Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
9
, No
.
3
,
Ma
rch
201
8
,
pp.
69
0
~
695
IS
S
N:
25
02
-
4752
,
DOI: 10
.11
591/
ijeecs
.
v9.i
3
.
pp
690
-
695
690
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Harmon
ic Load
Diagnosti
c Techn
iqu
es
an
d Meth
od
ologies:
A
Review
A.
S
. Hussi
n
1
, A.
R.
Abdul
la
h
2
, M.H
. Jopri
3
, T
.
S
ut
ikn
o
4
,
N.M.
S
aad
5
, We
ihown Tee
6
1,2,3,6
Cent
er
for Robotic
s
and
In
dustria
l
Autom
ation
(CeRIA)
,
Fa
cul
t
y
of
El
e
ct
ri
c
al
Engi
ne
eri
ng
,
Univer
siti
Te
kn
i
kal
Malay
s
ia Mel
ak
a
,
Ma
l
y
asi
a
4
Depa
rtment of
El
e
ct
ri
ca
l
Eng
in
ee
ring
,
Univ
ersitas Ahm
ad
Dahlan (UAD
),
Yog
y
a
kar
ta,
Indon
esia
5
CeRIA,
Fa
cul
t
y
of
E
lectr
oni
c a
nd
Com
pute
r
En
gine
er
ing, Univers
it
i Te
knik
al Mal
a
y
si
a
Mel
aka
,
Malay
s
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
un
9
, 201
6
Re
vised
N
ov
2
0
, 2
01
6
Accepte
d
Dec
11
, 201
6
Thi
s
pape
r
will
re
vie
w
on
the
exi
sting
techni
ques
and
m
et
ho
dologi
es
o
f
har
m
onic
lo
ad
dia
gnostic
s
y
s
tem
.
The
inc
r
ea
si
ngl
y
amount
of
har
m
oni
c
produc
ing
lo
ad
used
in
pow
er
s
y
st
em
are
the
m
ai
n
co
ntri
buti
on
in
quant
if
y
ing
eac
h
har
m
onic
disturba
nc
e
eff
ec
ts
of
th
e
m
ult
iple
har
m
onic
produc
ing
loa
ds
and
i
t
b
ec
ame
ver
y
important.
L
it
er
at
ur
e
pro
poses
two
diffe
re
n
t
t
ec
hniq
ues
and
m
et
hods
on
the
har
m
onic
source
id
entificat
ion
und
e
r
the
soft
computing
techniq
ue
class
ifi
catio
n.
The
adva
n
ta
ges
an
d
disadva
nt
age
s
of
har
m
onic
loa
d
ide
nti
fi
cation
t
ec
hniqu
es
and
m
et
hods
are
discussed
in
th
is
pape
r
.
In
the
pr
oposed
m
et
hod,
the
issue
on
th
e
har
m
onic
cont
ribution
is
det
ermine
and
t
ra
nsform
ed
to
a
dat
a
cor
r
elati
o
n
ana
l
y
sis
.
Sever
al
te
chn
iq
ues
to
ide
n
ti
f
y
the
source
s
of
har
m
onic
sign
als
in
el
e
ct
ri
c
power
s
y
stems
are
desc
rib
ed
and
re
vie
wed
base
d
on
pre
vious
pape
r
.
Com
par
at
ive
st
udie
s
of
the
m
et
hods
are
a
lso
done
to
e
val
ua
te
t
h
e
per
form
anc
e
of
ea
ch
te
chn
ique
.
How
eve
r,
witho
ut
suffic
i
ent
inf
orm
at
ion
in
thi
s
inc
onsistent
envi
ronm
ent
on
t
he
prope
rt
y
of
the
power
s
y
ste
m
,
ac
cur
a
t
e
har
m
onic
produ
ci
ng
loa
d
d
ia
g
nosis
m
et
hods
are
important
and
furth
er
inve
stigations
in
thi
s r
ega
rd
assum
es
gre
at i
m
pli
c
at
ion
.
Ke
yw
or
d
s
:
Diag
nosti
c
Har
m
on
ic
Har
m
on
ic
loa
d
P
ower
s
y
stems
Copyright
©
201
8
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
:
A.
S
. Hus
sin
,
Ce
nter fo
r
R
obotics an
d Ind
ust
rial
A
ut
om
ation
(CeR
IA),
Faculty
of Elec
tric
al
Engineer
ing
,
Unive
rsiti
Teknikal M
al
a
ysi
a Me
la
ka,
Ma
ly
asi
a
Em
a
il
:
adibah
.
s.hussin
@
gm
ail.co
m
1.
INTROD
U
CTION
Power
qual
it
y
disturba
nces
a
re
bec
om
ing
m
ajo
r
iss
ues
to
al
l
el
ect
rici
t
y
con
s
um
ers
with
the
ra
pid
grow
t
h
of
el
ec
tric
al
equ
ipm
ent
us
e
d
in
our
te
chnolo
gies
[
1].
Thes
e
distu
rb
a
nces
ca
n
af
fect
a
lot
of
se
ns
it
ive
loads
that
are
connecte
d
to
the
p
ower
sy
stem
wh
ic
h
le
ad
to
ha
r
dw
a
r
e
fail
ur
e
an
d
m
al
fu
nctio
n
[
2]
.
The
m
on
it
or
ing
of
t
he
powe
r
qual
it
y
sign
al
s
are
sti
ll
us
ing
the
conve
ntion
al
te
chn
i
qu
e
s
that
are
base
d
on
vis
ual
of
vo
lt
age
a
nd
current
wa
ve
for
m
s
[3
]
.
The
di
agnosis
of
th
ese
disturba
nc
es
are
v
ery
di
ff
ic
ult
an
d
it
req
ui
re
eng
i
neer’s
ex
pe
rtise
and
kn
owle
dge
in
m
a
ny
el
ect
rical
a
reas
[4
]
.
S
om
e
wav
ef
orm
dis
tortio
ns
that
usual
ly
aff
ect
the
po
w
er
qual
it
y
sign
al
are
har
m
on
ic
and
inter
ha
r
m
on
ic
distor
ti
on
s
[5
]
.
I
de
ntific
at
ion
of
ha
r
m
on
ic
load
ha
s
bee
n
in
creasin
gly
im
po
rtant
du
e
to
the
increas
e
le
vel
of
ha
r
m
on
ic
distor
ti
on
pr
ese
nt
in
powe
r
syst
e
m
s.
The
identific
at
ion
and
cl
assifi
cat
ion
of
ha
rm
on
ic
loads
base
d
on
pow
er
s
yst
e
m
m
easur
e
m
ents
unde
niably
a
di
ff
ic
ult
ta
sk
[
6]
.
The
fr
e
que
nt
ly
chan
gi
ng
na
ture
of
ha
rm
on
ic
an
d
t
he
un
certai
nty
assoc
ia
te
d
with
the
ha
rm
on
ic
c
har
act
eri
zat
ion
of
eve
n
com
m
on
ly
us
ed
loa
ds
,
m
ake
the
us
e
of
conve
ntion
al
m
et
hods
ou
t
dated
[7
]
,
[
8].
Re
duci
ng
a
nd
c
ontr
olli
ng
su
c
h
ha
rm
on
ic
s
ha
ve
bee
n
a
m
ajo
r
c
oncer
n
of
powe
r
e
ng
i
ne
er
s
for
m
any
ye
ar
s.
Re
search
es
predict
that
ha
r
m
on
ic
le
vels
i
n
po
wer
syst
e
m
are
go
in
g
to
increase
in
t
he
fu
tu
re
[9
]
co
ns
ide
rin
g
the
conve
nien
ce
broug
ht
to
e
ver
y
day
li
fe
by
the
us
e
of
th
ese
nonlinea
r
load
s.
Ha
rm
on
ic
s
are
her
e
to
sta
y
re
gardless
of
the
il
l
eff
ect
s
associat
ed
with
the
pr
ese
nce
of
ha
rm
on
ic
s
are
we
ll
do
cum
ented
in
the
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
Ha
r
m
on
ic
Loa
d Dia
gnos
ti
c
T
echn
i
qu
e
s
and Me
thod
ologies: A
Revi
ew
(
A.
S.
H
us
sin
)
691
li
te
ratur
e.
