TELKOM
NIKA
, Vol.14, No
.1, March 2
0
1
6
, pp. 21~2
8
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.2602
21
Re
cei
v
ed Se
ptem
ber 2, 2015; Re
vi
sed
Jan
uar
y 19, 2
016; Accepte
d
Jan
uary 28,
2016
An Optimized Method of Partial Discharge Data
Retrieval Tech
nique for Phase Resolved Pat
t
ern
1,3
Eka Putra Waldi,
1,4
Aulia,
1
Ariadi Haz
m
i,
2
Hairul Abral,
3
S
y
ukri Arief,
4
M. H. Ahm
a
d
1
Departme
n
t of electrica
l
Engi
neer
ing, An
dal
as Un
iv
ersit
y
,
251
63 Pa
dan
g,
W
e
st Sumatra INDONESIA
2
Departme
n
t of Mechan
ical E
ngi
neer
in
g, Andal
as Un
iv
ersit
y
, 25
163 P
ada
ng, W
e
st Sumatra INDONES
I
A
3
Departme
n
t of Chemistr
y
,
A
n
dal
as Univ
ersi
t
y
, 25
163 P
ada
ng, W
e
st Sumatra INDONES
I
A
4
Institute of High Voltag
e an
d High C
u
rre
nt, F
a
cult
y
of El
ec
trical Eng
i
ne
eri
ng,Un
iversiti T
e
kno
l
og
i Mal
a
ysia,
81310 Johor B
ahru, Johor, MALAYSIA.
Corresp
on
din
g
author, emai
l: eka
w
a
l
d
i
2@
ya
hoo.com
A
b
st
r
a
ct
T
he
measur
e
m
e
n
t of
ph
ase
reso
lved
p
a
rti
a
l
disch
arg
e
re
quir
e
s a
gi
ga
ntic
me
mory
ca
p
a
city to
store all t
he w
a
ves of the P
D
test re
sults. This li
mitatio
n
e
v
entua
lly h
i
n
d
e
r
s the testing. I
t
is necess
a
ry to
opti
m
i
z
e
o
n
th
e asp
e
ct of
memory stora
g
e
capa
bil
i
ty
to
reduc
e the st
orag
e re
quir
e
me
nts. In li
ght
of
forego
ing,
the
partia
l
d
i
sch
ar
ge
data
retriev
a
l tec
hni
qu
es
w
i
th sa
mpl
i
ng
r
a
te
meth
ods
w
e
re
used
to
de
tect
the p
eak of
par
tial d
i
schar
ge
as w
e
ll
as the
PD constitu
ent
repres
entat
iv
e
s
. T
he opti
m
i
z
ation
proc
ess
w
a
s
perfor
m
e
d
by
usi
n
g
inte
gra
t
ion
of osc
ill
o
scope
a
nd
La
bVIEW
softw
are. T
he
parti
a
l
d
i
schar
ge
d
a
ta
recordi
ng c
an
be e
a
sily co
nfi
ned to th
e po
i
n
ts of t
he parti
al disc
har
ge o
ccurrenc
e on
ly
. As results, the
storage
po
ints
w
e
re red
u
ced
by taki
ng w
a
ve
mag
n
itu
de
associ
ated w
i
t
h
PD, ther
eby
resulti
ng
in
more
repres
entative
data.
T
her
efor
e, this
opti
m
i
z
e
d
meth
od w
a
s
abl
e to r
e
d
u
ce
the file si
z
e
of t
he test r
e
sults
up
to 99 perce
nt o
f
PD origin
al si
z
e
th
ereby
d
e
c
r
easi
ng the us
age of h
a
rd dis
k
storage.
Ke
y
w
ords
: par
tial disc
harg
e
, phas
e resolv
ed
, sampl
i
n
g
rate
, peak detect, record
ing, L
abV
IEW
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Polyethylene
is wid
e
ly used in m
any
elec
tri
c
al a
pparatuses
a
nd devices
as hig
h
voltage in
sul
a
ting mate
ria
l
due to it
s
excelle
nt ele
c
tri
c
al a
nd m
e
ch
ani
cal p
r
o
pertie
s
. Usu
a
lly,
polymers a
r
e
use
d
togeth
e
r
with
othe
r in
sulatin
g
mate
rial in
high vo
ltage in
sulati
on, su
ch
a
s
a
i
r-
polymer
com
posite in
sul
a
tion. The di
el
ectri
c
st
re
n
g
th of the air i
s
le
ss tha
n
that the poly
m
er.
Also, the
di
scha
rge
event
s p
r
o
bably
start withi
n
th
e air a
nd
pa
rtially bri
dge
the in
sul
a
tion
betwe
en
con
ducto
rs with
re
spe
c
t to
the ap
pli
ed
voltage. It is gen
erally
known a
s
p
a
rtial
discha
rge (P
D). The PD causes the pro
g
re
ssive
d
e
te
rioration of polymer su
rfa
c
e and eventu
a
lly
leadin
g
to the electri
c
al bre
a
kd
own to the alte
rnatin
g curre
n
t (AC)
appa
ratu
se
s and devices [
1
].
