TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 8, August 201
4, pp. 5910 ~ 5917
DOI: 10.115
9
1
/telkomni
ka.
v
12i8.562
4
5910
Re
cei
v
ed
Jan
uary 12, 201
4
;
Revi
sed
Jun
e
13, 2014; A
c
cepted
Jun
e
25, 2014
Fault Location of Distribution Network Containing
Distributed Generations
Zou Bi-c
han
g
*, Zhouhon
g
Schoo
l of Elect
r
onics a
nd Info
rmation of Yan
g
tz
e Univ
ersit
y
, Jingzho
u, 434
023, Hu
be
i, Ch
ina
*Corres
p
o
ndi
n
g
emai
l: zoubic
han
g1
23@
126
.com
A
b
st
r
a
ct
It makes the
topol
ogy of di
stribut
io
n netw
o
rk mor
e
co
mplex to cont
ai
n a lot of dis
t
ributio
n
gen
eratio
ns (
D
Gs) in it. F
o
r
this re
ason,
a
fault
lo
cati
on
appr
oach
su
ita
b
le
to th
e d
i
stributi
on
netw
o
r
k
contai
nin
g
DG
s is pr
op
osed.
F
i
rstly, accor
d
in
g to th
e re
sults of p
o
w
e
r
flow
calcu
l
ati
on, a
data
bas
e o
f
voltag
e sa
g is
establ
ishe
d; th
en th
e p
o
w
e
r
qua
lity i
n
fo
r
m
a
t
ion
is
acq
u
ir
e
d
by pow
er qu
ality mo
nitors and
correlati
on
an
a
l
ysis o
n
co
llect
ed n
o
d
a
l v
o
lta
ge d
a
ta
and
v
o
ltag
e sa
g d
a
ta of the
esta
bli
s
hed
data
bas
e
is
perfor
m
e
d
; a
n
d
the
n
th
e
nod
e w
i
th th
e
mat
c
hin
g
ext
ent v
a
lu
e
most c
l
os
e to
1
is d
e
ter
m
i
n
e
d
as
the
faul
t
poi
nt. Simul
a
ti
on res
u
lts of I
EEE 33-b
u
s sy
stem s
how
tha
t
the pro
pos
ed
appr
oac
h is
e
ffective an
d us
ing
the prop
ose
d
appr
oach th
e mis
j
ud
g
m
e
n
t can be pr
eve
n
t
ed. F
i
nal
ly, the results of the ana
lysis o
n
t
h
e
influ
ences
of d
i
fferent lo
ad
mode
ls, transiti
o
n resist
a
n
ces,
conn
ectio
n
an
d disc
on
nectio
n
of DGs o
n
t
h
e
prop
osed
appr
oach sh
ow
that t
he propos
ed
appr
oach
poss
e
sses better ro
bustness a
nd r
e
lia
bi
lity.
Ke
y
w
ords
:
distrib
u
tion
ne
tw
ork,
distribu
ted g
ener
atio
n
(DG), fault
l
o
catio
n
, pow
e
r
qua
lity
mo
ni
tor,
correlation analysis, voltage sag
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
In rece
nt years, alon
g with
the gradu
ally
depletion of the conventi
onal ene
rgy su
ch a
s
oil, coal, n
a
t
ural g
a
s
pa
rticula
r
ly on
earth
, the
distrib
u
ted g
eneration
s
(DG
)
which
are
rep
r
e
s
ente
d
by solar en
ergy, wi
nd energy, fuel cell,
cog
ene
rati
on
and othe
r forms of distrib
u
t
ed
power supply
have been attracting mo
re and mo
re
attention and taken a
s
an ideal way of
comprehensi
ve utilizat
ion
of existing resources by the elec
tri
c
power utilities
in many countries
and re
se
archers aroun
d
the worl
d.
Ho
wever, a
great of DG
s bein
g
co
n
necte
d into th
e
distrib
u
tion
n
e
twork ma
ke
s the
conve
n
tional
di
stribution network
b
e
come
multiple
po
wer
sou
r
ces
confi
guratio
ns fro
m
single p
o
wer so
urce
, wh
ich ma
ke
s the power qu
ali
t
y of distribution
netwo
rk, p
o
wer flow a
nd fa
ult levels take
great chan
ge
s, thus di
re
ctly affect
s the fault locatio
n
of
distrib
u
tion n
e
twork an
d relay prote
c
tio
n
[1-2
]. The scale an
d co
nfiguratio
n of modern po
wer
distrib
u
tion
n
e
twork are la
rge
an
d
com
p
licate
d
,
with
the im
prove
m
ent of
dist
ri
bution
netwo
rk
automation,
a variety of i
n
telligent
ele
c
tri
c
al a
pp
lia
nce
s
su
ch
a
s
fee
der terminal u
n
it (F
TU),
remote te
rmi
nal unit (RT
U
) h
a
ve bee
n use
d
, the failure in
such distrib
u
tion
netwo
rk
wo
uld
prod
uce h
u
g
e
am
ount
s of
ala
r
m info
rm
ation, th
e
co
nne
ction
and
quit
of many
DG
s ma
ke
s
the
data
mo
re complex
an
d variable, whi
c
h
fu
rther
i
n
cre
a
sed the
difficulty for the op
erato
r
s in
quickly ha
ndli
ng the fail
ure.
