In
te
r
n
ation
a
l Jou
rn
al
o
f Po
we
r
Elec
tron
ic
s an
d
D
r
ive S
y
stem
(IJ
PED
S
)
V
o
l.
10, N
o.
3, S
ep 2019,
pp.
1
6
8
7
~1
6
9
3
ISSN: 2088-
8694,
DOI
:
10.11591
/ijpeds.
v10.
i
3.pp1687-1693
1687
Jou
rn
a
l
h
o
me
pa
ge
:
ht
tp:
//i
a
e
score
.
com
/
j
o
u
r
na
l
s
/
i
n
d
e
x
.
p
hp/IJ
PED
S
Fault detection and classifica
tion in wind turb
ine by u
sing
artificial
neur
a
l netw
ork
N. F. Fad
zail, S
.
Mat
Z
a
li
School of E
l
ectric
a
l
System E
ngi
neering Univers
iti Malaysia P
er
l
i
s
,
M
alay
s
i
a
Art
i
cl
e In
fo
ABSTRACT
A
r
tic
le hist
o
r
y
:
Re
ce
i
v
e
d
Ju
l
2
7,
201
8
Re
vise
d Jan
1
2
, 2019
Ac
ce
p
t
ed
M
ar 1
3
,
2
019
Wi
nd
t
u
r
b
i
ne
i
s
on
e
of
t
h
e
p
res
e
nt
r
enew
abl
e
e
ne
rg
y
s
o
u
r
ces
t
ha
t
has
b
ecom
e
th
e
m
o
st
p
opular.
Th
e
o
p
erat
ion
a
l
and
m
a
i
n
ten
a
nc
e
cos
t
i
s
con
t
i
nu
ously
in
creasin
g,
e
sp
ecial
ly
f
or
w
in
d
gener
a
t
o
r.
E
arly
f
au
lt
d
et
ectio
n
is
v
ery
im
p
o
rt
a
n
t
t
o
o
p
t
imi
s
e
th
e
operat
i
o
n
al
a
n
d
m
aint
enan
ce
c
o
s
t
.
T
he
g
o
a
l
o
f
t
h
i
s
pro
j
ect
i
s
t
o
s
t
u
d
y
f
ault
det
ecti
o
n
an
d
clas
si
ficati
o
n
fo
r
a
wi
n
d
t
urbi
ne
(
WT
)
by
u
si
ng
artif
i
c
i
al
n
eu
ral
n
e
tw
ork
(AN
N
).
I
n
th
is
p
roject
,
a
s
i
ngle
ph
ase
fau
lt
was
placed
a
t
9
MW
d
ou
bly
-
f
e
d
in
du
cti
on
g
e
nerat
o
r
(DF
I
G)
W
T
in
M
A
TLAB
Sim
u
l
i
n
k
.
The
WT
w
as
t
ested
un
de
r
d
i
ff
erent
co
nditi
on
s
,
i.e
.
,
no
rm
al
c
ond
ition,
f
ault
at
P
hase
A
,
Ph
ase
B
a
n
d
P
h
as
e
C.
T
he
s
i
mula
t
i
o
n
resu
lts
w
ere
us
e
d
a
s
i
npu
ts
i
n
t
h
e
A
N
N
m
o
d
e
l
f
o
r
t
r
a
i
ni
ng.
T
hen
,
a
ne
w
se
t
of
dat
a
w
as
t
aken
u
nder
d
iff
e
rent
c
onditions
a
s
inp
u
ts
f
or
A
N
N
f
au
lt
c
lassifi
er.
Th
e
t
a
rg
et
o
u
t
puts
of
A
N
N
f
ault
cl
ass
i
fier
w
ere
set
as
‘
0’
o
r
‘
1’,
ba
se
d
o
n
t
he
f
a
ult co
nd
ition. Res
ul
ts ob
t
ain
e
d
sho
w
ed
th
a
t t
h
e A
N
N
f
a
ul
t cl
a
s
s
i
f
ier output
s
h
a
d
fol
l
owe
d
t
h
e
t
a
r
g
e
t
ou
tpu
t
s
.
I
n
c
o
nc
lu
sion
,
the
WT
f
a
u
l
t
d
e
t
ecti
o
n
and
clas
sificati
on m
e
th
od
by usin
g
ANN
were s
u
ccess
f
u
lly
dev
elop
ed
.
K
eyw
ord
s
:
A
r
tificia
l N
e
ura
l
N
etw
o
rk
(AN
N
)
D
o
u
b
l
y
-F
e
d
Induc
t
i
o
n
Generator
(DFIG)
Fa
ul
t
D
e
tec
tio
n
W
i
nd
Tu
r
bi
ne
(W
T
)
Co
pyri
gh
t © 2
019 In
stit
u
t
e
of Advanced
En
gi
neeri
n
g
an
d
S
c
ien
ce.
All
rights
res
e
rv
ed.
Corres
pon
d
i
n
g
Au
th
or:
Noor
F
az
li
a
n
a
bt
F
adz
a
i
l
,
S
c
hoo
l
o
f
Ele
c
t
rica
l
S
y
stem
Eng
i
n
eer
ing,
Uni
v
ersi
ty
M
al
ays
i
a
Per
l
is,
K
a
m
pus P
auh P
u
tra,
026
00 A
r
au, P
e
r
li
s,
Mala
y
sia.
Em
ail:
faz
lia
na
fadz
a
i
l@
u
n
i
m
a
p
.
e
du.m
y
1.
I
N
TR
OD
U
C
TI
O
N
Re
ne
w
a
b
l
e ene
r
gy
so
urce
s, espec
i
a
ll
y w
i
nd e
n
e
r
g
y
i
s p
r
ese
n
tl
y
t
he mos
t po
pu
lar
tec
h
n
o
lo
gy as ther
e
w
e
r
e
m
ore
tha
n
2
82.4
8
G
W
in
sta
lled
ca
pac
i
t
y
a
t
t
h
e
en
d
of
2
0
1
2
[1-4]
.
T
here
i
s
a
ne
ed
f
or
a
n
ear
ly
f
a
u
lt
detec
t
i
on
in
t
h
i
s
incre
a
sin
g
l
y
pop
ula
r
t
e
c
h
no
l
o
g
y
,
since
t
h
e ea
rly
fa
ul
t
de
te
c
t
i
o
n
in
W
T
ca
n
he
l
p
t
o
re
du
c
e
t
he
cos
t
for
e
ffe
c
tive
m
a
int
e
na
nc
e a
nd oper
a
ti
on
[1,
5-12].
I
n
li
t
e
r
at
ure
, t
h
e
re are
seve
r
al pu
b
l
ica
t
ions o
n fa
ul
t de
t
e
c
t
io
n i
n
v
e
st
iga
t
i
o
n
in
W
Ts. In [1
3], cond
i
tio
n
mo
n
ito
rin
g
a
nd
f
a
u
lt
d
e
t
ec
ti
on
i
n
WT
b
ased
o
n
DF
IG
b
y
usin
g
fu
z
z
y
l
ogi
c
were
p
rese
nte
d
.
It
w
a
s
f
o
c
used
o
n
fa
ul
ty
s
h
o
r
t
c
i
r
c
u
it.