The PD
curre
n
t activity is also influe
nced
by
the curren
t generated from
the so
urce, whe
r
eby th
e
greate
r
the
flow of re
so
urces, the
h
i
gher
th
e cu
rre
nt of the
PD. Like
wi
se, un
der t
he
environ
menta
l
con
d
ition
s
such
as the
hi
gh hu
midi
ty, the PD
activi
ty are affe
cte
d
eno
rmo
u
sl
y.
The hig
h
hu
midity would
accele
rate t
he proc
ess
of degradatio
n. Furthe
rmo
r
e, the repeti
t
ion
activities
of P
D
at
differe
nt diele
c
tri
c
val
ues
would
al
so
expedite
the d
e
g
r
adati
on p
r
o
c
e
s
s [
1
-3]
and thereby gene
rating a
numbe
r of PD data. The
s
e PD data su
ch as P
D
pattern co
uld be
a
result of furt
her
con
d
ition
monitorin
g
and dia
gno
st
ics
of elect
r
i
c
al e
quipm
e
n
t [5-7] su
ch
as
medium volta
ge und
erg
r
o
u
nd and ove
r
h
ead cable
s
.
There a
r
e
two ba
si
c p
o
ssi
b
ilities fo
r
re
cognition
of P
D
. Th
e first p
o
ssibility is th
e time-
resolved recognition fo
r measuri
ng th
e actual
s
h
a
pe of the ch
arge
displacement withi
n
the
defect in
nan
ose
c
o
nd
scale. Literally, there i
s
a
direct rel
a
tion
sh
ip between t
he phy
sic in
the
defect
and
shape
of
sign
al [8]. The
seco
nd
po
ssi
b
ility is the
ph
ase
rel
a
ted
reco
gnition. T
h
is
method u
s
e
s
the cla
ssi
c d
i
scharge d
e
te
ctor a
nd the
pattern that
mostly occu
rs in 50 o
r
60
Hz
sine
wave
or
PD ph
ase re
solved
(PDP
R). T
he
sha
p
e
of si
ngle
pu
lse
s
is not
rel
e
vant, only th
eir
relative high
and ph
ase an
gle are co
nsi
dere
d
whic
h t
he Time
cont
ant of PD co
n
v
entional sen
s
or
(TSPDCS) sh
ould be in o
r
d
e
r of ~1
s [9-11].
Curre
n
tly, the samplin
g rat
e
spe
ed of the
analo
g
-to
-
di
gital conve
r
te
r (ADC)
coul
d
be in
giga sam
p
lin
g
pe
r se
con
d
(GS/s)
a
nd suppo
rted by
very so
phi
sticated r
eal
-time sto
r
ag
e
(RAM)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 21 – 28
22
thereby attain
ing a P
D
pul
se d
a
ta an
d
saving th
e da
ta con
c
u
r
rentl
y
. A PD pulse is
rea
d
by the
data a
c
qui
siti
on sy
stem
(DAQ), in
whi
c
h
req
u
ire
s
a
tri
gger value to
store
the P
D
data in
a seri
es
of
memo
ry-re
l
ated sequ
en
ce
s with
a re
feren
c
e
mag
n
itude o
r
pha
se
of PD. Th
e ra
w P
D
d
a
ta
from the inte
rfacing
circuit
were tra
n
sfe
r
red in
to th
e compute
r
me
mory, and d
a
ta storage
system
for furthe
r an
alysis
and al
so
coul
d be
recalled
as
n
eede
d. [11-1
6
]. As ADC
spe
ed in
crea
se
d
rapidly, the
ability of digital osc
illoscopes is also increa
sing as well. Thus, the sampling
rate
can
rea
c
h up to 2
0
GS/s and followed by a large reco
rd length [17]. These data we
re obtain
ed from
the refe
ren
c
e
in whi
c
h the
one-tim
e
col
l
ection
of dat
a ca
n gen
era
t
e a data file
of 40 MB. It is
impossibl
e to
re
co
rd all th
e PD
data. If the data
we
re re
co
rde
d
b
a
se
d on
a ti
me interval
of
one
se
con
d
in an
hour, then th
e size of the data file woul
d become 14
4 GB.
Figure 1. Testing and Imitation of PD
set
up usi
ng Hi
gh
Voltage Amplifier
Usi
ng a
la
rge
sam
p
ling
rat
e
can
rea
d
a
PD p
e
a
k
p
r
eci
s
ion.
Ho
wever, the
am
ount of
data (mem
ory used) b
e
co
mes large wh
en the readi
n
g
was follo
wed by a wave of applicati
ons
PD wave readings. So it takes a long time when
transferring and storing
data
to a PC. This will
make th
e interval bet
wee
n
mea
s
urem
ents to
be g
r
eat a
s
well.
Therefo
r
e,
need to opti
m
ize
sampli
ng rate
and optimi
z
i
ng the use of
data to be st
ored. Thi
s
co
uld be a p
o
ssible la
rge
si
ze
thus re
qui
ring
data colle
ctio
n techni
que
s
and
software help to overcome this p
r
ob
lem.
