Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
6, N
o
. 5
,
O
c
tob
e
r
201
6, p
p
. 2
437
~244
6
I
S
SN
: 208
8-8
7
0
8
,
D
O
I
:
10.115
91
/ij
ece.v6
i
5.1
074
4
2
437
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
Insight on Effecti
v
en
ess of
Frequently E
x
ercis
e
d P
Q
Classification Techniques
B. De
vi
Vigh
n
e
shw
a
ri
1
, R. Neela
2
1
Dept of
Electrical & Electron
i
cs Engg, the Ox
fo
rd College of En
gineer
ing, Bang
alore, Ind
i
a
2
Dept of
Electrical & Elect
ron
i
cs Engg, Annamalai Univ
ersity
C
h
idambaram,
In
dia
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Apr 6, 2016
Rev
i
sed
Ju
l 25
,
20
16
Accepted Aug 10, 2016
The growing demands of global consumer
m
a
rket in green
ener
g
y
s
y
s
t
em
have open
e
d the
doors for many
technolog
ies as well as various sophisticated
ele
c
tri
cal d
e
vi
ce
s
for both comm
ercial
and dom
es
tic us
age. H
o
wever, wit
h
the incr
eas
ing d
e
m
a
nds
of energ
y
and
be
tte
r qualit
y of servi
c
e
s
, there is
a
significant in
crease in non-
lin
earity
in lo
ad distribution
causing potential
effect on
the Po
wer Quality
(PQ). The
harmfu
l eff
ects on
PQ are various
events e.g. sag, s
w
ell, harmonics etc th
at causes significant
amount of sy
stem
degradation. Th
erefore,
this paper di
sc
usse
s va
rious signific
a
nt re
sea
r
ch
techn
i
ques per
t
aining to
the PQ distur
bance clas
sification s
y
s
t
em
introduced
b
y
the
authors
i
n
the p
a
s
t
and
anal
yz
es
i
t
s
eff
e
ctiv
enes
s
s
cal
e i
n
term
s
of
res
earch
gap
.
T
h
e pap
e
r d
i
s
c
us
s
e
s
s
o
m
e
of th
e frequ
entl
y
ex
ercis
e
d
P
Q
classifi
cat
ion t
e
c
hniques from
th
e m
o
st relev
a
nt
liter
a
tur
e
s in or
der to h
a
ve
more insights of
the
techn
i
ques.
Keyword:
Energy
Power qu
ality
Power system
PQ classi
fication
PQ d
e
tectio
n
R
e
newa
bl
e
Copyright ©
201
6 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
B.
De
vi Vig
h
n
e
shw
a
ri,
Asst
. Pr
of:
De
pt
o
f
El
ect
ri
cal
& Electronics
Engg.
The
O
x
f
o
r
d
C
o
l
l
e
ge o
f
E
ngi
ne
eri
n
g
B
a
ngal
o
re,
I
n
d
i
a
Em
a
il:d
ev
io
xfo
r
d
@
g
m
ail.co
m
1.
INTRODUCTION
W
i
t
h
the usa
g
e of m
odern technologies the cons
um
er
market for e
n
ergy has also
unde
rwent a
si
gni
fi
ca
nt
rev
o
l
u
t
i
o
n i
n
l
a
st
5 y
ears. Wi
t
h
t
h
e evol
ut
i
on
of sm
art
-
g
r
i
d
base
d el
ect
ri
cal
sy
st
em
,
i
t
i
s
ant
i
c
i
p
at
ed t
h
at
cons
um
ers do
get
a
n
un
di
st
o
r
t
e
d c
u
rre
nt
as
wel
l
as v
o
l
t
a
ge i
n
c
ont
i
n
u
ous
m
ode.
Unfortun
ately with
t
h
e larg
e
n
u
m
b
e
r
o
f
satu
rated
m
a
rket
of power electronics
hav
e
also
in
trod
uced
sop
h
i
s
t
i
cat
ed d
e
vi
ces an
d co
n
t
rol
l
e
r sy
st
em
. Int
r
od
uct
i
o
n of s
u
ch a
dva
n
ced t
echn
o
l
o
gi
es not
o
n
l
y
en
hance
s
the c
u
stom
er expe
rience
but
also in
c
r
ease
d
y
n
am
i
c
dem
a
nds
of
t
h
e
cu
st
o
m
ers [
1
]
.
One
of
t
h
e
i
m
port
a
nt
t
h
i
n
g
to
und
erstand
i
s
th
at Power Qu
ality i.e. PQ
plays a v
e
ry
imp
o
rtan
t
ro
le in
g
l
ob
al co
n
s
u
m
er m
a
rk
et.
PQ b
e
ars
m
u
l
tip
le p
e
rcep
tio
ns for d
i
fferen
t typ
e
s o
f
users. Fo
r
an
exa
m
p
l
e co
n
s
u
m
er u
tilities co
n
s
id
er PQ in
term
s
o
f
reliab
ility o
f
th
e system
wh
ereas dev
i
ce
man
u
f
act
u
r
i
n
g fi
rm
co
n
s
id
er PQ as
stand
a
rd
for stream
li
n
e
d
and
ethical powe
r
distribution se
rvices [2].
Ho
wev
e
r, th
e prime ob
j
e
ctiv
e
o
f
PQ
fo
r
ev
er
y co
n
s
u
m
e
r
is
to
e
n
su
r
e
an
o
p
t
i
m
i
zed u
s
age
o
f
c
u
r
r
ent
,
v
o
l
t
a
ge
, a
n
d
f
l
uct
u
at
i
o
n i
n
f
r
e
que
nci
e
s.
W
i
t
h
t
h
e i
n
cl
usi
o
n
o
f
vari
ou
s
fo
r
m
s of
n
on-
lin
ear
l
o
ad
o
n
th
e
po
wer
d
i
str
i
bu
tio
n n
e
twor
k, it
gives rise
to
operational
problem
of the el
ectrical
devi
ces
. Thi
s
r
e
sul
t
s
i
n
vari
o
u
s p
r
o
b
l
e
m
s
e.g. sw
el
l
,
sag, u
nde
r-
v
o
l
t
a
ge,
o
v
er
-v
ol
t
a
ge,
ha
rm
oni
cs et
c [3]
.
At
prese
n
t, the
r
e is also
m
u
ch usa
g
e of
rene
wable ene
r
gy
sy
st
em
as well
as advance
d
po
wer t
r
a
n
sm
i
ssi
o
n
t
echn
o
l
o
gi
es.
Usag
e
of
suc
h
m
e
t
hods
hi
ghl
y
i
n
creases
t
h
e
n
o
n
-
l
i
n
ea
ri
t
y
i
n
t
h
e
sy
st
em
as wel
l
as
gi
ve
ri
se t
o
vari
ous i
s
s
u
es
rel
a
t
e
d t
o
re
gu
l
a
t
i
on of
vol
t
a
ge. D
u
e t
o
th
is, th
e co
m
p
lex
ity asso
ciat
ed with the planning and
ope
rat
i
o
ns i
n
c
r
eases i
n
t
h
e
net
w
or
k o
f
el
ect
ri
ci
t
y
suppl
y
.
The sy
st
em
al
so wi
t
n
esse
s si
gni
fi
ca
nt
l
e
vel
o
f
harm
oni
cs as
wel
l
as po
wer
si
nus
oi
d
s
o
w
i
n
g t
o
u
n
e
xpect
e
d
al
t
e
rat
i
ons i
n
t
h
e cont
i
nge
n
c
i
e
s i
n
t
h
e net
w
o
r
k
,
disturba
nces of load, etc. All these
are also called as PQ disturba
nces
events and it works
at the cost
of
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 5
,
O
c
tob
e
r
20
16
:
243
7
–
24
46
2
438
per
f
o
r
m
a
nce degra
d
at
i
o
n. It
al
so resul
t
s
i
n
perm
anent
an
d i
rre
versi
b
l
e
dam
a
ge of an
el
ect
ri
cal
devi
ce [4]
.
Sect
i
on
1
di
scusses a
b
out
t
h
e i
n
t
o
r
o
duct
i
o
n. Sect
i
o
n
2
di
scusses a
b
out
t
h
e exi
s
t
i
n
g sy
st
em
whi
l
e
Sect
i
o
n
3
di
scuss
e
s a
b
o
u
t
t
h
e resea
r
c
h
g
a
p.
Fi
nal
l
y
, Se
ct
i
on
4 m
a
kes
sum
m
ary
of t
h
e pa
per
.
1.
1.
Back
ground
Th
e
term
Po
wer
Qu
ality
o
r
p
o
p
u
l
arly k
n
o
wn
as PQ is
frequ
en
tly u
s
e
d
in
p
o
w
er electro
n
i
c for
assessi
ng t
h
e
superi
o
r
i
t
y
of
t
h
e si
gnal
being gene
rat
e
d
by
t
h
e sy
st
em
. It
can be
t
e
chni
cal
l
y
defi
ned as a
si
gni
fi
cant
pro
cess t
o
cont
rol
vari
o
u
s el
ect
rom
a
gneti
c operat
i
ons i
n
si
de
a sy
st
em
of p
o
wer el
ect
roni
cs. A
syste
m
o
f
po
wer electro
n
i
cs are said
to
d
e
liv
er qu
ality in
p
o
w
er if th
ere i
s
an
assu
rity o
f
un
d
e
v
i
ated
cu
rren
t
,
vol
t
a
ge, as well
as frequency
si
gnal
s
i
n
t
h
e p
o
wer sy
st
e
m
s.
When a syste
m
experi
ences unforeseen fluctuation
of t
h
e
cur
r
ent
or e
v
en a
v
o
l
t
a
ge fr
om
t
h
e usual
feat
ures,
i
t
can l
ead t
o
seri
ous
conse
quences
t
o
t
h
e
po
wer
sy
st
em
. It
can even res
u
l
t
i
n
sy
st
em
shut
do
wn
or m
a
y
produce i
rreve
rsi
b
l
e
dam
a
ge. The pri
m
e reason for suc
h
fluctuation is d
u
e to uninterrupted a
lterations of the supply of power. P
o
we
r Quality can
also be defined as a
t
e
r
m
t
h
at
can focus o
n
t
h
e eff
ect
i
v
eness of t
h
e po
wer su
ppl
y
wi
t
h
hi
ghl
y
sy
nchro
n
i
zed vo
l
t
a
ge and st
ream
li
ned
cu
rren
t
.
