
.
Figure 6. Str
u
c
t
ur
e of Fuzzy
N
eur
al N
e
t
w
ork
C
l
ass
i
fier
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
0
46
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 596
9 –
5975
5974
Adaptation of
weight ve
cto
r
of percept
ro
n;
1
(7)
Whe
r
e;
1
1
1
2
(8)
Traini
ng
data
used to
trai
n ne
ural
n
e
twork cl
assifie
r
for
cla
ssification F
a
ult t
a
ke
n o
n
different fault con
d
ition
s
. The first, singl
e fault c
onditi
ons fault to ground. While the se
con
d
, triple
fault con
d
itio
ns fault to g
r
o
und. Figu
re 7
sho
w
s
gradi
ng fault with i
n
put elem
ent, whi
c
h is volt
age
and current in
fault conditio
n
s in sy
stem.
Figure 7. Cla
ssifi
cation Fa
ult
4. Discus
s
ion
In this pa
pe
r a co
mbine
d
stru
cture wa
s ge
neralized
Fuzzy Sets
and
Neu
r
al
Network
techni
que
s to classificatio
n
of faults in system of
prot
ection. The
cl
assificatio
n
has bee
n
don
e
in
this pap
er is
cla
ssifi
cation
fault linear Neural
Network by usin
g a
formula. Th
e Fuzzy Neu
r
al
Network i
s
a
b
le to
gene
ra
lize
com
m
uni
cation
fa
ilure
s that
occu
r
on the
condit
i
on o
r
o
p
eration
failure relays
and ci
rcuit breakers on tra
n
smi
ssi
on lin
e. Fuzzy Sets are u
s
ed to i
d
entify the fault,
while th
e Ne
ural
Network is u
s
ed to
cl
assify f
ault. This
con
d
itio
n ca
n ma
ke
a different
circuit
topology i
n
Neural
Netwo
r
k training.
Ho
wever,
topolo
g
y ch
ang
es n
e
ce
ssary
to t
r
ain th
e
Neu
r
al
Network and
monitor com
pone
nts of system that
one are
a
re
co
nfiguratio
n o
c
curs. Propo
s
e
d
hybrid i
n
telligent technique is al
so very possible incorporation
of
some qualitat
ive aspects of
probl
em
bein
g
solved. T
h
i
s
i
s
o
n
e
of th
e maj
o
r
adva
n
tage
s
of Fu
zzy Ne
ural Network.
It sho
u
ld
be me
ntione
d that mainte
nan
ce fo
r the
pro
p
o
s
ed
m
odel i
s
u
s
e
d
as a
n
in
dep
e
n
dent
cla
ssifi
ers
for ea
ch
of th
e differe
nt co
mpone
nts
of
monito
ri
ng
sy
stem. Th
e p
r
opo
s
ed
mod
e
l
ca
n e
a
sily b
e
use
d
to resol
v
e the powe
r
system p
r
ote
c
tion sectio
n.
5. Conclu
sion
This pap
er prese
n
ted hybri
d
intellige
n
t
tech
ni
qu
e
for cla
ssifi
cation
of
faulty
com
pone
nts
in power
syst
em. System comp
one
nts i
n
po
wer
sy
st
em protectio
n
are a
s
a
set
of training
used
by Ne
ural
Network. T
h
e
Neu
r
al
Net
w
ork
wa
s tr
ai
ned to
p
r
od
u
c
e
an
e
s
timate of
deg
r
e
e
of
membe
r
ship of
syste
m
co
mpone
nt
s.
T
he stru
cture of
propo
sed
hy
brid
intellig
ent techniq
u
e
i
s
descri
bed
be
tween F
u
zzy
Sets and
Neural
Net
w
or
k. Fuzzy Set
s
are u
s
ed
as id
entificati
on
system
s failu
re tran
sie
n
t protectio
n
a
s
n
eural
n
e
two
r
k fault gradin
g
algorithm. P
o
we
r protecti
on
system
s
sim
u
lated u
s
in
g
ETAP softwa
r
e to
provid
e
data fault, t
he Fu
zzy rul
e
s d
e
si
gne
d
and
impleme
n
ted
usin
g MATLA
B
. Value of 1 indicate
s
if fault occu
rs a
nd value of 0
indicate
s n
o
-
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fuzzy
Neural Network
for Cl
a
ssifi
cation
Fault in Protection Syste
m
(Azriye
nni)
5975
fault occurs.
This te
chni
qu
e ha
s be
en t
e
sted
usi
ng
a
real
system
Riau. Th
e te
st results
sho
w
ed
that diagn
osi
s
ha
s
bee
n a
c
hieve
d
u
s
in
g Fu
zzy
Ne
u
r
al Net
w
o
r
k
wi
th output of
Neu
r
al
Net
w
ork.
Finally, the
concept Fu
zzy
Ne
ural
Network
can
be
p
r
opo
se
d a
s
a
l
ternative to
solve i
s
sue
of
failure
s that occur in p
r
ote
c
tion system.
Referen
ces
[1]
Souza, JCS et
al. Alarm proc
e
ssing i
n
el
ectri
c
al po
w
e
r s
y
st
ems throug
h a
neur
o fuzz
y
ap
proac
h.
IEEE
Transactio
n
in
Pow
e
r Deliv
ery
. 2004; 19(
2): 537-5
43.
[2]
Yu, W
& Li, X. F
u
zz
y
id
entific
ation us
in
g fuz
z
y
n
eura
l
net
w
o
rks
w
i
t
h
stab
l
e
lear
nin
g
al
go
rithms.
