Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
9
, No
.
2
,
Febr
ua
ry
201
8
,
pp.
380
~
386
IS
S
N:
25
02
-
4752
, DO
I: 10
.11
591/
ijeecs
.
v9.i
2
.
pp
380
-
386
380
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Classific
atio
n
of
The NTEV P
ro
b
lems on
the
Com
mercial
B
uil
din
g
Mohd
Abdul
Tali
b Mat Y
u
so
h,
S
aidatul
Ha
bs
ah A
sma
n,
Z
uha
il
a M
at Yasin
, Ahm
ad F
arid
Ab
idi
n
Univer
siti
Te
kno
logi
MA
RA (
Ui
TM),
40450
Sha
h
Alam,
Se
la
ngo
r,
Mal
a
y
s
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
9
, 2
01
7
Re
vised
Dec
2
0
, 2
01
7
Accepte
d
Ja
n 4
, 201
8
Neutr
al
to
E
art
h
Volta
ge
(NTE
V)
is
one
of
po
wer
qual
i
t
y
(PQ
)
proble
m
s
in
the
comm
erc
ial
buil
ding
th
at
ne
ed
to
be
r
esolved.
The
cl
assifi
cation
of
th
e
NTEV
proble
m
s
is
a
m
et
hod
to
ide
nti
f
y
the
source
t
y
p
es
of
dist
urba
nce
i
n
al
l
evi
a
ti
ng
the
proble
m
s.
Thi
s
pape
r
pre
sents
t
he
cl
assifi
ca
t
ion
of
NTEV
source
in
th
e
co
m
m
erc
ia
l
buil
d
i
ng
which
is
known
as
the
har
m
oni
c,
loose
te
rm
ina
t
ion,
an
d
li
ghtni
ng.
Th
e
Euc
li
d
ea
n
,
Cit
y
b
loc
k,
and
Cheb
y
she
v
var
ia
b
le
s
for
K
-
Nea
rest
Ne
ighb
or
(K
-
NN
)
cl
assif
y
ing
ar
e
be
ing
uti
lized
in
orde
r
to
ide
n
ti
f
y
the
b
est
per
for
m
anc
e
for
cl
assi
f
y
ing
the
NTEV
proble
m
s.
The
n,
S
-
Tr
ansform
(ST)
is
app
lied
as
a
pre
-
p
roc
essing
signal
to
ext
ra
ct
th
e
desire
d
f
eature
s
of
NTEV
p
ro
ble
m
for
c
la
ss
if
ie
r
input
.
Furth
ermore,
the
per
form
anc
e
of
K
-
NN
var
ia
ble
s
is
val
ida
t
ed
b
y
using
the
conf
usion
m
at
rix
and
l
ine
ar
reg
r
e
ss
ion.
The
c
la
s
sific
a
ti
on
r
esults
show
tha
t
all
the
K
-
NN
var
ia
b
le
s
c
apa
b
l
e
to
ide
n
ti
f
y
the
NTEV
probl
ems
.
W
hil
e
the
K
-
NN
result
s
show
tha
t
the
E
ucl
id
ea
n
and
Ci
t
y
blo
ck
var
i
able
s
are
well
per
fo
r
m
ed
rat
her
tha
n
th
e
Cheb
yshev
var
i
abl
e
.
How
eve
r,
th
e
Cheb
y
shev
var
i
abl
e
is
stil
l
rel
i
abl
e
as
th
e
conf
usion
m
at
r
i
x
show
s
m
inor
m
iscl
assific
a
ti
on
.
The
n
,
the
li
ne
ar
reg
r
ession
outpe
rform
ed
t
he
per
ce
nt
age
clos
e
to
a
per
f
ec
t
val
ue
whi
ch
is hun
dre
d
p
erce
nt.
Ke
yw
or
d
s
:
K
-
Near
e
st
N
ei
ghbor
(K
-
N
N)
cl
assifi
er tools
Neu
t
ral to E
art
h Vo
lt
age
(PQ
)
Power Q
ualit
y (PQ)
S
-
T
ran
s
f
or
m
(S
T)
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
:
Mohd
Abd
ul T
al
ib Mat
Yusoh
,
Un
i
ver
sit
i Te
knol
og
i M
ARA
(U
iTM
)
,
40450 S
hah A
l
a
m
, S
el
ango
r,
Ma
la
ysi
a
Em
a
il
:
m
oh
d.
a
bdul.tal
ib@
gma
il
.co
m
1.