I
n
[5],
f
a
u
lt
det
e
c
t
i
on
i
n
W
T
w
a
s
foc
u
se
d
o
n
ne
u
r
al
n
e
t
w
o
r
k
.
Th
e
fa
u
lt
de
tec
t
i
on
w
a
s
base
d
o
n
c
urre
nt
s
ig
na
tur
e
a
n
a
ly
sis.
F
au
lt
d
e
t
e
c
t
i
on
i
n
W
T
by
u
sin
g
a
rti
f
i
c
i
a
l
n
e
u
r
al
n
et
wo
rk
(
ANN)
b
as
ed
o
n
SCADA
d
a
t
a
a
n
a
ly
si
s
was
p
r
op
o
s
ed
i
n
[1
].
I
n
[7
],
c
o
n
d
it
ion
mo
ni
tor
i
ng
for
WT
g
ene
r
at
or
b
y
usin
g
tem
p
era
t
ur
e
t
r
end
a
n
a
l
y
s
i
s
w
as
d
e
v
el
o
p
ed.
T
h
e
u
n
b
al
a
n
ce v
o
lta
ge
f
aul
t
of
d
oub
l
y
f
ed
a
syn
c
h
r
o
nou
s
g
e
nera
t
o
r
WT
w
as
s
t
u
die
d
i
n
[1
4].
Me
a
n
w
h
i
l
e
i
n
[
15
-18],
fa
u
lts
d
ia
gn
os
is
by
u
si
ng
a
-p
rio
r
i
knowl
e
dg
e-b
a
sed
ANF
IS
for
WT
w
as pr
opose
d
.
The
r
e
sea
r
ch
w
as
f
oc
use
d
o
nl
y
on
fau
l
t
d
e
te
ct
io
n
an
d
n
o
t
o
n
f
a
u
lt
classification.
Moreover,
m
os
t
rese
arc
h
d
e
v
e
l
ope
d
fau
l
t
de
te
c
tio
n
by
u
s
i
ng
a
n
art
i
fic
i
a
l
i
n
t
el
li
gent
a
p
p
r
oach
l
i
k
e
f
u
zzy
an
d
ANN,
b
ased
o
n
curr
ent
as
t
he
i
np
u
t
.
A
n
a
r
t
i
f
ic
i
a
l
i
n
te
l
lige
n
c
e
(
A
I)
m
e
tho
d
l
i
ke
f
uz
z
y
i
s
v
e
ry
c
om
plic
a
t
e
d
d
ue
t
o
t
h
e
r
e
qu
ired
rule
t
o
deve
l
o
p
the
fuz
z
y
sys
t
e
m
.
A
I
m
e
a
ns
t
he
a
bi
l
i
t
y
o
f
a
ma
chi
n
e
to
p
erfor
m
s
im
i
l
ar
f
u
n
c
t
i
o
ns
t
h
a
t
c
har
a
c
t
erise
the
hu
m
a
n
th
o
ugh
t
[1
9].
A
I
b
ec
ome
s
t
he
m
os
t
po
p
u
lar
me
tho
d
f
or
f
au
l
t
d
e
t
ec
tio
n
d
u
e
t
o
m
a
n
y
a
d
v
a
n
t
ag
es
a
s
co
mp
a
r
ed
t
o
t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN: 2088-
8694
Int J
P
o
w
El
e
c
&
D
ri S
yst
,
V
ol.
10,
N
o.
3
, S
e
p
2
0
1
9
:
168
7
– 1
693
1
688
con
v
e
n
t
i
ona
l
f
a
ul
t
dia
g
no
st
i
c
a
ppro
a
c
h
es
[
1
3
].
A
NN
i
s
on
e
of
t
h
e
A
I
m
e
t
h
ods
a
nd
i
t
i
s
a
ty
pe
o
f
ne
t
w
ork
b
e
ca
use
it
s
e
e
s
t
h
e
nod
e
s
a
s
‘a
rt
i
f
i
c
i
a
l
n
e
u
r
on
s
’
[
19
-22
]
.
Th
er
e
are
t
w
o
t
ypes
o
f
ANN,
w
h
i
ch
a
re
f
eedfo
r
w
a
rd
an
d
f
eed
back
ANN.
T
h
e
f
ee
d
f
o
r
ward
ANN
d
o
e
s
n
o
t
con
t
ai
n
any
con
n
e
c
ti
on
b
e
t
we
e
n
i
np
ut
a
nd
o
utp
u
t
,
wh
ile
the
fe
ed
bac
k
A
N
N
contai
ns
c
on
nec
t
i
o
n
be
t
w
ee
n
i
n
p
u
t
a
n
d
o
u
t
p
ut
.
In
s
tat
i
c
p
u
r
p
o
s
e
t
h
e
f
e
ed
fo
r
w
ard
ANN
i
s
usu
a
l
l
y
use
d
a
nd
i
n
d
y
n
am
ic
pur
pose
t
h
e
fe
e
dbac
k
A
N
N
i
s
use
d
[
2
4
]
.
ANN
st
ru
ctu
r
es
c
o
n
s
i
st
o
f
i
n
put
l
ay
er
,
hi
dde
n la
ye
r a
nd o
u
t
p
u
t
l
a
y
e
r,
a
s show
n i
n
F
ig
ure
1.
F
i
gur
e 1.
A
NN
struct
u
re
In
t
hi
s
p
r
oj
ec
t
,
f
a
u
l
t
d
e
t
e
cti
o
n
an
d
c
l
a
s
sifi
cat
ion
f
o
r
WT
w
as
d
e
velop
e
d
by
u
s
i
ng
ANN
mo
d
e
l
in
MA
TLA
B,
b
as
ed
o
n
t
h
ree
p
h
a
se
c
urre
nt
s,
D
C
v
o
lta
ge
,
an
d
i
n
d
u
c
t
i
o
n
ge
ne
ra
t
o
r
spe
e
d
a
s
t
h
e
m
o
d
e
l
i
n
pu
t.
A
9
M
W
W
T
t
h
a
t
u
s
e
a
d
e
t
a
i
l
e
d
D
F
I
G
m
o
d
e
l
w
a
s
s
i
m
u
l
a
t
e
d
w
i
t
h
n
o
r
m
a
l
a
nd
fa
u
lt
co
n
d
i
tio
ns.
A
sin
g
l
e
pha
se
fa
ul
t
w
a
s
inser
t
e
d
i
nto
the
WT
m
ode
l
for
fa
ul
t
a
t
P
ha
se
A
,
P
h
a
s
e
B,
a
n
d
P
hase
C
.
The
simu
la
ti
on
re
s
u
l
t
s
of
three
p
h
ase
c
u
r
r
ents, DC
v
o
l
ta
ge,
in
duc
t
i
o
n
g
ene
r
at
or speed,
a
c
ti
ve
p
ow
e
r
,
and re
act
i
v
e po
w
e
r
w
e
r
e
obt
a
i
ne
d.
Th
en
,
t
h
e
ANN
mo
d
e
l
was
d
e
v
e
lo
p
e
d
.
