Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
4, N
o
. 4
,
A
ugu
st
2014
, pp
. 46
3
~
47
0
I
S
SN
: 208
8-8
7
0
8
4
63
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
ANFIS Based Approach to Esti
ma
te
Re
mna
n
t
Life
o
f
Po
we
r
Transformer by Predicting Furan Contents
Priyesh
Kum
a
r Pande
y, Z
a
kir Hus
a
in,
R.K.
Jari
a
l
Departem
ent
of
Ele
c
tri
cal
Eng
i
n
eering
,
Na
tiona
l
Institute
of
Te
ch
nolog
y Ham
i
rpu
r
,
Him
achal Prades
h,
India
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Ja
n
3, 2014
Rev
i
sed
Jun
11,
201
4
Accepted
Jun 25, 2014
A Condition mo
nitoring
and diagnostic pl
ay
ing important
role in
estimating
remnant Lif
e
o
f
power transf
ormer. Concen
tration of
furan
conten
t in
transformer oil
can b
e
a promising fact
or for
in
direct measurement of
the
aging of
tr
ansformer insulation. The oi
l
ge
ts
c
ontam
inat
ed m
a
inl
y
du
e t
o
prolonged oper
a
tion
and ag
ein
g
. The present resear
ch pap
e
r
introduces
Adaptive Neuro
F
u
zz
y
infer
e
nc
e s
y
s
t
em
(ANF
IS
) techniqu
e t
o
correl
a
t
e
furanic compounds
obtained b
y
High
Performance Liquid Chro
matograp
h
y
(HPLC) test to
rem
n
ant lif
e o
f
the power tr
a
n
sform
e
r. The
results are
obtain
e
d b
y
con
ducting HPLC test at
TIFAC-CORE lab, NIT Hamirpur on
thirte
en sam
p
le
s of Power transform
e
r oil ta
ken from
Himacha
l
Stat
e
Ele
c
tri
c
it
y
Boar
d Ltd
,
Shim
la
(I
ndia).
Keyword:
Ada
p
tive Ne
uro
Fuzzy
Degree of
polymerization
Fu
ran co
m
p
oun
d
s
In
fere
nce sy
ste
m
(ANFI
S
)
Li
qui
d C
h
r
o
m
a
t
o
g
r
ap
hy
Re
m
n
an
t life
Tran
sf
orm
e
r oi
l
Copyright ©
201
4 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
:
Zakir Husain
Depa
rtem
ent of Elect
ri
cal
E
n
gi
nee
r
i
n
g,
Natio
n
a
l
In
stitu
te of Techn
o
l
o
g
y
Ham
i
rp
u
r
,
Ham
i
rpur (H.P), Him
achal
Prades
h, India
Em
a
il: zah
u
s
ain
2
@g
m
a
i
l
.co
m
1.
INTRODUCTION
Th
e so
lid
i
n
sulatin
g
m
a
terial
s u
s
ed
i
n
tran
sform
e
rs
are mo
stly m
a
d
e
u
p
o
f
cellu
lo
se
materials v
i
z.
i
n
sul
a
t
i
n
g
pap
e
rs,
p
r
ess
boa
rd
s, a
n
d
so
o
n
.
These i
n
s
u
l
a
t
i
n
g
m
a
t
e
ri
al
s
m
u
st
have
hi
gh
di
el
ect
ri
c st
ren
g
t
h
a
n
d
hi
g
h
t
e
nsi
l
e
st
ren
g
t
h
as
pe
r
IS/
I
EC
st
a
nda
rd i
n
res
p
ect
t
o
co
ncer
ne
d
KV
A cl
ass
of
po
wer t
r
an
sf
orm
e
r.
Ho
we
ver
,
t
h
es
e
m
a
t
e
ri
al
s get
grad
ual
l
y
de
gra
d
e
d
d
u
e
to electrical and therm
a
l s
t
ress that decreas
es the
Th
erm
a
l stress h
a
nd
ling
capab
ility an
d
the d
i
electric streng
th
o
f
t
h
e
cellu
lo
se m
a
te
rial in
a co
mp
o
s
ite
in
su
lation
syste
m
o
f
th
e transform
e
r. Du
e to th
e d
e
grad
ation
of cellu
lo
sic
material in
th
e form
o
f
weaken
ing
o
f
in
term
o
l
ecular b
o
n
d
i
ng
, th
ere is sub
s
tan
tial d
ecrease in
th
e ten
s
ile
stren
g
t
h
o
f
t
h
e in
su
lation
syste
m
co
m
p
o
s
ed of t
h
e cellu
l
o
sic
material. Th
e t
r
an
sfo
r
m
e
r m
a
y g
e
ts
failed
up
on
witn
essi
n
g
rated or
b
e
low
rated
sh
ort-circu
it fo
rce [1
,
2
]
.