T
he
increase
d
in
t
he
non
-
li
near
l
oa
d
use
s
m
ay
di
stortin
g
the
ste
ady
sta
te
of
th
e
current
wav
e
form
s
and
ac
volt
age.
It
is
in
ret
urn
beco
m
es
a
po
t
entia
l
prob
le
m
to
the
re
gula
r
perform
ance
of
po
wer
syst
e
m
.
This
threat
m
ay
resu
lt
in
both
c
urr
ent
an
d
volt
ag
e
to
ap
pear
as
a
non
-
sin
us
oi
da
l
wav
ef
orm
at
a
giv
e
n
locat
ion.
The
har
m
on
ic
disto
rtion
issue
is
beco
m
ing
m
ore
i
m
po
rta
nt
in
powe
r
syst
em
[10].
It
is
usual
ly
no
t
possi
ble
to
separ
at
e
the
c
on
t
r
ib
ution
s
or
to
ide
ntify
the
s
ource
of
the
ha
rm
on
ic
distor
ti
on
by
j
ust
lo
ok
i
ng
a
t
the
wav
e
f
or
m
s.
The
i
ncr
easi
ngly
am
ou
nt
of
ha
rm
on
ic
pro
du
ci
ng
loa
d
use
s
in
t
he
power
syst
em
no
wa
days
a
re
m
aking
t
he
ha
r
m
on
ic
pr
od
ucing
l
oad
dia
gnost
ic
issue
bec
om
e
m
or
e
com
ple
x.
It
is
ver
y
i
m
po
rtant
to
de
velo
p
te
chn
iq
ues
a
nd
m
et
ho
ds
to
m
easur
e
t
he
ha
r
m
on
ic
distur
ba
nce
of
al
l
the
har
m
on
ic
pro
duci
ng
loa
d
in
powe
r
qu
al
it
y
m
anag
e
m
ent
par
ti
cul
arly
wh
e
n
t
here
is
a
ha
rm
on
ic
distu
rb
a
nce
in
po
wer
syst
e
m
.
Me
asur
ing
each
har
m
on
ic
im
pa
ct
s
of
the
se
ve
r
al
har
m
on
ic
producin
g
loa
ds
is
beco
m
ing
m
or
e
im
po
rtant
especial
ly
right
after
the
pap
e
r
pro
posin
g
the
"re
ward
an
d
puni
sh
m
ent
pr
ogra
m
"
[1
1].
W
ide
sp
rea
d
us
e
of
har
m
on
ic
pro
duci
ng
loads
a
re
m
aking
the
har
m
onic
s
disturba
nce
bec
om
e
an
essenti
al
par
t
of
the
po
wer
qua
li
ty
netwo
r
ks
[
12
]
.
Var
i
ou
s
regula
ti
on
s
li
ke
the
I
EC
61000
a
nd
IEEE
St
d.519
-
19
92
[13]
res
tric
t
the
power
util
it
ie
s
to
op
erate
pro
per
ly
w
it
hi
n
the
certai
n l
im
it
s o
f
dist
or
ti
on.
Ah
m
ad
et
.al
[
14]
re
viewe
d
se
ver
al
of
the
ex
ist
ing
syst
em
s,
m
et
ho
ds
a
nd
t
echn
i
qu
e
s
that
are
us
ed
to
m
on
it
or
po
wer
qual
it
y.
The
pap
e
r
is
m
ai
nly
fo
c
us
on
th
e
powe
r
qu
al
i
ty
m
on
it
or
in
g
syst
e
m
s
wh
ic
h
are
com
po
sed
of
va
rio
us
too
ls
,
com
m
un
ic
at
ion
li
nk
s,
s
of
t
war
e
et
c.
that
wo
r
k
tog
et
he
r
as
one
co
her
e
nt
syst
e
m
.
They
are
de
ve
lo
ping
an
unde
rstan
ding
ab
out
the
power
qual
it
y
m
anag
e
m
ent
in
the
area
of
powe
r
in
du
st
ry.
Power
qu
al
it
y
m
et
er
place
m
e
nt
te
chn
i
qu
es
t
hat
wer
e
pr
e
se
nted
a
nd
the
ba
sic
idea
of
ea
ch
m
et
ho
d
or
s
yst
e
m
are
discusse
d
in
order
to
ha
ve
an
un
der
st
and
i
ng
a
bout
it
s
i
m
po
r
ta
nce
and
r
ole.
T
he
com
par
ison
s
of
the
te
chn
iq
ues
is
m
ade in
term
s o
f
their m
erit
s an
d dem
erit
s.
Supr
iy
a an
d N
a
m
biar
[
12
]
on
the o
t
her
ha
nd r
evie
wed
t
he h
arm
on
ic
iden
ti
f
ic
at
ion
tech
niques in
t
heir
pap
e
r.
Si
ng
le
po
i
nt
m
e
tho
ds
and
m
ulti
po
int
m
et
ho
ds
ar
e
pro
p
os
ed
as
t
he
two
m
ai
n
te
c
hn
i
qu
e
s
of
ha
r
m
on
ic
so
urce
ide
ntific
at
ion
in
their
pap
er
.
They
pro
vid
e
a
detai
le
d
su
r
vey
of
these
m
et
hods
al
ong
with
their
adv
a
ntage
s
an
d
disa
dvanta
ge
s.
Ba
sed
on
th
ei
r
rev
ie
w,
t
he
y
con
cl
ud
e
d
th
at
har
m
on
ic
sta
te
est
i
m
a
ti
on
us
in
g
the
m
ulti
po
int
m
et
ho
ds
a
re
bette
r
su
it
e
d
f
or
har
m
on
ic
s
ource
ide
ntific
at
ion
.
Howe
ve
r,
without
s
uff
ic
ie
nt
inf
or
m
at
ion
on
the
to
polo
gy
of
the
pow
er
syst
em
,
correct
ha
rm
on
ic
source
ide
ntific
at
ion
m
et
hods
a
re
essenti
al
and i
nv
e
sti
gations i
n
this
r
e
gard as
su
m
es g
re
at
si
gn
i
ficance.
Id
e
ntific
at
ion
of
ha
rm
on
ic
lo
ad
s
ource
i
n
c
om
plex
powe
r
netw
orks
ass
um
es
gr
eat
si
gnific
ance
f
or
determ
ining
th
ese
com
m
itters
of
ha
rm
on
ic
s.
Even
t
ually
,
to
eff
ect
ively
detect
the
har
m
on
ic
m
i
ti
gation
m
eans
so
that the e
ntire po
wer
syst
e
m
is n
ot poll
uted.
No
resea
rc
her
s
has fou
nd
any stan
dard on
de
fini
ng
t
he m
et
hod
of
ide
ntifyi
ng
t
he
dom
inant
har
m
on
ic
disturbance
[15].
Wh
en
the
locat
io
n
of
the
do
m
inant
har
m
on
ic
so
urce
s
is
unkn
own
to
the
util
it
ie
s,
sever
al
te
c
hnic
al
pro
blem
s
will
com
e
up
.
These
pr
ob
le
m
s
m
ay
include
th
e
m
al
fu
nctio
ning
of
the
in
strum
ents,
reducti
on
i
n
the
li
fe
of
the
el
ect
ric
al
equ
ipm
ent
and
oth
e
r
fe
w
c
reati
ng
op
e
rati
onal
pr
ob
le
m
s
in
the
identific
at
io
n
m
e
tho
ds
of
powe
r
netw
ork.
Acc
ur
at
e
te
c
hn
i
qu
e
s
f
or
ha
rm
on
ic
so
urce i
den
ti
fi
cat
ion
a
re esse
ntial
u
nde
r
s
uc
h
ci
rc
um
sta
nces.
Fig
ure
1
s
hows
the
b
l
ock d
i
agr
am
o
f
t
he
e
xisti
ng
te
chn
iq
ues
and
m
e
tho
ds
on th
e iden
ti
ficat
io
n o
f harm
on
ic
lo
ad.
Figure
1. Bl
oc
k diag
ram
f
or
t
he
cl
assifi
cat
io
n of ha
rm
on
ic
load ide
ntific
at
ion m
et
ho
ds
2.