PD data storage at stocha
stic
way
s
for a long time with a high sa
mpling rate requires
a large
mem
o
ry. Thi
s
p
r
o
b
lem n
eed
s t
o
be
optimi
z
e
d
so that
the
measurement
s
can
be
carried
out in a long time. In this research, the st
orag
e of PD
measurement
data
is optim
i
zed
by utilizi
ng
the oscillo
sco
pe ca
pability in combi
natio
n with
the La
bVIEW software to minimi
ze the mem
o
ry
usa
ge.
2. Rese
arch
Metho
d
2.1. PD measuremen
t
Sy
stem
Figure 1
sho
w
s the
sch
e
m
a
tic di
ag
ram
of t
he experi
m
ental setu
p for
PD
me
asu
r
eme
n
t
system
wa
s u
s
ed. T
he 5
0
Hz A
C
voltag
e gen
er
ate
d
by a fun
c
tion
gene
rato
r (te
k
troni
k AF
G3
252
Arbitra
r
y Wav
e
form G
ene
rator)
we
re
a
m
plified at 2
0
00-fold
by u
s
i
ng the
po
wer amplifier (Mo
del
20/20
C; Tre
k
. Inst.)
whi
c
h wa
s p
a
sse
d
throu
gh th
e sp
ecim
en.
This
high v
o
ltage am
plifier
derived
from
functio
n
ge
n
e
rato
r
wa
s to
tally c
ontrolle
d by a
pe
rso
nal
comp
uter (PC) via
US
B
cabl
e. The high voltage so
urce we
re me
asu
r
ed th
rou
gh the output
signal voltag
e of high voltage
sou
r
ce amplif
ier (with a ratio of 1: 2000 v
o
lts). PD
p
u
lses were dete
c
ted by PD d
e
tector
(Haef
ely
Type 9231
).
Cou
p
ling
-
cap
a
citor (Ck)
was
used a
s
a
voltage divid
e
r
so th
at th
e voltage
do
es
not
rise
on th
e i
m
peda
nce of
the PD. Bef
o
re th
e PD
d
e
tector was
use
d
, the
cali
bration
wa
s
done
with a calib
ra
tion signal
ch
arge
(Ha
e
fely Type 9218)
durin
g the offline con
d
ition
.
PC controll
ed
all the syste
m
unde
r La
b
V
IEW prog
ra
m. The da
sh
ed line in Fi
gure
1 was
referring to t
h
e
calib
ration in
strum
ents d
u
r
ing the offlin
e state.
It could be don
e as imitation from the PD si
gnal
in whi
c
h p
o
sse
ss th
e sa
me magnitu
d
e
with the p
r
ogra
mmed P
D
from the
a
r
bitra
r
y functi
on
gene
rato
r.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
9
30
An Optim
i
zed
Method of Partial Di
scha
rge Data
Retri
e
val Te
ch
niq
ue for… (E
ka
Putra Waldi
)
23
Dotted line re
ctang
ular in
Figure 1 is p
a
rt of
the PD calib
ration th
at shoul
d
be
done at
the time of
m
easure
m
ent
equipm
ent in
the off state.
Calib
ration
is an imitatio
n
of PD
event t
hat
has the
sam
e
ma
gnitud
e
at ea
ch
even
t. PD
calib
ra
t
o
r i
s
an
imita
t
ion of
a
real
PD
events.
PD
calib
rato
r co
nsi
s
ts of
a capa
citor and
a
p
u
lse so
ur
c
e
, us
u
a
lly it w
o
rks
in on
e
po
la
r
i
ty to th
e
distan
ce bet
wee
n
pulses
is con
s
tant. Besid
e
s the
s
e
ways, PD cali
brato
r
can al
so be do
ne with
the help of the arbit
r
ary functio
n
gene
rator in
com
b
ination with
the LabVIEW software.
PD
magnitud
e
an
d locatio
n
of the inci
dent i
s
very rand
om
so it is ha
rd t
o
determi
n
e t
he value of th
e
ben
chma
rk PD. This
will
significa
ntly affect t
he level
of efficiency
of readi
ng t
he num
ber
a
n
d
magnitud
e
of PD.
(a)
(b)
Figure 2 Phased resolved
PD simul
a
tio
n
, (a) volt
age
appli
c
ation a
nd imitation o
f
PD, (b) one
of
imitation PD waveform
By using the
arbitrary fun
c
tion gen
erator, va
lue and p
u
lse
pha
s
e p
o
sition a
s
co
nstitutin
g
an imitation
PD ca
n be d
e
termin
ed, e
v
en the num
ber
of PD
on
any positive
polarity voltage
sou
r
ces o
r
th
e ne
gative p
o
l
arity can
be
prog
ram
m
ed
simultan
eou
s
l
y
. If the num
b
e
r
and
the
pe
ak
value PD is
alrea
d
y kno
w
n at each p
o
l
a
rity wave
so
urce, then the effici
en
cy of the number of
PD readi
ng
s pe
r p
h
a
s
e
and th
e p
e
a
k
valu
e
can
be
calcul
ated. Example
s
of
imitatio
n P
D
measurement
re
sult
s a
r
e
shown in
Fig
u
r
e
2a. In
this
Figu
re,
th
ere
a
r
e 30 PD
pulses for ea
ch
positive an
d negative pol
a
r
itie
s, wh
erea
s PD waveforms are sh
own in Figure 2b
.