Thu
s
u
lti
m
a
tel
y
, th
e life
o
f
trans
f
orm
e
r is seve
rely affected by t
h
is
wea
k
in
term
o
l
ecu
lar b
ond
ing
o
f
co
m
p
o
s
ite in
su
l
a
tio
n
system
material. Wh
en
degrad
ation o
f
p
a
p
e
r in
su
lation
occurs, t
h
e cel
lulose m
o
lecular chai
ns
gets
shorter a
n
d che
m
ical product
s
suc
h
as
fura
nic com
p
ounds are
p
r
od
u
c
ed
and
g
e
t d
i
sso
l
v
ed
i
n
th
e o
il. Oil-i
m
p
r
eg
n
a
ted
cellu
lo
se is d
e
grad
ed
d
u
e
to
three
m
a
in
mech
an
ism
s
;
hy
d
r
ol
y
t
i
c
deg
r
adat
i
o
n,
oxi
da
t
i
v
e deg
r
adat
i
o
n (py
r
ol
y
s
i
s
mech
an
ism
)
an
d
th
erm
a
l d
e
g
r
adatio
n
.
Con
c
en
tratio
n
of
f
u
ra
n c
o
nt
en
t
i
n
t
r
a
n
sf
orm
e
r
oi
l
reve
al
s an
i
ndi
rect
m
easurem
ent
of t
h
e a
g
i
n
g
of
t
r
an
sf
o
r
m
e
r. Fu
ran
co
nt
ent
measu
r
em
en
t co
n
s
ists
o
f
m
e
a
s
u
r
i
n
g fiv
e
fu
rfu
rals
d
i
sso
l
v
e
d in
tran
sfo
r
m
e
r o
il th
at d
i
ffer
in
th
eir stab
ility i.e.
2-
fu
rf
u
r
al (
2
-F
AL)
,
2
-
F
u
r
f
u
r
o
l (
2
-F
OL
),
5
-
Hy
d
r
oxy
m
e
thy
l
-2
-f
ur
f
u
ral
(5
-HM
F
),
5
-
M
e
thy
l
-2-
F
u
r
f
u
r
a
l (5
-
M
E
F) a
n
d
2
-
Acet
y
l
ene (
2
-
A
C
F
)
.
Gene
ra
t
i
on
of
t
h
i
s
f
u
ra
ni
c c
o
m
poun
d
de
pen
d
s
up
o
n
t
h
e
l
e
v
e
l
of
d
e
terioration
of
cellu
lo
sic
in
su
latio
n
system
[3, 4
,
5
,
6, 7
,
8].
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
JECE Vo
l. 4
,
N
o
. 4
,
Au
gu
st 2
014
:
46
3
–
47
0
46
4
Fur
a
ns a
r
e t
h
e
m
a
jor
pr
o
d
u
c
t
s
whi
c
h em
anat
es fr
om
degra
d
at
i
on o
f
pape
r
i
n
sul
a
t
i
o
n i
n
t
r
ansf
o
r
m
e
r
oil and the m
o
st prom
inent com
pone
nt of pa
per de
com
position in oil is 2-FAL. Inc
r
ease in Fura
n
conce
n
t
r
at
i
o
n i
n
t
r
ans
f
orm
e
r oi
l
i
ndi
cat
es d
ecreases i
n
t
h
e
t
e
nsi
l
e
st
rengt
h an
d t
h
e de
gr
ee of p
o
l
y
m
e
rizat
i
on
(DP) of th
e p
a
p
e
r. Furan
level in
a tran
sformer can
b
e
correlated with
paper DP, and there
b
y, an in-s
ervice
assessm
ent
of t
h
e
m
echani
c
al
st
rengt
h o
f
t
h
e pape
r i
n
s
u
l
a
t
i
on ca
n be
est
i
m
a
t
e
d. D
e
Pabl
o
rep
o
rt
ed t
h
e
rel
a
t
i
on bet
w
ee
n 2-F
A
L
an
d d
e
gree
o
f
pol
y
m
eri
zat
i
on (D
P
V
)
based
o
n
vi
sc
osi
t
y
of
oi
l
as
[
9
]
,
7100
8.88
2
V
DP
FAL
(1
)
Whe
r
e,
2
-
F
A
L
i
s
t
h
e f
u
r
f
u
ral
conce
n
t
r
at
i
o
n i
n
m
g
/
kg
or
p
p
m
of oi
l
.
Tabl
e
1 s
h
o
w
s t
h
e c
o
r
r
el
at
i
on
bet
w
een
2-
fu
ralde
h
y
d
e
(2
-FA
L
) a
nd
DP
with res
p
ect to the s
o
lid
i
n
s
u
l
a
t
i
on e
x
t
e
nt
of
dam
a
ge dep
i
ct
i
ng si
g
n
i
f
i
cant
o
n
pape
r i
n
sulation
dielectric an
d m
echanical properties.
Tabl
e 1.
C
o
r
r
el
at
i
on bet
w
ee
n 2-F
A
L
an
d DP
.
2-
FAL
(p
p
m
)
DP value
Health condition
0-0.1 1200-7
0
0
Healthy
Insulation
0.
1-
1.
0 700-
45
0
M
oder
a
te
Deter
i
oration
1-
10
450-
25
0
E
x
tensive
Deter
ioration
>10
<250
E
nd of life
New
p
a
per und
er
n
o
rm
al ag
i
n
g
con
d
ition
produ
ces Furfural at a rate o
f
1
.