DI
SCUS
SION OF
M
E
THOD
S
2.1. Neur
al
Ne
tworks
An
a
da
ptive
pe
rcep
ti
on
a
ppr
oach
i
n
ne
ur
al
netw
orks
ha
ve
been
act
ivel
y
te
ste
d,
us
ed
and
a
ppli
ed
su
ccess
fu
ll
y
fo
r
ha
rm
on
ic
est
i
m
ation
in
a
po
w
er
syst
em
.
The
adv
a
nt
ages
of
ap
pl
yi
ng
arti
fici
al
neu
ral
netw
orks
are
their
abili
ty
to
recog
nize
non
-
li
near
functi
on
s
,
adap
ta
ti
on
to
dif
fer
e
nt
envi
ronm
ents
and
hi
gh
no
ise
tolera
nce
.
These
net
wor
ks
ha
ve
to
be
trai
ne
d
pro
per
l
y
fo
r
accu
rate
identific
at
ion
of
ha
rm
on
ic
sources
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vol
.
9
,
No.
3
,
Ma
rc
h
201
8
:
690
–
695
692
This
m
e
tho
d
of
loa
d
ide
ntific
at
ion
hel
ps
in
non
-
i
ntrus
ive
har
m
on
ic
so
urce
detec
ti
on
with
ou
t
hu
m
an
i
nterv
e
ntio
n.
T
o
m
ake
init
ia
l
est
i
m
at
es
of
ha
rm
on
ic
sourc
es
in
a
po
wer
syst
e
m
with
nonlinea
r
l
oads,
ne
ur
al
netw
orks
a
re
a
ble
to
be
us
e
a
nd
r
el
at
ively
fe
w
pe
rm
anen
t
ha
rm
on
ic
m
easur
in
g
instr
um
ents.
The
est
im
ates
can
then be
us
e
d w
it
h
the stat
e est
i
m
ation
t
o
f
i
nd addit
ion
al
unknow
n
s
ources
.
In
[
16]
, to
m
ake in
it
ia
l
estim
a
te
s o
f
har
m
on
i
c so
urces i
n
a p
owe
r
syst
e
m
with non
li
nea
r
loads
, n
eu
ral
netw
orks
a
re
app
li
ed
.
F
or
f
ur
t
her
im
pr
ov
es
the
m
easurem
ents,
the
i
niti
al
est
i
m
a
te
s
are
t
hen
us
ed
as
ps
e
udom
easurem
ents
for
harm
on
ic
sta
te
est
i
m
a
ti
on
.
By
usi
ng
a
sim
ulatio
n
te
sts,
it
shows
that
the
t
raine
d
neural
netw
ork
s
are
a
ble
to
pr
oduce
acc
epta
ble
est
im
a
te
s
fo
r
va
ryi
ng
ha
r
m
on
ic
so
urces.
I
n
a
dd
it
io
n,
t
he
sta
te
est
i
m
at
or
will
gen
e
rall
y
pu
ll
these
est
i
m
a
t
es
cl
os
er
to
th
e
cor
rect
va
lu
es.
Eve
ntu
al
ly
,
the
process
c
a
n
be
su
ccess
fu
ll
y
m
on
it
or
ed
a
nd
ide
ntifie
d
t
he
"sus
pecte
d"
ha
rm
on
ic
source
that
ha
d
not
pr
e
vious
ly
been
m
easur
ed
.
H
oweve
r,
the
neural
netw
orks
m
us
t
"l
earn
"
to
a
sso
ci
at
e
the
av
ai
la
ble
power
netw
ork
data
pa
tt
ern
s
with
patte
rn
s
of
ha
rm
on
ic
so
urce
be
hav
i
or.
T
o
le
ar
n
t
hi
s
behavi
or
,
s
yst
e
m
op
erati
ng
data
a
nd
da
ta
are
ob
ta
ine
d
from
tem
po
rar
y
ha
rm
on
ic
so
urce
m
on
it
or
s
at
known
s
ource
s.
The
n,
th
e
neural
net
wor
ks
will
est
i
m
at
e
the
har
m
on
ic
so
urc
es
based
on
e
xperie
nce
in
the
sa
m
e
way
an
exp
e
rience
d
operat
or
in
fer
s
ps
e
u
do
m
easur
em
ents f
r
om
av
ai
la
ble d
at
a f
or co
nve
ntion
al
sta
te
es
tim
a
ti
on
.
Sr
ini
vasan
[
17]
proposes
a
neural
ne
tw
orks
base
d
a
ppr
oach
f
or
no
n
-
intru
si
ve
har
m
on
ic
s
ource
identific
at
ion.
The
a
pp
li
cat
io
n
of
ne
ur
al
ne
tworks
base
d
m
od
el
s
on
noni
ntr
us
ive
si
gnat
ure
ide
ntific
at
ion
without
hu
m
an
inte
rv
e
ntio
n
in
in
vestigat
ed
in
t
his
pa
pe
r.
Ne
ur
al
netw
orks
are
trai
ne
d
to
e
xtract
im
portan
t
featur
e
s
from
t
he
input
curre
nt
wav
e
form
i
n
this
approac
h
to
un
i
quel
y
i
den
ti
fy
va
rio
us
ty
pes
of
loads
us
ing
their
disti
nct
ha
rm
on
ic
sig
na
tures
.
T
her
e
f
ore,
the
no
n
-
in
va
sive
loa
d
dia
gnos
ti
c
will
be
i
m
po
rta
nt
in
f
utu
r
e
powe
r
-
qu
al
it
y
m
on
it
or
i
ng
and
en
ha
nce
m
ent
syst
e
m
s.
The
functi
on
an
d
cl
assi
f
ic
at
ion
ap
pro
xim
a
ti
on
capab
il
it
ie
s
of
arti
fici
al
neura
l
netw
orks
ha
ve
bee
n
us
ed
in
powe
r
-
qu
al
it
y
stud
ie
s,
fa
ult,
a
nd
ha
rm
on
ic
s
s
ource
cl
assifi
cat
ion
[
18
]
–
[
22
]
.
Me
a
nwhile
,
the
s
upp
ort
vecto
r
m
achine
m
od
el
has
s
how
n
a
hi
gh
pote
ntial
in
powe
r
har
m
on
ic
s
rela
te
d
patte
rn
rec
ogniti
on
[
23]
,
[24]
.
O
ne
the
m
ajo
r
be
ne
fits
that
are
der
iv
ed
from
the
par
al
le
l
structu
re
of
ar
ti
fici
al
neu
ral
netw
orks
is
their
abili
ty
to
adap
t
to
di
ff
e
re
nt
env
i
ronm
ents,
their
high
no
ise
tolerance
an
d
abili
ty
to
recogn
iz
e
non
-
li
ne
ar
f
unct
ions.
T
he
diag
nosis
of
ha
rm
on
ic
pr
oduci
ng
loa
ds
by
us
i
ng
the
app
li
cat
io
n
of
arti
fici
al
neu
ral
net
works
has
bee
n
ve
rifi
ed
befo
r
e
with
gr
eat
su
cce
ss.
The
us
e
f
ul
asp
ect
s
of
su
c
h
an
ap
plic
at
ion
are
discu
ssed
by
Var
a
da
n
an
d
Ma
kra
m
in
[25]
.
I
n
the
pa
per,
it
sh
ow
s
t
he
su
cces
s
in
the
i
m
ple
m
entat
io
n
eve
n
th
ough
the
cho
ic
e
of
arch
it
ect
ur
e
,
the
to
po
l
og
y
a
nd
t
he
sel
ect
ion
of
featu
res
of
th
e
neural
netw
orks
a
re
s
pec
ulati
ve
decisi
ons
base
d
on
en
gi
neer
i
ng
ju
dg
e
m
ent.
The
kn
own
bac
k
-
pro
pa
gatio
n
te
chn
iq
ue
an
d
a
s
up
e
rv
ise
d
le
arn
in
g
proce
dure
is
us
e
d
a
nd
ad
opte
d
in
the
netw
ork
t
ra
inin
g.
H
owe
ver,
t
o
exam
ine
the
ne
ur
al
netw
orks
,
the
pr
e
-
pr
oce
ssing
ra
w
data
is
nee
ded
fro
m
the
powe
r
s
yst
e
m
bef
or
e
bein
g
input t
o
t
he net
works.