T
he
A
N
N
w
as
u
sed
t
o
s
tud
y
t
h
e
dy
na
mic
of
W
T
s
y
s
t
em
f
r
o
m
t
h
e
in
put
a
nd
o
u
t
p
u
t
pat
t
ern
c
o
lle
c
t
ed
f
r
o
m
the
da
ta
set.
T
he
m
ode
l
l
ed
ANN
in
pu
t
con
s
ist
e
d
o
f
f
iv
e
i
npu
ts,
wh
ich
w
e
r
e
t
hre
e
p
h
a
se
c
urr
e
nt,
D
C
v
o
lta
ge
a
nd
i
n
d
u
c
t
i
on
g
e
n
e
ra
tor
sp
e
e
d
.
M
e
a
n
w
hi
l
e
,
the
o
u
tp
ut
f
o
r
t
ra
i
n
i
ng
p
r
o
c
ess
o
f
m
od
ell
e
d
ANN
co
n
s
i
s
t
e
d
o
f
o
ne
o
u
t
pu
t
w
h
i
c
h
was
act
i
v
e
po
wer
o
r
r
eact
i
v
e
p
o
w
er.
Two
ANN
mode
l
s
w
ere
d
e
vel
o
ped,
w
h
i
c
h
w
er
e
for
act
i
v
e
a
nd
rea
c
t
i
v
e
p
ow
er
m
o
d
e
ls.
Th
en
,
t
h
e
out
put
r
e
s
pon
se
s
f
r
o
m
ANN
t
r
ai
ni
ng
p
r
o
c
ess
mod
e
l
w
e
re
c
o
m
pared
to
t
h
e
t
a
r
g
e
t
ou
t
p
u
t
t
o
ev
alu
a
t
e
t
h
e
ANN
t
r
ai
ni
ng
mo
d
e
l
perform
ance
.
The
c
o
m
p
aris
o
n
w
as
m
ade
b
y
c
a
l
c
u
la
t
i
n
g
t
he
r
o
o
t
m
e
ans
square
e
rror
(R
MSE)
v
al
ue
.
Whe
n
RMS
E
v
a
l
ue
s
w
e
re
s
atis
fa
ct
o
r
y
for
tra
i
ni
n
g
,
a
few
da
ta
w
ere
tak
e
n
f
or
t
h
e
ANN
f
a
u
lt
class
i
f
i
e
r
.
The
ANN
fa
ul
t
c
l
ass
i
f
i
e
r
t
a
r
get
o
u
t
p
u
t
w
as
s
e
t
a
s
‘0’
or
‘
1’,
base
d
o
n
fau
l
t
c
o
n
d
i
t
i
on.
The
ANN
f
a
u
lt
clas
si
fie
r
w
as
deve
l
ope
d w
i
th
di
ffe
re
nt da
t
a
fr
om all c
ond
i
t
i
o
n
s
to
s
ee
t
he
p
e
rf
o
r
man
ce.
T
h
e
ANN f
a
ul
t cl
assi
f
i
er mo
d
el
was
val
i
d
at
ed w
ith
a
ne
w
set
o
f da
t
a
fr
o
m
differ
ent c
o
ndi
t
i
o
n
s
a
nd
f
a
ul
t
resi
st
an
ce
s.
2.
METHODOLOG
Y
A
9
M
W
w
i
n
d
farm
t
ha
t
use
a
de
tai
l
e
d
m
od
e
l
o
f
a
D
o
u
b
l
y
-F
ed
I
ndu
ction
Generator
(
D
FI
G
)
d
riven
by
a
w
i
n
d
tur
b
i
n
e
is
s
h
o
w
n
i
n
F
i
gur
e
2.
F
irst,
the
W
T
w
as
s
imu
l
ate
d
u
nde
r
norm
a
l
co
nd
i
tio
n
to
g
et
t
he
si
m
u
lat
i
on
r
e
s
u
l
t
o
f
t
h
re
e
p
h
a
se
c
ur
rents,
(
pu)
;
D
C
v
o
lta
g
e
,
(
V
)
;
ind
u
ct
ion
ge
ne
rator
s
p
ee
d,
(
pu);
ac
t
i
ve
pow
er,
P
(
p
u
)
;
a
n
d
re
a
c
ti
ve
p
ow
e
r
,
Q
(pu).
Later
,
a
t
hr
e
e
ph
ase
f
a
u
l
t
was
i
n
sert
ed
b
e
t
ween
l
i
n
e
B575
and
D
F
IG
w
ind tur
b
i
n
e.
The
f
a
u
l
t
s w
e
re
s
i
m
ula
t
e
d
f
or
f
au
l
t
at
P
hase A
, P
h
a
se
B
,
and
Phase C.
The
sim
u
l
a
ti
on
r
e
s
ul
ts
f
or
n
orma
l
a
nd
fa
ul
t
c
o
n
d
it
i
o
n
s
o
f
WT
m
o
d
e
l
were
u
s
e
d
t
o
d
ev
e
l
op
t
h
e
ANN
mode
l.
T
he
A
N
N
m
ode
l
w
a
s
de
vel
o
ped
t
o
d
etec
t
t
h
e
ty
pe
s
o
f
f
a
u
lt
.
T
h
e
A
N
N
mode
l
u
s
e
d
b
a
c
k
p
r
opa
gat
i
o
n
a
s
i
t
was
mo
re
e
ffi
ci
en
t
i
n
p
ro
du
ci
ng
t
h
e
t
arg
e
t
v
a
l
u
es
a
s
o
u
tp
u
t
v
a
l
ue
s
[2
4].
Mor
e
ove
r
,
b
ack
p
ro
pa
ga
tio
n
doe
s
not
r
e
q
u
i
re
t
he
e
xa
ct
f
o
r
m
of
a
n
a
l
y
t
i
c
a
l
f
u
n
c
t
i
o
n
o
n
w
hi
ch
t
he
m
o
d
el
s
ho
uld
be
b
u
ilt
[
1].
Tw
o
A
N
N
mode
l
s
w
er
e
de
vel
o
ped i
n
m
ode
ls,
w
h
ic
h
w
e
re
AN
N
t
raini
ng m
ode
l
a
n
d A
N
N
f
aul
t
c
l
a
s
s
if
i
e
r m
odel.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t
J
P
o
w
Elec
&
D
r
i
S
y
st
I
S
S
N
:
2088-
86
94
Fa
ul
t
de
tec
t
i
o
n
and cl
assi
fic
a
t
ion i
n
w
i
n
d
tur
b
i
n
e by
us
i
ng
a
r
t
ific
ia
l
ne
ural
n
e
t
w
ork (N. F.
Fa
dza
i
l)
1
689
Figure
2.
A
9 MW win
d
farm
wi
th
t
hr
ee
p
h
ase
fau
l
t [2
5
]
2.
1.
A
N
N
t
r
a
i
ning
mo
d
el
Th
e
ANN
train
i
ng
m
od
el
w
as
d
ev
elop
ed
b
y
u
s
in
g
M
A
TLAB
.
T
his
mo
d
e
l
ha
d
mu
lt
i
-
in
pu
t.