PD sig
nal me
asu
r
em
ents
were pe
rform
ed us
i
ng
cylin
der-plate el
ectrode
s. The d
i
amete
r
of the condu
ctor cylinde
r a
nd t
he diamet
er of the con
ducto
r plate
wa
s app
roxi
mately 0.5 cm and
5 cm
respe
c
tively. Polymer films were
placed
on
th
e cond
ucto
r
plate, an
d
cylinder ele
c
tro
d
e
s
were pla
c
ed
over the film at a given gap. The
length
of the gap betwee
n
the polymer surfa
c
e
and the
cylin
der tip
wa
s a
pproxim
ately 0.01 mm. Al
so, two types
of con
n
e
c
tors were u
s
ed.
The
first
wa
s the
co
nne
cto
r
Numb
er 1,
whe
r
e
t
he cylinder wa
s con
n
e
c
ted
to
the high
vo
ltage
electrode a
n
d
the groun
d condu
ctor pl
ate. The
se
co
nd co
nne
c
tor was lin
ke
d in
a way oppo
site
to the first
conne
ctor,
wh
ere th
e cond
uctor-cy
linde
r con
n
e
c
ted t
o
ground,
an
d the cond
uctor
plate was co
n
necte
d to the
high voltag
e. The L
D
PE
film with the
thi
c
kne
ss
80 µ
m
wa
s pl
aced
on
the grou
ndin
g
electro
de.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 21 – 28
24
2.2. Gener
a
l
con
cept.
Figure 3
sh
o
w
s the
readi
ng of the
co
rre
sp
ondi
ng
PD voltage
waveform. Where
i
s
comp
osed of two part
s
, the first part is a
function
of th
e oscillo
scop
e, where the analo
g
sign
al
is
conve
r
ted to a digital syste
m
. Most of the gene
ra
ted
data (record l
ength) d
epe
n
d
on the valu
e of
the sele
cted
sampli
ng
rate
, so th
at the
cha
r
a
c
teri
stics of th
e sen
s
or a
r
e
re
quire
d a
s
a
refe
re
nce
for determi
ni
ng the value of the selecte
d
sampli
ng ra
te. The seco
nd is the det
ermin
a
tion of the
value of PD
magnitud
e
. It is analy
z
ed
with pha
se
correspon
ding
to the prog
ram logi
c so t
hat
only the PD magnitud
e
a
nd pha
se val
ues (P
D ph
a
s
e of occu
rre
n
ce a
nd PD
magnitud
e
) a
r
e
saved o
n
the
PC. Commu
nicatio
n
between o
scill
oscope an
d the
PC is u
s
ing t
he TCIP in which
serve
s
a
s
the
acqui
sition
control an
d dat
a delivery sta
r
ter.
Figure 3. Analog to digital PD wavefo
rm
synch
r
oni
zati
on pro
c
e
ss
(a)
(b)
Figure 4. PD measurement
at 200 MS per se
co
nd sa
mpling rate, (a) voltage ap
plicatio
n and
(b)
Suitable PD d
a
ta
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
An Optim
i
zed
Method of Partial Di
scha
rge Data
Retri
e
val Te
ch
niq
ue for… (E
ka
Putra Waldi
)
25
The digitali
za
tion pro
c
e
ss
of partial discha
rge i
s
det
ecting the p
e
a
k value o
b
tained
from the
PD sensor (Haefel
y type
923
1)
can
be
read
b
y
usin
g o
s
cill
oscop
e
T
e
ktronix DP
O 5
1
04.
The o
s
cillo
scope
also
fun
c
tions
as th
e i
n
terface
fo
r L
abVIEW
software to
co
ntrol the m
e
a
s
u
r
ing
unit and to p
r
oce
s
s the si
g
nals afte
r the
digitization
p
r
ocesse
d ha
s finished. T
h
e analo
g
to the
digital process of samplin
g
rate is
re
quired to get
the
optimum p
e
rforman
c
e
of tool so that th
e
repe
atability of PD (loopin
g
) mea
s
u
r
em
ents doe
s no
t take a long time. This proce
s
s is divid
ed
into two part
s
, namely pro
s
e digitizatio
n
, and si
gnal p
r
oce
s
sing.
Figure 4(a
)
d
epict
s a voltage so
urce si
g
nal
in the form of sine wa
ve obtained from the
voltage probe
, wherea
s Fi
gure
4b illu
st
rates the PD
pulse-sha
ped
sign
al obtai
n
ed from th
e
PD
detecto
r. Both of these si
gnal
s are
synch
r
oni
ze
d in time. There were seve
ral events of PD
durin
g the first wave of the voltage so
urce
. The sam
p
li
ng rate was 2
00 MS/se
c
on
d.