Ra
y
and
S
ubudhi
[
26
]
a
re
m
ai
nly
fo
c
us
es
on
e
xp
l
oiti
ng
t
wo
com
pu
ta
ti
onal
intel
li
gen
c
e
te
chn
i
ques;
arti
fici
al
neu
ra
l
networks
an
d
ev
olu
ti
ona
ry
com
pu
ta
ti
on
te
chn
i
qu
e
s
in
ha
rm
on
ic
s
so
ur
ces
identific
at
ion
i
n
powe
r
syst
em
.
An
accu
rate
e
stim
ation
of
ha
rm
on
ic
s
in
di
storted
po
wer
syst
e
m
cur
re
nt
or
volt
age
si
gnal
is
i
m
po
rtant
to
ef
fici
ently
desig
n
filt
ers
f
or
ha
rm
on
ic
s
el
i
m
inati
on
.
A
ne
w
est
i
m
ation
al
gor
it
h
m
fo
r
est
i
m
at
ing
har
m
on
ic
c
on
t
e
nts
in
a
pow
e
r
syst
em
sign
al
co
ntam
inate
d
with
no
ise
is
presente
d
in
the
pa
per
.
It
is
es
sentia
l
for
desig
ning
filt
ers
f
or
el
im
inati
ng
a
nd
reducin
g
t
he
e
ff
ect
s
of
ha
rm
on
ic
s
in
a
po
wer
syst
em
af
te
r
th
e
est
i
m
ation
of the
har
m
on
ic
c
on
te
nt of t
he p
ow
e
r
syst
em
s
i
gn
al
by
us
in
g
t
he pr
opos
e
d
al
gorithm
.
On
t
he
oth
er
ha
nd,
Ja
nan
i
a
nd
Him
avathi
[
27
]
proposes
the
ne
ural
net
w
orks
base
d
ap
proac
h
f
or
the
identific
at
ion
of
var
i
ou
s
harm
on
ic
so
ur
ces
pr
ese
nt
in
an
e
le
ct
rical
instal
l
at
ion
.
T
he
ha
r
m
on
ic
so
ur
ce
dev
ic
e
s
are
ide
ntifie
d
in
this
m
et
ho
d
by
usi
ng
thei
r
disti
nct
'
har
m
on
ic
sig
natu
re
s'
extracte
d
from
the
input
current
wav
e
f
or
m
.
W
i
th
increase
in
the
nu
m
ber
of
loads
an
d
the
ir
com
bin
at
ions,
the
c
om
plexity
wil
l
be
increase.
Ther
e
f
or
e,
s
uc
h
a
uto
m
at
ed
non
-
intr
us
i
ve
ha
rm
on
ic
load
dev
ic
e
i
de
ntific
at
ion
m
ay
help
in
m
on
it
ori
ng
an
d
enh
a
ncin
g
po
wer
qual
it
y.
Finall
y,
the
perf
or
m
ance
of
a
neural
netw
ork
s
to
a
la
rg
e
e
xt
ent
de
pends
upon
t
he
ty
pe
of arc
hite
ct
ur
e
us
e
d
a
nd
their lea
r
ning a
lgorit
hm
.
2.2. Fuz
z
y
Log
ic
A
f
uzzy
log
ic
is
a
m
et
ho
d
to
com
bin
e
the
natu
ral
even
ts
with
the
im
pr
eci
sion
ass
ociat
ed
with
the
com
pu
ta
ti
on
al
powe
r
of
the
c
om
pu
te
r
to
pro
du
ce
highly
fle
xib
le
,
rob
us
t
and
i
ntell
igent
r
easo
ning
syst
em
s
.
It
consi
sts
of
m
ath
em
atical
streng
th
to
t
he
certa
in
em
ulati
on
of
li
nguisti
c
an
d
at
tribu
te
s
ass
oc
ia
te
d
with
hum
an
cogniti
on.
The
theor
y
of
f
uzz
y
log
ic
is
based
upon
the
f
un
ct
ion
s
of
pe
rsuasi
on
a
nd
c
ogniti
ve
processe
s
an
d
the
no
ti
on
of
r
el
at
ive
gr
ade
d
m
e
m
b
ersh
ip.
T
he
util
it
y
of
fuzzy
set
s
li
es
in
their
abili
ty
to
m
od
el
un
ce
rta
in
or
vague
data
wh
i
ch
a
re so
of
te
n enco
unte
re
d
in
r
eal
li
fe.
Me
thod
ba
sed
on
f
uzzy
lo
gi
c
for
har
m
on
i
c
source
ide
ntific
at
ion
is
presented
by
Ba
ns
hwa
r
a
nd
Chan
del
[
28
]
.
By
us
ing
the
f
uzzy
IF
-
TH
EN
infer
e
nce
r
ule
s,
ha
rm
on
ic
sources
a
re
detec
te
d.
The
n,
t
he
ou
t
pu
ts
of
the
f
uzzy
rul
e
base
s
are
then
de
-
f
uzzified
to
reco
ve
r
cri
sp
valu
es
to
id
entify
har
m
on
ic
so
urces.
By
us
i
ng
the
powe
r
qu
a
li
ty
analy
zer,
t
he
ha
rm
on
ic
sign
al
s
a
nd
s
pe
ct
ru
m
s
are
ob
t
ai
ned
wh
ic
h
a
ct
as
an
in
pu
t
in
th
e
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
Ha
r
m
on
ic
Loa
d Dia
gnos
ti
c
T
echn
i
qu
e
s
and Me
thod
ologies: A
Revi
ew
(
A.
S.
H
us
sin
)
693
pro
po
se
d
m
et
ho
d.
To
i
den
ti
f
y
the
har
m
on
ic
s
sources
i
n
powe
r
syst
e
m
,
the
te
chn
i
qu
e
util
iz
es
the
cur
re
nt
a
s
well
as
vo
lt
ag
e
sp
ect
r
a
of
l
oa
ds
.
Va
rio
us
e
xam
ples
are
done
to
dem
on
s
trat
e
the
po
te
nt
ia
l
of
the
pro
po
s
ed
te
chn
iq
ue.
The
a
ppli
cat
ion
of
f
uzzy
lo
gic
to
power
qu
al
it
y
issue
is
al
so
pr
opos
e
d
by
Naw
i
et
.al
[
29
]
.
It
pap
e
r
descr
i
bes
on
how
the
s
ources
of
har
m
on
ic
s
detect
ed
by
us
i
ng
f
uzzy
set
s
and
I
F
-
T
HE
N
infe
ren
ce
s
ru
le
s
.
It
is
well
known
t
ha
t
the
har
m
on
i
cs
can
create
unwa
nted
im
pacts
on
power
syst
e
m
s
and
to
e
ve
nt
ually
to
the
end
-
us
ers
.
Ma
nufa
ct
ur
es,
cust
ome
r
an
d
el
ect
ric
al
util
it
ie
s
have
al
ways
bee
n
bother
ed
by
har
m
on
ic
distor
ti
on
pro
blem
s
in
power
syst
em
.
B
y
app
ly
ing
the
app
li
cat
io
n
of
Fast
Fo
uri
er
T
ran
s
f
or
m
,
the
har
m
on
ic
signa
ls
and
sp
ect
r
um
are
determ
ined.
F
o
r
these
cases,
a
fu
zzy
infe
rence
syst
e
m
is
e
xp
e
rim
ental
l
y
i
m
ple
m
ented
t
o
sho
w
the
ge
ner
al
pr
ocedu
res
of
ho
w
to
use
this
theo
ry.
I
n
this
appr
oach,
it
app
ears
t
hat
the
fu
zzy
set
s
the
ory
can
play
an
im
po
rt
ant
r
ole
in
dia
gnos
i
ng
pow
er
qual
it
y
disturbance
s.
T
he
refor
e
,
it
can
offe
r
insi
gh
ts
t
owa
rd
s
the
sat
isfact
ion
of
the n
ee
ds o
f
c
ust
om
ers,
m
anu
factur
e
rs
a
nd
ut
il
i
ti
es.
The
dia
gnos
is
of
har
m
on
ic
pro
du
ci
ng
lo
a
d
is
i
m
po
rta
nt
in
orde
r
to
gi
ve
a
pro
per
gu
i
dan
ce
f
or
m
itigati
on
act
i
on.