T
h
e
in
p
u
t
s
f
or
A
NN
t
r
a
i
n
i
n
g
m
odel
w
e
r
e
t
hr
ee
phase
c
ur
r
e
nt
s,
(
pu)
;
D
C
vol
ta
ge,
(
V)
;
and
i
nduc
ti
o
n
ge
ner
a
t
o
r
spee
d,
(
p
u
)
.
Th
e
o
u
t
p
u
t
f
o
r
ANN
t
r
ainin
g
mod
e
l
was
activ
e
po
wer,
P
.
Th
en,
th
e
ANN
mo
d
e
l
was
rep
eated
f
o
r
r
eacti
v
e
p
o
w
er
,
Q
,
a
s
an
output
.
Th
e
ANN
train
in
g
m
ode
l
w
a
s
deve
lo
pe
d
fo
r
four
c
o
n
d
i
t
i
o
n
s
,
w
h
i
c
h
w
e
r
e
n
o
r
m
a
l
,
f
a
u
l
t
a
t
P
h
a
s
e
A
,
P
h
a
s
e
B
,
a
n
d
P
h
a
se
C
.
Fig
u
r
e
3
sh
o
w
s
th
e
ANN
train
i
ng
m
odel.
T
he
A
N
N
tr
a
i
n
i
n
g
m
ode
l
c
o
nsi
s
t
e
d
of
5
i
n
p
u
ts
a
nd
1
ou
tp
u
t.
T
h
e
h
i
d
de
n
l
a
yer
w
a
s
set
to
4
0
a
nd
t
h
e
o
u
tp
ut lay
er
w
as set to
1
.
The
p
e
r
f
or
ma
nce
o
f
A
N
N
t
r
a
in
in
g
m
o
del
w
a
s
eval
ua
te
d
b
y
c
om
par
i
n
g
th
e
tar
g
et
a
n
d
th
e
ANN
m
odel
o
u
t
p
u
t
by
us
ing
R
M
S
E
v
al
ue.
The
tr
a
i
ni
n
g
p
r
o
c
e
sses
w
e
r
e
r
epe
a
ted
un
ti
l
the
R
M
SE
v
a
l
ue
s
wer
e
sa
tisfa
c
t
or
y.
T
he
t
r
a
in
i
n
g
pr
o
c
esses
w
e
r
e
d
one
t
o
ge
t
t
h
e
bes
t
A
N
N
s
t
r
u
c
t
u
r
e
a
n
d
s
e
l
e
c
t
t
h
e
i
n
p
u
t
o
f
A
N
N
fau
l
t
class
i
fier.
F
i
gur
e
3.
A
NN
t
r
a
in
in
g
m
ode
l
2.
2.
A
N
N
f
a
ult
clas
sifier mo
d
el
Later
,
a
f
ew
d
ata
w
e
re
t
aken
f
o
r
ANN
f
a
u
l
t
d
e
t
e
ctio
n
f
o
r
W
T
.
Fi
gu
re
4
s
ho
ws
t
h
e
ANN
f
au
lt
c
l
a
ssif
i
e
r
m
o
d
e
l.
T
he
s
truc
tur
e
o
f
ANN
f
a
u
l
t
c
lass
i
f
ier
was
se
l
ec
t
e
d
as
5
i
np
ut
s,
4
0
hi
dd
en
l
a
y
ers,
3
o
u
t
pu
t
lay
e
rs and
3
o
utputs.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN: 2088-
8694
Int J
P
o
w
El
e
c
&
D
ri S
yst
,
V
ol.
10,
N
o.
3
, S
e
p
2
0
1
9
:
168
7
– 1
693
1
690
F
i
gure
4.
A
NN
fa
ult c
l
ass
i
f
i
e
r
m
odel
The
i
n
p
u
t
s
of
A
N
N
f
a
u
l
t
c
la
ssifier
w
er
e
thre
e
ph
a
s
e
curre
nts,
(
pu);
D
C
v
ol
ta
ge,
(
V
)
;
a
n
d
in
duc
t
i
on
ge
ne
rator
spee
d,
(
p
u
)
.
M
e
a
n
w
h
i
l
e
,
t
h
e
o
u
t
p
u
t
s
f
o
r
A
N
N
f
a
u
l
t
c
l
a
s
s
i
f
i
e
r
w
e
r
e
r
e
presente
d
a
s
X
,
Y
,
a
n
d
Z
.
T
h
e
v
a
l
u
e
o
f
e
a
c
h
o
u
t
p
u
t
w
a
s
s
e
t
a
s
‘
0
’
o
r
‘
1
’
,
b
a
s
e
d
on
fa
u
lt
c
o
n
d
i
t
i
on.
T
a
b
l
e
1
s
how
s
t
h
e
t
a
rge
t
ou
tpu
t
f
or
A
NN
f
a
u
lt
cla
ssi
fi
er
m
odel.
I
t
cons
ist
e
d
o
f
f
our
c
on
d
i
t
i
o
n
s
,
i.e.,
norm
a
l,
f
a
u
l
t
a
t
p
h
ase
A
,
P
h
a
se
B
,
and
P
h
ase
C.
Th
e
ANN
fau
l
t
cl
assi
f
i
er
w
as
v
ali
d
a
t
e
d
w
it
h
t
h
ree
di
fferen
t
f
au
l
t
r
esista
nce
v
a
lue
s
,
i.e.,
0.001
Ω,
0
.
0
1
Ω
a
n
d
0
.
1Ω
.
Th
e
p
e
rfo
rman
ce
o
f
ANN
f
a
u
l
t
s
c
l
a
ssi
fi
er
m
o
d
e
l
w
a
s
eva
l
ua
ted
by
c
o
m
p
ar
ing
t
h
e
targe
t
o
u
t
p
ut
o
f ANN f
a
u
l
t
s
cl
a
ssi
f
i
e
r
mo
d
e
l
by
u
si
n
g
di
ff
er
ent
d
a
t
a
f
or
all c
o
n
d
it
io
ns.
Ta
b
l
e
1.
Tar
get out
p
u
t
for
A
N
N
faul
t c
l
a
ssifi
e
r
C
ondition
X
Y
Z
No
r
m
a
l
0
0
0
Fa
ult a
t
Ph
a
s
e
A
1
0
0
Fa
ult a
t
Ph
a
s
e
B
0
1
0
Fa
ult a
t
Ph
a
s
e
C
0
0
1
3.
RESULT
S
3.1.
Resu
l
t
f
o
r
A
NN
tra
i
ni
ng
m
odel
F
i
gure
5 sh
ow
s the
t
r
ai
nin
g
r
es
u
lt for
re
activ
e
pow
er,
Q
(pu)
f
or
f
ault a
t
P
h
a
se A
a
nd
F
i
gu
re
6
show
s
the tra
i
ni
ng
resul
t
f
or
a
c
t
i
v
e
p
o
w
e
r,
P
(
pu
)
for
fa
ult
at P
hase
B
.
The
solid
l
i
n
e
re
presente
d
the ta
r
g
et o
u
t
p
u
t,
wh
il
e t
h
e
d
a
shed
l
in
e
rep
r
esen
t
e
d
th
e ANN
mo
d
e
l
out
put
.