So the
study
on both
the
rapid a
nal
ysi
s
and p
r
o
c
e
s
s
of the upl
oad
ed
data si
gnal
s
and the di
stri
bution net
wo
rk fault lo
ca
ti
on ha
s g
r
eat
significan
c
e
in isolatin
g the
fault, restori
n
g powe
r
serv
ice, improving powe
r
sup
p
ly reliability, ensu
r
ing the
quality of user
power supply
,
etc. About d
i
stri
butio
n net
work fault, do
mestic a
nd fo
reign
schola
r
s put forward
a
variety of fa
ult diagn
osi
s
method
s [3
-10]: tradi
tion
al matrix me
thod is sim
p
l
e
, but the fa
ult
toleran
c
e i
s
n
o
t high, if the real-tim
e informatio
n di
st
ortion o
r
information is n
o
t intact, it is ea
sy
to cause misjudgme
n
t; the artificial intelligen
ce
met
hod ha
s goo
d fault tolerance, but nee
ds
relatively
co
mplex
mod
e
l building, and fault
locati
o
n
efficien
cy is n
o
t high eith
er;
up today, m
o
st
of the algorit
hms in fault
locatio
n
are l
i
mited to
sim
p
le ra
diant o
peratio
n mod
e
of the sing
le
power
supply
,
for
compl
e
x distri
bution system
cont
ai
ning
DG, they lack appli
c
ability. Therefore,
it is nece
s
sary to study more ada
ptable
di
s
t
ribution network
fault loc
a
tion method.
At present, the po
we
r q
u
a
lity monitori
ng(PQ
M)
d
e
v
ices [1
1] are develo
p
ing
into the
monitori
ng system that continuo
us
ly
measures th
e PQ at mult
iple points.
These monit
o
ring
system
s
will provide su
ch functi
on
s a
s
high-sp
eed
communi
catio
n
with an Inte
rnet net
wo
rk
and
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fault Locatio
n of Distrib
u
tion
Net
w
ork Containin
g
Dist
ributed G
ene
rations (Zou B
i
-ch
ang
)
5911
statistical an
alysis th
roug
h uninterru
pted mea
s
ur
eme
n
t
s
an
d
dis
p
la
y w
i
th
a w
e
b
br
ow
ser
.
It
colle
cts,
com
pare
s
, a
nd
an
alyze
s
the
inf
o
rmatio
n t
hat has
gath
e
re
d at
lots of
mo
n
i
toring site
s.
In
this pa
per, b
a
s
ed
on the
sy
stem voltage
sag
ch
ara
c
te
ristics, the cha
nge
s of
the p
o
we
rflow
wh
e
n
the faults o
c
curs an
d th
e i
n
formatio
n
collecte
d
by
th
e po
we
r q
ual
ity monitors,
the faults in t
he
distrib
u
tion n
e
twork
with DGs
can b
e
de
tected an
d lo
cated in
stantl
y
through
co
rrelation
analy
s
is
in order to improve the adaptability of
distribution net
work fault locat
i
on.
2. Correlation Anal
y
s
is
Correl
ation a
nalysi
s
is an i
m
porta
nt met
hod of
sig
nal
pro
c
e
ssi
ng, t
he e
s
sen
c
e
o
f
whi
c
h
is judgi
ng th
e simila
rity of two sign
als in time
domain. At pre
s
ent, the ap
plicatio
ns of
the
correl
ation a
nalysi
s
in
po
wer system are
mainly
concern
ed i
n
f
ault line
sele
ction, fault
p
hase
sele
ction, tra
n
sie
n
t differe
ntial prote
c
tio
n
and indu
cti
on motor fault
diagno
si
s, etc.
From a
statist
i
cal poi
nt of view ,co
r
relati
on anal
y
s
is i
s
the cha
r
a
c
te
rizatio
n
of co
rrelation
degree b
e
twe
en two
or
mo
re vari
able
s
,
assum
e
that
x(t)
、
y(t) a
r
e
two finite en
ergy real
sign
al
waveform, th
e line
a
r
simi
larity deg
re
e
of two
wavef
o
rm
ca
n b
e
descri
bed
by
the
correlat
ion
coeffici
ent
ρ
[12-14].
22
()
()
d
()
d
(
)
d
xt
y
t
t
x
tt
y
t
t
(1)
The co
rrelatio
n coeffici
ent after discretization is:
1
0
11
22
00
()
(
)
()
()
N
n
xy
NN
nn
xn
y
n
x
ny
n
(2)
The matching
degre
e
of co
rrel
a
tion an
alysis is d
e
fine
d as:
1
11
1
22
2
00
0
||
|
(
)
(
)
[
(
)
(
)
]
|
NN
N
xy
nn
n
xn
y
n
x
n
y
n
(3)
Theo
retically in Equatio
n (3) the tim
e
sh
ould b
e
infinit
e
, while
doin
g
co
rrelation
analysi
s
,
the Equation
(3) i
s
still effe
ctive in finite l
ength data
windo
w, and correlation a
n
a
lysis m
a
tchi
ng
degree
value interval i
s
am
ong [0,
1].If
= 1
,
the two
waves exi
s
t in
positive lin
e
a
r
relations
h
ip; If
=-1, th
en
there i
s
a n
e
g
a
tive linea
r
re
lationship b
e
twee
n two
wa
ves. If
=0, the
two wave
s ha
ve no relation
ship.
The
comm
onl
y recogni
ze
d
view is that if
ρ
<0.3
there i
s
no
correlati
on an
d 0.3
ρ
<0.