T
her
e
f
or
e,
the
a
bili
ty
of
fu
zzy
l
og
ic
is
app
li
ed
as
i
ntegr
at
in
g
syst
em
to
ide
ntify
va
rio
us
har
m
on
ic
pro
duci
ng
l
oad
s
of
tho
se
dist
urbance
pa
rtic
ularly
of
ha
rm
on
ic
.
It
is
do
ne
by
a
da
pting
the
f
uzzy
set
s
and
IF
-
TH
EN
infer
e
nce
r
ul
es.
As
far
as
the
har
m
on
ic
s
load
ide
ntifi
cat
ion
is
co
nc
ern
e
d,
t
he
te
chn
i
qu
e
dev
el
op
e
d
by
us
in
g
the
fu
zz
y
log
ic
m
e
tho
d
by
stud
yi
ng
t
he
beh
a
vior
a
nd
char
act
erist
ic
of
t
he
distu
r
ba
nce
at
the d
at
a
pro
du
ced a p
ro
m
isi
ng
resu
lt
.
3.
COMP
A
RA
TI
VE ST
U
DIES
OF
ME
THOD
S
Othe
r
than
cl
as
sifyi
ng
the
loa
d
diag
nosti
c
ap
proac
h,
it
is
i
m
portant
to
com
par
e
in
div
i
du
al
te
chn
iq
ues
and
dif
fe
ren
t
cat
egories.
C
om
par
ison
of
s
om
e
of
the
m
et
hods
us
e
d
in
har
m
on
ic
l
oa
d
diag
no
sis
ha
s
bee
n
at
tem
pted
by
sever
al
re
searc
he
rs.
T
he
m
os
t
com
pr
ehe
ns
ive
and
ea
rlie
st
co
m
par
ison
s
wa
s
m
ade
by
W
il
li
s
and
Northc
ote
-
Gr
e
en
[
30
]
.
T
hey
did
the
diag
nosis
m
et
ho
d
co
m
par
ison
te
sts
on
14
loa
ds
.
Me
anwhil
e,
D
ash
[
31]
al
so
di
d
the
dia
gnos
is
c
om
par
ison
m
et
ho
d
on
seve
ral
ne
ur
al
netw
orks
a
nd
t
he
f
uzzy
lo
gic
base
d
m
et
ho
ds.
Li
u
[32]
com
par
e
d
three
ot
her
te
c
hn
i
qu
e
s
ne
ural
netw
orks
,
f
uz
zy
lo
gic
a
nd
a
utoreg
ressive
m
od
el
s
on
t
he
basis
of
the
si
m
ulati
on
stud
y.
It
is
co
nclu
ded
in
the
pap
er
that
the
neu
ral
net
wor
ks
an
d
fuzzy
l
og
ic
are
m
uch
bette
r
com
par
ed
to
a
utoreg
ressive
m
od
el
s.
To
est
ablish
the
s
upe
rior
it
y
if
their
pro
po
se
d
diag
no
sti
c
m
et
ho
ds,
m
any
researc
hes
pro
vid
e
oth
er
li
m
i
te
d
com
par
at
iv
e
data
ov
e
r
the
lim
it
ed
nu
m
ber
of
previ
ou
sly
publishe
d
m
eth
ods.
Mbam
al
u
and
El
-
Ha
war
y
[
33]
has
done
th
e
com
par
ison
of
thei
r
aut
or
e
gr
essi
ve
m
od
el
with
the
Bo
x
-
Jen
kins
m
et
ho
d.
Me
an
wh
il
e,
Will
is
[34]
m
akes
a
com
par
iso
n
betwee
n
thei
r
sim
ulatio
n
m
et
hod
with
tw
o
othe
r
si
m
ulati
on
m
e
t
hods
.
The
rece
nt
com
par
iso
ns
of
t
he
di
ff
e
rent
load
diag
no
si
s
m
e
tho
ds
are
need
e
d
t
o
pro
vi
de
a
chall
eng
i
ng
oppo
rtu
nity
fo
r
fu
t
ur
e
resear
ch.
W
it
h
the
wide
va
riet
y
of
as
su
m
ptions
an
d
ob
j
ect
iv
es,
th
e
un
li
m
i
te
d
po
ss
ibil
it
y
of
m
a
tch
in
g
an
d
m
ixing
of
diff
e
rent
co
m
po
ne
nts
of
var
io
us
m
e
thods
is
achievabl
e
.
Ba
sed
on
the
s
ta
ti
sti
ca
l
ro
bus
t
m
et
ho
d
[
35]
the
el
ect
ric
load
dia
gnos
is
of
diff
e
re
nt
te
chni
qu
es
a
re
com
pare
d
by
us
in
g
the
s
hort
te
rm
load
di
agnostic
.
Ot
he
r
m
e
tho
d
li
ke
t
he
dia
gnos
e
th
e
load
in
sm
art
gr
id
is
discusse
d
by
Z.
Aun
g
[
36]
a
nd ide
ntifyi
ng
the loa
d by c
onsiderin
g
t
he
m
et
eorolo
gy f
act
or
s
is e
xp
al
ai
ne
d by Y
.Jin
[
37]
.
The
a
pp
li
cat
io
n
of
a
rtific
ia
l
ne
ur
al
netw
orks
m
e
tho
d
t
o
the
pro
b
le
m
of
ha
rm
on
ic
load
di
agnosis
has
so
m
e
pr
act
ic
al
issue
a
nd
they
are
discusse
d
in
[
25
]
.
Ba
sed
on
the
pa
per
,
it
is
sh
ow
n
that
in
the
i
de
ntific
at
io
n
process
,
the
m
agn
it
ude
an
d
ha
rm
on
ic
orde
r
of
the
i
nject
ed
har
m
on
ic
c
urre
nt
are
the
only
two
crit
ic
al
fea
tures
.
Fu
rt
her
m
or
e,
the f
eat
ure
s
pac
e
is
not
se
pa
rated b
y
sim
ple
lin
ear
bo
unda
ries
f
ro
m
the
nee
d
of
a h
id
de
n
l
ay
er
i
n
the
im
ple
m
entat
ion
s
howe
d.
The
c
om
plex
hype
r
-
s
urface
s
existe
nce
in
the
feat
ur
e
s
pace
that
discr
i
m
inate
betwee
n
t
he
cl
asses
of
loa
ds
j
ust
ifie
s
the
use
of
a
rtific
ia
l
ne
ur
al
netw
orks
f
or
this
a
ppli
cat
ion
.
The
c
hoic
e
of
the
final
net
w
ork
in
te
rm
s
of
it
s
arc
hitec
ture
an
d
t
opolog
y
was
e
xam
in
ed
on
t
he
basi
s
of
se
ver
al
fa
ct
or
s
,
trai
ning
tim
e
a
nd
c
orrect
cl
assifi
cat
ion
bei
ng
the
m
ai
n
iss
ues
of
the
deci
sion
process.
Table
1
il
lustr
at
e
the
detai
le
d nu
m
ber
s
of
t
he
a
rtic
le
s that
us
ed
n
e
ural
n
e
t
works a
nd fuzzy
lo
gic
m
et
ho
d.
Table
1.
List
of Lit
eratu
re
for
H
a
rm
on
ic
Lo
a
d Id
e
ntific
at
ion
Neu
ral
N
etwo
rk M
eth
o
d
Fu
zzy
Log
ic M
eth
o
d
[
1
0
,1
1
,20
,21
]
[
2
2
,2
3
]
1.
Neural
Netw
orks:
Ne
ural
ne
tworks
is
one
of
the
m
os
t
current
ef
fecti
ve
cl
assifi
cat
ion
m
et
ho
ds.
I
ts
natu
ral
sp
ee
d
a
nd
ve
rsati
li
ty
a
re
the
a
dv
a
nta
ges
of
c
hoos
i
ng
ne
ural
netw
ork
i
n
the
data
cl
assifi
cat
ion
.
It
can
ha
ndle
the
non
-
li
near
a
nd
m
ulti
-
var
ia
bl
es
data
set
s.
Bi
tt
er
et
.al.
[38],
disc
us
se
d
crit
ic
al
cases
i
n
intru
si
ons
li
ke
sp
am
,
wo
rm
being
re
so
l
ved
by
ne
ur
al
net
w
orks.