7
m
g
/
g
o
f
p
a
p
e
r /h
and
as
t
h
e agi
n
g
o
f
t
r
ans
f
orm
e
r oc
curs
t
h
i
s
rat
e
of
ge
nerat
i
o
n
i
s
su
bst
a
nt
i
a
l
l
y
i
n
crease
d
b
y
05m
g/
g
of
pape
r,
i
ndi
cat
i
n
g e
x
t
e
nsi
v
e
de
gree
o
f
det
e
ri
orat
i
o
n a
n
d
en
d
o
f
l
i
f
e a
p
p
r
oxi
m
a
t
e
ly
15 t
o
2
0
y
e
a
r
s
[
12]
.
Wh
en
2
-
FAL
v
a
lu
e is greater th
an
1
0
and
DP test rev
eals a v
a
lu
e less than
250
, th
e in
su
latin
g
p
a
per
is con
s
id
ered to
h
a
v
e
lo
st all
its
m
ech
an
ical streng
th
, and th
e tran
sformer life
has
rea
c
hed
i
t
s
en
d
o
f
l
i
f
e
cri
t
e
ri
a. Fu
ran
s
conce
n
t
r
at
i
o
n i
s
m
easured
by
Hi
g
h
-Pe
r
f
o
rm
ance Li
qui
d C
h
r
o
m
a
t
ogr
aphy
(
H
PLC
)
or
Ga
s
C
h
r
o
m
a
t
ograp
hy
-M
ass S
p
ect
rom
e
t
r
y
(GC
/
M
S
) base
d
o
n
Am
erican Soci
ety for Testi
n
g and Materials
(ASTM
D5
8
3
7
,
Sta
nda
rd
Sp
ecificatio
ns
fo
r M
i
ne
ra
l
In
su
latin
g Oil i
n
Electrical
App
a
rat
u
s).
This researc
h
pape
r presents
a
novel
approa
ch fo
r estim
ati
n
g th
e rem
n
an
t
life
o
f
po
wer
tran
sform
e
r
u
s
ing
fu
ran
con
cen
t
r
atio
n
i
n
tran
sform
e
r o
il. Resu
lts
show that there is
a good
co
rrel
a
t
i
on bet
w
een
f
u
ra
n
co
n
c
en
tration
lev
e
l in
o
il tran
sfo
r
m
e
r an
d
its esti
mate
d li
fe. The rest
o
f
t
h
e pape
r ha
s been o
r
ga
ni
z
e
d as
f
o
llow
s
: Sectio
n
I
I
d
i
scu
sses th
e f
u
r
a
n
analysis an
d
th
e test co
n
d
u
c
ted o
n
th
e tr
an
sfo
r
m
e
r
o
il.
Sectio
n
III
di
scuss
e
s p
r
o
p
o
se
d AN
FI
S m
odel
t
o
est
i
m
ate rem
n
ant
l
i
f
e
of t
r
a
n
sf
o
r
m
e
r
s
. R
e
sul
t
s
and
di
scussi
o
n
are gi
v
e
n
in
section
IV an
d fi
n
a
lly co
n
c
lu
sion
s are
d
r
awn in
sectio
n
V
o
f
t
h
e
p
a
p
e
r.
2.
FU
RAN
A
NALY
S
IS
BY
HPLC
Fu
ran
is
u
s
ed to
d
e
term
in
e
th
e d
e
grad
ati
o
n
of th
e
so
li
d
in
su
latio
n lik
e
p
a
p
e
r,
p
r
essb
oard
etc.
Acco
r
d
i
n
g t
o
ASTM
D
-
5
8
3
7
t
h
e
fu
ra
n a
n
al
y
s
i
s
i
s
pe
rf
orm
e
d
by
t
h
e
Hi
g
h
Per
f
o
r
m
a
nce Li
q
u
i
d
C
h
r
o
m
a
t
ograp
hy
(
H
PLC
)
. T
h
e l
a
b
set
u
p
f
o
r
pe
rf
orm
i
ng
Hi
g
h
Pe
rf
orm
a
nce Li
qui
d C
h
r
o
m
a
t
ograp
hy
(
H
PLC
)
to
estim
a
t
e fu
ran
co
n
t
en
ts is sh
own
in Figure 1
.
Fi
gu
re
1.
Lab
s
e
t
up
f
o
r
H
P
L
C
t
o
fi
nd
f
u
ra
n
com
pou
n
d
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
ANFIS B
a
sed
Approac
h
to E
s
timate
Remnan
t Life o
f
Po
wer Tra
n
s
f
o
rmer
b
y
Pred
icting
… (Za
k
ir Hu
sain
)
46
5
HPLC
has m
o
bi
l
e
phase an
d
st
at
i
onary
pha
se and t
h
e det
ect
or use
d
t
o
d
e
t
ect
furan
s
i
s
UV
vi
si
bl
e
detector.
T
h
e detector detect
s
speci
fi
c
wa
v
e
l
e
ngt
h
an
d
p
eak a
b
so
r
b
anc
e
of
t
r
an
sf
orm
e
r oi
l
.
Fu
ran
c
ont
e
n
t
measu
r
em
en
t co
nsists o
f
m
e
asu
r
i
n
g
fiv
e
furfu
r
als
d
i
sso
l
v
ed
in
tran
sformer o
il th
at d
i
ffer in
their st
ab
ility
.