5
i
s
for low degree
c
o
relation;0.5
ρ
<
0
is for moderate c
o
relation;0.8
ρ
<1 fo
r high d
egree
correl
ation.
3.Fault Loc
a
t
ion for Complex Distribu
tion Ne
t
w
o
r
k
s
3.1. Fault Lo
cation
Algorithm Proces
s
Voltage
sag i
s
referred to
as the
po
we
r voltage RM
S value ab
ru
ptly drop
ped
to 90% ~
10% of the ra
ted voltage a
m
plitude in a
sho
r
t time
, the typical du
ra
tion wa
s 0.5
~ 30
wave cy
cle
[15]. The pri
m
ary ca
use of voltage sag is certai
n
bran
ch current transi
ent
incre
a
se in
th
e
system.Sho
rt
-ci
r
cuit faults,
swit
ch o
peration, the on
-and
-off of transfo
rme
r
an
d ca
pa
citor
set,
sud
den
cut o
f
DG
an
d la
rge in
du
ction
motor
st
a
r
tin
g
, etc can
ca
use
voltage
sag [16].
Wh
e
n
a
fault occu
rs i
n
the
feede
r,
voltage
sa
gs propag
ate
p
r
esenting
diff
erent
mag
n
itude
s a
nd
ph
ase
angle for e
a
c
h feed
er no
de. The nea
rer the n
ode
is to the point of the fault, the gre
a
ter
magnitud
e
the voltage sag
.
Based on the above ch
ar
acteri
stics, the powe
r
quali
t
y (PQ) monito
r
installatio
n
at the root n
ode
s sa
mple
s the dat
a i
n
formation of
different DGs in and out
of
netwo
rk ope
ration, acco
rdi
ng to the
net
work to
pol
og
y data, the pre- a
nd d
u
rin
g
-
fault po
we
rflow
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 591
0 –
5917
5912
data can b
e
obtaine
d, the datab
as
e
of root n
o
d
e
s voltag
e sags
are
built
up throug
h
the
simulatio
n
of
each no
de fa
ult online,the
n
the voltage
amplitude
co
rrel
a
tion mat
c
hin
g
de
gre
e
of
root no
de i
s
cal
c
ulate
d
a
nd by comp
aring
with the
databa
se ,th
e optimum
matchin
g
poi
nt is
extracted, so
the point of
fault can be
determi
ned.
The step
s of the fault loca
tion algorithm
is
s
h
ow
n
in
F
i
gu
r
e
1
.
Figure 1. Flow Ch
art of Fa
ult Locatio
n
The fun
c
tion
above the
dot
ted line: 1) to
cal
c
ul
ate pre
-
fault
data
of each
feed
er such as
voltage, cu
rre
n
t and po
we
rflow etc.(i
ncl
u
ding DGS in
or out of the
netwo
rk); 2) t
o
esta
blish th
e
databa
se
of
spe
c
ified
no
d
e
s volta
ge
sag by
si
mul
a
ting differe
nt nod
es fault
s
, calculate t
h
e
transitio
n re
si
stan
ce.The fu
nction b
e
lo
w
the dott
ed lin
e is to analy
z
e the co
rrelati
on between t
h
e
magnitud
e
of
the
sp
ecifie
d
nod
es voltag
e sag
and
th
e relevent
da
ta group
s i
n
t
he voltag
e
sa
gs
databa
se
an
d
so
rt the
mat
c
hin
g
d
egree
se
que
nc
e, th
e no
de m
a
tching
deg
ree
whi
c
h i
s
nea
res
t
to 1 is ce
rtain
l
y the faulted point.
3.2. Data
Ac
quisition of the Feed
ers a
nd Sy
nchronization
The p
r
op
ose
d
algo
rithm
use
s
voltag
e
and
cu
rre
nt mea
s
ureme
n
ts at the fe
eder ro
ot
node
s a
s
wel
l
as th
e volta
ge me
asure
m
ents
at the
node
s
whi
c
h
DG
s a
r
e i
n
stalled alo
ng t
h
e
feeder. P
r
e
-
a
nd d
u
rin
g
-fau
lt voltage a
n
d
current
ph
aso
r
s mu
st b
e
re
co
rd
ed a
t
the feed
er
root
node
s. T
he
d
u
ring
-fault vol
t
age
sag
ma
gnitude
s
aqui
re
d
at
root
no
des (wh
e
re
the p
o
wer qu
alty
mornito
r
s a
r
e
installed
)
are
used by the
al
gorith
m
in o
r
de
r to find the faulted poin
t
.
Since the
dat
a re
co
rde
d
b
y
FTU an
d the PQM i
s
multifariou
s
a
nd di
sorde
r
ly, and the
time which DGs a
r
e interconne
cted i
s
random, in o
r
der to refle
c
t
the same fa
ult event, all
the
data mu
st be
stri
ctly sync
h
r
onized. Fo
r this p
u
rp
ose, GPS mean
s i
s
u
s
ed to p
r
o
v
ide the unifo
rm
high-accu
ra
cy clock for the whol
e
distri
buted net
works n
eed
ed m
o
rnito
r
ing to
make
su
re th
at all
the re
co
rde
d
data is
stri
ctl
y
synch
r
oni
zed, the data
on both
sid
e
s can
also be
comm
uni
cate
d in
real time by the fiber com
m
unication chann
el [17].