IEEE
T
r
ansactio
n
on
F
u
zz
y Syst
em
. 2004; 12(
3): 411-4
19.
[3]
F
l
auzi
no R A
n
, et al. H
y
bri
d
i
n
telli
ge
nt archit
ecture
for fault identifi
catio
n
i
n
po
w
e
r
distri
bu
tion s
y
stem.
IEEE
. 2012; 1-
6.
[4]
Yusuf AA, et a
l
. A novel fa
ult
feature e
x
trac
tion sch
eme fo
r po
w
e
r tra
n
s
m
ission
lin
e fa
ult dia
g
n
o
sis.
IEEE Africon.
201
1; 13-1
5
.
[5]
Kamel
T
S
et a
l
. Adva
nce
d
d
i
stance
protecti
on sc
heme
for
lo
ng tr
ansmis
s
ion
li
nes
in
el
ectric p
o
w
e
r
s
y
stems usi
ng
multipl
e
class
i
fi
ed ANF
I
S Netw
o
r
ks.
IEEE
. 2009. 1-5.
[6]
Aziz AMS et
al. Appl
icatio
n
of ANF
I
S in
high
imp
e
d
a
n
ce faults det
ection a
nd cla
ssificatio
n
i
n
d
i
stri
bu
ti
o
n
ne
tw
orks.
IEEE
. 2011.
[7]
Li F
K
et al.
Real-ti
m
e clas
sificatio
n
of sports
mov
e
me
nt using a
d
a
p
t
ive clusteri
ng.
International
confere
n
ce o
n
compl
e
x. Intelli
gent an
d soft
w
a
re inte
nsive s
y
stem. 20
12.
[8]
Ismail MM, Hasan MAM.
Distance re
lay pr
otection for sh
ort and l
ong tra
n
s
m
iss
i
on
lin
e
. Procee
din
g
s o
f
intern
ation
a
l co
nferenc
e on m
ode
lli
ng id
entifi
c
ation & contr
o
l (ICMIC). 2013
.
[9]
Azri
yen
n
i, Mus
t
afa W
M
. Performanc
e N
eur
o-F
u
zz
y for P
o
w
e
r
S
y
stem
F
ault L
o
cati
on
.
International
Journ
a
l of Eng
i
neer
ing a
nd T
e
chno
logy
. 2
013
; 3(4): 497-50
1
.
[10]
Azri
yen
n
i, Mus
t
afa W
M
, Sukma YD, Dam
e
EM. A
pplic
ati
on of Back
pro
pag
atio
n Ne
ur
al Net
w
o
r
k for
F
ault Locati
o
n
in T
r
ansmission L
i
ne
150
kV
. Internationa
l Journ
a
l
of Electrical a
nd Electro
n
i
c
Engi
neer
in
g (IJEEE).
2013; 2(
4): 21-30.
[11]
Pang
C, et.al.
F
a
st Distanc
e R
e
lay
Sche
me
fo
r De
te
cting S
y
mmetrical Fault During
Po
w
e
r S
w
ing.
IEEE
T
r
ansaction on Po
w
e
r D
e
liver
y
.
2
010; 2
5
(4). 220
5-2
2
1
2
.
[12]
Upe
ndar
J, et
al.
Com
p
reh
e
n
sive
Ad
aptive
Dista
nce
Re
l
a
yin
g
Sc
heme
for Par
a
l
l
el
T
r
ansmissi
o
n
Lines.
IEEE Tr
ansaction on Power Deliv
ery.
201
1; 26(2): 10
39-1
051.
[13]
Lee H,
et
al. Novel Protecti
on
Sch
e
me
w
i
th
the Sup
e
rco
nducti
ng P
o
w
e
r Cabl
es a
nd
F
ault Curr
ent
Limiters T
h
rou
gh RT
DS T
e
st in Iche
on S
u
b
s
tation.
IEEE Transaction on Applied S
u
per
c
onductivity
.
201
2; 22(3).
[14]
Jian
g K, Sing
h C. Ne
w
mo
dels a
nd co
n
c
epts fo
r po
w
e
r s
y
stem re
li
abil
i
t
y
eval
uati
on inc
l
u
d
in
g
protectio
n
s
y
st
em failur
e
s.
IEEE Transactions on power sy
stem
. 20
11; 26
(4): 1845-
18
55
.
[15] Z
hang JF
, Smith JS,
W
u
QH. Morphol
og
ical
undec
imated
w
a
vel
e
t decom
positi
on for fau
l
t location o
n
po
w
e
r transmis
s
ion li
nes.
IEEE Transactions
on circuits and system
s
. 200
6
;
53(6): 139
5-1
402.
[16]
Lee H, Ju
ng C,
Chon
g SS, et al
. Nove
l prote
c
tion schem
e
w
i
t
h
the su
perc
ond
uctin
g
po
w
e
r cabl
es an
d
fault curr
ent l
i
miters thro
ug
h RT
DS test in Ic
heo
n s
ubstatio
n
.
IEEE
T
r
ansacti
on
s on
ap
pli
e
d
superc
ond
uctiv
i
ty.
2012; 22(
3)
: 1-5.
[17]
F
onseca W
A
D
S
, et al. Simultaneo
us fault section estim
a
ti
on an
d pr
otect
i
ve devic
e fail
u
r
e detectio
n
usin
g perc
enta
ge va
lues
of th
e protectiv
e
d
e
v
ices al
arms.
IEEE Transaction on power system
s
. 2
013;
28(1): 17
0-1
8
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.