He
re
por
ts
that
dataset
char
act
e
risti
cs,
su
c
h
as
f
or
m
at,
siz
e
a
nd
dim
ensio
nalit
y
are
ve
ry
crit
ic
al
in
order
to
m
odel
a
su
cces
sf
ul
ne
ur
al
netw
ork
.
Fo
r
har
m
on
ic
dev
ic
e
ide
ntific
at
ion
,
the
harm
on
ic
com
po
ne
nts
of
i
nput
c
urren
t
wa
veform
are
us
ed
a
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vol
.
9
,
No.
3
,
Ma
rc
h
201
8
:
690
–
695
694
so
urces
of
valuab
le
in
form
ation
.
Howe
ver,
the
par
am
et
e
r
of
t
he
phas
e
ang
le
of
th
e
currents
ca
n
com
pr
om
ise
the
ha
rm
on
ic
sou
rce ide
ntific
at
ion
as s
how
n b
y [39].
2.
Fu
zzy
lo
gic:
B
y
app
ly
in
g
th
e
IF
-
THE
N
ru
le
s
,
f
uzzy
lo
gic
is
a
pote
ntial
te
c
hn
i
qu
e
to
c
ope
with
decisi
on
-
m
aking
strat
eg
ie
s.
It
ca
n
pro
vid
e
a
li
nguist
ic
represe
ntati
on
a
nd
can
s
ol
ve
the
no
n
-
li
near
pr
ob
le
m
s.
Fu
zzy
log
ic
cl
assifi
ers
are
pr
opos
e
d
by
Liu
et
.al.
[4
0],
w
her
e
f
uzzy
syst
e
m
e
m
p
loye
d
to
evaluate
th
e
po
te
ntial
threa
ts.
The
res
ults
show
t
hat
f
uz
zy
syst
e
m
cou
ld
dec
rease
th
e
false
al
arm
rate
and
pro
vide
bette
r
e
valuati
on of t
he po
te
nt
ia
l t
hr
eat
s.
4.
CONCL
US
IO
N
A
re
view
of
t
he
ha
rm
on
ic
load
diag
nosis
te
chn
iq
ues
a
nd
m
et
ho
ds
a
re
discu
ss
ed
a
nd
the
possible
directi
on
of
f
urt
her
re
searc
h
are
a
naly
zed
i
n
this
pa
pe
r.
B
ased
on
this
,
t
wo
di
ff
e
ren
t
m
et
hods
unde
r
t
he
s
of
t
com
pu
ti
ng
te
c
hn
i
qu
e
s
to
det
erm
ine
the
ind
ividu
al
c
on
t
rib
utions
of
m
ultip
le
har
m
on
ic
load
s
are
res
ea
rch
e
d.
Eve
n
th
ough
the
te
chn
i
ques
and
m
et
ho
ds
for
the
diag
no
sti
cs
hav
e
been
us
e
d,
t
he
al
gorithm
s
us
e
d
to
determ
ine
the
har
m
on
ic
loa
d
of
m
ulti
ple
har
m
on
ic
so
urc
es
rem
ai
n
to
be
ver
ifie
d
due
to
it
s
accuracy
le
vel.
Ther
e
f
or
e,
the f
ut
ur
e
ste
p
that
need
to b
e
ta
ke
n
in
t
his
resea
rch
is
t
o
de
vel
op
a
bette
r
t
he
or
et
ic
al
sup
por
t
to
the
te
chn
iq
ues
and
m
e
tho
ds a
nd get
ti
ng
a
data
processin
g
m
et
ho
d.
Diff
e
re
nt
so
ft
com
pu
ti
ng
te
chn
i
qu
e
s
nam
ely;
neu
ral
netw
ork
an
d
fu
zzy
log
ic
hav
e
be
en
us
e
d
t
o
har
m
on
ic
loa
d
diag
nosti
c.
Th
e
adv
a
ntage
s
a
nd
disad
va
ntag
es
of
t
hese
m
eth
ods
a
nd
te
c
hn
iqu
es
a
re
prese
nted
and
discuss
e
d.
It
can
be
co
nclud
e
d
f
ro
m
the
works
repo
rted
so
far
that
de
m
and
on
the
diagnostic
te
chn
i
qu
e
s
base
d
on
soft
com
pu
ti
ng
m
et
hods
are
gaini
ng
hi
gh
ly
bene
fici
al
for
thei
r
eff
ect
ive
us
e
.
The
resear
ch
ne
ed
to
be
rep
la
ci
ng
a
nd
s
hiftin
g
fro
m
the
old
m
e
t
hods
an
d
te
ch
ni
qu
es
with
a
ne
wer
an
d
m
or
e
accurate
on
e
.
Th
us
,
a
furthe
r
re
sear
ch
on
this
is
su
e
us
in
g
a
n
ad
van
ce
d
di
gital
sign
al
proces
sin
g
s
uc
h
as
Tim
e
-
Fr
equ
e
ncy
Distrib
ution (T
FD
)
is
need
e
d du
e
to
ac
hie
ve
high acc
ur
acy
and reli
able
res
ult o
f harm
on
i
c load dia
gnos
i
s.
ACKN
OWLE
DGE
MENTS
This
resear
ch
is
su
pp
or
te
d
by
Advan
ce
d
Di
gi
ta
l
Sign
al
Pr
oc
essing
La
bora
tory
(ADS
P
Lab).
Sp
eci
al
than
ks
al
s
o
to
the
Fac
ulty
of
Ele
ct
rical
E
ng
i
neer
i
ng
an
d
E
ng
i
neer
i
ng
Tech
nolo
gy
of
U
niv
e
rsiti
Tek
nik
a
l
Ma
la
ysi
a
Me
lak
a
(U
TeM
),
Ce
nter
for
R
obot
ic
s
and
I
ndus
t
ria
l
A
uto
m
at
io
n
(CeR
IA)
of
UTeM,
Mi
nistry
of
Higher
E
ducat
ion
Ma
la
ysi
a
(MO
HE)
a
nd
Mi
nistry
of
S
ci
ence,
Tec
hnology
an
d
I
nn
ov
at
io
n
(MOST
I)
f
or
giv
in
g
their
c
oope
rati
on
an
d
fun
ding
for
thi
s
researc
h
wit
h
grant
num
ber
06
-
01
-
14
-
SF
00119
L
0002
5.
Their
su
pp
or
t i
s
gra
te
fu
ll
y ac
kn
ow
l
edg
e
d.
REFERE
NCE
S
[1]
N.
A.
Abidull
ah,
A.
R.
Abdulla
h,
N.
H.
Sham
sudi
n,
N.
H.
T.
H.
Ahm
ad,
and
M.
H.
Jopri,
“Real
-
tim
e
power
quali
t
y
signals
monitoring
system,
”
Proce
ed
ing
-
2013
IEE
E
Studen
t
Conf.
Res.
Dev.
S
COReD
2013,
no.
Dec
ember
,
pp.
433
–
438,
2015
.
[2]
A.
Rahi
m
Abdulla
h
,
N.
H.
T
.
H.
Ahm
ad,
N.
A.
Abidull
ah
,
N.
H.
Sham
sudi
n,
and
M.
H.
J
opri,
“
Perform
anc
e
Eva
lu
at
ion
of
R
ea
l
Pow
er
Qualit
y
Disturb
anc
es
Anal
y
sis
Us
ing
S
-
Tra
nsform
,
”
A
ppl.
Me
ch.
Mat
e
r.
,
vol.
752
–
753
,
pp.
1
343
–
1348
,
2015.
[3]
N.
H.
T.
Huda
,
A.
R.
Abdulla
h
,
and
M.
H.
Jopri,
“
Powe
r
qualit
y
signals
det
e
cti
on
using
S
-
transform
,
”
Proc.
201
3
IEE
E
7th
Int. Power
Eng
.
Optim.
Conf.
PEOCO
2013,
pp
.
552
–
55
7,
2013
.
[4]
N.
H.
H.
and
A
bidul
la
h
,
N.
A,
Abdulla
h,
A.
R.
,
Zuri
_Sha’
ame
r
i,
A.
,
Sham
sudin,
N.H.,
Ahm
ad
and
M.
H.