3.3. Load Mo
deling
Usually, load
s
con
n
e
c
ted
at a
distri
b
u
ti
on fee
der pre
s
e
n
t si
g
n
ificant
ran
d
o
m an
d
uncertainty.
Based
on th
e
i
r complex p
o
we
r d
e
man
d
. The
s
e lo
a
d
s
can
be th
ree, two
or
si
ngle-
pha
se. In thi
s
p
ape
r, the
Polynomial
st
atic lo
ad
mo
dels with
con
s
tant p
o
wer,
con
s
tant
cu
rrent
and co
nsta
nt impeda
nce which is
called
comp
re
hen
si
ve load mode
l (CLM
) are forme
d
in cert
ain
ratio as
follows
:
00
00
2
0
2
0
[(
)
(
)
]
[(
)
(
)
]
UU
p
pp
UU
UU
qq
q
UU
P
Pa
b
c
QQ
a
b
c
(
4
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fault Locatio
n of Distrib
u
tion
Net
w
ork Containin
g
Dist
ributed G
ene
rations (Zou B
i
-ch
ang
)
5913
1
1
pp
p
qq
q
ab
c
ab
c
(5)
Whe
r
e
U
is v
o
ltage ap
plie
d to the load,
P
、
Q
are the
corre
s
p
ondin
g
active an
d
rea
c
tive
power respe
c
tively at
U
;
U
0
is the lo
ad nomi
nal voltage,
P
0
、
Q
0
are the correspon
ding
rated
a
c
tive
an
d r
e
a
c
tive
po
we
r
r
e
sp
ec
tively.a,b,c defin
e th
e ratios of
ea
ch type
of l
o
ad o
c
cupi
ed i
n
the power.
p
,
q
are the sub
s
cript sig
n
s of
active and re
active po
wer
respe
c
tively.
Each coeffici
ent
reflect
s
the static cha
r
a
c
terist
ic of lo
ad, its value
is differ
ent
for differe
nt load g
r
ou
p, the
coeffici
ents a
r
e the pa
ram
e
ter identifica
t
ion obje
c
ts o
f
load model.
3.4. Po
w
e
r Fl
o
w
Calcula
t
i
on
Larg
e
amou
n
t
s of
DG
s
acce
ss to th
e di
stri
butio
n
net
work furthe
r
aggravate th
e o
r
iginal
asymmet
r
y of it, which ma
ke
s the conv
entional p
o
wer flow
algo
rithm of three
-
p
hase dist
ribut
ion
netwo
rk
un
su
itable. After the conn
ectio
n
of DG
s, the
accu
racy
of
the po
we
r flo
w
calculation
has
a gre
a
t influe
nce
on the a
c
cura
cy of fa
ult location.
Literatu
re [18
]
propo
se
d a
fast algo
rith
m
whi
c
h
can
ca
culate t
he lo
op net
wo
rk p
o
we
r flo
w
co
ntaining
DG
s but is unfit f
o
r th
ree
-
ph
ase
unbal
an
ced
power flo
w
ca
culatio
n
; In view
of
th
e
complex d
i
stributio
n
n
e
t
work with DGs,
literature [19]
prop
osed a
n
i
m
prove
d
po
wer flo
w
alg
o
rit
h
m, whi
c
h
pe
rforma
nces b
e
tter in d
ealin
g
with loo
p
an
d
unbal
an
ced t
h
ree
-
p
h
a
s
e p
o
we
r flo
w
cal
c
ulatio
n cont
aining
DG
s. I
n
this
pap
er,
an
improve
d
back/forward sweep algo
rith
m[19] is use
d
to calcul
ate
powe
r
flow of the distrib
u
tion
netwo
rk cont
aining DG
s.
3.5.
Voltage Sag Analy
s
is During
-Fa
u
lt
A simple con
f
iguration of
power di
strib
u
tion
network is sho
w
n in
Figure 2, if the fault
occurre
d
at point b, the voltage sa
g
(RM
S
value) refle
c
ted at nod
e a is:
ab
a
a
b
UU
U
(
6
)
ab
a
b
b
UZ
I
(7)
Whe
r
e
a
U
is pre-f
ault voltage a
t
a.
ab
U
is voltage
variation bet
wee
n
pre
-
faul
t and duri
ng-f
ault;
I
b
is
the fault c
u
rrent,
ab
Z
is line impeda
nce be
tween n
ode a
and b.
a
b
c
Loa
d
Lo
a
d
Tr
a
n
s
i
t
i
on r
e
si
s
t
or
Figure 2. Structure of Si
mpl
e
Distri
bution
Network
In pra
c
tical
calcul
ation, in orde
r to accu
ra
tely extract
the transi
ent
voltage sag
values,
the Kalman filtering metho
d
[20] is use
d
to ex
tract the voltage sa
g eigenvalu
e
; it is suitable for
the sho
r
t-te
rm voltage sa
g disturban
ce
analysi
s
with
rapid voltage
chan
ge
s.
3.6. Simulation Test
In order to verify the effectiveness of
the proposed al
gorithm, an I
EEE33 node
system
[21] is use
d
to simulate th
e fault and
analysi
z
e it. The types of DG, install
a
tion location a
nd
cap
a
city a
r
e
sho
w
n
in T
a
ble 1, th
e d
e
tailed
param
e
t
ers of the
sy
stem
are
refe
red t
o
literatu
r
e
[21].