Alth
ou
gh
th
e ter
m
PQ so
un
d
s
m
o
re related
to
p
o
w
er bu
t tech
n
i
call
y
sp
eak
in
g it is
mo
re ass
o
ciated
with voltage quality ra
ther than electric current
or
power. It is
because
power is a term
tha
t
is rela
ted to
ori
g
i
n
al
ener
gy
fl
ow
al
on
g wi
t
h
t
h
e am
ount
of c
u
rre
nt
t
h
at
i
s
req
u
i
r
ed t
o
m
a
i
n
t
a
i
n
t
h
e
st
abl
e
l
o
ad. Ta
bl
e 1
sh
o
w
s th
e categ
o
r
ies o
f
v
a
riou
s sig
n
i
fican
t ev
en
ts stu
d
i
ed
in
ex
istin
g
liter
a
tu
res. Th
e p
r
i
m
e ca
teg
o
r
ies
are A.
Short
du
rat
i
on fl
uct
u
at
i
on, B
.
Lon
g
durat
i
o
n fl
uct
u
at
i
on,
C. Transient, D. V
o
ltage Im
balance, E.
Waveform
di
st
ort
i
ons.
Tab
l
e
1
.
Catego
ry
o
f
Ev
en
ts i
n
Power
Qu
ality
Category
Voltage Measu
r
e
m
ent
T
i
m
e
Per
i
od
A
Swell
M
o
m
e
ntary
1.
1-
1.
4 pu
30 cycles
-
3
s
e
c
T
e
m
por
ary
1.
1-
1.
2 pu
3 sec-
1
m
i
n
I
n
stantan
e
ous
1.
1-
1.
8 pu
0.
5sec-
30
cy
cles
Sag
M
o
m
e
ntary
0.
1-
0.
9 pu
30 cycles
-
3
s
e
c
T
e
m
por
ary
0.
1-
0.
9
pu
3sec-
1
m
in
I
n
stantan
e
ous
0.
1-
0.
9 pu
0.
5cycles
– 30
cycles
I
n
ter
r
uption
T
e
m
por
ary
<0.
1
pu
3sec-
1
m
in
M
o
m
e
ntary
<0.
1
pu
0.
5cycle-
3
sec
B
Under voltage
>1
m
i
n
0.
8-
0.
9 pu
Sustained,
I
n
terr
up
tion
>1
m
i
n
0.
0
pu
Over
voltage
>1
m
i
n
1.
1-
1.
2
pu
C Oscillatory
High
Fr
equency
5
μ
sec 0-
4
pu
L
o
w-Frequency
0.
3-
50
m
s
ec
0-
4pu
M
e
diu
m
-Frequency
20
μ
sec 0-
8pu
Im
puls
i
ve
M
illisecond
>1
m
s
ec
-
M
i
cr
osecond
50-
1
m
sec
-
Nanosecond
<50nsec
-
D Voltage
i
m
balance
0.
5-
2%
Stead
y
State
E
Noise
-
Stead
y
Stat
e
Notching
-
Stead
y
Stat
e
Har
m
onics
-
Stead
y
Stat
e
Studies towards PQ is not new and it dates
back m
o
re tha
n
a decade yea
r
s b
ack; howe
v
er, there is an
i
n
creasi
ng i
n
t
e
rest
on t
h
i
s
t
opi
c owi
ng t
o
va
ri
ous si
gni
fi
cant
reasons. At
present
t
i
m
e
,
the
m
odern de
v
i
ces of
el
ect
ricit
y
use
m
i
crocont
rol
l
e
rs i
n
order t
o
m
e
et
perform
ance of appl
i
cat
i
on. B
u
t
it
al
so resul
t
s
i
n
vari
ous
critica
l
d
e
g
r
adatio
n
to
ward
s
q
u
a
lity o
f
p
o
wer. Ex
ten
s
ive u
s
ag
e
o
f
sop
h
i
sticated
(o
r h
y
b
r
id) cap
acito
rs o
r
com
p
lex
m
o
t
o
r dri
v
es f
r
eque
nt
l
y
encount
ers break d
o
w
n
s
wi
t
h
i
n
creasi
ng l
o
ad o
f
p
o
w
e
r. It
al
so resu
l
t
s
i
n
syste
m
malfu
n
c
tio
n
th
at serio
u
s
ly d
e
g
r
ad
es t
h
e p
o
wer
q
u
a
lity. Mo
reo
v
e
r, t
h
e p
r
esen
t era o
f
con
s
u
m
er
mark
et
uses gri
d
as t
h
e
pri
m
e basi
s of
po
wer di
st
ri
bu
t
i
on sy
st
em
t
h
at
has hi
gher i
n
t
e
rcon
nect
ed su
b-sy
st
em
s. Hence, a
s
m
aller am
ount of power system
degradation will cost th
e entire grid system
to b
ear
the cost of inferior
perform
a
nce
with respect to power e
fficiency and stabi
lity. W
ith the
usage
of the
m
odern day electrica
l
syste
m
, various proble
m
s
surfaces e.g. flickering, fluctu
ating voltages, distortions in
wa
veform
s
etc.
Hence,
th
e b
e
st way to
so
lv
e su
ch
issu
es o
f
po
wer qu
alit
y is
to
p
e
rfo
r
m
id
en
tifica
t
i
o
n
of th
e p
o
wer sig
n
a
ls fo
llo
wed
b
y
cl
assi
fyi
ng goo
d t
o
bad si
gnal
s
. Hence, i
t
i
s
im
port
a
nt
t
o
i
n
vest
i
g
at
e effecti
v
el
y about
t
h
e PQ di
st
urbanc
es for
si
gni
fi
cant
und
erst
andi
ng t
h
e
pro
b
l
e
m
s
and t
h
ereby
rect
i
f
y
i
ng i
t
.
There has been an
ext
e
nsi
v
e am
ount
of
in
v
e
stig
atio
n
to
ward
s th
is
p
r
o
b
l
em
. Th
e sig
n
i
fican
t p
r
o
cesses ad
op
ted
in
en
h
a
n
c
ing
th
e
q
u
a
lity o
f
th
e
p
o
w
er
are segm
enta
tion, feature e
x
traction,
artific
ial in
te
ll
ig
en
ce, etc. In th
is
p
r
o
cess, seg
m
en
tatio
n
an
d
featu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
In
si
g
h
t
on
Effectiven
ess o
f
Freq
u
e
n
tly Exercised
PQ Cla
s
sifica
tio
n
Techn
i
q
u
e
s (B.
Devi
Vig
h
n
e
sh
wa
ri)
2
439
ex
tractio
n
is q
u
ite co
mmo
n
i
n
all.
Basic
a
l
l
y
,
seg
m
en
tat
i
on, feature extraction are the
ess
e
ntial
steps to design
any classifier s
y
ste
m
for cl
assi
fy
i
ng PQ eve
n
t
s
. The st
andar
d
PQ cl
assi
fi
er desi
gn i
s
sh
ow
n i
n
Fi
g
u
re 1
[
5
]
.
Figure
1.
Standard PQ Classifi
er
Design
The i
nput
t
o
thi
s
cl
assi
fi
er desi
gn i
s
a waveform
wit
h
di
st
urbances, w
h
i
c
h i
s
passed on t
o
bl
ock of
segm
ent
a
ti
on t
h
at
essent
i
a
l
l
y
cat
egori
ze t
h
e
dat
a
i
n
t
h
e form
of st
at
i
c
and no
n-st
at
i
on sect
i
on. A si
gni
fi
cant
am
ount
of feat
ure i
s
ext
r
act
ed fr
om
t
h
e wavefo
rm
s (wi
t
h
event
)
. A l
a
rge am
ount
of
i
n
form
ati
on can be
ex
tracted
fro
m th
e sta
tic s
i
g
n
a
l j
u
st to
carry out
com
p
arative anal
y
s
i
s
of
di
fferent
fo
rm
s of PQ di
st
ur
b
a
nces.
Th
e po
in
t-to
-po
i
n
t
is th
e freq
u
e
n
tly ad
op
ted
tech
n
i
qu
e
to com
p
are the statistical
val
u
es of
t
h
e di
st
urbe
d
waveform
s and pure
signal. Methods
e
.
g. Regression fram
ework,
Kalm
an Filter, Short-Term
ed F
ourier
Transf
orm
s
are freque
nt
l
y
adopt
ed. Sim
i
l
a
rly
,
feat
ure ext
r
act
i
on i
s
respon
si
bl
e for i
d
ent
i
fy
i
ng and sepa
rat
i
ng
si
gni
fi
cant
features f
r
om
t
h
e
si
gnal
i
n
or
der
t
o
un
derst
a
nd
t
h
e pot
ent
i
a
l
si
gnal
.
The j
o
b o
f
cl
assi
fi
er
i
s
t
o
categ
o
r
ize o
r
cl
assify v
a
rio
u
s
t
y
p
e
s o
f
p
o
wer
q
u
a
lity d
i
stu
r
b
a
n
ces. At presen
t, th
ere are two
fo
rm
s o
f
classifier
vi
z. cl
assi
fi
er based on
st
a
tistical approach and classifi
er ba
sed o
n
det
e
rm
ini
s
t
i
c
approach
. St
at
i
s
t
i
c
al
approach
[6
] is u
s
ed
i
n
a situ
atio
n
o
f
availab
i
li
ty fo
r massiv
e
d
a
taset
wh
ile d
e
termin
isti
c ap
p
r
o
a
ch
[7
] is u
s
ed
fo
r limi
t
ed
size of dataset.