Figure 5 sho
w
s th
e assu
mption of a
waveform tha
t
is on the
small dotted b
o
x from
Figure 2. To get the perfe
ct waveform, a high sam
p
li
ng rate was u
s
ed. In this st
udy, a sampli
ng
rate of 10
G
S
/s wa
s u
s
e
d
. By taking t
he pe
ak val
u
e of a pul
se,
the pul
se m
agnitud
e
valu
e at
90% and the
maximum value (pe
a
k) wa
s mea
s
u
r
ed
as sho
w
n in Figure 5(a
)
. Value above
90%
of the maximum magnitu
d
e
of pulse
wa
veform of
the first and se
cond half-cycl
e has a p
e
rio
d
of
time along lin
e AB, which states the val
ue of SR
= 1
/ AB is calle
d the peak
sampling rate of
90% (90% S
R
).
Figure 5. PD pro
c
e
ss from
analo
g
to digi
tal, (a) anal
og
signal a
nd (b
) digital sig
nal
Figure 6. Det
e
ction effici
en
cy of PD num
ber pe
r ph
ase function of
sampli
ng rate
The digital
signal processi
ng is a p
r
o
c
ess
to obtain
a digital data from the a
nalog
sign
al by sa
mpling the an
alog si
gnal di
scretely
und
e
r
the peri
od o
r
frequ
en
cy sampling d
o
m
a
in.
The
Nyqui
st theory
provid
es th
e rule th
at the fre
que
ncy
sampli
ng
sh
ould
be
at lea
s
t two tim
e
s
the maximum
freq
uen
cy of
the related
si
gnal
or hal
f
of the
sign
al p
e
r
iod. T
he
ma
ximum value
of
sampli
ng p
e
ri
od (time
bet
wee
n
poi
nt A and B
)
shou
ld be e
qual t
o
0.5 of the
perio
d in
ord
e
r t
o
achi
eve the
value above
90% of the peak m
agni
tude. Then,
the digitize
d
values a
r
e
the
intersectio
n
betwe
en the
sign
al curve
and the
Ny
qui
st p
e
rio
d
that are
se
e
n
as th
e vertical
dashed line
s
i
n
Figure 5(b
)
.
Although th
e
pea
k of P
D
si
gnal
ha
s be
e
n
dete
c
ted
u
s
ing a
samplin
g techniq
ue
b
a
se
d
on Nyqui
st theory. Ho
weve
r, anothe
r sig
nal ca
n also
be dete
c
ted i
n
stea
d of the real pe
ak of
PD
durin
g the de
tection proce
ss that
can cause the PD sign
al to
be counted mo
re
than one p
u
lse,
resulting i
n
a
n
in
corre
c
t v
a
lue. In th
e
actual
mea
s
urem
ent, the
un
wante
d
si
gnal
s
can
not
be
avoided. T
h
u
s
, it is
cruci
a
l to verify the
me
asure
d
sig
nal
s th
at can
be
p
e
rform
ed
usi
n
g
0
20
40
60
80
10
0
12
0
14
0
1
000
00
0
1
0
000
00
0
1
000
00
000
10
00
000
000
1
000
00
000
00
Sa
m
p
lin
g
r
a
t
e
(
s
am
p
l
e
/
se
c)
E
f
f
i
s
i
e
n
c
y
D
e
t
e
c
t
i
o
n of
P
D
pe
a
k
(
%
)
A
bi
l
i
ty
of Sa
m
p
l
i
n
g r
a
te
to r
ead
m
o
r
e
th
an
90
%
PD
m
a
x
N
e
gati
v
e p
o
l
a
r
i
ty
Po
s
i
ti
v
e
p
o
l
ar
i
t
y
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 21 – 28
26
LabVIEW. La
bVIEW is sof
t
ware
with th
e buildu
p
fun
c
tion
s that ca
n be used to detect the pe
ak
value of
a
wave
co
unte
d
a
s
the
PD sig
nal i
n
th
e process of
rea
d
ing
the
data f
r
om t
he
oscillo
scope.
If any, som
e
additio
nal
logic fu
nctio
n
is
req
u
ire
d
to enabl
e th
e LabVIE
W
for
recogni
zin
g
the nature of PD si
gn
al an
d PD pha
se
-resolve
d.