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 591
0 –
5917
5914
Figure 3. IEEE 33-bu
s System
Table 1. Lo
ca
tion, Type and Cap
a
city of DG
Nodes
T
y
pes of
DG
Capacit
y
6
Double-fed w
i
n
d
machines
P=100 kW
,
Q=10
0 kvar
12
Photovoltaic cell
P=300 kW, I=10 A
24
Fuel cells
P=100 kW, U=12.66 kV
29
Constant speed t
y
pe
asynchronous m
o
tor
P=200 kW, s=0.033
PS
:
P,Q
are
the a
c
tive power
and
rea
c
tive
power of DG re
sp
ecti
vely
,
I
is the rated
cur
r
e
n
t
,
U
is the rated voltage, s is the slip
spe
ed.
The more of the numbe
r
of powe
r
qua
lity moni
tors
(PQM) in
stall
ed, the highe
r of the
pre
c
isi
on of the algo
rithm in this pape
r, but w
hen the
mornito
r
ing p
o
ints re
ach a certai
n numb
e
r,
lo
c
a
tion
acc
u
r
a
c
y
w
ill r
eac
h
the
limit
va
lu
e
ow
ing to
th
e in
tr
ins
i
c ch
ar
ac
te
ris
t
ics
o
f
p
o
we
r
distrib
u
tion f
eede
rs. A
s
a
rule of thu
m
b, the insta
llation point
s sho
u
ld be
a
b
le to dete
c
t
the
variation
s
of the voltage
s a
nd cu
rr
ents of
each fee
der,
in this exam
pl
e, the node
s at 2, 5, 11, 19
,
28 we
re in
st
alled the PQ
monitors. Whe
n
the fault occurred
at the point whe
r
e a
DG
is
con
n
e
c
ted, the DG will be
cut off the network in
st
antly and the fault can b
e
easily
located by the
PQ monito
r, so this
kin
d
of
faults
will be i
gnored
i
n
this pape
r. Be
ca
use
the o
pera
t
ion of DG i
s
at
rand
om, therefore the no
rmal ope
ratio
n
data of
the netwo
rk
wi
th DG or wit
hout sh
ould
be
recorded a
n
d
transmitted t
o
the databa
se for
st
ora
g
e
. Here, the load mod
e
l used for sim
u
lat
i
on
is the com
p
rehen
sive loa
d
model a
s
depi
cted in
section 3.2. T
he most co
mmon faults are
singl
e-p
h
a
s
e
d
at site,
so
in this
pap
er,
only si
ngle
-
pha
se fault
s
are
simul
a
te
d and
an
alyzed.
Assu
me that
a fault occu
rre
d at node
25 fo
r pha
se A, the transition
re
sist
ance is 2
, the
duratio
n time
is 0.2s, the result
s of the volt
age sa
gs
at nearby no
des a
r
e
cal
c
u
l
ated and
sho
w
n
in Figure 4, the results
comparison betwee
n esti
mation and
simulation(obtained by PSS/E
softwa
r
e
)
is li
sted in Tabl
e 2.
0.
0
0.
4
0.
8
1.
2
0
1
2
3
4
t/
s
V/
p
u
5
n
o
de
29
no
d
e
7
n
o
de
24
n
o
de
25
Figure 4. Cal
c
ulatio
n Re
su
lt of Voltage Sag
Table 2. Co
m
pari
s
on An
alysis of the Results
of Voltages S
ag at 1.2s
nodes
estmation values /pu
simulation values /pu
29 0.75
0.74
14 0.53
0.52
25 0.02
0.01
7 0.41
0.42
Acco
rdi
ng to
Figure 4 a
nd
Table 2
,
th
e
magnitud
e
of
the voltage
sag at no
de 2
5
is the
large
s
t, the
ot
her no
des’
si
mulation
and
analysi
s
re
sul
t
s of th
e volta
ge
sag
a
r
e
ro
ughly the
sa
me
as the e
s
timated re
sults
of the algorit
hm, wh
ich shows the voltage sa
g an
alysis mo
del
is
confo
r
me
d to the actual si
tuations.
(PS: the estimate
d values a
r
e cal
c
ulate
d
according to th
e
algorithm, yet the s
i
mulation val
ues
are
obtained by PSS/E s
o
ftware).
Whe
n
the fa
ult occurre
d
at node
28, the DG’ b
r
ea
ker at no
de
29 tripp
ed of
f, which
indicated tha
t
the faulted
point whi
c
h i
s
too cl
o
s
e to DG install
a
tion position
may affect the
operation of t
he DG. Fu
rth
e
r a
s
sume th
at faults
o
c
cu
r at othe
r n
o
d
e
s, data
b
a
s
e
of voltage sa
g is
built up
by
co
llecting
the v
a
lue
s
of volta
ge
sag
throu
gh the
po
we
r quality m
onit
o
r
after
33-no
de
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fault Locatio
n of Distrib
u
tion
Net
w
ork Containin
g
Dist
ributed G
ene
rations (Zou B
i
-ch
ang
)
5915
failure
is sim
u
lated .Still a
s
sume
that n
ode
25
oc
cu
rs a
sho
r
t ci
rcuit fault, com
pare
the
re
al-time
acq
u
isitio
n of the powe
r
q
uality monitor of voltage sag
s
with th
e voltage sa
g grou
ps in t
h
e
databa
se a
n
d
make the
correlation m
a
tching de
gre
e
ca
lcul
ation, the result is sh
own in Fig
u
re
5.
