scieee Science in your language
[en] (orig)

Transient fault area location and fault classification for distribution systems based on wavelet transform and Adaptive Neuro-Fuzzy Inference System (ANFIS)

Abstract

A novel method to locate the zone of transient faults and to classify the fault type in Power Distribution Systems using wavelet transforms and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) has been developed. It draws on advanced techniques of signal processing based on wavelet transforms, using data sampled from the main feeder current to extract important characteristics and dynamic features of the fault signal. In this method, algorithms designed for fault detection and classification based on features extracted from wavelet transforms were implemented. One of four different algorithms based on ANFIS, according to the type of fault, was then used to locate the fault zone. Studies and simulations in an EMTP-RV environment for the 25kV power distribution system of Canada were carried out by considering ten types of faults with different fault inception, fault resistance and fault locations. The simulation results showed high accuracy in classifying the type of fault and determining the fault area, so that the maximum observed error was less than 2%.

Read accessible full text

Transient fault area location and fault classification for distribution systems based on wavelet transform and Adaptive Neuro-Fuzzy Inference System (ANFIS)

Author: Khaleghi, Ali
Publisher: Vysoká škola báňská - Technická univerzita Ostrava
Year: 2018
DOI: 10.15598/aeee.v16i2.2563
Source: https://dspace.vsb.cz/bitstreams/c81aaab3-bc8e-4e31-9dcc-92e5ee60039b/download
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
T ansien Faul A ea Loca ion and Faul
Classi ica ion o Dis ibu ion Sys ems Based on
Wa ele T ans o m and Adap i e Neu o-Fuzzy
In e ence Sys em (ANFIS)
Ali KHALEGHI 1, Mahmoud OUKATI SADEGH 1,
Mahdi GHAZIZADEH-AHSAEE 2, Ali eza MEHDIPOUR RABORI 3
1Depa men o Elec ical and Elec onics Enginee ing, Facul y o Elec ical and Compu e Enginee ing,
Uni e si y o Sis an and Baluches an, Zahedan, Daneshgah Boule a d, I an
2Depa men o Elec ical Enginee ing, Facul y o Enginee ing, Uni e si y o Zabol, Zabol, I an
3Depa men o Elec ical Enginee ing and Compu e , Facul y o Enginee ing,
Shahid Bahona Uni e si y, Pajoohesh Squa e, Ke man, I an
[email p o ec ed], ouk[email p o ec ed], [email p o ec ed], a.mehdip[email p o ec ed]
DOI: 10.15598/aeee. 16i2.2563
Abs ac . A no el me hod o loca e he zone o an-
sien aul s and o classi y he aul ype in Powe Dis-
ibu ion Sys ems using wa ele ans o ms and Adap-
i e Neu o-Fuzzy In e ence Sys ems (ANFIS) has been
de eloped. I d aws on ad anced echniques o sig-
nal p ocessing based on wa ele ans o ms, using da a
sampled om he main eede cu en o ex ac impo -
an cha ac e is ics and dynamic ea u es o he aul
signal. In his me hod, algo i hms designed o aul
de ec ion and classi ica ion based on ea u es ex ac ed
om wa ele ans o ms we e implemen ed. One o
ou di e en algo i hms based on ANFIS, acco ding
o he ype o aul , was hen used o loca e he aul
zone. S udies and simula ions in an EMTP-RV en-
i onmen o he 25 kV powe dis ibu ion sys em o
Canada we e ca ied ou by conside ing en ypes o
aul s wi h di e en aul incep ion, aul esis ance and
aul loca ions. The simula ion esul s showed high ac-
cu acy in classi ying he ype o aul and de e mining
he aul a ea, so ha he maximum obse ed e o was
less han 2 %.
Keywo ds
Adap i e Neu o-Fuzzy In e ence Sys em (AN-
FIS), elec ical dis ibu ion sys ems, aul
classi ica ion, aul de ec ion, aul loca ion,
wa ele ans o ms.
1. In oduc ion
Faul loca ion is a key issue in he p o ec ion o powe
sys ems and accu a e and swi aul loca ion p ocesses
educe expec ed ene gy ha will no be supplied, in-
c ease sys em e iciency and p omo e cus ome sa is-
ac ion wi h he powe dis ibu ion sys em. Imple-
men a ion o aul loca ion algo i hms in powe sys-
ems needs o conside bo h ansmission and dis ibu-
ion ne wo ks. P olonged aul co ec ion p ocesses in
powe sys ems may cause i epa able damage, and con-
sequen ly, apid aul de ec ion and co ec ion in hese
sys ems is o he u mos impo ance. Measu emen s
o ol age, cu en , powe and equency in ansmis-
sion lines can be made wi h high p ecision, allowing
he exac aul loca ion o be de e mined quickly and
imely ac ion aken o esol e he p oblem. A a i-
e y o algo i hms has been p esen ed in he li e a u e
and a numbe o hem ha e been applied o p ac ical
ne wo ks [1], [2] and [3].
In dis ibu ion ne wo ks, each eede o a dis ibu ion
subs a ion co e s a la ge a ea and, unlike ansmission
ne wo ks, i is no a s aigh line bu a he a line com-
posed o se e al la e als. In addi ion, each eede in-
cludes a a ie y o dis ibu ion ans o me s. The e-
o e, aul loca ion in dis ibu ion ne wo ks is mo e
challenging, cos ly, and less accu a e han o ans-