A good e
x
am
ple of cl
assifier based on statistical approach
will be Support
Vector Machine and
Neu
r
al Netwo
r
k
[8
], wh
ile ex
a
m
p
l
e fo
r classifier fo
r d
e
termin
isti
c ap
p
r
o
ach
will b
e
fu
zzy lo
g
i
c and
ru
le-b
ased
expert
sy
st
em
[
9
]
.
The fi
nal
st
age of P
Q
cl
assi
fi
cati
on t
echni
que i
s
deci
si
on
m
a
ki
ng. The n
e
xt
sect
i
on di
scusses
about the essential
techniques used
for e
n
hancing the power quality classi
fication processes. The process of
featu
r
e
ex
tractio
n
p
l
ays a
c
r
itica
l
ro
le
in
PQ class
i
fica
tio
n
p
r
o
cess as tech
n
i
cal
mean
in
g
o
f
featu
r
e is
di
st
urbances i
n
PQ event
cl
assi
fi
cati
on. Therefore, ex
t
r
act
ed feat
ure gi
ves t
h
e di
rect
info
rm
at
i
on about
t
h
e
i
d
ent
i
f
i
cati
on of t
h
e di
st
urba
nces and henc
e i
s
used for cl
assi
fyi
ng vari
ous event
s
of
PQ. Fi
gure 2
gi
ves
pi
ct
ori
a
l
represent
a
t
i
on of st
an
dard cl
assi
fi
cati
on o
f
exi
s
t
i
ng
process
of
feat
ure ext
r
act
i
on i
n
PQ.
Fig
u
r
e
2
.
Standar
d
Classif
i
catio
n of
Feature
Extraction i
n
PQ Classification
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 5
,
O
c
tob
e
r
20
16
:
243
7
–
24
46
2
440
The exi
s
t
i
ng m
e
t
hods
of cl
assi
fi
cat
i
on of feat
ure are:
i.
Fouri
e
r T
r
ansf
orm
It is h
i
g
h
l
y su
it
ab
le fo
r static s
i
g
n
a
l an
d
wh
en th
ere is a n
eed
to
cap
tu
re sp
ectru
m
at p
a
rti
c
u
l
ar lev
e
l o
f
freq
u
ency
. One
of the specific form
s of Fouri
e
r Transfo
r
m
is
also
ca
lled
as
Sh
o
r
t-Term Fo
u
r
ier Tran
sfo
r
m
th
at
sp
lits th
e s
i
g
n
a
l in
to
d
i
min
u
tiv
e stat
ic frag
m
en
ts. Hen
ce, Sh
o
r
t-Term Fo
u
r
ier Tran
sform
ev
a
l
u
a
tes th
e
si
nusoi
dal
fre
q
u
ency
as wel
l
as phase co
nt
ent
s
of l
o
cal
si
g
n
al
fragm
ent
s
. Usi
ng m
ovi
ng
wi
nd
ow,
Sh
ort
-
Term
Fouri
e
r
Trans
f
orm
capt
u
res i
n
fo
rm
at
i
on about
si
gnal
f
r
a
m
es. It
i
s
sui
t
a
bl
e t
echni
que
for
dy
nam
i
c and
no
n-
st
ati
c
si
gnal
ov
er const
a
nt
si
ze of
wi
nd
ow.
ii.
W
a
v
e
let Tran
sfo
r
m
Thi
s
m
e
t
hod
perf
orm
s
dil
a
ti
on of t
h
e si
gnal
prot
ot
y
p
e
funct
i
on an
d
t
h
ereby
perf
orm
s
si
gnal
decom
posi
t
i
ons on vari
o
u
s l
e
vel
s
. It
uses grou
p t
h
eory
rep
r
esent
a
t
i
on as
wel
l
as square-i
nt
egral
funct
i
on t
o
p
r
ov
id
e
d
e
fin
itiv
e in
fo
rm
atio
n
of freq
u
e
n
c
y an
d
tim
e fo
r a p
a
rticu
l
ar sig
n
al. Discrete Wav
e
let Tran
sfo
r
m an
d
C
ont
i
nuo
us
Wavel
e
t
t
r
ansfor
m
are t
h
e t
w
o f
r
eque
nt
l
y
used t
echni
ques i
n
P
Q
cl
assi
fi
cati
on p
r
ocess.
iii.
S-Trans
f
orm
Thi
s
form
of t
echni
que i
s
desi
gned
by
i
n
t
e
gr
at
i
ng t
h
e feat
ures of wa
vel
e
t
transf
orm
s
and Short
-
Te
rm
Fouri
e
r
Trans
f
orm
.
It
uses t
i
m
e-seri
es analysis for assessing the
real an
d im
aginary contents from
the spectra.
An i
n
t
e
rest
i
ng fact
about
t
h
e S-t
r
ansf
orm
i
s
t
h
e usage of
a t
y
pi
cal
patt
ern di
rect
l
y
represent
s
a speci
fi
c event
of
di
st
urbance i
n
PQ. Thi
s
t
echn
i
que i
s
al
so fo
un
d t
o
be i
n
t
e
g
r
at
ed wi
t
h
vari
ous f
o
rm
s of o
t
her t
echni
ques
(e.g
.
artifici
a
l in
te
ll
i
g
en
ce) to
en
h
a
n
ce its fu
n
c
tio
nali
ty.
iv
.
Hilbert Huang
Transform
Thi
s
process p
e
rform
s decom
posit
i
on of the si
gnal
t
o
g
e
nerat
e
t
h
e knowl
e
dge
-based
i
n
form
at
i
on
about the am
plitude and
frequency of a signa
l
. The tec
hnique
m
a
kes use of
em
piri
cal
m
ode decom
positio
n and
arranges
t
h
e f
r
e
quency
i
n
des
cendi
ng
o
r
der.
The
decom
pos
ed
sig
n
a
l is su
bj
ected
to
Hilb
ert tran
sfo
r
m
to
g
e
t
m
o
re p
r
ecise in
fo
rm
at
io
n
ab
out th
e ev
en
t clas
sificat
io
n
in
power qu
ality.
Table 2.
E
f
fectiveness
of PQ Classification
No PQ
event
E
x
isting T
e
chniques
F
F
T
D
F
T
S
-
T
H
H
T
1 Sag
95
98.
67
100
100100
2 Swell
98
99.
33
100
95
3 Har
m
onic
100
99.
33
100
100
4 Flicker
89
98.
67
100
100
5 Notch
-
97.
33
83
95
6 Spike
-
-
77
98
7 T
r
ansient
100
98.
67
100
98
Tabl
e 2
hi
ghl
i
ght
s t
h
e ef
fect
i
v
eness
of t
h
e
e
x
i
s
t
i
ng PQ
det
ect
i
on t
echni
ques i
n
t
e
rm
s of percent
a
ge
,
where it can be seen that
S-t
r
ansfo
r
m
s
was foun
d t
o
posse’s
bet
t
e
r perform
ance t
h
an Hi
l
b
ert
Huang Tra
n
sf
orm
(HHT
). The si
gni
fi
cant
advant
age of usage
of Short
-
Te
rm
Fouri
e
r Trans
f
orm
i
s
t
h
at
it
can be used fo
r st
ati
c
sig
n
a
ls an
d
it is
q
u
ite easier to
i
m
p
l
e
m
en
t. Th
e b
e
n
e
fit o
f
u
s
i
n
g
HHT is its c
a
p
a
b
ilit
y o
f
cap
tu
rin
g
featu
r
es fro
m
wavef
o
rm
s of
di
st
ort
e
d t
y
pes. It
can al
so
p
r
od
uce q
u
adra
t
u
re signal that can
direc
tly be used for eval
uating
phase and am
pl
i
t
ude of a si
gnal
.
The ben
e
fi
t
of usi
ng S
-
transform
is its inherent charecteristics of sim
p
le
conve
rsi
on
fro
m
tim
e t
o
frequency
d
o
m
a
in an
d t
h
en t
o
Fou
r
i
e
r Freq
uency
Transf
o
r
m
(FFT). Si
m
i
l
a
rly
,
W
a
vel
e
t
Transf
orm
can pro
v
i
d
e preci
se rep
r
e
s
ent
a
t
i
on of
fre
quency
a
nd t
i
m
e for
f
u
r
n
i
s
hi
n
g
bet
t
e
r res
o
l
u
t
i
on o
f
PQ cu
rves.
Li
k
e
wi
se, t
h
e req
u
i
rem
e
nt
of hi
g
h
er
po
wer
qual
i
t
y
can al
so be achi
e
ved usi
n
g
Gab
o
r T
r
ansf
o
r
m
.
I
t
h
a
s th
e p
o
t
en
tial cap
ab
il
ity
o
f
p
r
od
u
c
ing
ou
tco
m
e with
ma
x
i
m
a
l s
i
g
n
a
l-to
-n
o
i
se ratio
with
su
p
e
rio
r
resolu
tio
n
of the signals.
Hence, i
t
can be seen t
h
at
t
h
ere i
s
an exi
s
t
e
nce of vari
ous s
t
andard m
e
t
hods respo
n
si
bl
e for carry
i
n
g
out
feat
ure e
x
t
r
act
i
ons of
a si
gnal
.
The
next
sect
i
on di
scus
ses about
ot
her
freq
u
ent
l
y
exerci
sed t
echni
qu
es t
o
enhance t
h
e P
Q
di
st
ur
bance
det
ect
i
on and c
l
assi
fi
cat
i
on pr
ocess.
1.
2.
The Problem
The m
odern c
ons
um
er
m
a
rket
of po
we
r di
st
ri
but
i
o
n ha
s wi
t
n
esse
d a pa
radi
gm
shi
f
t
of con
s
um
er
to
ward
s th
e
n
e
w serv
ice p
r
o
v
id
ers in
case the PQ is not found satisfact
ory
b
y
th
eir ex
isti
n
g
serv
ice prov
id
ers.
It becom
e
s therefore
highly
m
a
ndatory
for th
e serv
ice p
r
ov
id
er to
furn
ish
h
i
gh
er deg
r
ee
o
f
PQ in
th
eir
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
In
si
g
h
t
on
Effectiven
ess o
f
Freq
u
e
n
tly Exercised
PQ Cla
s
sifica
tio
n
Techn
i
q
u
e
s (B.