0.
0
0.
4
0.
8
5
15
25
t/
s
md
/
p
u
Figure 5. Matchin
g
De
gre
e
Result at No
de 20 when F
ault
The Figu
re 5 sho
w
s that the cal
c
ulation
corr
el
ation m
a
tchin
g
deg
re
e at node 25 is 0.98,
most cl
ose to
1, so by the
matchin
g
re
sults
,the nod
e
25 ca
n be in
ferre
d a
s
the
point of failure,
whic
h is
c
o
nsis
tent with the fac
t
s
.
4. Applicabilit
y
Anal
y
s
is
on the Algorithm
4.1. Influenc
e of the L
o
a
d
Model on the Algorith
m
Environme
n
tal facto
r
s,
lo
ad flu
c
tuatio
n, DG
a
c
cess a
nd
exit, distributio
n tra
n
sformer
con
n
e
c
tion m
ode,
a
nd so
on will
affe
ct the
fault
l
o
cation
re
sults.in
orde
r to
verif
y
the al
gorith
m
has better applicability and robustness, l
e
t’s take
IEE
E
33 node
system as
an ex
ample,
selecti
n
g
different loa
d
model, tra
n
sition resi
st
ance, DG in
terco
nne
ction
and exit
to
evaluate the
algorithm’ applicability in this paper.
The me
ntion
ed loa
d
mod
e
l
in se
ction
2
is comp
reh
e
n
s
ive loa
d
mo
del (CLM
), th
e model
has reflected
the field
operation more accurately, i
n
order to illustrate thi
s
, we use
the const
ant
impeda
nce lo
ad model
(CI
L
D) to an
alyze and compa
r
e. The model
is as follo
ws:
2
0
0
()
U
SS
U
(8)
Whe
r
e
S
0
is the rated
com
p
lex power,
S
is the power
of the con
s
ta
nt impedan
ce
load model.
Let a
si
ngle
-
pha
se fa
ult
occur at
nod
e 25
,the fa
ult location
result
s u
nde
r two
loa
d
model
s are shown in Figu
re 6.
0.
3
0.
5
0.
7
10
20
30
No
d
e
Md
0
0.
9
CI
L
M
CL
M
Figure 6. Fau
l
t Location
Re
sult with Diffe
rent Loa
d Mo
del
Acco
rdi
ng to
Figure 6, the
value of the
matchin
g
de
gree
(Md
)
d
r
ops
a little when the
con
s
tant i
m
p
edan
ce
loa
d
model
(CILM
)
i
s
u
s
e
d
,b
ut it do
es not
a
ffect the
dete
r
minatio
n of
the
fault location
result, since the matchi
ng
degree at no
de 25 is
still the bigg
est.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 591
0 –
5917
5916
4.2. Influenc
e of the Tr
an
sition Re
sistance
Setting the transitio
n resi
stance
as 0.1,
2,15,200
respectively, still let a phase A fault
occur at nod
e 25,by simul
a
tion analysi
s
, the loca
tion
result
s unde
r 4 circum
sta
n
ce
s are sho
w
n
in Figure 7.
No
d
e
0.
3
0.
5
0.
7
0.
9
0
10
20
30
Md
0.
1
2
15
200
Figure 7. Fau
l
t Location
Re
sult with
Diffe
rent Tran
sition Re
sista
n
ce
Acco
rdi
ng to Figure 7, wh
en a fault occurrs
, the co
rrelation m
a
tching de
gre
e
s vary a
little with different tran
sition
resi
stan
ce o
w
ing to
the variation of the
fault
currents, of course, the
existen
c
e of
DG i
s
su
re to
cau
s
e
som
e
influenc
e on
the fault cu
rrents, ho
weve
r, no matter t
he
transitio
n re
sistan
ce i
s
la
rge o
r
small,
it
does n
o
t affect the correlation m
a
tchin
g
deg
ree
analysi
s
, so t
he algo
rithm
has
stron
g
e
r
ability to
resist the influence of transition
resi
stan
ce
s.
4.3. Influenc
e of the
DG’
Conn
ection
If the fault occurre
d
at nod
e 25, shift the DG
fro
m
nod
e 29 to no
de
26, the fault locatio
n
result is sh
own in Figure 8.
0.
0
0.
4
0.
8
5
15
25
No
d
e
Md
Figure 8. Fau
l
t Location
Re
sult
wh
en DG
in and out the Network
Acco
rdi
ng to
the figure, whe
n
the DG is
discon
n
e
cted fro
m
the network, the fault
locatio
n
re
sul
t
is out of question; wh
en the DG i
s
con
necte
d, owin
g to the DG’s voltage supp
ort,
the voltage sag magnitu
de
at node 26 is relatively sm
aller, so the
correlation mat
c
hin
g
deg
ree
at
25 is sli
ghtly redu
ced, yet it does
n
o
t affect the fault location result.
If the ob
serve
d
an
d me
asured voltag
e da
ta is mi
ssed
o
r
failed
to b
e
received fo
r v
a
riou
s
rea
s
on
s, a
n
alternative m
e
thod of
addi
ng fault
cu
rre
nt inform
atio
n a
s
a
com
p
ensation
crite
r
ion
can b
e
used
to prevent misjud
gme
n
t, i.e whether
the fault curre
nt is co
ntinuou
s or n
o
t is
con
s
id
ere
d
a
s
the j
udgm
ent criterio
n
.When
only
a si
ngle fa
ult occu
rs
at a feed
er, t
he
discontin
uou
s current zone
is definitely the fault zone [
22].