INTROD
U
CTION
High
Ne
utral
t
o
Eart
h
V
oltag
e
(N
T
EV
)
in
t
he
com
m
ercial
bu
il
di
ng
is
on
e
of
t
he
co
nce
r
ns
in
Power
Qu
al
it
y
(PQ)
di
sturb
a
nces
du
e
to
hazar
dous
to
the
hum
ans,
anim
al
s,
el
ect
ric
an
d
el
ect
r
on
ic
a
ppli
ance
s,
a
nd
el
ect
rical
netw
orks
syst
em
[1
]
,
[
2]
.
Ma
ny
r
esearche
rs
ha
ve
discu
ssed
th
e
high
NTE
V
in
the
com
m
ercial
bu
il
di
ng
is
due
to
i
m
pr
oper
wirin
g,
poor
groun
ding
sy
stem
,
nonlinea
r
loa
d,
an
d
ca
ble
dam
aged
[3
–
6]
.
Howe
ver,
the
discuss
i
on
is
st
il
l
in
assum
ption
t
o
dete
rm
in
e
the
source
ty
pes
of
NTE
V
as
in
[
7]
an
d
ha
ve
not
been p
r
ov
e
d b
y any arti
fici
al
intel
li
gen
t t
ech
niques.
Norm
al
l
y,
the
so
urce
ty
pes
of
PQ
distu
rb
a
nc
e
in
the
c
omm
ercial
buil
ding
can
be
ide
ntifi
ed
by
us
i
ng
the
cl
assifi
cat
ion
te
c
hn
i
qu
e
[
8]
.
N
ow
a
days,
a
lot
of
cl
assi
fier
to
ols
are
c
om
bin
ed
with
the
sig
nal
proc
essin
g
te
chn
iq
ue
t
o
cl
assify
the
P
Q
di
sturb
a
nces
i
n
the
com
m
ercial
bu
il
di
ng
[9
–
11]
.
The
sig
nal
processi
ng
is u
ti
li
zed
to
ext
ract
the
desire
d
in
f
or
m
at
ion
wh
ic
h
is
app
li
ed
as
a
n
i
nput
f
or
the
cl
assifi
er
to
ols
[
8]
.
T
hen,
s
om
e
of
th
e
disturba
nces
i
ntr
oduce
oth
e
r
pro
blem
s
in
sign
al
processi
ng
due
to
lim
it
at
ion
an
d
ca
nnot
be
a
naly
zed
to
extract
the
des
ired
in
form
at
i
on
acc
urat
el
y
[12],
[
13]
.
He
nc
e,
an
a
ppr
opr
ia
te
var
ia
ble
in
cl
assifi
er
to
ol
an
d
sign
al
proces
sing
te
ch
nique
ne
ed
t
o
be
util
iz
ed
t
o
ide
ntify
the
t
ypes
of
NTE
V
pro
bl
em
in
the
c
omm
ercial
bu
il
di
ng.
T
he
NTE
V
occurre
nce
need
s
to
be
a
war
e
of
due
to
haza
rdo
us
to
hum
ans,
anim
al
s,
el
ec
tric
al
app
li
anc
es, a
nd elec
tric
al
n
et
work syst
em
[14]
.
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
Cl
as
sif
ic
ation
of T
he
NTE
V P
ro
ble
ms
on T
he
Co
mm
erci
al
Buildin
g
(
M
ohd
A
bdul
Ta
li
b
Mat Y
usoh
)
381
The
m
ai
n
co
nt
ribu
ti
on
of
th
is
pa
per
is
to
cl
assify
the
s
ource
ty
pe
s
of
NTE
V
pro
bl
e
m
s
in
the
com
m
ercial
bu
il
ding
du
e
to
the
t
ri
plen
ha
rm
on
ic
an
d
t
ran
sie
nt.
T
he
transient
ca
n
be
div
ide
d
i
nto
t
w
o
cat
egories;
tra
ns
ie
nt
due
to
the
lo
os
e
te
rm
i
nation
an
d
t
ra
ns
ie
nt
due
to
the
li
ghtnin
g
strike.
T
hen,
th
e
K
-
Near
est
Neig
hbor
(
K
-
N
N)
is
sel
ect
ed
as
the
cl
assifi
er
to
ol
and
has
bee
n
te
ste
d
with
dif
f
eren
t
var
ia
bles,
wh
ic
h
are
E
uclidean
,
Ci
ty
blo
ck,
a
nd
Che
bys
hev
.
The
differe
nt
var
ia
bles
of
K
-
N
N
a
re
util
iz
ed
to
i
de
ntify
the
best
perform
ance
in
cl
assify
ing
the
NTE
V
pro
ble
m
.