Th
e ANN
m
odel ou
t
p
u
t
s
in
F
igure
5 c
a
p
t
ure
d
t
he
t
a
rg
e
t
o
ut
p
u
t
w
e
l
l
.
Th
e g
r
aph
o
f
ANN mo
d
e
l
i
n
F
i
g
u
r
e 6
w
a
s o
s
ci
llated but
t
h
e RM
S
E
value was sti
ll low.
F
i
gure
5. Rea
ct
ive p
o
we
r,
Q
for
f
aul
t
at
p
h
as
e
A
F
i
gure
6.
A
ctive
pow
e
r
, P
f
or fa
u
l
t
a
t phas
e
B
Ta
b
l
e
2
s
h
ow
s
the
re
sul
t
o
f
R
M
S
E
v
al
ue
o
f
A
N
N
m
ode
l
for
tra
i
ni
n
g
p
u
r
po
se
w
i
t
h
di
ff
ere
n
t
fa
ult
con
d
i
t
i
on
s
for
acti
v
e
a
n
d
rea
c
ti
v
e
p
ower
.
All
t
h
e
R
M
S
E
v
a
l
ue
s
w
e
r
e
l
o
w
.
T
h
e
h
i
g
h
e
s
t
R
M
S
E
v
a
l
u
e
w
a
s
f
o
r
ac
t
i
ve
pow
er
f
or
n
o
r
ma
l
c
o
n
d
i
t
ion,
w
hi
c
h
w
a
s
0
.
0
2
2
6
.
T
he
l
ow
e
s
t
v
a
l
u
e
o
f
R
M
S
E
w
a
s
f
o
r
r
e
a
c
t
i
v
e
p
o
w
e
r
f
o
r
fa
ul
t a
t
P
ha
se C w
hi
c
h
w
as
0
.001
4.
0
0.
02
0.
04
0.
06
0.
08
0.
1
0.
12
0.
14
0.
16
0.
18
0.
2
-0.
2
-0.
1
0
0.
1
0.
2
0.
3
0.
4
0.
5
ti
m
e
s (
s
e
c
)
R
e
a
c
t
i
v
e
p
o
w
e
r
, Q
(
p
u
)
T
a
rget
O
ut
pu
t
A
N
N
M
o
del
Out
p
ut
0
0.
0
2
0.
04
0.
06
0.
0
8
0.
1
0.
1
2
0.
14
0.
16
0.
1
8
0.
2
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
ti
m
e
s
(
s
e
c
)
A
c
ti
v
e
p
o
w
e
r
, P
(p
u
)
Ta
r
g
e
t
O
u
t
p
u
t
AN
N
M
ode
l Output
Evaluation Warning : The document was created with Spire.PDF for Python.
Int J
P
o
w
E
l
e
c
&
D
ri S
yst
IS
S
N
:
2088-
86
94
F
aul
t de
t
e
ct
i
on
and c
l
ass
i
f
i
c
a
t
i
o
n
i
n
w
i
n
d
t
u
r
b
i
n
e
by
usi
ng
art
i
f
i
c
i
a
l
ne
ur
al
net
w
ork
(N.
F. Fa
dza
i
l)
1
691
Tab
l
e
2.
R
MS
E
val
u
e
for
A
N
N
tra
i
nin
g
m
od
e
l
C
ondit
i
on
R
MS
E
Ac
tive
pow
e
r
,
P
R
ea
ct
ive
powe
r
,
Q
Norm
a
l
0
.
0226
0
.
0082
Fa
ul
t at
pha
s
e
A
0
.
0041
0
.
0016
Fa
ul
t at
pha
s
e
B
0
.
0108
0
.
0039
Fa
ul
t at
pha
s
e
C
0
.
0047
0
.
0014
3.2.
Res
ults f
or A
NN
f
a
ul
t clas
sifier mo
d
el
A
f
ter
t
h
e
tra
i
ni
ng
p
roce
ss
w
a
s
s
a
tis
fa
ct
ory,
t
he
A
N
N
fau
l
t
clas
si
fi
er
w
as
d
ev
elo
p
e
d
.
T
he
ANN
f
a
u
l
t
class
i
fier
w
as
t
e
s
te
d
a
n
d
va
li
da
t
e
d
wi
t
h
d
iffer
e
nt
f
a
u
lt
res
i
st
a
n
ce
val
u
es.
Tab
l
e
3
s
h
o
w
s
the
re
s
u
lts
o
f
ANN
fa
ul
t
c
l
ass
i
fier
f
or
a
l
l
c
o
n
d
i
t
i
o
n
s
w
i
th
d
i
f
fe
rent
f
a
u
lt
r
esis
t
a
n
ces
.
T
h
e
r
esult
s
s
ho
wed
th
at
ANN
f
a
u
l
t
cl
assi
f
i
er
fo
l
l
ow
e
d
the
t
a
r
ge
t
o
u
tp
ut.
Tab
l
e 3.
Resu
l
t
s
o
f
v
a
l
i
da
ti
o
n
for A
N
N
fa
u
l
t
c
lass
ifier
mode
l
CON
D
I
T
I
O
N
F
a
ult R
e
sista
n
c
e
(
ohm
)
TAR
G
ET
A
N
N
OUT
PUT
X
Y
Z
X
Y
Z
Norm
a
l
none
0
0
0
8
.
07E-0
4
2.
03E
-
0
4
0.
0003
Fa
ult at
pha
s
e
A
0.
001
1
0
0
0
.
9997
2
.
68E
-
0
4
1.
54E-0
4
0.
01
1
0
0
1
.
0000
2
.
08E
-
0
6
9.
62E-0
6
0.
1
1
0
0
1.
0000
4
.
43E
-
0
4
2.
47E-0
4
Fa
ult at
pha
s
e
B
0.
001
0
1
0
0
.
0002
0
.
9997
1
.
34E-0
4
0.
01
0
1
0
4
.
56E-1
4
1
.
0000
0
.
00E+
00
0.
1
0
1
0
3.
36E-0
4
0
.
9996
5
.
30E-0
6
Fa
ult at
pha
s
e
C
0.
001
0
0
1
4
.
02E-0
7
1.
21E
-
0
4
0.
9997
0.
01
0
0
1
1
.
19E-1
0
3.
29E
-
1
3
1.
0000
0.
1
0
0
1
5.
72E-0
4
2.
84E
-
0
4
0.
9988
4.
DISC
USSION
Th
e
ai
m
of
t
his
re
se
arc
h
i
s
t
o
d
e
v
elo
p
a
f
a
ult
d
e
t
e
cti
o
n
and
c
l
a
s
si
fi
e
r
m
ode
l
fo
r
W
T
b
y
usi
n
g
AN
N.
The
re
sults
o
b
t
a
i
ne
d
for
trai
ni
n
g
p
r
o
cess
s
how
e
d
t
hat
th
e
A
N
N
t
r
a
ini
ng
m
ode
l
ou
tpu
t
f
o
l
low
e
d
t
h
e
t
a
rge
t
o
u
t
p
ut
s
e
x
c
e
l
l
e
n
tly
,
e
x
cept
fo
r
no
rmal
c
o
n
d
i
ti
on
.