Fu
rfu
r
a
ls ar
e co
m
p
o
s
ed
o
f
2-f
u
r
a
ld
eh
yd
e,
5-
h
y
d
r
o
x
y
m
e
th
yl-
2
-f
ur
ald
e
h
y
d
e
, 2-
fur
o
ic acid
, fu
rf
ur
yl alco
ho
l, 5-
m
e
t
h
y
l
-2-f
ural
dehy
de
an
d
2
-
a
cet
y
l
fu
ra
n c
o
m
poun
ds.
2
-fu
ral
d
e
h
y
d
e
(
2
-F
AL)
i
s
t
h
e
m
o
st
st
abl
e
com
p
o
u
n
d
f
o
r
m
easurem
ent
;
ho
we
ver
,
al
l
ot
her
fu
ran c
o
m
pou
n
d
s, e
x
cept
2
-fu
rf
u
r
y
l
al
cohol
, a
r
e det
ect
e
d
i
n
i
n
s
u
l
a
t
i
ng
oi
l
up
to
140
̊
C.
Each
fura
nic
com
pound ha
s
a cha
r
acteris
tic
m
a
x
i
m
u
m lig
h
t
ab
sorb
an
ce
o
ccurri
n
g
with
in th
e
i
ndi
cat
ed ra
n
g
e
s of wa
vel
e
n
g
t
h
s
.
To det
ect
t
h
e speci
fi
c w
a
vel
e
n
g
t
h
f
o
r
f
u
ra
ni
c com
pou
nds
, use
of v
a
r
i
abl
e
wavel
e
ngt
h
or
di
o
d
e a
rray
d
e
t
ect
or i
s
nec
e
ssary
. T
h
e s
p
e
c
i
f
i
c
wavel
e
n
g
t
h i
s
sel
ect
ed
t
o
y
i
el
d m
a
xim
u
m
abs
o
rbance
for each c
o
m
pound. Res
u
lts s
h
ow that a
b
s
o
rba
n
ce a
n
d
wa
vel
e
ngt
h inc
r
ease
s
by a si
gnificant a
nd
easily obse
r
va
ble m
a
rgin as
deterioration a
n
d contam
inat
ions
of
t
r
an
sf
or
m
e
r oi
l
i
n
c
r
eas
es, whic
h
a
r
e reflected
b
y
th
e
fu
ran
con
cen
t
r
atio
n level in
th
e
o
il.
The e
x
peri
m
e
nt
ha
s
been
pe
rf
orm
e
d o
n
ne
w, m
e
di
um
and
hi
g
h
l
y
age
d
t
r
ans
f
o
r
m
e
r oi
l
.
I
f
t
h
e
oi
l
w
h
ich
is n
o
t
in co
mmissio
n
o
r
o
f
ab
ou
t 1
year
ag
e and
if
the f
u
r
a
n
con
t
en
t
is v
e
r
y
lo
w
in
r
a
ng
e 0-
0.1ppm th
en
it is categ
o
r
ized
as
n
e
w
o
il. Th
e
o
il wh
ich
is tested
af
ter 6 to
7
years
and
t
h
e furan con
t
en
t is in
ran
g
e
1-7pp
m
th
an
it is categ
o
r
ized
as m
e
d
i
u
m
ag
ed
o
il. Fin
a
lly, It h
a
s b
e
en
fo
und
th
at in
case, th
e
o
il sam
p
le
tested
after
9
to
10
years
o
r
m
o
re an
d fu
ran
co
n
t
en
ts are g
r
eater th
an 10
pp
m
it is c
a
teg
o
rized
as
h
i
gh
ly ag
ed
an
d it is
requ
ired
to b
e
m
o
n
ito
red
.
3.
TR
AN
SFORM
E
R LIFE ES
TIMATION TECHNIQUE
In t
h
e p
r
ese
n
t
wo
rk
,
Ada
p
tiv
e neu
r
o-
fuzzy
i
n
fe
rence
sy
stem
(AN
F
IS
)
ba
sed tech
n
i
qu
e i
s
p
r
op
osed
t
o
esti
m
a
te
th
e re
m
n
an
t life o
f
p
o
w
er transfo
r
m
e
rs an
d
its co
rrelation
to th
e fu
ran
con
t
en
ts presen
t
in
th
e
t
r
ans
f
o
r
m
e
r oi
l obt
ai
ne
d by
H
P
LC
spect
ral
r
e
sp
onse
.
The
d
e
t
a
i
l
of t
h
e t
echni
que
use
d
i
s
gi
ve
n bel
o
w a
nd t
h
e
resu
lts are
d
i
scu
ssed
i
n
later section
.
In ANFIS sug
e
n
o
typ
e
fu
zzy
lo
g
i
c m
e
th
od
is
u
s
ed
which
is,
co
m
p
u
t
atio
n
a
lly effectiv
e and work
s
well with
o
p
tim
isa
tio
n
and
ad
ap
tiv
e tech
n
i
qu
es.
Wav
e
leng
th
(n
m)
and
the m
a
xim
u
m abs
o
rption (nm
)
are th
e i
n
put
s
f
o
r t
h
e
AN
FIS
m
odel
,
w
h
i
c
h a
r
e
r
ecor
d
e
d
f
o
r
d
i
ffere
nt
t
r
ans
f
o
r
m
e
r oi
l
sam
p
l
e
s. The
cor
r
es
po
n
d
i
n
g
AN
FIS
si
m
u
l
a
ti
on m
odel
de
v
e
l
ope
d i
s
s
h
ow
n i
n
Fi
g
u
r
e
2.