Fu
rthe
r,
S
-
Transf
or
m
(S
T
)
is
c
hosen
to
e
xtract
the
feat
ures
of
NTE
V
pro
ble
m
accord
in
g
to
the
sta
ti
sti
cal
analy
sis
al
go
rit
hm
s.
ST
te
chni
qu
e
is
sel
ect
ed
du
e
to
t
he
cap
abili
ty
in
proc
essin
g
t
he
sig
nal
with
ou
t
m
issi
ng
a
ny
inform
at
ion
featur
e
s,
var
ia
bles
wi
ndow
s
cal
e,
an
d
ou
t
pe
rfor
m
wav
el
et
tran
sf
or
m
(
W
T)
a
nd
sh
ort
-
ti
m
e
fo
ur
ie
r
tra
nsfo
r
m
(S
TFT
).
Fin
al
ly
,
the
per
f
orm
ances
of
K
-
NN
a
re
ob
s
er
ved
ba
se
d
on
the
di
ff
e
r
ent
va
riables
by
us
in
g
c
onf
usi
on
m
at
rix
a
nd
li
near
re
gr
es
sion.
Acc
ordin
g
to
t
he
conf
us
io
n
m
a
trix
and
li
near
re
gr
essi
on
resu
lt
s,
the
perform
a
nce
of
K
-
NN
c
la
ssifie
r
too
ls
is
analy
sed
bas
ed
on
the p
e
rce
ntage
s of acc
ur
acy
,
m
isc
la
ssific
at
i
on and t
he rel
at
ion
s
hip
betwe
en
the
tar
get and
ou
t
pu
t
res
ults.
2.
RESEA
R
CH MET
HO
D
To
cl
assify
th
e
NTE
V
pro
bl
e
m
s
in
the
c
omm
ercial
bu
il
din
g,
seve
ral
m
et
ho
ds
s
uc
h
as
the
ST
op
e
rati
on, K
-
N
N
cl
assifi
er
to
ol
shou
l
d
be
f
ollow
e
d
2.1.
Th
e S
-
Tr
an
s
fo
rm
(ST
)
Theor
y
ST tech
nique i
s a tim
e
-
fr
eque
ncy pre
-
proces
sing si
gnal
, wh
ic
h
prese
nts th
e res
ult i
n
c
omplex
nu
m
ber
s
pectr
um
. Th
en
, S
T
al
so
produces
the
resu
lt
in
S
-
m
at
rix
w
hich
c
on
sist
t
he
m
ulti
ple num
ber
s of
colum
ns
and
r
ow
s
. The
ST
of signal
x(
t)
ca
n be
def
i
ned as
foll
ow
[
13
]
:
ft
j
f
t
e
e
t
x
f
f
S
2
2
)
(
2
2
)
(
2
|
|
)
,
(
(
1
)
Accor
ding to
(
1
)
, let
kT
and
NT
n
f
, the
d
isc
rete
ST is
g
ive
n by:
0
,
,
2
1
0
2
2
2
2
n
e
e
NT
n
m
X
NT
n
kT
S
mk
N
m
m
n
m
n
(
2
)
wh
e
re,
k,
m
, n
=
0,
1,
…, N
-
1
T=sam
pling
in
te
rv
al
N=total
of sam
pling p
oin
t
Accor
ding
to
t
he
ST
resu
lt
,
t
he
feat
ur
es
of
NTE
V
pro
ble
m
are
extracte
d
by
us
in
g
the
sta
ti
sti
ca
l
analy
sis
te
chn
i
qu
e
,
that
incl
ude
the
sta
nd
a
rd
dev
ia
ti
on
(
3
)
,
m
ean
(
4
)
,
var
i
ance
(
5
)
,
s
kewness
(
6
)
,
kurto
sis
(
7
)
,
an
d
t
otal
har
m
on
ic
distor
ti
on
(THD)
(
8
)
.
F
igure
1
s
hows
the
operati
on
to
cl
assify
the
NTE
V
pro
ble
m
in
the
com
m
ercial
buil
din
g
incl
ud
e
d
ST,
sta
ti
sti
cal
analy
sis,
and
K
-
N
N
cl
assifi
er
to
ols.
Th
en,
the
al
go
rithm
s
of
sta
ti
sti
cal
an
alysis are
giv
e
n as
[
15
]
:
2
1
1
2
)
(
1
1
m
a
x
M
i
j
ij
Y
Y
M
(
3
)
M
j
ij
Y
M
Y
1
1
m
a
x
(
4
)
M
i
j
ij
Y
Y
M
1
2
2
)
(
1
1
m
a
x
(
5
)
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.