Alth
oug
h
th
e
ANN
mo
d
e
l
fo
r
t
r
ai
n
i
ng
p
ro
ces
s
i
n
n
o
r
mal
c
o
n
d
i
t
i
o
n
w
a
s
o
s
c
i
l
l
a
t
e
d
b
u
t
t
h
e
R
M
S
E
w
a
s
s
t
i
l
l
l
o
w
.
T
h
e
p
e
r
f
o
r
ma
nce
of
t
rai
n
in
g
pro
c
ess
c
a
n
be
s
e
e
n
b
y
ca
l
c
u
l
a
tin
g
t
h
e
R
M
S
E
v
a
l
ues
.
T
he
n
ea
rest
R
MS
E
va
lue
to
z
e
r
o
va
l
ue
w
a
s
t
he
h
i
g
he
st
m
ode
l
per
f
orm
a
nce
.
There
f
ore
,
t
he
t
rai
n
i
ng
proc
e
s
s
pe
rfor
m
a
n
ces
d
e
p
en
de
d
on
the
RM
S
E
.
F
r
om
t
he
r
esults
o
bta
i
ned,
a
ll
t
he
R
M
S
E
v
a
l
u
e
s
f
o
r
t
r
a
i
n
i
n
g
w
e
r
e
n
e
a
r
t
o
z
e
r
o
.
A
f
t
e
r
t
h
e
A
N
N
t
r
a
i
n
i
ng
mo
de
l
pr
oc
ess
was
successf
ully
d
e
v
e
lo
p
e
d
,
t
h
e
ANN
f
a
u
l
t
classif
i
er
w
as
d
ev
elop
ed
b
y
u
s
in
g
th
e
sam
e
i
npu
t
a
nd
struc
t
ure.
T
he
A
N
N
f
a
u
l
t
class
i
fier
w
a
s
t
e
s
te
d
a
n
d
v
a
li
d
a
te
d
u
nde
r
differ
e
nt
f
a
u
lt
r
e
s
i
s
ta
n
c
es
a
n
d
c
o
n
d
i
tio
ns.
The
r
e
su
lt
s
o
b
ta
i
n
ed
f
rom
ANN
f
a
u
l
t
cl
assi
f
i
er
s
h
o
w
e
d
t
h
a
t
ANN
faul
t
cl
as
sif
i
er
f
o
l
l
o
wed
t
h
e
ta
rg
et
o
u
t
p
u
t
e
x
a
ct
ly
f
or
d
i
f
fere
nt
d
at
a
u
n
d
e
r
al
l
cond
it
io
n
s
.
Th
u
s
,
t
h
e ANN
mo
d
e
l
fo
r fault
d
e
t
e
c
t
i
o
n
w
as p
ro
v
e
n t
o
h
a
v
e a g
ood p
e
rf
o
r
ma
n
c
e
.
5.
CONCL
U
S
ION
I
n
conc
lusi
o
n
, th
e
A
N
N
m
odel w
a
s less compl
i
ca
t
e
d a
s
it
di
d n
o
t
re
q
uire the
e
xac
t
f
orm
of an
a
ly
tic
a
l
f
u
n
c
t
i
on
o
n
whi
c
h
th
e
mo
d
e
l
sh
oul
d
b
e
buil
t
t
o
d
e
v
e
l
o
p
t
h
e
mo
d
e
l
.
M
o
r
eov
e
r,
t
h
e
d
ev
el
op
ed
ANN
mo
del
h
a
d
show
n
g
o
o
d
e
fficie
n
c
y
,
ba
sed
o
n
t
he
out
p
u
t
a
n
d
t
a
r
g
et
v
al
ues
o
f
ANN
f
a
ul
t
cl
as
si
fi
er
m
odel
were
exa
c
t
l
y
the
sa
m
e
.
In
c
o
n
c
l
usi
on,
t
he
f
a
u
l
t
dete
c
t
i
o
n
an
d
cla
s
s
i
fica
tio
n
m
e
th
od
of
W
T
b
y
u
s
i
ng
A
N
N
w
e
r
e
suc
cessfu
l
l
y
de
v
e
l
ope
d.
ACKNOW
LEDG
E
MEN
T
S
The
a
u
t
h
or
w
o
u
l
d
l
i
k
e
to
a
ckn
o
w
l
e
d
ge
t
he
s
u
p
port
from
the
F
u
n
d
a
m
e
nta
l
R
es
ea
rch
Grant
S
c
hem
e
(
F
R
G
S
)
under
a
g
r
a
n
t
n
u
m
b
er
o
f
F
R
G
S
/
1
/2
0
1
7
/
TK
10
/
U
N
I
MA
P
/
0
2/
10
f
rom
t
h
e
M
i
ni
st
ry
o
f
H
i
g
h
er
E
duca
t
i
on M
a
la
ys
ia.
REFE
RENCES
[1]
Zh
en-Yo
u
Zhan
g
a
nd
K
e-S
h
eng
W
a
ng,
"
W
i
n
d
t
u
r
b
i
n
e
f
au
lt
d
et
ecti
o
n
b
as
ed
o
n
SCAD
A
dat
a
a
n
a
lys
i
s
usi
ng
AN
N,"
in
Ad
van
ces i
n
M
a
n
u
fa
ctu
r
i
n
g
, v
ol.
2
(
1),
pp.
70-7
8
,
2
0
1
4
.
[2]
Gl
ob
a
l
W
ind
E
n
ergy
C
ou
nci
l
,
"G
lob
a
l
Win
d
Rep
ort
|
Gw
ec
,
"
2
0
1
6
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN: 2088-
8694
Int J
P
o
w
El
e
c
&
D
ri S
yst
,
V
ol.
10,
N
o.
3
, S
e
p
2
0
1
9
:
168
7
– 1
693
1
692
[3]
E.
T
echn
o
l
ogy
,
"Redu
c
i
n
g
cost
s
of
e
m
e
rg
in
g
ren
e
wabl
e
en
e
r
gy
t
ec
h
nol
og
ie
s
-
an
a
nal
y
s
i
s
o
f
t
h
e
dyn
am
ic
dev
e
lo
pm
ent
with
w
i
nd
po
wer
as
case s
t
u
dy Sti
n
e
G
r
enaa J
e
nsen
,"
T
echn
o
lo
gy
,
vol.
2
,
p
p
.
17
9-20
2,
2
0
04.
[4]
British
W
i
nd E
n
ergy Associa
t
i
on, "
W
i
nd
turbi
n
e
technology bri
ef
in
g s
h
eet,
"
Br.
Wi
nd
En
ergy
Ass
o
c.
, 2
00
5
.
[5]
Raed
K
I
brahi
m
,
Jann
is
T
aut
z
-Wein
e
rt
a
nd
S
im
on
J
W
a
t
s
o
n
,
"
N
eura
l
ne
twork
s
f
o
r
w
ind
tu
rb
ine
fa
ult
de
te
c
t
io
n
via
c
u
rre
nt sig
na
ture
a
n
a
ly
si
s,
"
P
r
e
s
e
n
te
d
a
t
t
he
W
in
d
Euro
pe
S
u
m
m
i
t,
H
am
burg,
2
7-29th
S
e
pt
2
0
16.