Fi
gu
re 2.
A
N
F
I
S net
w
or
k [
1
0
]
The o
u
t
p
ut
o
f
t
h
e m
odel
i
s
t
h
e fura
n c
ont
e
n
t
s
and c
o
r
r
es
po
ndi
ng est
i
m
at
ed rem
n
ant
l
i
f
e. The m
odel
is built using t
h
e gra
phical user interface tool
provide
d
by MATLAB
[11]. Each i
n
put
is fuzzi
fi
ed
into
ten
set
s
[
12]
of
Ga
ussi
an
com
b
i
n
at
i
on m
e
m
b
ershi
p
f
unct
i
o
n
(
M
F)
go
ver
n
e
d
by
t
h
e e
q
uat
i
o
n
(2
)
1
(,
,
,
)
e
x
p
2
m
xc
fx
c
s
m
s
(2)
whe
r
e
‘c’
is ce
ntre,
‘s
’ is
wi
dth a
n
d ‘m
’ is F
u
zzificatio
n
fa
ct
or.
T
h
e
fu
nct
i
on
set
devel
o
p
e
d
fo
r eac
h
i
n
p
u
t
i
s
a
com
b
i
n
at
i
on
of
s a
n
d
c
param
e
t
e
rs.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 4
,
N
o
. 4
,
Au
gu
st 2
014
:
46
3
–
47
0
46
6
Fi
gu
re
3.
A
N
F
I
S m
odel
t
o
est
i
m
a
t
e
fura
n c
o
nt
en
ts and
rem
n
an
t life of
power transfo
r
m
e
rs.
Resu
lt o
f
fu
zzi
fi
cation
from
each input is then applied
w
ith fuzzy
operator in
the a
n
t
ecedent a
nd
rel
a
t
e
d t
o
t
h
e
con
s
eq
ue
nce,
by
I
F
-
A
N
D
-
T
HEN
a
ppl
i
cat
i
o
n
m
e
t
hod
[1
3
-
1
4
]
.
T
h
e
o
u
t
p
ut
o
f
s
u
gen
o
m
odel
i
s
obt
ai
ne
d
by
t
h
e wei
g
ht
ed
av
e
r
age
(
W
A
)
of
t
h
e si
ngl
et
o
n
s i
n
t
h
e
o
u
t
p
ut
s.
4.
R
E
SU
LTS AN
D ANA
LY
SIS
The
di
scussi
on
can
be m
a
de i
n
se
veral
A t
o
t
a
l
num
ber o
f
t
h
i
r
t
een
p
o
we
r t
r
ans
f
orm
e
r oi
l
sam
p
l
e
s are
taken
from
Himachal State
Electricity
Board
Ltd,
S
h
im
la (I
n
d
ia)
f
o
r t
h
e
researc
h
a
n
d the
stu
d
y
pu
rp
ose
.
These tra
n
sformers are of
differe
n
t
m
a
nufac
turing com
p
ani
e
s’ viz.
Gene
ra
l Electric Co
mpany, C
r
om
pton and
Grea
ves a
n
d C
o
m
put
i
ng Ta
b
u
l
a
t
i
ng R
e
c
o
r
d
i
ng
(C
TR
)
ha
v
i
ng
di
f
f
ere
n
t
i
n
st
al
l
a
t
i
on y
ear,
M
VA
rat
i
n
gs
vari
es
fr
om
5-5
0
M
V
A, di
f
f
ere
n
t
K
V
rat
i
ngs
vari
e
s
6
.
6
-
1
3
3
KV
. Usi
n
g Hi
g
h
Pe
rf
orm
a
nce
Li
q
u
i
d
C
h
r
o
m
a
t
o
g
r
ap
hy
(HPLC) th
e tran
sfo
r
m
e
r o
il sam
p
les were tested
at TI
FAC
-
C
O
R
E
l
a
b, N
I
T Ham
i
rpur (
H
.P) a
nd t
h
e
o
b
t
a
i
n
ed
resul
t
s
are gi
v
e
n i
n
Tabl
e 2.
Whe
n
t
h
e fu
r
a
n co
nt
ent
of th
e tran
sform
e
r o
il is
m
o
re
th
an
10p
p
m
o
r
th
e
wav
e
len
g
t
h
is
m
o
re th
an
43
0n
m
th
at
m
e
an
s tran
sf
o
r
m
e
r has
en
d
hi
s l
i
f
e cr
iteria.
The ANFIS m
odel
devel
ope
d
usi
n
g Matlab
gra
phical user i
n
fe
re
nce is
with t
h
e
accuracy
of more
tha
n
98%
.
Tabl
e
2. R
e
s
u
l
t
o
f
H
P
LC
a
n
d
AN
FIS
m
odel
Tran
.
No.
W
a
velength
(n
m
)
Absorbance
(n
m
)
Furan
(p
p
m
)
ANFIS
output
Cor
r
e
spondin
g
Re
m
n
ant Life
(yea
rs)
1 321.
31
1.
240
0.
00
0.
076
40.
0
2 334.
40
1.
280
0.