2
,
Fe
br
uary
201
8
:
3
80
–
386
382
3
2
1
1
2
1
3
)
(
1
1
)
(
1
m
a
x
M
j
j
ij
M
j
j
ij
Y
Y
M
Y
Y
M
s
(
6
)
2
1
2
1
4
)
(
1
1
)
(
1
m
a
x
M
j
j
ij
M
j
j
ij
Y
Y
M
Y
Y
M
k
(
7
)
1
2
1
2
2
m
a
x
ij
M
j
ij
Y
Y
T
H
D
(
8
)
Figure
1
.
Th
e
Cl
assifi
cat
ion
of N
T
EV
Pro
bl
e
m
Ba
sed on t
he
e
qu
at
io
n
i
n
(
3
)
-
(
8
)
, th
e
r
es
ult o
f
N
TE
V feat
ure
s can
b
e
obtai
ne
d
a
nd u
ti
li
zed
as an
in
pu
t
for K
-
NN cl
assifi
er.
2.2. K
-
Ne
arest Neig
hb
or (K
-
NN)
K
-
NN
ca
n
be
cat
egorized
as
the
Near
est
Ne
ighbor
(
NN)
f
a
m
ily
[16]
.
K
-
NN
is
reli
able
and
easy
to
i
m
ple
m
ent
fo
r
cl
assifi
cat
ion
by
us
ing
non
-
pa
ram
et
ric
and
la
zy
le
arn
in
g
te
chn
i
qu
e
[17]
.
B
asi
cal
ly
,
the
K
-
NN
cl
assifi
cat
ion
op
e
rates
base
d
on
the
distan
ce
m
et
ric
of
input
cl
assify
in
g
an
d
te
sti
ng
sam
ples.
Then,
the
perform
ance
of
K
-
N
N
al
s
o
de
pends
on
the
sel
ect
ed
num
ber
of
K.
T
hus,
the
num
ber
of
K
is
sel
ect
ed
ba
sed
on
the
best
values
bef
ore
analy
zi
ng
the
perform
ance
of
K
-
N
N
us
ing
the
different
va
riables
.
This
pap
e
r
s
how
s
th
e
te
sti
ng
for
diff
e
re
nt
va
riables
of
K
-
NN
in
orde
r
to
i
de
ntify
the
be
st
perform
ance
of
K
-
N
N
in
cl
assify
in
g
the
N
TEV
pro
blem
.
The
va
riables
of
K
-
N
N
su
c
h
as
t
he
E
uclidean
(
9
)
,
C
it
y
blo
ck
(
10
)
,
Chebys
he
v
(
11
)
a
r
e
us
e
d
to
a
naly
ze
the
NT
EV
pro
blem
s
in
the
com
m
ercial
b
uildin
g.
T
he
m
at
he
m
at
ic
a
l
al
gorithm
s
are
give
n
belo
w
[18]
:
(
9
)
(
10
)
k
i
i
i
y
x
1
2
)
(
k
i
i
i
y
x
1
|
|
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
Cl
as
sif
ic
ation
of T
he
NTE
V P
ro
ble
ms
on T
he
Co
mm
erci
al
Buildin
g
(
M
ohd
A
bdul
Ta
li
b
Mat Y
usoh
)
383
1
2
1
2
,
x
m
a
x
y
y
x
(
11
)
3.
RESU
LT
S
AND A
N
ALYSIS
Figure
2
s
hows
the inp
ut of cl
assifi
cat
ion
w
hi
ch
util
iz
ed o
n diff
e
re
nt v
a
ria
bles
of
K
-
N
N
c
la
ssifie
r
too
ls.
T
he num
ber
o
f sam
pl
es that
hav
e
b
e
en use
d
as i
nput
an
d t
est
ing da
ta
are 124 a
nd
126, res
pect
ive
ly
.
Ba
sed on t
he s
a
m
ples d
at
a,
K
-
N
N
a
naly
sis i
s u
ti
li
zed to i
de
ntify t
he best c
at
egories o
f K
-
NN v
a
riables i
n
cl
assify
ing
t
he NTE
V pro
blem
s.