Devi
Vig
h
n
e
sh
wa
ri)
2
441
services
. T
h
e i
ssues
for PQ cl
assification is
basically two
typ
e
s
v
i
z. i) ev
en
ts an
d ii) stead
y
state fl
u
c
tuatio
n
s
.
Ev
en
ts ar
e estimated
b
y
abnor
m
a
l
ity in
th
e cu
rr
en
t and
v
o
ltag
e
. Stead
y
-
state f
l
u
c
tu
ation is esti
m
a
ted
by th
e
measure
by which t
h
e c
u
rrent as well as
voltage can di
ffe
r
fr
om
t
h
e st
an
dar
d
val
u
e
al
o
n
g
wi
t
h
di
st
o
r
t
i
ons
. It
al
so
c
onsi
d
ers
am
ount
o
f
di
st
ur
ba
nce bet
w
e
e
n
t
h
e
p
h
ases
(
h
arm
oni
c, v
o
l
t
a
ge vari
at
i
o
n, di
st
ort
i
o
n
)
.
1.
3.
Prop
osed
S
t
u
d
y
Th
erefo
r
e, it is v
e
ry m
u
ch
i
m
p
o
rtan
t th
at sign
ifican
t ev
en
ts o
f
PQ classifi
catio
n
shou
ld
b
e
surv
eilled
effectively.
Howeve
r, t
h
ere
has bee
n
a
volume of
resear
c
h
work
being already carried
ou
t with
i
n
th
e research
com
m
uni
t
y
. It
i
s
fo
un
d t
h
at
e
x
i
s
t
i
ng
p
o
we
r
sy
st
em
s are hi
ghl
y
c
o
m
p
l
e
x and
m
a
ssi
ve, whi
c
h m
a
kes t
h
e
dat
a
analysis
m
o
st
com
p
lex and unreliable. T
h
erefore, eve
n
wit
h
pre
s
ence
of
massive research arc
h
ives, st
anda
rd
work toward
s
PQ ev
en
t classificatio
n
still misses ou
t
fro
m
th
e literatu
res.
Th
ere is a
n
e
ed of evo
l
v
i
ng
up with
a co
st
eff
ectiv
e so
lu
tion tow
a
rd
s
r
e
liab
l
e and
ef
f
ectiv
e an
alysis
o
f
ev
en
t
d
e
tection
i
n
PQ
study and
i
nvest
i
g
at
es t
h
e un
derl
y
i
n
g
ope
rat
i
o
ns res
p
o
n
si
bl
e
fo
r sy
stem
perform
ance en
ha
nce
m
ent. The process of
feat
ure
e
x
t
r
act
i
o
n
an
d
cl
assi
fi
cat
i
on a
r
e t
h
e
m
o
st
cri
t
i
cal
phases
w
h
i
c
h
re
qui
res m
o
re a
m
ount
s o
f
at
t
e
nt
i
o
n
s
t
o
u
n
d
e
rstand
m
o
re ab
ou
t
d
i
sturb
a
n
c
es in PQ. Basically
, PQ is assessed un
d
e
r
v
a
riou
s co
nd
itio
ns,
wh
ich
h
a
s
b
een
less i
n
v
e
stig
ated
. Th
erefore, th
is
p
a
p
e
r is an
attem
p
t
to
u
n
d
e
rstan
d
th
e m
o
st freq
u
e
n
tly p
r
acticed
techniques
for
PQ classi
fication in
powe
r el
ectronics
2.
REVIEW
IN
G
E
X
ISTI
NG
TECHN
IQ
UES
Thi
s
sect
i
on di
scusses about
t
h
e vari
ous e
x
i
s
t
i
ng
t
echni
ques
t
h
at
was i
n
t
r
oduced f
o
r t
h
e
pur
pose
of
carry
i
ng o
u
t
cl
assi
fi
cat
i
on of
PQ di
st
ur
bance
s
i
n
po
wer el
ect
roni
cs.
2.
1.
Artificial Neu
r
al
Ne
tw
or
k
The usage
of
art
i
f
i
c
ial
neural
net
w
ork
i
s
p
l
ay
ed a huge
cont
ri
but
o
r
y
r
o
l
e
i
n
vari
o
u
s
sy
st
em
of
com
put
i
ng wor
l
d e.g. appr
oxi
m
a
ti
on of fu
nc
t
i
on, pat
t
e
rn recogni
t
i
on, cl
us
t
e
ri
ng, opt
im
i
z
at
i
on et
c. Out
of al
l
t
h
ese pot
ent
i
a
ls, appro
x
i
m
at
ion of f
unct
i
o
ns as wel
l
as
class
i
fica
tion charecteristic
s
has been used in
i
nvest
i
g
at
i
ng PQ cl
assi
fi
cati
on p
r
ocess.
One of t
h
e st
andar
d
wo
rks
of
PQ cl
assi
fi
cat
i
on was f
o
u
nd t
o
be carri
ed o
u
t
by
M
onedero
et
al
. [10]
.
The app
r
oac
h
consi
d
ers si
g
n
a
l
as an i
nput
whi
c
h i
s
su
bje
c
t
e
d t
o
pre-
pr
o
cessi
ng fol
l
o
w
e
d by
ne
ural
n
e
t
w
ork
classifi
er. Th
e
tech
n
i
q
u
e
(Figu
r
e 3) h
a
s th
e
cap
ab
ilit
y to
cl
assify elec
trica
l
d
i
stu
r
b
a
n
ce in
th
e fo
rm
o
f
vo
ltag
e
,
frequency,
and
harm
onics. T
h
e accuracy in
detection of
disturbance was found to
be 98.73% over events of
sag, swel
l
,
un
d
e
r-v
ol
t
a
ge, and
over
-
v
o
l
t
a
ge.
Fi
gu
re
3
A
p
p
r
oache
s
of M
o
n
e
der
o
et
al
.
[
1
0
]
Dast
fan an
d Z
a
deh [
11]
hav
e
used S
-
t
r
ans
f
orm
as
wel
l
as neural
net
w
ork
.
The
st
udy
has nea
r
l
y
sim
i
l
a
r approach wi
t
h
di
ffere
nce of ap
pl
y
i
ng S-t
r
ansf
orm
aft
e
r prepr
o
ces
si
ng t
h
e si
gnal
.
The t
echni
qu
e has
al
so used back
pro
p
agat
i
on
neural
net
w
or
k a
l
gori
t
h
m
consi
d
eri
ng t
h
e cas
e st
udy
of 3 t
y
pes of event
s
e.g. i
)
h
a
rm
o
n
i
c an
d
sag
,
ii) sag and
flick
e
r, an
d iii) flick
e
r and
h
a
rm
o
n
i
c. Th
e ex
p
e
ri
m
e
n
t
at
io
n
was carried ou
t o
v
e
r
IEEE 34
bus
standard. The overall accuracy of detec
tio
n was found to be arou
nd 97-98% for 1000-2000
epochs res
p
ect
ivel
y
.
Sim
i
l
a
r line of i
nvest
i
g
at
i
on was al
so
carri
ed out
by
M
i
shra et
al
. [
12]
, wh
o have
used S-
t
r
ansform
and neural
net
w
o
r
k
for pe
rfo
rm
i
ng cl
assi
fi
cat
i
on of PQ co
nsi
d
e
r
i
ng 1
1
cases of event
s
. The a
u
t
hors
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 5
,
O
c
tob
e
r
20
16
:
243
7
–
24
46
2
442
have use
d
feed-forward algori
t
h
m
. Th
e outcome of the study was assessed
using accuracy param
e
ters for 3 and
4 feat
ures t
h
at
were f
o
u
nd t
o
be aro
u
n
d
9
5
-
9
7%.
Vi
skado
u
r
o
s et
al
. [13]
ha
ve
used
neural
ne
t
w
ork
al
ong
w
i
t
h
wavel
e
t
-
based app
r
oach
i
n
o
r
der t
o
id
en
tify th
e P
Q
ev
en
ts. Th
e ap
p
r
o
ach
is n
early s
i
milar
to
that presented 3 years back by Mishra et a
l
.
[12]
,
Sim
ilar version of study along with us
age of em
pirical
m
ode deco
m
positio
n
was seen in the study of Manjula
an
d
Sarm
a [1
4
]
. Th
e au
tho
r
s
hav
e
u
s
ed
p
r
o
b
a
b
ilist
i
c n
e
u
r
al n
e
two
r
k
.
Memo
n
et al. [15
]
hav
e
d
e
v
e
lop
e
d
a feed
-
forward training
m
echanis
m
for cl
assi
fi
cat
i
o
n pur
pose. T
h
e prim
e
i
n
t
e
nt
i
o
n of t
h
e st
udy
was t
o
m
i
nim
i
ze t
h
e
t
r
ai
ni
ng eff
o
rt
s and c
o
r
r
ect
l
y
i
d
ent
i
f
y
t
h
e source o
f
powe
r
di
st
urbances.
Al
t
hou
gh,
t
h
e
wor
k
i
s
m
o
re i
n
cl
i
n
ed
t
o
wards
usi
ng
wavel
e
t
t
r
ansfo
r
m
s
but
neural
net
w
or
k
has a
ssi
st
ed t
o
enhance t
h
e com
put
ati
onal
perfo
r
m
ance
of t
h
e st
udy
. The essent
i
a
l study
cont
ri
but
i
on was t
o
us
e radi
al
basi
s funct
i
on and m
u
lt
i
-
l
a
y
e
red percept
r
on.
The outcome of the study
was evaluate
d with respect to static and non-
static signals to find accuracy of
97.
45% i
n
det
ect
i
on rat
e
.
Most recently,
the work carri
ed out
by Rodriguez et al.