47
0.
393
36.
4
3 342.
80
1.
390
0.
97
0.
861
27.
4
4 359.
10
1.
440
1.
43
1.
120
15.
9
5 363.
80
1.
510
3.
12
3.
060
6.
75
6 388.
90
1.
620
4.
21
4.
160
3.
24
7 397.
80
1.
701
5.
33
5.
210
2.
5
8 410.
21
1.
721
8.
28
8.
25
1.
84
9 424.
58
1.
791
9.
22
9.
07
1.
26
10
442.
60
1.
810
13.
08
13.
3
0.
602
11
457.
20
1.
874
14.
01
14.
00
0.
438
12
460.
80
1.
901
15.
00
14.
80
0.
0
13
472.
10
1.
971
16.
78
16.
80
0.
0
Co
m
p
arin
g
th
e resu
lts ob
tain
ed
in
Tab
l
e 2
by th
o
s
e li
m
its
i
n
Tab
l
e
1
,
th
e t
r
an
sfo
r
m
e
r life esti
m
a
tio
n
can
b
e
o
b
t
ai
n
e
d
f
r
o
m
th
e w
a
velen
g
t
h of
t
h
e t
r
an
sfo
r
m
e
r
o
il
sp
ectr
a
l
r
e
sponse as sh
own
i
n
Tab
l
e
3
.
Tabl
e 3.
C
o
r
r
el
at
i
on bet
w
ee
n wavel
e
ngt
h
a
n
d DP
W
a
velength
(n
m
)
DP value
Significance on pa
per
insulation
290-
33
0
1200-
7
0
0
Healthy
I
n
sulation
330-
34
5
700-
45
0
M
oder
a
te
Deter
i
oration
345-
43
0
450-
25
0
E
x
tensive
Deter
i
oration
>430
<250
E
nd of life
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
ANFIS B
a
sed
Approac
h
to E
s
timate
Remnan
t Life o
f
Po
wer Tra
n
s
f
o
rmer
b
y
Pred
icting
… (Za
k
ir Hu
sain
)
46
7
The m
e
m
b
ersh
i
p
f
unct
i
o
n
pl
ot
f
o
r t
h
e i
n
p
u
t
wavel
e
n
g
t
h
a
nd a
b
s
o
r
b
a
n
ce
use
d
i
n
ANF
I
S
m
odel
i
s
sho
w
n i
n
Fi
gu
r
e
3 an
d Fi
g
u
r
e
4 res
p
ect
i
v
el
y
.
Al
l
t
h
e m
e
m
b
ershi
p
functions are fuzzified
in
to
ten
sets so th
at
the de
sired
out
put is
satisfied
according t
o
the input a
n
d
out
put. T
h
e rules
devel
ope
d t
o
predict furans
a
n
d to
esti
m
a
te re
m
n
an
t life is shown in
Figu
re
5
.
Th
e th
ree
di
m
e
nsi
onal
g
r
a
p
h
bet
w
een
w
a
vel
e
n
g
t
h
,
a
b
s
o
r
b
a
n
ce,
f
u
r
a
n con
t
en
ts
an
d r
e
m
n
an
t life is show
n in
Fig
u
r
e
7
an
d Figu
r
e
8 r
e
sp
ectively.
Fi
gu
re 4.
M
e
m
b
ers
h
i
p
f
u
nct
i
o
n pl
ot
fo
r wave
l
e
ngt
h
Fi
gu
re 5.
M
e
m
b
ers
h
i
p
f
u
nct
i
o
n pl
ot
fo
r
a
b
s
o
r
b
ance
.
Fi
gu
re
6.
R
u
l
e
s
de
vel
o
ped
t
o
f
i
nd t
h
e
fu
ran
c
ont
e
n
t
s
a
n
d
re
m
n
ant
l
i
f
e of
t
r
ansf
o
r
m
e
rs.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 4
,
N
o
. 4
,
Au
gu
st 2
014
:
46
3
–
47
0
46
8
Fi
gu
re
7.
3
D
g
r
ap
h
bet
w
ee
n
wavel
e
ngt
h, a
b
sor
b
a
n
ce a
n
d
f
u
ra
n c
o
nt
ent
s
o
f
t
r
a
n
sf
o
r
m
e
r.
Fi
gu
re 8.
3
D
g
r
ap
h bet
w
ee
n wavel
e
ngt
h,
a
b
sor
b
a
n
ce
a
n
d
r
e
m
n
ant
l
i
f
e of
po
we
r
t
r
a
n
sf
o
r
m
e
r.
5.
CO
NCL
USI
O
N
The pre
s
ent
re
search
pa
pe
r
i
n
t
r
od
uces a n
ovel
A
d
apt
i
ve
ne
ur
o f
u
zzy
i
n
fe
rence
sy
st
e
m
(ANFI
S
)
ap
pro
ach
t
o
correlate th
e esti
mated
tran
sformer rem
n
an
t
life and
t
h
e furan
resu
lts
with
t
h
e
h
e
lp
o
f
sp
ectru
m
an
alysis show
n in
Fi
g
u
r
e
8
.
Fi
gu
re 8.