(a)
(b)
(c)
Figure
2
: T
he plot cl
assifi
cat
ion i
nput
featu
r
es:
(
a)
F
1
a
nd
F2
,
(b) F
3
a
nd
F4
,
and
(c) F
5 and F
8
3.1. Per
f
orm
ance
of
K
-
NN
Va
ri
ab
le
s B
ase
d on
Diff
ere
n
t
K
Values
The K
-
N
N
cl
as
sific
at
ion
us
es
50 it
erati
ons wi
th a d
i
ff
e
ren
t
nu
m
ber
of K.
Figure
3
sho
ws
each
di
ff
e
ren
t
values
of
K
produce
the
dif
fe
ren
t
re
su
lt
s
of
K
-
NN
va
riable
s.
The
best
values
of
K
f
or
va
riables
Eu
cl
idean,
Ci
ty
bl
ock
,
an
d
C
he
bysh
e
v
a
re
between
1
unti
l
4.
T
he
nu
m
ber
o
f
K
betwee
n
1
unti
l
4
sho
ws
the
identic
al
res
ults
with
re
sp
ect
t
o
their
pe
rfor
m
ance.
Ba
se
d
on
the
best
value
of
K,
the K
-
NN
is
ana
ly
zed b
y
us
in
g
the
con
fu
si
on m
at
rix
te
chn
i
qu
e
and line
ar
regressio
n.
(a)
(b)
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.
2
,
Fe
br
uary
201
8
:
3
80
–
386
384
(c)
Figure
3
Per
for
m
ance of
K value
for K
-
NN ty
pes (a)
Eu
cl
id
e
an,
(
b)
Ci
ty
bloc
k,
(c) C
he
bys
hev
3.2. C
on
f
usio
n
Matri
x Resul
ts
Figure
4
s
how
s
the
re
su
lt
of
con
f
us
io
n
m
at
rix
w
hich
has
be
en
te
ste
d
by
usi
ng
t
he
K
-
NN
cl
assifi
er
of
the
di
ff
e
ren
t
va
riables
s
uc
h
a
s
Eucli
de
an
,
C
it
y
blo
ck,
a
nd
Chebys
hev.
T
he
cl
asses
1,
2,
and
3
re
pr
e
sen
ts
the
NTE
V
pro
ble
m
s
du
e
to
the
har
m
on
ic
,
lo
ose
connecti
on,
a
nd
li
ghtni
ng,
r
especti
vel
y.
Ba
sed
on
the
fig
ur
e
,
th
e
nu
m
ber
s
of
sa
m
ples
fo
r
cl
ass
1,
cl
ass
2,
a
nd
cl
ass
3
a
re
i
de
ntica
l
42.
The
resu
lt
of
var
ia
ble
E
uclidea
n
sh
ows
that
the
cl
ass
1,
cl
ass
2,
an
d
c
la
ss
3
are
33.
3%
.
The
re
su
lt
of
K
-
N
N
f
or
va
r
ia
ble
Ci
ty
blo
ck
seem
s
si
m
i
lar
with
var
ia
ble
Eucli
dean
.
T
he
n
f
or
the
var
ia
ble
Chebys
hev
shows
the
resu
lt
s
for
cl
ass
1,
cl
ass
2,
an
d
cl
as
s
3
a
re
33.3%,
31%
,
and
33.
3%,
re
spe
ct
ively
.
The
correct
num
ber
cl
assifi
es
for
var
ia
ble
Cheby
sh
e
v
are
42,
39,
a
nd
42 for cl
ass
1,
cl
ass 2
,
and cla
ss 3, r
e
sp
ect
ive
ly
.
Ov
e
rall
resu
lt
s
sh
ow
that
the
K
-
NN
cl
assifi
er
us
i
ng
the
E
uclidean
a
nd
Ci
ty
blo
ck
va
r
ia
bles
are
the
best r
e
su
lt
s
on
cl
assify
ing
t
he NTE
V pro
blem
s w
it
h
100% a
ccur
acy
.
(a)
(b)
(c)
Figure
4
.
Th
e
K
-
NN Co
nfusi
on Mat
rix f
or
va
riable (
a)
Euc
li
dean
(b)
Ci
ty
b
loc
k (c)
C
he
by
sh
ev
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
Cl
as
sif
ic
ation
of T
he
NTE
V P
ro
ble
ms
on T
he
Co
mm
erci
al
Buildin
g
(
M
ohd
A
bdul
Ta
li
b
Mat Y
usoh
)
385
3.3. Re
gressio
n R
es
ults
Table
1
el
ab
orat
es
the
perf
or
m
ance
of
K
-
N
N
base
d
on
the
diff
e
ren
t
var
ia
bles
by
us
in
g
li
near
regressio
n.