[16] have used feedfo
rward ne
ural network
for i
d
e
n
t
i
f
i
cat
ion a
nd cl
assi
fi
cat
i
on of m
u
l
tipl
e
form
s of P
Q
di
st
ur
bances
. The st
u
d
y
ha
s al
so im
pl
ement
e
d
adapt
i
v
e l
i
n
ear net
w
o
r
k
fo
r es
t
i
m
a
ti
ng harm
oni
cs and i
t
s
di
s
t
ort
i
ons. The
a
ssessm
e
nt
of t
h
e st
udy
was c
a
rri
ed
out on real-ti
m
e hardware with accur
acy of 90%. The study has also perform
ed a co
m
p
arative analys
i
s
with 6
ex
istin
g
tech
n
i
q
u
e
s (Fu
zzy C mean
s, Kal
m
a
n
filter, S-tran
sfo
r
m
with
n
e
u
r
al n
e
two
r
k
,
S-tran
sfo
r
m
with
b
i
n
a
ry
feat
ure
m
a
t
r
i
x
, S-t
r
ansform
wi
t
h
m
odul
ar neural
net
w
or
k
,
and fuzzy
art
m
a
p consi
d
er
17 cases of an event
.
Jasper et
al
. [1
7]
have
used
b
ack pr
opa
gat
i
on t
echni
q
u
e fo
r enha
nci
ng t
h
e po
wer
qual
i
t
y
of a sh
unt
i
n
vert
er
.
The aut
h
ors
h
a
ve at
t
e
m
p
ted t
o
o
v
ercom
e
t
h
e i
ssues of
conve
nt
i
onal
p
u
l
s
e wi
de m
odul
at
i
on usi
n
g
neu
r
al
net
w
or
k.
Su
n
d
aram
[18]
ha
ve use
d
t
h
e si
m
i
l
a
r concept
di
scussed
by
R
odri
g
uez et
al
. [16]
.
Al
t
hou
gh t
h
e
accuracy of 97% is achieved but the pape
r significant la
cks justificatio
n and evide
n
c
e
of its outcomes as
com
p
ared t
o
ori
g
i
n
al
versi
on of R
odri
g
uez et
al
. [16]
, whi
c
h has
m
a
t
h
em
at
i
c
al
and em
p
i
ri
cal evi
d
ence behi
nd
i
t
s
out
co
m
e
. Sim
i
l
a
r cat
egory
of st
u
d
y
wa
s also carried
out by W
i
n et al. [19].
2.
2.
Supp
ort Vec
t
or
Mac
h
ine
Support
vector
m
achine is ba
sically
a supervi
s
ed l
earni
ng
t
echni
que t
h
at applies statistical learning
mechanis
m
.
I
t
is a precise
techni
que f
o
r c
a
rry
i
ng fo
r b
o
t
h l
i
n
ear and no
n-l
i
n
ear cl
assi
fi
cati
on as wel
l
as
regression. In the line of rese
arch
w
o
r
k
t
o
w
a
rd PQ
cl
assi
ficat
i
on, t
h
e st
udy
con
duct
e
d
by
Shaky
a
an
d
Si
ng
h
[20]
has bee
n
fo
un
d t
o
i
n
corp
orat
e su
p
port
vect
or m
achi
n
e fo
r i
d
ent
i
f
y
i
ng t
h
e
pro
b
l
e
m
a
ti
c area of
classification. The study has
also used fuzz
y class
i
fier. Figure
4 show
ca
ses t
h
e approac
h
of S
h
aky
a
an
d Si
ng
h
[2
0
]
, wh
ich
is also
fou
n
d
to
b
e
co
mmo
n
ap
p
r
o
ach
in
u
s
in
g
artificia
l in
te
ll
ig
en
ce.
Fi
gu
re 4.
A
p
pr
oache
s
of Sha
k
y
a
and
Si
n
g
h
[
20]
Li
n et
al
. [21]
have al
so appl
i
e
d usi
ng i
n
t
e
gr
al
m
odell
i
ng of
wavel
e
t
s
wi
t
h
supp
ort
vect
or
m
achi
n
e i
n
orde
r t
o
carry
out
cl
assi
fi
cation. T
h
e
out
co
m
e
shows t
h
e
present
e
d sy
st
em
t
o
consum
e
l
e
ss t
h
an 10
seconds
o
f
t
r
ai
ni
ng t
i
m
e
.
Kocam
an et
al
. [22]
have use
d
joi
n
t
im
pl
ement
a
t
i
on of wa
vel
e
t
based supp
ort
vect
or m
achi
n
e.
The aut
h
or
ha
ve use
d
di
scret
e
wavel
e
t
t
r
ansfo
r
m
and p
e
rform
ed t
h
e
cl
assi
ficati
on
usi
ng s
u
p
p
o
r
t
vect
or
machine. Similar line of resea
r
ch has
al
so
be
en carri
ed o
u
t
by
M
i
l
c
hevski
et
al
. [23]
. Usi
ng t
h
e schem
e
of t
r
ee-
based su
pp
ort
vect
or m
a
chi
n
e (Fi
gure 5
)
, t
h
e aut
hor
have
cl
assi
f
i
ed vari
ous PQ
di
st
urb
a
nce event
s
(n
orm
a
l
,
swel
l
,
sag, out
a
g
e, harm
oni
c, et
c)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
In
si
g
h
t
on
Effectiven
ess o
f
Freq
u
e
n
tly Exercised
PQ Cla
s
sifica
tio
n
Techn
i
q
u
e
s (B.
Devi
Vig
h
n
e
sh
wa
ri)
2
443
Fi
gu
re
5.
A
p
pr
oache
s
of M
i
l
c
hev
s
ki
et
al
.
[
2
3]
Usage
of
deci
si
on t
r
ee an
d s
u
pp
ort
v
ector machine was als
o
seen in th
e st
udy of Ray et
al. [24]. T
h
e
aut
hors ha
ve used a hardware
-based app
r
oac
h
consi
d
er
i
n
g wi
nd ener
gy
syst
em
. The out
com
e
of t
h
e st
udy
was
found to posse
s 99%
of the
a
ccuracy in detection of PQ disturban
ce eve
n
ts. The recent
work carried
out by
Naderi
an an
d
Sal
e
m
n
i
a
[25]
, where t
h
e
aut
hor
have
j
o
i
n
t
l
y
used su
pp
ort
vect
or
m
achi
n
e wi
t
h
Gabo
r
t
r
ansform
s
. The out
com
e
of t
h
e st
udy
was a
ssessed usi
ng
m
ean absol
u
t
e
error
.
2.
3.
Wavelet-base
d Techniqu
es
It
i
s
foun
d t
h
at
70%
of t
h
e t
echni
que
wi
del
y
used fo
r anal
y
s
i
s
f PQ cl
assi
fi
cati
on pr
ocess i
s
carri
e
d
out
usi
ng wa
v
e
l
e
t
-
based t
r
ansform
s
. He et
al
. [26]
ha
ve used sel
f
orga
ni
zi
ng l
earni
ng
array
wi
t
h
w
a
vel
e
t
t
r
ansform
for
perf
orm
i
ng PQ cl
assi
ficati
on.
Howe
ver,
wa
vel
e
t
s
are
m
a
inl
y
preferre
d f
o
r pe
rfo
rm
i
ng
feat
ure
extraction. The
author
has pe
rform
e
d sim
u
la
tion on
7
e
v
ents
of PQ disturbances
with accom
p
lishing accuracy
of
94%.
Su
ja
and Jer
o
m
e
[27]
have
appl
i
e
d wavel
e
t
s
al
on
g wi
t
h
ne
ur
al
net
w
ork
fo
r
cl
assi
fi
cat
i
on of P
Q
di
st
urbances.
Usage of
wave
l
e
t
t
r
ansform
was al
so seen i
n
si
gni
fi
cant
wor
k
of Pani
g
r
ahi
et
al
. [28]
consi
d
eri
ng
11 case st
udy
of PQ di
st
ur
bances. The aut
hors ha
ve u
s
ed S-t
r
ansfo
r
m
and neural
net
w
or
k. Shar
eef and
M
oham
e
d [29]
prese
n
t
e
d a t
e
chni
que
t
o
a
n
al
y
s
i
s
PQ di
st
urb
a
nce usi
n
g i
m
age p
r
ocessi
ng
t
echni
ques.
Nat
h
an
d
M
i
shra [30]
ha
ve used
wavel
e
t
s
t
o
st
udy
t
h
e excl
usi
v
e
cases of
vol
t
a
ge s
a
g. Eri
s
t
i
and
Dem
i
r [31]
ha
ve used
l
ogi
st
i
c
t
r
ee alon
g wi
t
h
wave
l
e
t
s
for carry
i
n
g o
u
r e
x
t
r
act
i
on o
f
feat
ure.
R
o
y
and
Nat
h
[3
2]
have al
s
o
use
d
wavel
e
t
s
for cl
assi
fi
cat
i
on of
PQ di
st
urba
nces al
ong wi
t
h
n
e
ural
net
w
or
k.
The out
com
e
was fo
un
d t
o
achi
e
ve
arou
nd
95%
i
n
det
ect
i
on of P
Q
di
st
ur
bances
consi
d
eri
n
g 1
1
cases of eve
n
t
s
.
2.
4.
Swa
rm Intellig
ence
Swarm
in
tellig
en
ce h
a
s also
see
m
ed
to
p
l
ay
a v
ital ro
le in
o
p
tim
izat
io
n
prin
cip
l
e.
Using co
n
c
ep
ts of
an
t co
l
o
n
y
op
timizatio
n
an
d particle swarm
o
p
tim
izat
io
n
,
so
m
e
o
f
th
e
recen
t
research
wo
rk
h
a
s attem
p
ted
to
per
f
o
r
m
PQ cl
assi
fi
cat
i
on.
N
a
y
a
k an
d
Dash
[3
3]
ha
ve
use
d
pa
rt
i
c
l
e
swar
m
opt
i
m
i
zat
i
o
n f
o
r e
n
ha
nci
n
g t
h
e
ope
rat
i
o
n o
f
t
h
e cl
ust
e
r ce
nt
er
s al
on
g
wi
t
h
f
u
zzy
app
r
oac
h
.