Jopr
i,
“
Rea
l
-
Ti
m
e
Pow
er
Quali
t
y
Disturbances
Dete
c
t
ion
and
Cla
ss
ifica
t
ion
S
y
stem,”
World
Appl
.
Sci.
J
.
,
vol
.
32,
no.
8,
pp.
1637
–
1651
,
2014.
[5]
A.
R.
Abdull
ah,
N.
S.
Ahm
ad,
N.
Baha
r
i,
M.
Mana
p,
A.
Jidin
and
M.
H.
Jopri
,
"
Short
-
ci
rcui
t
analy
sis
of
vol
ta
ge
sour
ce
inv
erte
r
using
spec
trogram
,
''
in
IEE
E
Int
e
rna
ti
on
al
Confer
enc
e
on
l
ec
tr
ical
Mac
hine
s
and
Sy
stems
(ICEMS),
pp.
1808
-
1813
,
2013.
[6]
M.
H.
Jopri,
an
d
A.
R.
Abdullah,
"A
new
two
point
s
m
et
hod
for
ide
nti
f
y
dom
i
nant
har
m
o
nic
d
isturba
nc
e
using
fre
quency
and
p
hase
spec
trogram
,
''
In
te
rnationa
l
R
ev
i
ew
o
f
E
lect
rical
Engi
ne
erin
g
,
9(2)
,
453
-
459
,
2014.
[7]
A.
R.
Abdul
la
h,
N.
A.
Abidul
la
h
,
N.
H.
Sham
sudin,
N.
H.
H.
Ahm
ad,
and
M.
H
.
Jopri,
"P
erf
orm
ance
Veri
f
ic
a
ti
on
o
f
Pow
er
Quali
t
y
S
ign
al
s
cl
assifi
cat
ion
S
y
st
em.
''
In
Appl
ie
d
Me
chan
ic
s and
Ma
te
rial
s
,
Vol.
752,
115
8
-
1163,
2015
.
[8]
A.
R.
Abdul
la
h
,
N.
U.
R
.
B.
A.
H.
I.
R
.
A
.
H.
N
orddin,
N.
Z
.
Ab
idi
n,
A.
Am
an,
a
nd
M.
H.
Jopri
,
"
Leakage
curren
t
analy
sis
on
pol
yme
ric
and
no
n
-
poly
meric
insulat
ing
m
ate
ria
ls
using
ti
me
-
fr
eque
nc
y
distrib
uti
on,
'
'
In
I
EEE
Inte
rna
ti
ona
l
Co
nfe
re
nc
e
on
Pow
er
and
En
erg
y
(P
ECon),
pp
.
979
-
984,
2012
.
[9]
K.
D.
McBe
e
an
d
M.
G.
Sim
ões,
“
Eva
luating
the
long
-
term
impact
of
a
cont
inuo
usl
y
inc
r
ea
sing
har
m
onic
deman
d
on
fe
ed
er
-
l
eve
l
v
olt
a
g
e
distor
ti
on
,
”
IE
EE Tr
ans.
In
d.
App
l.,
vo
l. 50, no. 3, pp. 2142
–
2149,
2014
.
[10]
Z.
Yin
,
Y.
Sun,
and
T.
Yu
,
“N
ew
methods
ex
p
lorati
on
for
har
monic
sour
ce
id
ent
ification
te
ch
nologi
es,
”
DRP
T
2011
-
2011
4th
I
nt.
Conf
.
El
e
ct
r
.
Util
.
Dere
gu
l. R
estruc
t
.
Pow
er T
ec
hnol
.
,
pp.
399
–
402,
2011
.
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
Ha
r
m
on
ic
Loa
d Dia
gnos
ti
c
T
echn
i
qu
e
s
and Me
thod
ologies: A
Revi
ew
(
A.
S.
H
us
sin
)
695
[11]
A.
Mce
ac
h
ern
,
W
.
M.
G.
Rad
y
,
W
.
A.
Moncrie
f,
G.
T.
He
y
dt
,
an
d
M.
M.
Grana
g
han,
“
Reve
nu
e
a
nd
Harm
onic
s:
An
Eva
lu
at
ion
of
So
m
e
Propos
ed
Ra
te
Stru
ct
ure
s
,
”
I
EE
E
Tr
ans.
Pow
er
Deliv.,
vo
l. 10
,
no
.
1
,
pp
.
474
–
482,
1995
.
[12]
P.
Supri
y
a
and
T.
N.
P
adman
a
bhan
Nam
bia
r,
“
Revi
ew
of
har
m
onic
source
id
ent
ifica
ti
on
t
ec
h
nique
s,”
Int.
Rev.
El
e
ct
r.
Eng
.
,
vo
l
.
7
,
no
.
3
,
pp
.
45
25
–
4531,
2012
.
[13]
A.
R.
Abdull
ah,
G.
Z
.
Peng,
S
.
A.
Ghani
,
and
M.
H.
Jopri
,
“A
new
ve
c
tor
d
raft
method
for
harm
onic
sour
ce
det
e
ct
ion
at
poi
nt
of
common
coupl
ing,
”
Proc.
2014
IEE
E
8th
Int.
Pow
er
Eng.
Optim.
Conf.
PEOCO
2014,
no.
Marc
h,
pp.
110
–
114,
2014
.
[14]
B.
Alsa
yy
ed
Ah
m
ad,
H.
H
.
E
lShei
kh,
and
A.
F
adoun,
“R
evie
w
of
pow
er
qualit
y
monitoring
sy
stems,”
2015
In
t.
Conf.
Ind
.
Eng
.
Oper.
Man
ag.,
p
p.
1
–
8
,
2015
.
[15]
M.
H.
Jopri,
N.
A.
Abidull
ah
,
G
.
Z
.
Peng,
and
A.
R.
Abdull
ah,
“
A
new
two
point
s
m
et
hod
for
ide
nti
f
y
dom
inant
har
m
onic
distur
banc
e
using
fre
quency
and
pha
se
spec
trogr
am,
”
Int
.
R
ev.
E
lect
r.
Eng.
,
vo
l.
9
,
no.
2,
pp.
453
–
4
59,
2014.
[16]
K.
Hart
ana,
“
La
70
803,
”
Pow
er
,
vol.
5
,
no
.
4
,
pp
.
1098
–
1104,
199
0.
[17]
D.
Sriniva
san,
W
.
S.
Ng,
and
A.
C.
Li
ew,
“
Neura
l
-
ne
twork
-
base
d
signat
ure
re
cogni
t
ion
for
har
m
onic
source
ide
nti
f
icati
on
,
”
I
EE
E
Tr
ans.
Pow
er
Deliv.
,
vo
l. 21
,
no
.
1
,
pp
.
398
–
405,
2006
.
[18]
C.
Rese
ac
h
,
“
N
y
12345,
”
no.
3,
p
p.
229
–
232
,
198
3.
[19]
J.
W
ij
a
y
akul
aso
ori
y
a,
G.
Putrus,
and
P.
Minn
s,
“
El
ec
tric
po
wer
quali
ty
dist
urbance
cl
ass
if
i
cat
ion
using
self
-
adapti
ng
arti
fici
al
neural
net
wor
ks,
” IEE Proceed
ings
-
Gene
ra
t
ion, …, no. 1, 2002.
[20]
P.
Dash
and
D.
Sw
ai
n,
“
Pow
er
q
ual
ity
assess
m
ent
using
an
ada
ptive
neur
al
ne
twork,
”
Powe
r
Elec
tr
on
.
…,
pp.
770
–
775,
1996
.
[21]
M.
Rukonuzz
a
m
an
and
M.
N
aka
oka
,
“
Magnit
ude
and
phase
det
erminati
on
of
harm
onic
currents
by
adapti
v
e
le
arning
back
-
pr
opagati
on
neural
net
work,
”
Proc
.
IEE
E
1999
Int
.
C
onf.
Pow
er
El
e
ct
ron.
Drive
S
y
s
t
.
PEDS
’99
(Cat
.
No.99T
H8475), vol.
2
,
no
.
Jul
y
,
pp.
1168
–
1173
v
ol.
2,
1999.
[22]
S.
Santoso,
S.
Mem
ber
,
E.
J.
Pow
ers,
W
.
M.
G
ra
d
y
,
S.