Ac
cordin
g
to
the
ta
ble,
the
K
-
N
N
res
ults
f
or
va
riables
E
uclid
ean
a
nd
Ci
ty
bl
ock
a
re
100%
for
al
l
cl
asses.
That
m
eans
the
relat
ion
s
hip
bet
wee
n
the
ta
r
get
a
nd
the
ou
t
pu
t
a
r
e
li
near
pro
por
ti
on
al
.
Howe
ver,
th
e
var
ia
ble
Cheb
yshev
s
hows
t
he
res
ults
f
or
c
la
ss
1
an
d
cl
as
s
3
are
hundre
d
pe
rce
nt
li
near
.
T
hen,
the
va
riable
Chebys
hev
s
ho
ws
the
cl
ass
2
resu
lt
is
94.69
%
can
be
ex
pl
ai
ned
by
the
li
near
re
gressi
on
.
The
total
res
ults
of
K
-
NN
a
re
100%
,
100%
,
a
nd
98.23%
f
or
vari
ables
Eucli
dea
n,
Ci
ty
blo
ck,
and
Ch
ebys
he
v,
res
pecti
vely
.
The
var
ia
bles
Eucl
idean
an
d
Ci
ty
blo
ck
sho
w
the
resu
lt
s
are
si
m
il
ar
to
ea
ch
oth
e
r
an
d
represe
nt
as
t
he
best
perform
ance
in
analy
zi
ng
the
pro
blem
s
of
NTEV
du
e
t
o
th
e
har
m
on
ic
,
l
oos
e
te
rm
inati
on
,
a
nd
li
ghtni
ng
in
th
e
com
m
ercial
bu
il
din
g.
H
owev
er,
the
Che
bys
hev
resu
lt
is
sti
ll
dep
endable,
since
the
relat
ion
s
hip
betwee
n
the
ta
rg
et
a
nd the
ou
t
pu
t a
re cl
ose
to
the
hu
ndre
d perce
nt.
Table
1
. Per
for
m
ance of
K
-
N
N b
ase
d on the
d
if
fer
e
nt
var
ia
bles
Variables
Clas
s 1
(
%)
Clas
s 2
(
%)
Clas
s 3
(
%)
Total (%
)
Euclid
ean
100
100
100
100
City
b
lo
ck
100
100
100
100
Ch
eb
y
sh
ev
100
9
4
.69
100
9
8
.23
4.
CONCL
US
I
O
N
This
pa
per
is
pr
ese
nted
t
o
cl
assify
the
NT
EV
pr
ob
le
m
in
the
com
m
erci
al
bu
il
ding
by
us
in
g
the
K
-
NN
cl
assifi
er
t
oo
ls
.
T
he
K
-
NN
cl
assifi
er
t
oo
l
us
es
the
di
ff
e
ren
t
c
onti
nuou
s
va
riables
,
wh
ic
h
a
re
E
uc
li
dean
,
Ci
ty
blo
ck,
an
d
Che
bysh
e
v
.
The
ST
is
ap
plied
as
the
pr
opos
e
d
te
ch
nique
to
ext
ract
the
featu
res
of
NTE
V
pro
blem
,
wh
ic
h
is
then
util
iz
ed
as
cl
assifi
cat
ion
input.
T
he
pe
rfor
m
ance
of
K
-
N
N
cl
assifi
er
with
di
f
fer
e
nt
var
ia
bles
is
observ
e
d
ba
sed
on
the
co
nfusi
on
m
at
rix
and
li
near
re
gr
e
ssio
n
res
ults.
The
Eucli
dea
n
a
nd
Ci
t
y
blo
c
k
va
riables
p
rese
nted
t
he best res
ult t
o
cl
assify
the N
TE
V
pro
blem
s in
the co
m
m
ercia
l bu
il
ding c
ompare
d
to
the
Che
bys
he
v
va
riable.
Ev
en
th
ough
the
perform
ance
of
the
Ch
ebys
he
v
va
riable
is
lo
wer
c
om
par
ed
to
the
oth
e
r
va
riables
,
it
is
still
re
li
a
ble
as
the
co
nfusion
m
at
rix
sh
ows
m
ino
r
m
i
scl
assifi
cat
ion
wh
ic
h
is
2.4%.
Then,
the
li
near
re
gressi
on
s
hows
t
hat
the
per
ce
nt
age
of
Chebys
hev
resu
lt
cl
ose
to
per
fect
va
lue
wh
ic
h
is
hund
red
per
ce
nt.