R
e
l
a
t
i
on bet
w
ee
n Li
fe of
t
r
ans
f
orm
e
r
(
Y
ea
rs)
an
d Fur
a
n
co
nce
n
t
r
at
i
on Vs wa
vel
e
ngt
h (nm
)
‐
5
0
5
10
15
20
25
30
35
40
45
300
350
400
450
500
Life
of
transformer
(yea
rs)
Wavelen
g
th
(
n
m)
Life
of
transformer(years)
Furan
concent
r
at
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
ANFIS B
a
sed
Approac
h
to E
s
timate
Remnan
t Life o
f
Po
wer Tra
n
s
f
o
rmer
b
y
Pred
icting
… (Za
k
ir Hu
sain
)
46
9
It is fou
n
d
from co
m
p
arison
th
at all th
e resu
lts o
b
t
ai
n
e
d
fro
m
ANFIS mo
d
e
l is alm
o
st
eq
u
a
l to
t
h
e
r
e
su
lts
ob
tain
ed
b
y
H
P
LC. Th
e tr
an
sf
or
m
e
r
lif
e cr
iter
i
a en
d
s
w
h
en
t
h
e
w
a
v
e
len
g
t
h
is
m
o
r
e
th
an 430n
m
o
r
fu
ran c
o
nt
ent
s
i
n
creases m
o
re
t
h
an 1
0
ppm
. The m
odel
s
de
vel
o
ped
ha
ve a
ccuracy
o
f
m
o
re t
h
an 9
8
perc
ent
.
It
i
s
al
so
ob
vi
o
u
s
t
h
at
ap
p
r
oac
h
i
s
free
f
r
om
che
m
ical treatm
e
nt and
l
e
ss
t
i
m
e co
ns
um
i
ng.
ACKNOWLE
DGE
M
ENTS
The a
u
thors a
r
e thankful t
o
t
h
e Tec
h
nol
ogy
Info
rm
ation Forecasting
and Asses
s
m
e
nt Council and
C
e
nt
res
of R
e
l
e
vance
& E
x
c
e
l
l
e
nce (TI
F
A
C
-C
OR
E) at
N
I
T Ham
i
rpu
r
l
a
b, I
n
di
a f
o
r
p
r
o
v
i
d
i
ng t
h
e n
ecessary
facilities to
p
e
rfo
r
m
th
e ex
perim
e
n
t
s an
d
research
wo
rk
. Th
e au
t
h
ors
are also
th
ankfu
l
to
th
e
Hi
mach
al
Pra
d
esh Electri
city Board
(HPSEB),
I
ndi
a
f
o
r
p
r
o
v
i
d
i
n
g
t
h
e t
r
ans
f
orm
e
r oi
l
sam
p
l
e
s for e
x
peri
m
e
nt
.
REFERE
NC
ES
[1]
M
.
M
i
rzai
e, A.
Gholam
i and H. Ta
yyeb
i,
“
Thermal degra
dation of cellu
lose paper insulation in power
transformers
”, in proc. IEEE International Conf
e
r
ence
on S
o
l
i
d D
i
el
ectr
i
cs
,
ICS
D
'07.
2007
, pp
. 67
3-676.
[2]
R. D. Stebbins,
D. S. M
y
ers
an
d A. B
.
Shkolnik, “
Furanic com
pounds in dielectric
liquid
samples: Review and
update of diagn
ostic in
terpreta
tion
and estimatio
n of insula
tion a
g
eing
”, in
proc.
the 7th
International Conf
erence
on Properties an
d Applications
o
f
Diel
ect
ric
M
a
t
e
ria
l
s
,
2003
, pp
.
921-926 vol.3
.
[3]
M.
Ali,
C.
Eley,
A.
M.
Emsley,
R.
Hey
w
ood and X.
Xaio, “Measuring and un
derstanding th
e
ageing of kraft
insulating
pap
e
r
in power
transfo
r
m
e
rs”,
IEEE Electrical In
sula
tion Magazine
, vo
l. 12
, pp
. 28-34
,
1996.
[4]
M.
Ali,
C.
Eley,
A.
M.
Emsley,
R.
Hey
w
ood and X.
Xaio, “Measuring and un
derstanding th
e
ageing of kraft
insulating
pap
e
r
in power
transfo
r
m
e
rs”,
IEEE Electrical In
sula
tion Magazine
, vo
l. 12
, pp
. 28-34
,
1996.
[5]
R. D. Stebbins,
D. S. M
y
ers
an
d A. B
.
Shkolnik, “
Furanic com
pounds in dielectric
liquid
samples: Review and
update of diagn
ostic in
terpreta
tion
and estimatio
n of insula
tion a
g
eing
”, in
proc.
the 7th
International Conf
erence
on Properties an
d Applications o
f
Diel
ectric Materials, 2003
, pp
.
921-926 vol.3
.
[6]
T. K. Saha, “Review of m
odern dia
gnostic t
ech
niques for assessing insulati
on
condition in ag
ed transform
e
rs”
,
IEEE Transactio
ns on
Dielectrics
and Electrica
l
I
n
sulation,
vol. 1
0
, pp
. 903-917
,
2003.
[7]
A. M. Emsley
, X. Xiao, R. J. H
e
y
w
ood
and M.
Ali, “Degradatio
n
of cellu
lo
sic insulation in po
wer transformers.