(a)
(b)
Figure 7. Partial discha
rge
pha
se re
solv
ed for
(a
) con
nectio
n
type-1 and (b
) con
nectio
n
type-2
3. Results a
nd Analy
s
is
3.1. Reading
the error o
f
the differe
n
c
e
v
a
lue of sa
mpling rate w
i
th the d
u
mmy
PD signal
PD wave
s were dete
c
ted
by the sen
s
o
r
unde
r the P
D
duration a
pproxim
ately to one
microsecond,
wherea
s 90
% of the
peak value of SR is not more
t
han 10% of the time duration
PD. To get a peak valu
e above 90%
SR, it was a
s
certain
ed first wheth
e
r th
e sen
s
o
r
s h
a
v
e
detecte
d all
PD events, in
other
wo
rd
s
the num
b
e
r
o
f
PD rea
d
ing
s
pe
r ph
ase
voltage mu
st be
accurate. Th
e next step
was to d
e
tect
how m
any pe
ak valu
es th
a
t
was
equ
al to or
bigge
r th
an
90% SR. To
get better
re
sults at a val
u
e equ
al to o
r
bigge
r than
9
0
% of SR in t
he dete
r
min
a
tion
of samplin
g rate, it is nece
s
sary to kno
w
the ab
ility of an oscill
oscope to rea
d
the magnitu
de
PD
detecte
d by t
he
sen
s
o
r
. T
h
is
wa
s
don
e
by mea
s
u
r
in
g the P
D
pul
se imitatio
n t
hat con
s
ist
s
of 30
pulses P
D
at
ea
ch p
o
sitiv
e
and
ne
gati
v
e pola
r
it
y. Da
ta
co
lle
c
t
ion w
a
s
pe
r
f
or
me
d
te
n
time
s
fo
r
each additio
n
a
l mea
s
urem
ent sam
p
ling
rate. Teste
d
sampli
ng rate ran
g
ing fro
m
2 MS/s up
to
1.25 GS/s. Figure 5
sho
w
s the avera
g
e
efficien
cy readin
g
s nu
m
ber eve
r
y sin
g
le PD voltage
waveform a
p
p
licatio
n (on
e
cy
cle), it
a
ppea
rs t
hat t
he effici
en
cy of the
average
rea
d
ing
s
PD
imitation amo
unt wo
rth 10
0
%
at a sampli
ng rate
of
at least 1
0
MS p
e
r second. In
this pap
er, th
e
sampli
ng rate use
d
two time or mo
re,
as 25 MS perse
con
d
. If this sam
p
lin
g rate u
s
ing
on
voltage source with 5
0
Hz,
it can
re
co
rd
data a
r
e 5
0
0
,000 p
o
int, and d
a
ta for
PD in o
ne
wave
aplication ha
ve the maksi
m
um aroun
d 100
point
.
Furthe
r mo
re, the data reco
rd to P
C
can
reduc
e
up to 99%.
3.2. Application of PD Measur
e
ments
w
i
th Differ
ent Conne
ctors
PD readi
ng
s with 25 MS/s samplin
g rat
e
for a freque
ncy of 50 Hz
requi
re
d 0.4 million
points,
whereas the measuring
capacity of the o
scill
oscope
was up to 25 million poi
n
ts.
Therefore, th
e wave
s
were rea
d
in
real
time as
m
u
ch as
62.5
wa
ves. In this
study, the map
p
in
g
of the PD events wa
s only for one
se
con
d
(50
waves)
at intervals of
four se
con
d
s.
PD data
retri
e
val ha
s not
been
ca
pture
d
for a
sin
g
l
e
time only,
but it wa
s
ca
ptured
repetitively. Storing all PD
data req
u
ire
s
a gigant
ic st
orag
e system
but this process is almo
st
impossibl
e to
be reali
z
e
d
. To solve thi
s
probl
em, ca
pturing
and
storing the P
D
data co
uld b
e
done rand
o
m
ly at certai
n time intervals [8], [11-
1
3
], [17]. Given the limitations of exist
i
ng
memory, it is nece
s
sary to captu
r
e on
ly the PD
related wave
s to redu
ce the
use of larg
er
memory
with
out lo
sing
up
the valua
b
le i
n
formatio
n of
PD pa
ram
e
te
rs. T
he
LabVI
E
W
softwa
r
e,
in
this ca
se, wit
h
its available
feature
s
, can
r
edu
ce the file size of the test re
sult
s.
Figure 9(a
)
a
nd 9(b
)
sho
w
the ch
arge
magnitude
as a fun
c
tion
of phase a
n
g
le for
con
n
e
c
tion type-1 a
nd con
nectio
n
type-2 respe
c
ti
vely. The PD data was
store
d
on a PC in 50
cycle
s
of the
alternating v
o
ltage source. The PD
measurement
has resulted
in two different
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
An Optim
i
zed
Method of Partial Di
scha
rge Data
Retri
e
val Te
ch
niq
ue for… (E
ka
Putra Waldi
)
27
types of con
necto
rs p
o
ssessed the sa
me pattern
s.
In the positive polarity, PD activities h
a
ve
occurre
d
at zero
pha
se
an
gle of the vol
t
age ap
p
lication. It also
occurre
d
at a
n
egative ph
ase
angle, but the
magnitude of
PDs have re
mained at po
sitive values.
PD
ac
tivity
w
a
s
pr
ec
e
ded
b
y
a
c
h
arg
e
ca
rr
ier
(C
C)
in
je
c
t
io
n in
th
e
a
i
r
,
w
h
er
e
molecule
s were ioni
ze
d by the CC an
d amplified b
y
the high electri
c
field e
nhan
cem
ent. The
possibility of t
he exi
s
tence
of the
CC at t
he hi
gh
electric field was derived f
r
om
the ionization
of
air m
o
le
cule
s, but the
process ve
ry com
p
licate
d
.
Also
, CC obtai
ne
d from
the
el
ectro
de. T
h
is
CC
coul
d di
stort t
he
sha
pe
of the el
ectrode.