5. Conclusio
n
A new fault l
o
catio
n
al
gori
t
hm is propo
sed
by u
s
in
g
co
rrelation
matchin
g
d
e
g
r
ee
and
voltage sa
g cha
r
a
c
teri
stics in a com
p
l
e
x distri
butio
n netwo
rk
co
ntaining
seve
ral DG
s in this
pape
r. The
key factor th
at affects th
e a
l
gorithm
i
s
th
e accu
ra
cy a
nd complete
n
e
ss of the
no
de
voltage
sa
g databa
se whi
c
h can
b
e
g
uara
n
teed by
repe
atedly simulation
s. If the power fl
ow
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fault Locatio
n of Distrib
u
tion
Net
w
ork Containin
g
Dist
ributed G
ene
rations (Zou B
i
-ch
ang
)
5917
cal
c
ulation is more
accu
rate, DGs’
connection and exit will not
affect the accuracy of the fault
locatio
n
. As the intelligent
distrib
u
tion n
e
twork is
fu
rther imp
r
oved
, DGs will b
e
reali
z
ed the t
r
ue
sen
s
e
of "plu
g and
play", yet this will
al
so in
crea
se t
he complexit
y
of fault loca
tion, the met
hod
in this pape
r can p
r
ovide referen
c
e fo
r intelligent
dist
ribution n
e
twork fault lo
cat
i
on. In addition,
how to a
c
hi
eve the optimal
allocation of intelligent
mo
nitors i
n
order to make the
config
uratio
n
of
the monito
rin
g
point
s are smallest b
u
t wi
th hi
gh me
asurem
ent re
du
ndan
cy, is a
n
urge
nt nee
d
for
further study.
Referen
ces
[1]
Z
hou W
e
i, Z
hang Ya
o, Xia
Che
ngj
un, et al
.
Effect of distribute
d
ge
ne
ration o
n
rel
a
y protection
of
distrib
u
ted s
y
stem
.
Pow
e
r System Protecti
o
n
and C
ontro
l
. 201
0; 38(3): 1-
5(in C
h
in
ese).
[2]
Hele
n Ch
eun
g,
Haml
yn A, Lin W
ang, et al.
Netw
ork-integ
r
ated ad
aptiv
e
protection for
feeders w
i
th
d
i
stri
bu
te
d
ge
ne
ra
ti
on
.
IEEE Po
w
e
r and Energy
Soci
et
y
M
eeting–Conv
er
sion and Deliv
er
y
of Electrical
Energ
y
i
n
the 2
1
st
Centur
y, Pittsburg
h
, USA. 2008; 1-8.
[3]
Z
hang Z
h
ao. Gener
al matri
x
a
l
gorit
hm for dis
t
ributio
n s
y
st
e
m
fault locati
ng
.
Electric Power Automatio
n
Equi
p
m
ent
. 20
05; 25(5): 4
0
-4
3(in C
h
in
ese).
[4]
Z
hong
Xia
o
C
h
en, Ke
Cu
i,
Xi
ngYu
Ch
en, Y
anF
an
g
L
i
. Di
stributio
n n
e
t
w
ork fault
di
agn
osis meth
o
d
base
d
on
gra
nul
ar comp
utin
g-BP.
T
e
lko
m
nika-In
don
esi
a
n Jour
nal
of Electrical E
n
g
i
neer
ing.
20
13
;
11(3): 11
81-
11
88.
[5] Don
g
Ha
i
y
in
g
.
Substatio
n
fa
ult di
agn
osis
base
d
o
n
tim
e
seq
u
e
n
ce f
u
zz
y
petri
net
.
T
e
lkomnik
a
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2013; 1
1
(9): 4
861-
487
0.
[6]
Z
hao Y
an, H
u
Xu
eh
ao. Imp
a
c
t
s of distri
bute
d
g
ener
atio
n o
n
d
i
stribut
i
on s
y
stem v
o
lta
ge
sags
.
Pow
e
r
System
Technology
. 200
8; 32(14): 5-9(
in C
h
in
ese).
[7]
Lee
SJ, C
hoi
MS, Kang
SH,
et a
l
. An
i
n
tell
i
gent
an
d
effici
ent fa
ult l
o
cati
on
an
d d
i
a
gno
sis sch
eme f
o
r
radi
al distri
buti
on s
y
stems.
IEEE Trans on Power Deliv
ery
. 200
4; 19(2): 52
4-53
2.
[8]
Pereir
a RAF
,
Silva LGW
da, K
e
zun
o
vic M, et al.
Improved f
ault loc
a
tio
n
on
distributi
on fe
eders b
a
se
d
on matchi
ng d
u
rin
g
-fault volt
age sa
gs
. IEEE Trans on Power Delivery
. 2
009; 24(
2): 852
-862.
[9]
Pereir
a RAF
,
S
ilva
LGW
da, K
e
zun
o
vic M,
et
al.
L
o
catio
n
of
singl
e-l
i
ne-to-
g
roun
d fau
l
ts o
n
distrib
u
tio
n
feeders us
in
g voltag
e meas
u
r
ements
. T
he
Latin Amer
ica
T
r
ansmissi
on
and D
i
stributi
o
n Confer
enc
e.
Carac
a
s, Vene
zuel
a. 200
6; 15-18.