Part 2: formatio
n of furan products in insulating oil”,
IEE Proceedings Science,
Measurement and Technology,
vo
l.
147, pp
. 110-11
4, 2000
.
[8]
M. Arshad, “
R
em
nant life Estim
ation
Model Using Fuzzy
log
i
c f
o
r Power Tr
a
n
sforme
r Asse
t Ma
nagement”, Ph.D
Thesis, Cur
tin
University
of
Technolog
y
,
2005
.
[9]
ANFIS NETWO
R
K,
page 7,
©Copyright 2002
by
Piero
P. Bonisso
ne
[10]
MATLAB R2009b version 7
.
9.0
.
529.
[11]
A. Abu-Siada,
Lai Sin Pin, and
S
y
ed
Islam, “Remnant Life
Estimation of
Power Tr
ansformer
using Oil UV V
i
s
Spectral Respon
ses”, Pap
e
r
accepted for
presen
tation
at
I
E
EE PSCE 09
that held
in USA, March,
2009.
[12]
S. M. Islam, T. Wu, G.
Ledwich,
“A Novel Fuzzy
Logic Approach
to Tr
ans
f
ormer Fault Diagnosis”,
IE
EE
Transaction on
Dielectric
and Electrical
Insulation
, Vol.7 No. 2,
April 2000, pp.
177-186.
[13]
K. V. Kumar, S. S.
Kumar and
B. Praveena, “Soft Co
mputing
Based Fault Diagnosis”,
Interna
tional Journal
o
f
Computer and Electrical
Engin
e
ering (
I
JCEE)
,
Vol: 2, No: 4
,
pp
. 753-760
, Aug 2
010.
[14]
Sk
Moin Ah
med, Haitham Abu-Rub, Shad
y
S.
Refaat, Atif Iqbal, Diagnosis of
St
ator Turn-to-
Tur
n
Fault and Stator
Voltage Unbalan
ce Fau
l
t Using ANFIS,
IJECE
, V
o
l3 , No. 1, Feb
2013.
BIOGRAP
HI
ES OF
AUTH
ORS
Priy
esh Kumar
Pandey
gradu
a
ted in
Electrical Engineer
ing from At
m
i
y
a
inst
itut
e
of
techno
lo
g
y
and science Rajkot, Gujarat in
the
y
e
ar
2009.
He
worker
as lectur
er in
Atmiy
a institute of
techno
log
y
and
s
c
ienc
e Ra
jkot
for s
i
x m
onths
. Then
he work
ed in Gujar
a
t s
t
at
e el
ec
tric
it
y
corporation Ltd
for two
y
e
ars in the field of
operation and maintenan
c
e.
His
current res
e
arch
inter
e
sts include insulation cond
ition
assessment techniques, spec
ialising in
th
e
diagnosis of
solid insula
tion
in
transform
e
r
based
of
m
o
istu
re
a
sse
ssme
n
t.
He
is c
u
rre
ntly
pursuing his
M.Tech
in Con
d
ition Monitoring Control
and
Pr
otection of
Electrical Apparatus from NIT
Hamirpur, India
Email:
priy
eshpandey
88@gmail.com
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
JECE Vo
l. 4
,
N
o
. 4
,
Au
gu
st 2
014
:
46
3
–
47
0
47
0
R K Jarial received PhD from University
of
R
a
jashtan through MNIT, Jaipur
, India in 2007 and
Masters Degree from
National
Institute of t
e
chnolog
y
Kuruks
hetra, Har
y
an
a, India in 1993
res
p
ect
ivel
y.
He is
current
l
y
working as
As
s
o
s
i
ate P
r
ofes
s
o
r in Electri
cal Eng
i
ne
ering
Department of NIT Hamirpur, HP, India for the la
s
t
twent
y
ye
ars
.
He
has obtained One Patent
and has been instrumental in
the establis
hment of TIFAC-C
O
RE in Power
Transformer
Diagnostics at N
I
T Hamirpur. His current ar
eas of
interests ar
e Power electronic dr
ives and High
voltag
e
eng
i
neer
ing.
Zakir Husain to
ok his Bechelor
and Master de
grees in Ele
c
tr
i
cal Eng
i
nee
r
ing
from
Aligarh
Muslim
Universit
y
, Al
igar
, Indi
a. Subsequen
t
l
y
he did
his Ph.
D
from
Motilal
Nehru Nation
a
l
Institute of T
ech
nolog
y, Allah
a
b
a
d, Indi
a. He ha
s been em
plo
y
e
d
as Lectur
er in
Departm
e
nt of
Electri
cal Eng
i
n
eering
,
National Institute of
Tech
nolog
y
,
Ham
i
rpur (H.P.), India since 1989 and
is working as A
ssociate Professor in th
e sam
e
i
n
stitute. During
2011 and
200
12 he was on
Academ
ic pur
s
u
it and worked
as
Head, Dep
a
rt
ment of Electrical
Engineering, Colleg
e
of
Engineering, Salman Bin Abd
u
laziz Univ
ersi
ty
, Kingdom of
Saudi Arabia. His princip
a
l
res
earch
in
teres
t
is
P
o
wer s
y
s
t
em
and Drives
.
Email:
zahusain
2
@ y
a
hoo.com
Evaluation Warning : The document was created with Spire.PDF for Python.