The st
u
d
y
has
al
so m
odi
fi
ed
di
scret
e
S
-
t
r
a
n
sfo
r
m
fo
r feat
ure e
x
t
r
act
i
on al
o
ng
wi
t
h
k-m
eans clustering. Bis
w
al et al. [34]
ha
ve ad
o
p
t
e
d
t
h
e use
o
f
a
n
t
col
o
ny
optim
ization for clas
sifying the PQ. T
h
e st
udy
has als
o
used fuzzy
c
-
means
a
p
pro
ach
fo
r
e
vol
vi
n
g
u
p
wi
t
h
decision tree.
The enha
ncemen
t
of t
h
e pat
t
erns i
s
do
ne u
s
i
ng a
n
t
col
o
ny
opt
i
m
i
z
at
i
on.
The st
u
d
y
carr
i
ed o
u
t
by
Pari
zi
et
al
.
[3
5]
ha
s al
so
use
d
pa
rt
i
c
l
e
swarm
op
tim
iza
tio
n
along
with th
e S-tran
sfo
r
m
fo
r classifyin
g PQ
disturba
nces. T
h
e outcom
e is recorde
d
with accuracy of 92-97%. Similar
lin
e of st
udy
was also carried
out
by
M
a
jhi
et
al
. [
36]
.
St
u
d
y
carri
ed
o
u
t
by
Kum
a
rasaba
pa
t
h
y
and M
a
n
o
h
ara
n
[
3
7]
ha
ve u
s
ed a
n
t
c
o
l
o
ny
opt
i
m
i
zati
on al
on
g
wi
t
h
f
u
zzy
l
ogi
c
f
o
r
ad
dr
essi
ng
t
h
e
de
gr
aded
p
r
obl
em
s o
f
po
wer
q
u
al
i
t
y
. The st
udy
hav
e
desi
g
n
e
d
a
ne
w f
u
zzy
c
ont
r
o
l
l
e
r an
d
pe
rf
orm
e
d a si
m
u
lat
i
on st
u
d
y
on
M
a
t
l
a
b. Usa
g
e of
f
u
zzy
i
n
f
e
rence
sy
st
em
was d
o
m
i
n
ant
i
n
t
h
e s
t
udy
of
Nay
a
k
[3
3]
t
o
o.
2.
5.
Other
Cl
assifi
cati
on Tech
niques
The ot
her fre
q
u
ent
l
y
used cl
assi
fi
cati
on t
echni
que uses Ex
p
e
rt
’s sy
st
em
.
Ti
anrui
and Se
n [3
8]
have
p
r
esen
ted
a tec
h
n
i
qu
e to
au
to
mat
e
th
e p
o
w
er q
u
a
lity
co
n
t
ro
llin
g
m
ech
an
is
m. Ji
ap
ei e
t
a
l
.
[3
9
]
h
a
v
e
presen
ted
a
n
e
w inv
e
rter syste
m
with
b
e
tter co
n
t
ro
llin
g
cap
ab
ilit
y to
en
su
re po
wer qu
alit
y. Gu
o et
al. [4
0
]
h
a
v
e
dev
e
lo
p
e
d
a tech
n
i
q
u
e
to
ev
alu
a
te p
o
w
er q
u
a
lity u
s
in
g
statis
tica
l
co
rrelatio
n
-
b
a
sed tech
n
i
q
u
e
. The stu
d
y
co
n
ducted
b
y
Reaz et al. [41] has adopted a
unique usa
g
e
of integratin
g fuzzy logic, discrete wave
let transform
,
and
neural
network. The study outcom
e
was fou
nd to posses 98.19%
of accuracy in th
e detec
tion. Hu
ang and Lin [42]
have used m
a
x
i
m
u
m
likelihood m
e
thod
for developing a
classifi
cation technique. Mos
t
recently Goes et al.
[4
3
]
h
a
v
e
u
s
ed
d
a
ta
m
i
n
i
n
g
p
r
o
cess to
p
e
rfo
r
m class
i
fica
tio
n
of th
e po
wer
q
u
a
lity d
i
stu
r
b
a
n
ces.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 5
,
O
c
tob
e
r
20
16
:
243
7
–
24
46
2
444
Table 3 showc
a
ses the sca
l
e
of effectivenes
s of th
e existing approaches
used for PQ classifica
tion.
The o
u
t
c
om
e shows
SVM
i
s
t
h
e
m
o
st
preferred a
ppr
oach
fol
l
o
wed
by
A
r
t
i
f
i
c
i
a
l
Neural
Net
w
or
k.
Ho
wever
,
fu
zzy lo
g
i
c an
d swarm
in
te
ll
ig
en
ce b
a
sed
ap
pro
ach
still lag
s
b
e
h
i
n
d
in
terms o
f
classificati
o
n
.
Table
3. E
f
fectiveness
of
ot
he
r Classification Techniques
Facto
r
Other techniques
SVM
ANN
FL
SI
Data
m
i
ning
***
***
**
*
Knowledge Repres
entation
*
*****
***
***
Sustaining Uncer
t
ainty
****
***
***
***
Adaptability
***
***
** ***
Gener
a
lized Per
f
or
m
a
nce
*****
***
*
*
Lea
r
ning Abilit
y
****
****
*
***
3.
RESEARC
H GAP
Thi
s
secti
on di
scusses about
t
h
e si
gni
fi
cant
poi
nt
s t
h
at
are f
o
u
nd t
o
m
a
p i
n
t
e
r
m
s of research gap. T
h
e
vi
ews bri
e
fed
i
n
t
h
i
s
sect
i
on are deri
ved
fr
o
m
pri
o
r sec
tio
ns. Hen
ce, th
e
pro
m
in
en
t re
sea
r
ch ga
ps explored are:
3.1.
Inherent Issues in Transfor
m Techniques
The st
udy
fo
un
d t
r
ansf
orm
-
based t
echni
ques
e.g. FFT,
S-Tr
ansfo
r
m
s
, Short
-
Ter
m
Fouri
e
r
Transf
orm
et
c have frequ
ent
l
y
bei
ng used. B
u
t
,
Unf
o
rt
unat
e
l
y
, FFT-b
ased approache
s
are incapab
l
e
of processi
n
g
si
gnal
wi
t
h
l
a
rge vari
at
i
on. Short
-
Te
rm
Fouri
e
r Transfo
r
m
cannot
anal
y
ze non-st
at
i
c
si
gnal
s
. The usage of H
H
T i
s
rest
ri
cti
v
e t
o
onl
y
narr
ow
b
a
nd co
n
d
i
t
i
ons. S-t
r
ansf
orm
s
cannot
cat
er
up t
h
e
real
-t
im
e requi
rem
e
nt
s a
n
d
wavel
e
t
-
based
appr
oach i
s
pot
ent
i
a
ll
y
im
pacted by
t
h
e l
e
vel
of n
o
i
s
e. Even
Gabo
r-
based t
r
ansfo
r
m
approaches
are st
rongl
y
associ
at
ed wi
t
h
hi
gh com
put
at
ional
com
p
l
e
x
iti
es. Such i
ssues of t
r
ansfo
r
m
-
based cl
assi
fi
cat
i
o
n
schem
e
are not
fo
un
d t
o
be a
d
dressed i
n
a
n
y
exi
s
t
i
ng st
udi
es.
3.2.
Problems Ass
o
ciated wi
th A
NN-b
a
sed
Ap
proach
Al
l
t
h
e appr
oaches based
on
neural
net
w
or
k
are hi
ghl
y
de
p
e
ndent
o
n
t
h
e
si
ze of t
r
ai
ni
ng
dat
a
. M
o
re
the
training
data m
o
re is the
leve
l of accuracy. Although AN
N could offer better knowle
dge representation but
its generalized perform
a
nce pattern is not satisfactory.
There is a huge tra
d
e-off be
tween
accuracy and training
ti
me, wh
ich
is
sti
ll u
n
s
o
l
v
e
d in
th
e ex
ist
i
n
g
syste
m
.
Ano
t
her bi
gger
pr
obl
em
i
s
dat
a
het
e
rogenei
t
y
, whi
c
h
cannot
be e
ffec
t
i
v
ely
handl
ed
by
AN
N.
3.3.
Fewer Standard Outcomes
A rob
u
s
t ev
aluatio
n
o
f
ou
tcomes ass
i
sts
to
u
n
d
e
rstan
d
th
e effectiv
e wo
rk til
l d
a
te. Ou
r in
v
e
stig
atio
n
sh
o
w
s th
at 9
5
% o
f
th
e stu
d
i
es co
nd
u
c
ted
till d
a
te in
PQ
classifica
tio
n
are n
o
t
b
e
n
c
h
m
ark
e
d
.
Few to
n
a
me are
st
udy
of M
one
dero et
al
. [10]
, Dast
fan and Zadeh [1
1]
,
M
i
shra et
al
. [12]
, M
a
njul
a and Sarm
a [14]
, Lin et
al
.
[21]
, K
o
cam
an
et
al
. [22]
, M
i
lchevski
et
al
. [23]
, Na
deri
an a
nd Sal
e
m
n
i
a
[25]
, Suja
and
Jerom
e
[27]
, Nat
h
an
d
M
i
shra [30]
,
N
a
y
a
k and Das
h
[33]
, B
i
swal
et
al
. [34]
et
c.
3.4.
Less No
v
e
lty in Approa
ches
Our i
n
vest
i
g
at
ion f
o
u
nd t
h
at
45% o
f
t
h
e exi
s
t
i
ng st
udi
es
are sl
i
ght
enhancem
ent
of pri
o
r st
udi
es.
Unf
o
rt
u
n
at
el
y
,
t
h
e sli
ght
enhancem
ent
has n
o
t
y
i
el
d any signi
fi
cant
al
terat
i
ons i
n
t
h
e outcom
e
s. W
e
fou
nd t
h
at
st
udy
con
duct
e
d by
M
i
shra et
al
. [12]
has act
ual
l
y
been
replica
t
ed by othe
r research
ers
e.g. Viska
d
ouros
et al.