It was al
so
v
e
rified i
n
thi
s
experim
ent th
at the tip
of the
needl
e el
ectrode
wa
s defo
r
med
by
a di
scha
rge
[8]. O
ne
rea
s
o
n
for the
sig
n
ifican
t increa
se
in t
h
e
PD pul
se
discha
rge
time
wa
s du
e to th
e spre
ad
out
of ch
arg
e
di
st
ribution
on th
e surfa
c
e
of the
s
p
ec
imen
.
Con
n
e
c
tion
type-1 was a
cylindri
c
al co
ndu
ct
or that
wa
s co
nne
ct
ed to a high
voltage
sou
r
ce. Also,
the plate electrod
e wa
s conne
cted
to grou
nd, wh
ere the CC wa
s domin
ant at the
positive pola
r
ity. Thus, the
maximum po
sitive PD
(MP
D
P) value wa
s gre
a
ter tha
n
the maximum
negative PD (MPDN) valu
e. Also, followed by t
he numbe
r of positive PD (NPDP) wa
s hig
h
e
r
than the num
ber of ne
gative PD (NP
D
N). Mea
n
wh
il
e, the conn
e
c
tion type-2
wa
s a cylind
r
ical
con
d
u
c
tor
co
nne
cted to t
he g
r
ou
nd
condu
ctor
plat
e, then the
plate was co
nne
cted to t
he
sou
r
ces th
ere
b
y resulting i
n
the CC accumulat
ed
at the neg
ative polarity part. T
herefo
r
e, MP
DP
value wa
s sm
aller than MP
DN valu
e. Likewi
se, NP
DP
value wa
s lo
wer tha
n
the NPDN value
as
sho
w
n in Ta
b
l
e 1.
Table 1. The
Perform
a
n
c
e
of PD from different conne
ction types
PD quantities
Connection t
y
p
e
-
1
Connection t
y
p
e
-
2
Max
i
mum PD
positive (pC)
1280.64
8.6
Max
i
mum PD
negative (pC)
-870.505
12.4
Number of P
D
positive
7317
10138
Number of P
D
negative
5668
10686
4. Conclusio
n
As a
con
c
lu
sion,
acqui
sition p
e
a
k
o
p
timization
con
ducte
d
with
PD d
e
ci
sion
-wave
results
ha
s b
een
su
ccessf
ully ca
rrie
d
o
u
t. By ta
king
wave m
agnit
ude a
s
so
ciat
ed with P
D
, it may
redu
ce
the
storage
poi
nt
and th
ere
b
y
resulting
i
n
more
representative
data
.
Therefore,
by
desi
gning
a
L
abVIEW p
r
o
g
r
am fe
atures,
it wa
s
able
to
redu
ce
the
file si
ze
of th
e test
re
sult
s
up
to 97 percent
s of PD origi
n
al size there
b
y
decre
asi
ng
the usa
ge of hard di
sk sto
r
age.
Ackn
o
w
l
e
dg
ement
This re
sea
r
ch
wa
s sup
p
o
r
ted
by Dire
ktor
at Jend
ral
Pendidi
kan
Tinggi, Keme
nteria
n
Pendidi
kan
Na
sion
al, Penelitian
Hiba
h F
a
kulta
s
T
e
kni
k
Ta
hun
angg
aran
No.
030/UN.16/P
L
/HT/201
4, Indone
sia
Referen
ces
[1]
PHF
Morshuis.
Degra
dati
on
of solid
die
l
ect
r
ic
due to i
n
ter
nal p
a
rtial
disc
harg
e
: some throu
ghts on
progr
ess ma
d
e
an
d
w
h
er
e
to go
no
w
. IE
EE T
r
ansacti
o
n
on
di
el
ectric
and
el
ectrica
l
insu
lati
on
.
200
5;12(
5): 90
5-91
3.
[2]
Eka Putra Waldi.
Infl
ue
nce
of Partia
l D
i
sc
harg
e
on Br
e
a
kdow
n
of Air
-
Poly
mer
Co
mposite
. IEEE
Internatio
na
l C
onfere
n
ce o
n
Con
d
itio
n Mon
i
to
ring a
nd Di
ag
nosis (CMD). B
a
li. 20
12: 79
1-
974.
[3]
Waldi EP, Mur
a
kami Y, M Nagao.
Effect of hu
mi
dity o
n
b
r
eakd
o
w
n
of lo
w
density p
o
ly
ethyle
ne fi
l
m
due to
parti
al d
i
schar
ge
. Inter
natio
nal
Co
nfer
ence
on C
o
n
d
it
ion M
onitor
i
ng
and
Dia
gn
osis.
Bali. 2
008:
655-
658
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 21 – 28
28
[4]
Eka Putra W
a
ldi, Yos
h
i
nob
u
Murakam
i
, N
aoh
iro H
o
zum
i
, Masa
yuki
N
aga
o. Break
d
o
w
n
of Air-
Pol
y
mer C
o
mp
osite Insu
latio
n
du
e to Parti
a
l Disc
har
ge a
nd Influ
enc
e o
f
T
hermal Insu
latio
n
.