ACKN
OWLE
DGE
MENTS
This
wor
k was
su
pp
or
te
d
in
part by
U
niv
er
sit
i Teknolo
gi M
ara (UiTM
)
un
der the sc
hola
r
sh
ip
.
REFERE
NCE
S
[1]
D.
Dorr,
“
Dete
r
m
ini
ng
volt
ag
e
l
eve
ls
of
concern
for
hum
an
and
ani
m
al
r
esponse
to
AC
cur
r
ent,”
in
Pow
er
&
Ene
rgy
So
ciety
Gene
ral M
e
eting
,
2
009
.
P
ES
’09
.
IEE
E
,
2009
,
pp
.
1
–
6.
[2]
S.
Ravl
ić
and
A.
Marušić
,
“
Sim
ula
ti
on
Model
s
for
Vari
ous
Neutr
al
E
art
hin
g
Methods
in
Medium
Volta
ge
S
y
stems
,
”
Procedia
Eng
.
,
vol
.
10
0,
pp
.
1182
–
119
1,
2015
.
[3]
D.
Sall
es
et
al
.
,
“
As
sess
ing
the
Coll
ectiv
e
Harm
onic
Im
pa
ct
of
Modern
Reside
ntial
Lo
a
ds
-
2014;Part
I
:
Methodol
og
y
,
”
Powe
r Del
iv.
I
E
EE
Tr
ans.
,
vol.
27,
no
.
4
,
pp
.
19
37
–
1946,
Oct
.
2
012.
[4]
A.
M.
Vari
z,
J.
L.
.
Pere
ir
a,
S.
Carne
iro,
and
P.
G.
Barbosa
,
“
Harm
onic
ana
l
y
sis
of
the
power
distri
buti
on
Neutr
al
-
to
-
Earth
Vo
lt
ag
e
(NEV)
te
st
ca
se
using
four
-
wire
thr
ee
-
phase
har
m
onic
cur
ren
t
injection
m
et
hod,
”
in
IEE
E
Powe
r
&
Ene
rgy
So
ciety
Gene
ral M
e
eting
,
2009
,
pp
.
1
–
7.
[5]
J.
D.
Bouford
and
D.
J.
Hurle
y
,
“
A
Rec
om
m
ende
d
Standa
rd
for
V
olt
ag
es
Appea
ri
ng
Across
Public
l
y
Ac
ce
ss
ib
l
e
Surfac
es,
”
IE
EE
Tr
ans.
Powe
r Del
i
v.
,
vol
.
29
,
no
.
1,
pp.
48
–
51,
Fe
b.
2014
.
[6]
M.
A.
T.
M.
Yusoh,
A.
F.
Abidin,
Z.
M.
Yasin,
M.
F.
Ghani,
and
U.
Mat,
“
Modell
ing
Neutr
al
to
E
art
h
Volta
g
e
(
NTEV
)
on
the
Com
m
erc
ia
l
Bui
ldi
ng,
”
2016
IEEE
Conf
.
S
yst. P
r
oce
ss
Control
,
p
p.
91
–
96
,
2016
.
[7]
C.
Jiang
,
Q.
Shi,
Y.
Ti
an
,
X
.
L
i
,
S.
Li
n,
and
Y.
Li
u
,
“
Sensiti
vi
t
y
stud
y
on
th
e
stra
y
voltage
of
low
voltage
reside
nt
ia
l
n
et
w
orks,”
2016
IEEE
8th
Int.
Pow
er
El
ectron.
Moti
o
n
Control
Conf.
IPE
MC
-
ECC
E
Asia
2016
,
pp.
2571
–
2576,
201
6.
[8]
K.
Thi
rum
ala,
A.
C.
Um
ari
kar
,
a
nd
T.
Ja
in,
“
Estim
at
ion
of
singl
e
-
phase
and
three
-
phase
power
-
qu
al
ity
indices
using e
m
piri
c
al
wave
let
tr
ansfor
m
,
”
IE
EE
Tr
ans.
Powe
r De
liv.
,
v
ol.
30
,
no
.
1
,
pp
.
445
–
454,
2015
.
[9]
A.
As
uhai
m
i
Mo
hd
Zi
n,
M.
Saini,
M.
W
.
Mus
ta
fa,
A.
R.
Sulta
n,
a
nd
Rahi
m
uddin,
“
New
al
gorit
hm
for
det
ection
and
fau
lt
cl
assifi
ca
t
ion
on
p
ara
l
lel
tra
nsm
ission
line
using
DW
T
a
nd
BP
NN
base
d
on
Cla
rk
e’s
tr
a
nsform
at
ion,
”
Neurocomputi
ng
,
vol
.