Mem
ber
,
and
A.
C.
Parsons
,
“Power
Q
ual
ity
Disturban
ce
W
ave
form
Rec
o
gnit
ion The
or
et
i
ca
l
Foundati
on
,
”
Powe
r
,
vol
.
15
,
no.
1
,
pp
.
222
–
2
28,
2000
.
[23]
S.
Poy
hon
en,
M.
Negre
a,
a.
Ark
kio,
H.
H
y
ot
y
n
iem
i,
and
H.
Koivo,
“Fault
diagno
stic
s of
an
el
ec
tr
ic
al
machin
e
wit
h
multi
ple suppor
t
vect
or
cl
ass
if
i
ers,”
Proc.
I
EE
E
I
nte
rna
ti
na
l
S
y
m
p.
Int
el
l
.
Con
trol,
pp
.
373
–
378
,
2
002.
[24]
L.
S.
Moulin,
A
.
P.
A.
da
Silva,
M.
A.
El
-
Shark
awi,
and
R
.
J.
Marks,
“
Support
vec
tor
and
m
ul
ti
lay
e
r
per
ce
ptro
n
neur
al
ne
tworks
appl
ie
d
to
power
s
y
stems
tra
nsient
stabilit
y
an
aly
s
is
with
input
dimensional
ity
re
duct
ion
,
”
IE
E
E
Powe
r E
ng
.
So
c. Summer Me
e
t.
,
pp.
1308
–
1313
,
2002.
[25]
S.
Vara
dan
and
E.
B.
Makra
m
,
“
Prac
tical
co
nside
ra
ti
ons
in
the
ap
pli
c
at
ion
of
neur
al
net
works
to
th
e
ide
nt
ifi
c
at
ion
o
f
har
m
onic
lo
ads,”
E
le
c
tr.
Powe
r
Syst.
Re
s.
,
vo
l. 3
0,
no
.
2
,
pp
.
103
–
106,
1994
.
[26]
P.
K.
Ra
y
and
B
.
Subudhi,
“
Neuro
-
e
volutionar
y
appr
oac
h
es
to
power
sy
st
em
har
m
onic
s
esti
m
at
i
on,
”
Int.
J.
Elec
t
r.
Powe
r E
n
ergy
S
yst.
,
vol
.
64
,
pp
.
212
–
220,
2015
.
[27]
K.
Jana
ni
and
S.
Him
ava
thi,
“N
o
n
-
int
rus
iv
e
harm
onic
sour
ce
ide
n
ti
ficati
on
using
neural
ne
tworks,
”
2013
Int.
Conf
.
Com
put.
Pow
er,
E
ner
g
y
,
Inf. Co
m
m
un.
,
pp.
59
–
6
4,
2013
.
[28]
a.
Banshwar
and
a.
K.
Ch
ande
l
,
“Identi
ficati
on
o
f
harm
onic
sour
ce
s
using
fuzzy
l
ogic
,
”
2010
Jt.
I
nt.
Conf.
Pow
er
El
e
ct
ron. Drive
s
Ene
rg
y
S
y
st.
20
10
Pow
er
Indi
a,
pp.
1
–
7
,
2010
.
[29]
S.
M.
Nawi,
J.
Johari,
and
A.
F
.
Abidin
,
“A
fuz
zy
logi
c
applica
ti
on
for
id
entific
ati
on
of
harm
on
ic
s
disturbance
s
sour
ce
s,”
Natl. P
ower
Eng. Con
f.
PECon 2003
-
Proc.
,
pp.
27
–
31
,
2003
.
[30]
A.
K.
Singh,
I
bra
hee
m
,
S
.
Kh
at
oon,
M.
Mua
zz
am,
and
D.
K.
Chat
urv
edi,
“L
oad
forec
asti
ng
technique
s
a
nd
methodol
ogi
e
s:
A
rev
ie
w
,
”
ICPC
ES 20
12
-
2012
2nd
Int. Conf. Power,
Contr
ol E
m
bed.
S
y
st
.
,
no.
Dec
ember, 2012
.
[31]
P.
K.
DASH,
A.
C.
LE
IW
,
and
S.
RAH
M
AN
,
“A
compari
son
o
f
fuz
z
y
n
eur
al
n
et
works
for
the
gene
ra
ti
on
of
da
i
l
y
ave
ra
g
e and
p
eak l
oad
profi
le
s,
”
Int.
J. Sy
st
.
S
ci
.
,
vol.
26
,
no
.
11
,
p
p.
2091
–
2106
,
1
995.
[32]
K.
Li
u
et
a
l.,
“
Com
par
ison
of
ver
y
short
-
te
rm
load
fore
ca
st
ing
t
echnique
s,”
I
EEE
Tr
ans.
Powe
r
Syst.
,
vol
.
11
,
no.
2
,
pp.
877
–
882
,
19
96.
[33]
G.
A.
N.
Mbama
lu
and
M.
E
.
El
-
Hawar
y
,
“
Loa
d
Forec
asti
ng
Via
Suboptim
al
Seasona
l
Au
tore
gre
ss
ive
Models
and
Ite
ra
ti
ve
l
y
R
ewe
ight
ed
Least
Squ
are
s E
st
imati
on
,
”
IE
EE Tr
ans.
P
ower
Syst
.
,
vo
l.
8,
no
.
1
,
pp
.
343
–
348,
1993
.
[34]
H.
L.
W
illis,
M.
J.
Buri,
and
L.
A.
Finley
,
“
Fore
ca
sting
El
e
ct
ri
c
Dem
and
of
Distribut
ion
S
y
s
te
m
Planni
ng
in
Rur
al
and
Spars
ely
Po
pula
t
ed
Reg
ions,
”
IE
EE Tr
ans.
P
ower
Syst.,
vo
l.
10,
no
.
4
,
pp
.
20
08
–
2013,
1995
.
[35]
Y.
Chakhc
houk
h,
P.
Panci
a
tici,
and
L.
Mili
,
“
El
e
ct
ri
c
Loa
d
F
ore
ca
st
ing
Base
d
on
Stat
isti
c
al
Robust
Methods,
”
Powe
r Sy
st
.
I
EEE
Tr
ans.,
vo
l. 26
,
no
.
3
,
pp
.
982
–
991,
2011
.
[36]
Z
.
Aung,
M.
To
ukh
y
,
and
J.
W
illia
m
s,
“
Towa
rds a
cc
ur
at
e elect
r
icit
y
loa
d
for
ecasti
ng
in
sm
art
grids,
”
DBK
DA
2012,
…,
no
.
c
,
pp
.
51
–
57,
2012
.
[37]
Y.
Jin,
“
Mete
or
ologi
c
al
Fa
ct
ors
Considere
d
Lo
ad
Dec
oup
li
ng
Forec
asti
ng
Tec
hnique
,
”
J.
Adv.
Mate
r.
R
es
.
,
v
ol.
354
–
55,
pp
.
922
–
926,
2012
.
[38]
C.
Bit
te
r
,
D.
A.
El
iz
ondo
,
and
T.
W
at
son,
“
Ap
pli
c
at
ion
of
art
i
fic
i
al
neur
a
l
netw
orks
and
re
la
te
d
te
chni
qu
es
t
o
int
rusion
d
et
e
ct
i
on,
”
N
eural
N
etwor
ks
(
IJ
CNN)
,
2010
Int. Jt. Con
f.
,
pp.
1
–
8,
2010
.
[39]
S.
Vara
dan
and E.
B.
Makr
am,
“
¢
OW
|
R
Harm
onic
loa
d
ide
nt
if
ic
a
ti
on
and
de
te
r
m
ina
ti
on
of
loa
d
compos
it
ion
using
a
l
ea
st
square
s m
et
hod,
”
vo
l. 37,
1996.
[40]
L.
Li
u
,
P.
W
an,
Y.
W
ang,
and
S.
Li
u,
“
Cluste
ring
and
H
y
br
id
Gene
ti
c
Alg
orit
hm
base
d
Intrusion
Dete
ctio
n
Strat
eg
y
,
”
TEL
KOMNIKA
Indon
es.
J
.
Elec
tr.
E
n
g.
,
vol. 12, no. 1, pp. 762
–
770,
20
14.
Evaluation Warning : The document was created with Spire.PDF for Python.