IEEJ
T
r
ansactio
n
s o
n
F
unda
menta
l
s and Materi
als
. 2012; 13
2(11)
: 1039-1
0
4
4
.
[5]
Miofen Z
h
u, Guoji
n
C
h
e
n
, Mi
ng
Xu, T
i
ngtin
g Li
u, Do
ng
Xi
e, Yan
p
in
g Z
h
ang. Stu
d
y a
n
d
Mon
i
tori
ng
S
y
stem
for P
a
rtial
Disch
arge
of l
e
ctrica
l E
qui
pment.
T
e
lk
omnik
a
In
don
e
s
ian
Jour
na
l
o
f
Electrica
l
Engi
neer
in
g.
2009; 12: 1
273-
128
3.
[6]
Hung-cheng C, Feng-chang
G, Chun-
y
aao LA. Ne
w
m
e
t
hod base
on
ex
t
ension theor
y
for
partial
disch
arge p
a
ttern reco
gniti
on
.
WSEAS
TRANSACTIONS on SYSTEM
. 2008; 12(7): 1
402
-141
1.
[7]
Elpi
dio
R, T
e
resa MD.
Lea
n
mainte
nanc
e
mode
l to r
e
d
u
c
e
scra
p
s a
n
d
W
I
P in ma
nufa
c
turing
s
y
stem:
case study
in
po
w
e
r cable fact
or
y
.
WSEAS TRANSACTIONS on SYSTEM
. 2013; 1
2
(12):
650-
666.
[8]
PHF
Morshi
us,
F
H
Kru
e
g
e
r.
T
r
ansition from
st
reamer
to to
w
n
se
nd M
e
ch
anism
in
Di
el
e
c
tric Voi
d
.
J.
Phys
. 1990: 2
3
:
1562-1
5
6
8
.
[9]
F
H
Kreuger, E
Gulski, A Krivda.
Class
ificati
on of Partia
l Di
scharg
e
.
IEEE Transaction on Dielectric
s
and El
ectrical I
n
sul
a
tion
. 1
993
; 28(6): 917-9
8
3
.
[10]
E Gulski. Computer-a
ide
d
Me
asurem
ent of Partial Disc
har
ge in HV Eq
uip
m
ent.
IEEE Tr
ansaction on
Dielectrics and Elec
trical Insulation
. 19
93: 28
(6): 969-9
83.
[11]
E Gulski.
Dig
i
t
al An
alis
is
of Partia
l D
i
sch
arge
. IEEE
T
r
ansacti
on
on
Diel
e
ctrics
an
d El
ectrica
l
Insulation.
19
9
5
: 28(6): 82
2-8
37.
[12]
Paul
Sei
i
tz.
Pe
rma
n
e
n
t Di
gita
l Rec
o
rdi
n
g
of
PD i
m
puls
e
i
n
a
n
Impu
lse
ana
ly
z
e
r
. Proc.
5th H
aefe
l
y
S
y
mp. Stutgart
.
1992.
[13]
GC Monta
nari,
A C
ontin,
A
Caval
i
n
i
. Ra
nd
om
Sam
p
li
ng
and
Dat
a
Proc
essin
g
for
PD-
puls
e
H
e
ig
ht
and Sh
ap
e An
al
ysis.
IEEE Transacti
on o
n
D
i
electrics
and E
l
ectrical Ins
u
lat
i
on
. 20
00; 7(1):
30-39.
[14]
A Cava
lin
i, A Conti
n
, GC Mo
ntanar
i. Adva
n
c
ed PD
Interfa
c
e in On-F
i
e
ld
Measur
ement.
Part I: Noise
Rectio
n.
IEEE
Transaction on Dielectri
cs
and Electrical Insulation.
20
03; 10(2): 21
6-2
2
4
.
[15]
Ardila-
R
e
y
JA,
Martinez-T
arifa JM, Ro
bles
G, Rojas-Mor
e
no M, Albarrac
i
n R. A P
a
rtial
Disch
arg
e
s
acqu
isitio
n a
n
d
statistical
ana
l
y
sis soft
w
a
re.
Instru
me
ntation an
d Measur
e
m
ent
T
e
chno
lo
gy
Confer
ence
. 2
012: 16
70-
167
5.
[16]
J T
ang, J Z
h
o
u
, X Z
h
n
g
, F
Liu. A
transfor
m
er
Partia
l D
i
scharg
e
Me
as
ureme
n
t S
y
ste
m
Base
d o
n
F
l
uoresc
ent F
i
ber.
Energ
i
es
. 201
2: 490-
150
2.
[17]
T
Okamoto, H Suzuk
i
, N H
o
zumi, M Iked
a.
Pa
rtial
Disc
h
a
r
ge E
ndur
anc
e
Life
of Po
l
y
m
e
r Insu
latin
g
Materials
at Hi
gh T
e
mperatur
e.
Electrical En
gin
eeri
ng i
n
Ja
pan
. 19
99; 12
6
(
1): 15-22.
Evaluation Warning : The document was created with Spire.PDF for Python.