[13]
, M
a
nj
ul
a and Sarm
a [14]
. W
i
n et
al
. [19]
, Pani
g
r
ahi
et
al
. [28]
,
R
o
y
and Nat
h
[31]
et
c. The st
udy
cond
uct
e
d
by
Sun
d
aram
[18]
i
s
sim
i
l
a
r t
o
R
odri
g
uez et
al
.
[16]
. Si
m
i
l
a
rly
,
st
udy
co
nd
uct
e
d t
o
wa
rds
usa
g
e o
f
supp
ort
vect
or
m
achi
n
e has sim
i
l
a
r work e.
g.
Kocam
an et
al. [22]
an
d M
i
l
c
hevski
et
al
. [23]
. Li
kewi
se, St
udy
cond
uct
e
d by
P
a
ri
zi
et
al
. [44]
, M
a
jhi
et
al
. [3
6]
have o
r
i
g
i
n
a
l
l
y
used by
Nay
a
k and
Dash [
33]
.
Hence, these a
r
e the significant research gaps to
war
d
s t
h
e st
udy
of P
Q
cl
assi
fi
cat
i
on t
echni
ques.
Hence, the above m
e
ntioned issues should
be ad
d
r
essed ef
f
ectiv
el
y in
fu
tu
re.
4.
CON
C
LUSIO
N
& FUTU
RE
WORK
Fro
m
th
e d
i
scu
ssio
n
o
f
Power Qu
ality
(PQ) th
eo
ry an
d
ex
istin
g
research
wo
rk, it can
b
e
s
een
th
at PQ
di
st
urbances ar
e
m
o
re or l
e
ss a di
rect
repre
s
ent
a
t
i
on of n
on-st
at
i
c
si
gnal
s
. In t
h
e past
,
t
h
e
m
o
st
freq
u
ent
l
y
adopt
ed t
echni
ques are art
i
f
i
c
i
a
l
i
n
t
e
ll
i
g
ence, wavel
e
t
-
ba
sed approac
h
, Swa
r
m
i
n
t
e
l
l
i
g
ence-based ap
pr
oach et
c.
W
e
fo
un
d t
h
at
alm
o
st every
techni
que
has g
o
t
advant
ag
es
as well as p
o
t
e
n
tial
li
mi
tat
i
o
n
s
. Th
e p
a
p
e
r
has also
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
In
si
g
h
t
on
Effectiven
ess o
f
Freq
u
e
n
tly Exercised
PQ Cla
s
sifica
tio
n
Techn
i
q
u
e
s (B.
Devi
Vig
h
n
e
sh
wa
ri)
2
445
di
scussed t
h
e
r
e
search gap
b
r
i
e
fl
y
.
From
t
h
e ent
i
r
e i
nvest
i
g
at
i
on, i
t
can be
onl
y
sai
d
t
h
at
t
h
ere sh
oul
d
be
m
o
r
e
analysis on joi
n
t
m
echanis
m
.
Ou
r
fu
ture work
will b
e
fo
cused
o
n
d
e
v
e
lop
i
n
g
a
n
o
v
e
l alg
o
r
ith
m
fo
r PQ classifica
tio
n
in
o
r
d
e
r t
o
en
h
a
n
ce th
e
on
lin
e
m
o
n
ito
rin
g
syste
m
. We will in
it
iat
e
o
u
r
in
v
e
stig
ati
o
n
b
y
i
m
p
l
e
m
en
tin
g
th
e enhan
ced
versi
on
of
wav
e
l
e
t
and neural
net
w
or
k-
based ap
proac
h
.
The p
r
ime reaso
n
b
e
h
i
nd
th
is wi
ll b
e
to
wav
e
let an
d
neural
net
w
or
k-base
d ap
pro
ach are m
o
st
freq
u
ent
l
y
adopt
ed t
echni
que
and t
h
ere
b
y
com
p
arati
v
e anal
y
s
i
s
b
eco
m
e
s ea
sie
r
an
d
less co
mp
licated
to
in
i
t
i
ate wi
th
. Ou
r
fu
tu
re
d
i
rectio
n
will b
e
also
to
ward
s
d
e
sig
n
in
g
a
no
vel
cl
assi
fi
er base
d o
n
e
n
hanced
neu
r
al
net
w
or
k. T
h
e
wor
k
i
s
a
n
t
i
c
ipat
ed t
o
have
l
e
sser depen
d
e
n
cy
of
higher accurac
y
on size of training da
tabase. Our
work
will be towards
both single a
nd m
u
ltiple form
s of
d
i
stu
r
b
a
n
ces in PQ ev
en
ts.
Our fin
a
l d
e
stin
ati
o
n
will b
e
to
dev
e
lo
p
a fu
zzy-classifi
er with
en
h
a
n
ced
m
u
lti
lev
e
l
classification approach.
REFERE
NC
ES
[1]
C. P. Verhoef
,
et al.
,
“
C
us
tom
e
r experi
enc
e
cr
eat
i
on: Determ
inan
ts,
dy
na
mi
c
s
a
nd
ma
na
gement strategies,”
Journal
of reta
iling
, vo
l/issue: 85(1), pp.
31-41, 2009
.
[2]
J. B. Karmachar
y
a
, “Good Manu
factur
ing Prac
tices (GMP) for Medi
cinal Products,” 2014
.
[3]
R. Jangid and
P. A. K. Pa
rkh, “Reducing the Voltage Sag
and Swe
ll Problem in Distributio
n S
y
stem Using
D
y
namic Voltag
e
Restor
er with
PI C
ontroller,”
Retriv
ed 22nd
Oct, 2015.
[4]
T. Fang,
et al.
, “
P
roces
s
for elec
t
r
ica
l
l
y
conn
ec
tin
g ele
c
tri
cal d
e
vi
ces
us
ing a cond
uctiv
e anis
otrop
i
c m
a
teri
al
,” U.S
.
Patent, 1997
.
[5]
D. Saxena,
et a
l
.
, “
P
ower quali
t
y
ev
ent cl
as
s
i
fi
cat
ion: an over
v
iew and ke
y
i
s
s
u
es
,”
International Journal of
Engineering, Scie
nce and Techno
logy
, vol/issue: 2
(
3), pp
. 186-199
, 2010.
[6]
C. Thurau,
et a
l
.
, “
D
eterm
i
nis
t
ic CUR for Im
proved Large-S
c
a
l
e Data Anal
ys
is
: An Em
pirical S
t
ud
y
,
” in
S
D
M
,
pp. 684-695
, 20
12.
[7]
J. Solomon and B. Rock
, “Imaging spect
rometr
y
for
ear
th remote sensing,”
Scien
c
e
,
vol/issue: 228(4704), p
p
.
1147-1152, 198
5.
[8]
V. Kecm
an,
“
L
e
a
rning
and s
o
ft
com
pu
ting: supp
ort vector mach
ines,” n
e
ur
al networks, and fuzzy
logic models.
MIT press, 2001
.
[9]
C.
C. Le
e,
“Fuz
zy
l
ogi
c i
n
c
o
nt
ro
l s
y
s
t
em
s
:
f
u
zz
y
logic
con
t
roller
II,
”
S
y
stems, Man and C
y
bernetics, IEEE
Transactions
, vo
l/issue: 20
(2), pp
. 419-435
, 1990
.
[10]
Monedero,
et al.
, “VICARED:
A neural network based s
y
stem
for the
detection
of electrical d
i
sturbances in real
time,” in
Advan
ces in
Natur
a
l C
o
mputation,
Springer Ber
lin
Heid
elberg
, pp
. 147-
154, 2005
.
[11]
Dastfan and A.
S. Zadeh
,
“
I
ntellig
ent Power
Qualit
y
Monitor
i
ng b
y
using S-
Transform
and Neural Network
,
”
Retrieved, 10th
November, 2015
.
[12]
S.
Mishra
,
et al.
, “
D
etec
tion
and
clas
s
i
fi
ca
tion o
f
power qual
i
t
y
disturbances using S-tr
ansform and probabilistic
neural network
,
”
Power
De
li
ver
y
,
IEEE Transactions
, vol/issue: 2
3
(1), pp
. 280-28
7, 2008
.
[13]
G. Viskadouros,
et al.
, “Power Quality
Even
t Recognition and
Cla
ssification
Using a Wa
velet-Based Neur
al
Network,”
Int. J. of Susta
inable
Water and
Environmental S
y
stem
s
, vol/issue: 3(1)
, pp
. 45-52
, 201
1.
[14]
M. Manjula and
A. V.
R.
S
.
Sa
rma
,
“Asse
ss
me
n
t
of
power qu
ali
t
y
events
b
y
em
pi
rical mode d
eco
mposition based
neural network
,
” in Proceed
ing o
f
the Worl
d Con
g
ress on Engin
e
ering, pp. 4-6, 2
012.
[15]
A. P. Memon,
et al.
, “
T
im
e-Frequenc
y and Ar
tific
ia
l Neural
Networ
k Applications and Anal
ysis for Elec
tric
al
S
y
s
t
em
P
o
wer Qualit
y Dis
t
urb
a
n
ces
in M
A
TLAB,”
In
ternationa
l Journal of Inn
o
vative Techno
logy
and Exploring
Engineering (
I
JI
TEE)
, www. ijitee. org 3
,
R
e
tr
ieved, 10
th Nov, 2
015.
[16]
M. V. Rodriguez,
et al.
, “
D
ete
c
tion and cl
as
s
i
fi
cat
ion
of single and combined power quality
d
i
sturbances using
neural networks,”
Industrial Electronics
, I
E
EE Transactions
, vo
l/issue: 61(5), pp.
2473-2482, 201
4.
[17]
S. S. Jasper,
“
A
rtific
ia
l Neura
l
N
e
twork Control
l
e
d
DS
TATCOM
for Power Quality
Impr
ovement,” Retr
iev
e
d, 12
th
Nov, 2015.
[18]
Sundaram, Network, Lin
ear
Neural, “P
ower
Quality
Disturb
a
nce Classifica
tion Using Adaptive Linear Neur
al
Network (ADALINE) and
Feed
Forward Neural
Network (FFNN
)
,” 2015
.