[6]
Carlos
M
.
Pezzani,
J
ose
M
.
Bossi
o
,
A
r
i
el
M
.Ca
s
tell
ino,
G
uillerno
R
.
B
o
s
s
i
o
and
Cri
s
tian
H.
D
e
A
ngel
o
,
"A
P
LL-
bas
e
d
resam
p
l
i
ng
tech
niq
u
e
f
o
r
vi
bration
an
aly
s
i
s
i
n
v
a
riabl
e
-s
peed
w
ind
turbi
n
es
w
ith
P
M
S
G:
A
b
earing
f
a
ult
case,"
M
e
c
h
an
ica
l
S
y
st
ems
and
Sign
al Pr
oces
si
ng
,
vol.
85
,
pp
.
3
54-366
,
2
01
7.
[7]
P
e
ng
G
uo
,
Dav
i
d
In
f
i
e
l
d
and
Xiyun
Y
an
g,
"
Wi
n
d
t
urb
i
n
e
g
enera
t
o
r
c
ondition
monit
o
r
i
ng
u
s
i
ng
t
e
m
perat
u
re
t
rend
anal
ys
is
,
"
IEEE Tran
sa
c
t
io
ns
on
Sustainable
E
n
e
r
gy
, vo
l
.
3(
1) ,
p
p
.
1
24
-13
3
, 2
01
2.
[8]
L.
M
.
Pop
a
,
B
.
B
.
Jen
s
en
,
E
.
R
it
ch
ie
a
nd
I.
Bol
d
ea,
"
Con
d
i
tio
n
mon
ito
r
i
n
g
o
f
w
i
n
d
g
e
n
e
r
a
t
o
r
s,"
vo
l.
3
,
pp
.
18
39
-
1
8
4
6
,
2
003.
[9]
S.
Y
.
K
i
m, I
. H.
R
a
,
an
d S
. H.
Kim,
"Design
o
f w
i
n
d
t
urb
i
n
e
f
au
lt det
e
c
t
i
o
n
s
ystem b
a
sed
on
perf
o
rm
an
ce curv
e," in
6t
h Inter
natio
nal Co
nf
erence
on So
ft
Comp
u
tin
g
a
n
d
Int
e
llig
e
n
t S
y
st
ems
,
a
nd 13th
Inter
nati
onal Symp
os
ium
o
n
Ad
van
c
ed
Int
e
llig
e
nce S
y
stem
s,
S
C
IS/
I
SIS
20
12
,
pp.
2
03
3-2
036
,
201
2.
[10]
K.
K
im
,
G
.
P
art
h
asarat
hy
,
O.
U
l
u
y
o
l
,
W
.
F
o
sli
e
n,
S
.
S
h
en
g,
a
n
d
P
.
F
l
e
m
i
n
g
,
'
U
s
e
o
f
S
C
A
D
A
d
a
t
a
f
o
r
f
a
i
l
u
r
e
det
ectio
n
in
w
ind
tu
rb
i
n
es
,
'
i
n
AS
ME 20
11
5th
In
ter
n
a
t
i
onal Co
n
f
er
ence on
E
n
erg
y
S
u
s
t
ain
ability,
P
a
rts
A,
B
,
and
C
,
p
p
.
2
07
1–
20
7
9
,
20
11
.
[11]
M
.
L
i
t
o
n
H
o
ss
a
i
n
,
A
.
A
b
u
-
S
i
a
d
a
,
a
n
d
S
.
M
.
M
u
y
ee
n
,
'
M
e
t
h
o
d
s
f
o
r
a
d
vanced
w
ind
turbine
co
nd
it
ion
mo
n
itoring
a
nd
early di
a
gnos
i
s
:
A
l
iterature review,"
En
erg
i
es
,
vol.
1
1
,
n
o.
5
,
2
018.
[12]
A.
S
alem,
A.
A
bu-Si
ada,
a
nd
S.
I
sl
a
m
,
"Improved
condition
monit
ori
n
g
tech
ni
q
u
e
f
o
r
w
i
nd
t
u
rb
i
n
e
g
earbo
x
an
d
sh
aft stres
s
d
etecti
on,'
IET S
c
i.
Meas
.
T
echn
o
l.
,
vo
l
.
1
1,
n
o
.
4
,
p
p
.
431-4
3
7
,
2
017
.
[13]
Hi
chem
.
M
e
rabet,
T
ahar.
Bahi
a
nd
N
ou
ra
H
alem
,
'Co
n
d
i
t
i
on
m
o
n
ito
rin
g
a
nd
f
au
lt
d
etect
io
n
in
w
in
d
tu
rbi
n
e
bas
e
d
o
n
DFI
G b
y
t
he
f
u
zzy log
i
c
,”
Energ
y
Pr
oced
ia
, v
ol.
7
4
,
p
p
.5
18
-5
2
8
, 2
01
5.
[14]
Braul
i
o
Baraho
na,
N
i
co
laos
A
.
Cu
tul
u
lis,
An
ca
D
.,
Han
s
en
a
n
d
P
o
ul
S
o
r
ense
n
,
"
Unb
a
lan
ce
v
o
l
t
age
f
a
ults:
t
h
e
im
p
a
c
t
o
n
stru
ctu
r
al
l
o
a
ds
o
f
doub
ly
f
ed
a
s
y
n
c
hro
nou
s
g
e
nerato
r
w
i
n
d
tu
r
b
in
e
s
,
”
Wi
nd E
n
er
gy
,
vo
l
.
1
7
,
p
p.
1
1
2
3-
11
35
,
2
0
1
4
.
[15]
Bindi
Chen,
Peter
C.
M
atthe
w
s
an
d
P
e
t
e
r
J
.
T
a
v
n
e
r
,
"
W
i
n
d
t
u
r
b
i
n
e
pit
c
h
f
a
ult
s
p
ro
gno
si
s
using
a-p
r
iori
k
nowled
g
e-
bas
e
d
A
N
F
I
S,
”
E
x
pert Sys
t
ems
w
i
th
Applicat
i
o
n
s
,
v
o
l
.
4
0(17
),
pp.
6
86
3-6
876
,
201
3.
[16]
M
.
M
.
Ismail
an
d
A
.
F
.
Bend
ary,
"
P
r
otect
io
n
o
f
D
F
I
G
wi
nd
t
urbin
e
u
s
ing
fu
z
z
y
l
og
ic
c
on
tr
ol
,
"
Alexand
ri
a En
g. J.
,
v
o
l.
5
5
, n
o.
2,
pp
. 94
1
–
9
4
9
, 2
01
6.
[17]
M
.
S
chl
echting
e
n
an
d
I.
F
.
S
a
nto
s
,
"
W
in
d
tu
rb
i
n
e
co
nd
iti
o
n
mon
i
t
o
r
i
n
g
b
a
s
e
d
o
n
S
C
A
D
A
d
a
t
a
u
s
i
n
g
n
o
r
m
a
l
beh
a
vi
or m
od
els. P
art
2
:
Ap
p
licati
on exam
p
l
es
,
"
App
l
.
Soft Comput
.
J.
, v
ol
.
1
4
, n
o
. P
AR
T C,
pp
. 4
47
-46
0
,
2
01
4.