168
,
pp
.
98
3
–
993,
2015
.
[10]
O.
Ozgone
nel,
T
.
Yalcin,
I
.
Gune
y
,
and
U.
Kurt
,
“
A
new
cl
assific
ation
for
power
qual
ity
eve
n
ts
in
distri
buti
o
n
s
y
stems
,
”
Elec
tr.
Powe
r S
yst. Res.
,
vol. 95, pp. 19
2
–
199,
2013
.
[11]
J.
H.
A.
Mora
les
,
E.
A.
Orduña
,
and
C.
Rehtanz
,
“
Cla
ss
ifi
c
at
ion
of
li
gh
tni
ng
str
oke
on
tr
ansm
ission
li
n
e
using
m
ult
i
-
resolut
ion
ana
l
y
sis
and
m
a
chi
ne
learni
ng
,
”
El
e
ct
r.
Powe
r
E
nergy
Syst
.
,
vol
.
58,
pp
.
19
–
31
,
2
014.
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.
2
,
Fe
br
uary
201
8
:
3
80
–
386
386
[12]
A.
Rodriguez,
J.
A.
Aguado
,
F.
Marti
n,
J.
J.
Lo
pez
,
F.
Munoz
,
and
J.
E.
Rui
z,
“
Rule
-
base
d
cl
a
ss
ifi
ca
ti
on
of
power
qualit
y
di
sturbanc
es
using
S
-
tra
nsform
,
”
E
le
c
tr. Powe
r Sy
st
.
R
es.
,
vol
.
86
,
p
p.
113
–
121
,
201
2.
[13]
A.
F.
Abidin
,
A.
Moham
ed,
and
H.
Share
e
f,
“
Int
el
li
g
ent
detec
t
io
n
of
unstab
le
po
wer
sw
ing
for
c
orre
ct
dista
n
c
e
rel
a
y
oper
at
ion
using
S
-
tra
nsfor
m
and
neur
al
n
e
tworks,”
E
xpe
rt
Syst.
App
l.
,
vol.
38,
no.
12,
pp.
14969
–
14975,
2011.
[14]
C.
DeNardo,
“
Stra
y
and
cont
a
ct
volt
age
an
upd
at
e
on
I
EE
E
PE
S
P1695
W
orki
ng
Group
ac
ti
v
ities,”
in
Rural
El
e
ct
ric
Pow
er
Confe
renc
e
,
201
2,
pp
.
A3
-
1
-
A3
-
7.
[15]
N.
Huang,
D.
Xu,
X.
Li
u,
and
L.
Li
n
,
“
Pow
er
qual
ity
d
isturb
anc
es
c
la
ss
ifi
c
ation
base
d
on
S
-
tra
nsform
and
proba
bil
ist
ic ne
u
ral
n
et
work,
”
N
e
urocomputing
,
v
ol.
98
,
pp
.
12
–
23
,
2012
.
[16]
M.
A.
Jabba
r
,
B
.
L
.
Dee
ksha
tul
u
,
and
P.
Ch
andr
a,
“
Cla
ss
ifica
t
io
n
of
Hea
rt
Dise
a
se
Us
ing
K
-
Nea
rest
Neighbor
and
Gene
tic
Alg
o
rit
hm
,
”
Proce
d
i
a
Technol.
,
vol
.
10,
pp
.
85
–
94
,
2
013.
[17]
A.
Kum
ar,
R.
Bhat
nag
ar,
and
S.
Srivasta
va
,
“
ARS
kNN
-
A
k
-
NN
c
la
ss
ifi
er
using
m
ass
base
d
sim
il
ari
t
y
m
ea
sure,”
Proce
dia
Comput.
S
ci.
,
vol. 46, n
o.
Ic
ict
2014
,
pp
.
457
–
462,
2015.
[18]
R.
A.
Nugraha
en
i
and
K.
Mut
ij
ar
sa,
“
Com
par
at
iv
e
anal
y
s
is
of
m
ac
hine
le
a
rning
K
NN
,
SV
M,
and
ran
dom
fore
sts
al
gorit
hm
for
fa
ci
a
l
expr
ession
c
la
ss
ifi
c
at
ion
,
”
in
2016
Inte
rnatio
nal
Seminar
on
Appl
ic
a
ti
on
for
Technol
ogy
o
f
Information
and
Comm
unic
ati
on
(
ISe
mant
ic)
,
2016,
pp
.
163
–
168
.
Evaluation Warning : The document was created with Spire.PDF for Python.