[19]
N.
N.
K.
Win,
et al.
, “Comparison Of
Power
Quality
Disturb
a
nces
Cla
ssific
a
tion Ba
se
d
On Neural Network,
”
International Jo
urnal of S
c
ie
n
tific &
Technology
Research
, vol/is
sue: 4(7), 2015
.
[20]
D.
D.
Shaky
a
and S.
N.
Sing
h, “SVM based fault location
and cla
s
s
i
fi
cat
io
n us
ing fuzz
y
clas
s
i
fi
er for PQ
monitoring,” in
Power and Energy Society General Meeting-Con
ve
rsion and Delivery of Electrical Energy in th
e
21st Century
,
p
p
. 1-8
,
2008
.
[21]
W. M. Lin,
et al.
, “
D
ete
c
tion
and classifi
cat
io
n of m
u
ltiple powe
r-quali
t
y
d
i
sturbanc
es with wavele
t m
u
lticl
a
ss
SVM,
”
Power
D
e
li
ver
y,
IEEE Transactions
, vo
l/issue: 23(4), pp.
2575-2582, 200
8.
[22]
C. Kocam
an,
et al.
, “Classification of two co
mmon po
wer quality
d
i
sturban
ces using wavelet b
a
sed SVM,”
in
MELECON 2
010-2010 15th I
EEE Mediterr
anean Electrotechn
ical Conferen
ce
, pp. 587-591, 20
10.
[23]
D. Taskovski,
et al.
, “Classification of Power Qu
alit
y
Disturb
a
nces Using Wavelets
and Support Vector Mach
ine,”
Ele
k
tr
onika
ir
El
ektr
o
t
echni
ka
, v
o
l/issue: 1
9
(2), p
p
. 25-30
, 2012
.
[24]
P.
K.
Ray
,
et al.
,
“
O
ptim
al f
eatur
e and
dec
i
s
i
on tree
-bas
ed
cl
assification
of power qua
lity
d
i
sturbances
in
distributed
gen
e
r
a
tion
s
y
stems,”
Sustainable Energy, IEEE Transactions
, vol/issue: 5(1)
, pp
. 200-
208, 2014
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE
Vo
l. 6
,
N
o
. 5
,
O
c
tob
e
r
20
16
:
243
7
–
24
46
2
446
[25]
N. Sobhan and
A. Salemnia,
“Detec
tion
and
Classification
of Power-Qua
lity
Even
ts Using Discrete Gabo
r
Transform and
Support Vector Machine,”
The 6th Intern
ational
Power
Electronics D
r
ive Systems and
Technologies Co
nference
, 2015.
[26]
H.
He
a
nd J. Sta
r
zy
k, “A se
l
f-orga
n
i
z
i
ng le
a
r
ning a
rray
sy
ste
m
for powe
r
qua
lity
cl
a
ssi
fi
ca
t
i
o
n
ba
se
d on wa
ve
let
trans
f
orm
,
”
Pow
e
r Delivery, I
E
EE Transactions
, vol/issue: 21(1), pp.
286-295
,
20
06.
[27]
S.
Suja
a
nd J.
Je
rome
,
“Powe
r
si
gnal disturbance classificatio
n usi
ng wavelet based ne
ural n
e
twork,”
S
e
rb
ian
journal of electr
ical engineering
,
vol/issue:
4(1), p
p
. 71-83
, 2007
.
[28]
B. K. Panigrah
i,
et a
l
.
, “
A
M
ach
i
n
e Int
e
ll
igen
ce
Approach for
Cl
as
s
i
fica
tion of
P
o
wer Quali
t
y
Di
s
t
urbances
,
”
200
9.
[29]
H. Shareef and
A. Mohamed, “Power
quality
d
i
sturbance detection and visua
l
i
zat
ion util
izing
im
age proces
s
i
n
g
methods,” in
Pr
oceed
ing in
terna
tional mid
d
le
ea
st power con
f
ere
n
ce, Ciro, Eg
ypt
, 2010.
[30]
S.
Na
th,
et al.
, “
W
avelet b
a
s
e
d featur
e extr
act
ion
for clas
s
i
fic
a
tio
n of P
o
wer Qualit
y
Dis
t
urb
a
nc
es
,
”
Retri
e
ved 20th
Nov, 2015.
[31]
H.
Eristi
and Y.
Demir,
“The
feature selection b
a
sed power quality even
t
classification using wavelet transform and
logistic
m
odel
tr
ee,
”
Prze
gl
ą
d El
ektr
o
t
echni
cz
ny,
vol/issue:
88(7a)
, pp. 43-48, 201
2.
[32]
S. Ro
y
and S. N
a
th, “Classification of power qua
lity
disturb
a
nces
using
features o
f
signals,”
Intern
ational Journal
of Scientific and
Re
search Publications
, vo
l/issue: 2(11), 2012.
[33]
M
.
Na
yak and
S
.
Das
h
, “
F
uzzy De
ci
sion T
r
e
e
and Par
tic
le
Swarm
Optim
i
zation for Min
i
n
g
of Time Series
Data,”
In
ternational Journal of
Computer
Applications
, vol/issue: 17(7)
, pp
. 35-
41, 2011
.
[34]
B. Biswal,
et al.
, “
A
h
y
br
id an
t
colon
y
opt
im
iz
ation
techn
i
que
for power signal pa
t
t
ern clas
s
i
fi
cat
ion,”
Exp
e
rt
Systems with
Ap
plications
, vo
l/is
sue: 38(5), pp
. 6
368-6375, 2011
.
[35]
S
.
Has
h
em
ineja
d
,
et al.
, “Power quality
dis
t
urbance classif
i
ca
tion using S
-
transform and
hidden Marko
v
model,”
Electric Power Compon
ents and S
y
stems
, vol/issue: 40(1
0
), pp
. 1160-118
2, 2012
.
[36]
B. Majhi,
et al.
, “
C
lassific
a
tion
of Power Signals Using PSO
bas
e
d K-M
eans
Algorithm
and
F
u
zz
y C M
e
an
s
Algorithm,”
Jou
r
nal of
Agricultu
r
e and Life S
c
ien
ces
, vo
l/issue: 1(
1), 2014
.
[37]
N. Kum
a
rasabapath
y
and P.
S. Manoharan, “MATLAB Sim
u
lation of UP
QC fo
r Power Qualit
y
Mitigation Usin
g
an Ant Co
lon
y
B
a
sed Fu
zzy
Control
Techniqu
e,”
The Scientific W
o
rld Journal
, 20
15.
[38]
F. Tianru
i and
O. Sen, “Design of
a Distributio
n Network Power Quality
Monitoring S
y
stem based on Meter
i
n
g
Autom
a
tion S
y
stem
s and its Applic
ation
,
”
Indo
nesian Journal of Electrical
En
gineering and Computer Science
,
vol/issue: 10(7), 2012.
[39]
J.
Jia
p
e
i
,
et al.
, “A Control Strateg
y
for Sing
le-phase Grid-Co
nnect
ed Inver
t
er
with P
o
wer Qualit
y R
e
gul
ato
r
y
Function,”
Indo
nesian Journal o
f
Electrical
Eng
i
neering and
Co
mputer Science
, vol/issue: 12(1), 2014.
[40]
F. L. Guo
and
W. H. Feng, “Applic
ation of
Grey
Correlation
Degree
and
TOPSIS Method in Ev
aluation of Power
Qualit
y,
”
Indonesian Journal o
f
Electrica
l
En
g
i
n
eering and
Computer Science
, v
o
l/issue: 1
2
(6), 2
014.
[41]
M
.
B
.
I
.
R
e
a
z
,
et al.
, “
E
xp
ert s
y
s
t
em
for power q
u
alit
y
dis
t
urban
c
e cl
as
s
i
fier
,”
Po
wer
Deli
ver
y,
IEEE Transactions
,
vol/issue: 22(3), pp.
1979-1988
, 2007.
[42]
C. H. Huang and C. Lin, “
M
ultiple Chaos S
y
nc
hroniza
ti
on S
y
st
em
for Power Qualit
y Cl
assific
a
tion in a Power
Sy
s
t
e
m
,
”
The S
c
ientific World
Jo
urnal
, 2014.
[43]
A.
R.
T.
Góes,
et al.
, “Classification of Power Quality
Considering
Voltage SA
GS in Distribution Sy
stems using
KDD Process,
”
Pesquisa Opera
c
ional
, vol/issue: 35(2), pp. 329-
352, 2015
.
[44]
A.
Parizi,
et al.
,
“
P
ower Qualit
y Dis
t
urbance Cl
as
s
i
fica
tion US
ING S
-
Trans
f
orm
and F
u
zz
y
S
y
s
t
em
s
Oriented
b
y
PSO Algorithm,
”
International Journal on “Tech
nical and
Ph
ysical Problems o
f
Engineering
, vo
l/issue: 4(2)
, pp
.
153-161, 2012
.
BIOGRAP
HI
ES OF
AUTH
ORS
I am B.Devi Vighneshwari, I h
a
ve
completed my
BE in EEE
an
d ME in Power
S
y
stem. I h
a
ve
been done m
y
workshop/training on Digital
C
ontrol of Power Electro
n
i
c Equipment,
CAD
forEle
ctri
ca
l Drawing, Power Sy
s
t
em
Sim
u
lation
and Artific
ial Int
e
llig
enc
e
and Meta – Heuristi
c
Techn
i
ques App
lic
ations
in Power S
y
s
t
em
.
I am
Dr. R. Ne
e
l
a P
r
ofes
s
o
r of E
l
ec
tric
al
Engin
eeri
ng,
I ha
ve
c
o
mpl
e
te
d m
y
B.
E
in E
l
ec
t
r
ic
al
&
Ele
c
troni
cs
, M
.
T
ech in P
o
wer S
y
s
t
em
s
and P
h
D in
Ele
c
tri
c
a
l
. I h
a
ve te
chi
ng
exper
i
ence is more
than 22
ye
ars
an
d in R
e
s
ear
ch m
o
re th
an 12
ye
ar
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.