[18]
O.
N
ou
reld
een
a
n
d
I
.
H
a
m
d
an
,
"An
effi
cien
t
AN
F
I
S
crow
bar
p
r
o
t
ec
tion
for
d
f
i
g
w
ind
t
u
rbines
during
faul
t
s
,"
in
20
17
1
9
t
h
In
terna
t
i
onal Mi
ddle-E
a
st Power
S
y
st
ems Con
f
er
en
ce, ME
PCON 2
0
17
- Pr
oceed
ing
s
,
v
o
l
.
2
01
8-
F
e
bru
a
ry
,
p
p.
263–2
69
,
2
018
.
[19]
A.
M
el
lit,
S
.A.
Kal
o
girou,
L
.
H
o
nt
ori
a
a
nd
S
.
Shaari,
"
Artifici
al
i
nt
e
lligence
techniques
f
or
s
izing
phot
ovol
t
ai
c
syst
e
m
s: A revie
w,
”
Ren
e
wabl
e an
d S
u
sta
i
n
a
b
l
e
Energ
y
Revi
ews
,
vol
.
1
3
,
pp.
4
0
6
-419,
2
0
0
9
.
[20]
F
.
F
a
z
l
i
a
n
a
,
S
.
M
.
Z
a
l
i
,
R
.
N
o
r
f
a
d
i
l
a
h
a
n
d
M
o
h
d
A
l
i
f
I
s
m
a
i
l
,
"
D
yn
am
ic
m
o
d
el
o
f
di
st
rib
u
ti
on
n
et
wo
rk
cell
us
in
g
artif
ic
ial
neu
r
al
n
etw
o
rk
a
ppro
ach,"
2
0
1
6
In
tern
at
ion
a
l
Conference o
n
A
d
van
c
es i
n
E
l
ectrica
l, El
ec
tro
n
i
c
an
d
Sys
t
ems
E
n
g
i
neer
in
g
(
I
CAEES)
,
pp
. 48
4
-
48
7
.
[21]
F
.
S
.
P
a
n
c
hal
a
n
d
M
.
P
anchal
,
"Review
on
m
eth
ods
o
f
s
e
lectin
g
n
u
m
ber
of
h
i
dden
no
des
in
a
rtificial
neu
r
al
net
w
ork
,
"
In
t. J.
Co
mpu
t
. Sci
.
M
o
b. Co
mp
u
t
.
vol.
311
,
n
o.
11,
p
p
. 4
55
–46
4,
2
0
14.
[22]
T.
N
.
W
a
ng
,
C.
H
.
Ch
eng,
a
n
d
H
.
W.
C
hi
u
,
"
P
r
edi
c
tin
g
post
-
tre
a
t
men
t
s
urvi
vabil
i
t
y
o
f
patien
t
s
w
i
t
h
b
reas
t
cancer
us
in
g
artifici
al
n
eural
n
e
tw
ork
met
h
o
d
s,"
in
P
r
o
ceedin
gs
of
the Ann
u
a
l
Internat
io
nal Confer
ence o
f
th
e IEEE
En
gi
neeri
ng in
M
e
d
i
ci
ne
a
n
d
Biol
ogy
Soci
e
ty,
EMB
S
, p
p.
12
9
0
-1
29
3, 2
01
3.
[23]
F
.
S
. P
a
nchal
an
d
M
.
P
anchal
,
"Int
ernat
i
on
al
j
o
u
rnal
o
f co
m
p
ute
r
sc
ie
nc
e
a
n
d
mob
i
le
c
om
pu
ting
re
vie
w
o
n
me
tho
d
s
of
s
electin
g n
u
mber
o
f
hidd
en n
o
d
es i
n
art
i
fi
cia
l
n
eu
ral n
e
two
r
k
,
"
20
14
.
[24]
Nav
een
K
u
m
ar
C
h
a
nd
a
a
n
d
Y
o
n
g
F
u,
"
AN
N-based
f
a
u
l
t
cl
ass
i
fi
catio
n
an
d
lo
c
a
ti
on
i
n
m
v
dc
s
h
i
p
b
o
a
rd
p
o
w
er
syst
e
m
s,"
20
11
Nor
t
h Am
eri
c
an Power
Sym
p
o
s
i
u
m
,
p
p
. 1
- 7
.
[25]
Th
e M
a
thWork
s
,
I
nc, “W
in
d
F
a
rm
-DF
I
G
De
t
a
il
ed M
o
d
el”.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t
J
P
o
w
Elec
&
D
r
i
S
y
st
I
S
S
N
:
2088-
86
94
Fa
ul
t
de
tec
t
i
o
n
and cl
assi
fic
a
t
ion i
n
w
i
n
d
tur
b
i
n
e by
us
i
ng
a
r
t
ific
ia
l
ne
ural
n
e
t
w
ork (N. F.
Fa
dza
i
l)
1
693
BIOGRAPHI
E
S
OF
AUT
HORS
No
or
F
a
z
li
an
a
bt
F
a
d
za
il
recei
ved
the
B.E
n
g
a
n
d
M.
S
c
.deg
rees
f
ro
m
t
h
e
Univ
e
r
sity
M
al
aysi
a
Pe
rlis,
Ma
la
ysia
,
in
2
01
2
a
n
d
2013
,
re
sp
e
c
t
iv
e
l
y
,
a
nd
n
o
w
b
ec
om
e
a
part
tim
e
P
h
.D
.
stu
d
en
t
in
Un
iversit
y
M
alay
si
a
P
e
rlis,
all
in
e
lect
rical
e
n
g
ineerin
g.
S
h
e
i
s
c
u
r
r
e
n
t
l
y
a
L
e
c
t
u
r
e
r
w
i
t
h
t
h
e
Cent
re
f
or
D
ip
lo
m
a
S
t
udies
,
U
n
iv
ersi
ti
M
alay
sia
Pe
rli
s
,
Ka
n
g
a
r
,
Perli
s, Mal
ays
i
a
.
S
a
m
i
la
M
at
Z
al
i
r
eceiv
e
d
th
e
B.E
n
g
an
d
M
.
S
c
.
deg
r
ees
f
rom
the
Natio
nal
U
n
i
v
ers
ity
o
f
Ma
la
y
s
ia
,
Ma
la
y
s
ia
,
in
1
99
9
a
n
d
20
02
,
re
sp
e
c
t
iv
e
l
y
,
a
nd
t
h
e
P
h
.
D.
d
eg
ree
f
r
om
T
h
e
U
ni
versit
y
of
M
an
ches
ter,
M
anches
ter,
U
.K
.,
i
n
20
12,
a
ll
in
e
l
ectrical
e
n
g
ineerin
g.
S
he
i
s
curren
t
l
y
a
S
eni
o
r
Lect
urer
w
it
h
the
Sch
o
o
l
o
f
El
ectrical
S
yst
e
m
s
E
n
g
i
n
eerin
g,
U
ni
ve
r
s
it
i
M
a
la
y
s
ia
P
e
r
li
s,
K
a
n
ga
r
,
Perlis,
Mal
a
ysia.
Evaluation Warning : The document was created with Spire.PDF for Python.