[10]
Dai Sh
ua
ng
yi
n
,
Han Min
x
iao,
Yan W
enl
i. Voltag
e sag
ass
e
ssment for di
stri
butio
n net
w
o
rk contai
ni
n
g
d
i
stri
bu
te
d
ge
ne
ra
ti
on
.
Power System
Technology
. 201
1; 35(7): 145-
14
9(i
n
Chi
nese).
[11]
W
on DJ, C
hun
g IY, Kim JM, e
t
al. A n
e
w
alg
o
rithm to
locat
e
po
w
e
r-
qua
lit
y event s
ourc
e
w
i
t
h
im
prove
d
realiz
atio
n of di
stribut
ed mo
nit
o
rin
g
scheme.
IEEE Trans on Power Deliv
ery
. 2006; 21(
3): 1641-
164
6.
[12]
Li Se
n, Son
g
Guobi
ng, Ka
ng
Xi
ao
nin
g
, et a
l
. T
i
me-domain
fault li
ne s
e
le
ction b
a
se
d o
n
correl
a
ti
o
n
ana
l
y
sis in n
e
u
t
ral ind
i
rect gro
und s
y
stem.
Pow
e
r System P
r
otection a
nd C
ontrol
. 20
08; 3
6
(13): 15-
19.
[13]
Mao Pe
ng, Du
an Yu
qia
n
, Jia
ng Na.
A corr
e
l
atio
n an
al
ysis
base
d
metho
d
to detect fault
y
l
i
ne
un
de
r
singl
e p
has
e to gro
und fa
ult
occurred
in
d
i
stributi
on n
e
tw
o
r
k
w
i
th sm
a
ll curre
nt ne
utral gro
u
n
d
in
g.
Pow
e
r System
T
e
chno
logy.
2
004; 28(
2): 36-
39(i
n
Chi
nes
e).
[14]
Hou
Xin
g
u
o
, W
u
Z
hen
gg
uo,
Xia
Li,
et al.
Stator w
i
ndi
ng
fault
di
agn
osi
s
metho
d
of i
n
ductio
n
motor
base
d
on co
he
rence a
nalys
is
. Proceed
in
gs o
f
the CSEE. 2005; 25(4): 8
3
-8
6(in C
h
in
ese).
[15]
Xi
ao-
xi
na
ng
nin
.
Po
w
e
r Qualit
y Anal
ysis a
nd
C
ontro
l. Beiji
ng
: China El
ectric
Po
w
e
r Press. 200
4
.
[16]
Li Ya
n, Yu
Xi
n
m
ei, Xio
ng
Xi
n
y
i
n
, et al.
A su
rvey on c
a
lcu
l
a
t
ion a
nd a
n
a
l
y
s
is metho
d
s of
voltag
e sag
.
Pow
e
r System
T
e
chno
logy
. 2
004; 28(
14): 74
-78(in C
h
i
nese
)
.
[17]
Lü H
u
, Z
hong
Min
x
iu, W
a
n
g
Lipi
ng, et a
l
.
L
o
catio
n
al
gorit
hm usin
g sam
p
les s
y
n
c
hro
n
i
z
ed b
y
GPS fo
r
EHV transmiss
ion li
ne.
Auto
mation of Electri
c
Pow
e
r Systems
. 19
98; 22(
8
)
: 26-29(i
n
Chi
nese).
[18]
W
ang J
i
an
xun,
Li
u Qu
nfang,
Liu
Hu
iji
n.
F
a
st
an
d
direct
po
w
e
r fl
o
w
alg
o
ri
thm for d
i
stribu
tion
net
w
o
r
k
w
i
t
h
distri
buted
gener
atio
n.
Electric Power Au
tomation Equipment
. 201
1; 3
1
(2): 17-2
1
(in
Chin
ese).
[19]
Ding M
i
n
g
, Guo Xuefe
ng.
T
h
r
ee-p
hase
pow
er flow
for the w
eakly mesh
e
d
distrib
u
tio
n
n
e
tw
ork w
i
th the
d
i
stri
bu
te
d
ge
ne
ra
ti
on
. Procee
din
g
s of the C
SEE. 2009;
9(1
3
): 35-40 (i
n C
h
in
ese).
[20] Z
hang-
bo.
T
h
e
Extraction of t
he Ei
genv
al
ue
of the Vo
lta
ge
Sag a
nd th
e D
i
sturba
nce An
a
l
ysis
. China
electric p
o
w
e
r r
e
searc
h
institut
e, Beiji
ng. 20
0
4
.
[21]
Baran
ME, W
u
F
F
.
Net
w
ork
Rec
onfig
urati
on
in
Distrib
uti
on S
y
stems
fo
r Loss
Re
ducti
on
an
d L
o
a
d
Bala
ncin
g.
IEEE Trans on Power Delivery.
1
989; 4(2): 1
401
-140
7.
[22]
Che
n
Pe
ng, T
eng
Hu
an, T
e
n
g
F
u
sh
en
g. An
al
gorit
hm
for f
ault
locati
on
in
po
w
e
r
distrib
u
tion
net
w
o
r
k
und
er the c
ond
ition
of la
ck
of fault inform
atio
n.
Autom
a
tion
of Electric Pow
e
r System
s
. 20
03;
27(
10)
:
71-7
2
(in Ch
in
ese).
Evaluation Warning : The document was created with Spire.PDF for Python.