mission sys ems. Few s udies ha e explo ed his issue
[4], [5], [6] and [7].
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 155
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
Faul loca ion in a dis ibu ion ne wo k is aced wi h
he ollowing p oblems ha complica e he p ocess o
de e mining he aul loca ion:
•The wide expansion o dis ibu ion ne wo k eed-
e s and hei la e als.
•Di e en ypes o o e head and unde g ound ca-
bles, wi h a ying c oss-sec ional a eas and phase
con igu a ions, in di e en pa s o he dis ibu-
ion ne wo k.
•The p esence o dis ibu ion ans o me s in di e -
en pa s o he dis ibu ion ne wo k, wi h a ied
nominal capaci ies and loading ac o s.
•The exis ence o jus one da a logge o aul ol -
age and cu en a he beginning o he dis ibu-
ion ne wo k eede s.
The me hods o aul loca ion in powe sys ems a e
di ided in o wo majo ca ego ies: impedance and a -
elling wa es [5] and [7]. Howe e , as aul loca ion is
mo e di icul in dis ibu ion ne wo ks, gi en p oblems
such as se e al la e als, aul loca ion in dis ibu ion
ne wo ks is di ided in o wo majo pa s: (1) loca ing
he aul zone, (2) de e mining he exac loca ion o
aul . Fi s , he aul ed zone is exi ed om he ne -
wo k and hen exac aul loca ion is de e mined. The
pu pose o his pape is o in oduce new me hods o
de e mine he a ea o he aul ( aul ed zone).
Re e ence [8] es ima ed he aul zone using cu en
pa e ns, and e-close - use coo dina ion. Re e -
ence [9] used an algo i hm based on a ma ix o
loca e he aul zone. In his me hod, he a ays
we e made o bina y da a (0 and 1) ansmi ed
om Feede Te minal Uni s (FTU). Re e ence [10]
p oposed a synch onized ol age-based non-i e a i e
me hod by aking ad an age o he subs i u ion
heo em. By eplacing he aul ed line wi h a sui ably
adjus ed cu en sou ce injec ing he same amoun
o ansmission line cu en , an equi alen ne wo k
was es ablished. A wo-s age aul loca ion algo i hm
using Radial Basis Func ion (RBF) based Suppo
Vec o Machine (SVM) and Scaled Conjuga e G a-
dien (SCALCG)-based A i icial Neu al Ne wo k
(ANN) was p oposed in [11]. In he i s s age, he
magni udes o he undamen al ha monics o he
posi i e sequence ol age and cu en signals o he
aul ed phases we e inpu o RBF-based SVM o ge
an app oxima e aul a ea. In he second s age, he
SCALCG-based ANN was implemen ed o indica e he
p ecise aul loca ion using high equency cha ac e -
is ics. The impedance-based me hod p oposed in [12],
loca es he aul in a hie a chical manne , in which he
aul ed zone, aul ed line and aul poin a e loca ed in
u n. Re e ence [13] p oposed a mul i-objec i e op i-
miza ion me hod using a Non-Domina ed So ing Ge-
ne ic Algo i hm (NSGA) algo i hm o de e mine he
loca ion o aul s in he dis ibu ion sys em.
When a aul occu s in powe sys ems, as and ac-
cu a e aul classi ica ion ( o pos - aul analysis) and
es o a ion o he sys em o i s o iginal s a e a e o
he u mos impo ance. In many aul loca ion me h-
ods, in o ma ion abou he ype o aul is he basis
o aul loca ion, so he co ec classi ica ion o aul s
a ec s he p ecise de ec ion o he aul zone. Gi en
he impo ance o aul classi ica ion o elay pe o -
mance, many s udies ha e ocused on aul classi ica-
ion p oblems in he ansmission sys em [14], [15] and
[16]. S udies o aul classi ica ion in powe sys ems
a e di ided in o wo g oups: (1) designs ha u ilize
s eady s a e elec ical componen s [17], [18] and [19],
and (2) designs ha u ilize ansien elec ical compo-
nen s [20], [21] and [22]. Fo example, Re . [18] used an
algo i hm based on uzzy logic o de e mine he ype
o aul in adial unbalanced sys ems. Re e ence [20]
used a new app oach, based on wa ele ans o ms, o
iden i y and classi y he ype o aul by compa ing he
wa e o ms. Re e ence [22] de eloped a new me hod o
classi y he ype o aul , using Adap i e Neu o-Fuzzy
In e ence Sys ems (ANFIS). This me hod was based on
applying wa ele ans o ms o he aul cu en .
Since mos aul s occu ing in powe sys ems a e
ansien in na u e [23], in his pape , we p opose a new
algo i hm o de e mine he a ea o ansien aul in
dis ibu ion ne wo ks using ANFIS. Based on ea u es
ex ac ed om he main eede cu en , no el algo-
i hms o de ec ing and classi ying di e en ypes o
aul s a e p esen ed. This in o ma ion is hen used o
de ec he aul zone.
Fou algo i hms we e designed o de ec he aul
zone, one o each ype o aul (single-phase-
o-g ound, double-phase- o-g ound, phase- o-phase,
h ee-phase/ h ee-phase- o-g ound). The aul zone
was hen de e mined using ained ANFIS ne wo ks.
EMTP-RV is used o simula ions. Simula ions we e
ca ied ou in h ee s eps: (1) iden i ica ion o he aul ,
(2) classi ica ion o he aul ype, and (3) loca ion o
he aul ed zone. This me hod is less complex han
p e iously epo ed me hods as he e a e no long and
complex calcula ions in ol ed, and he aul ed zone is
de e mined p omp ly (app oxima ely 4 cycles). This
me hod also has highe accu acy, when compa ed wi h
o he s udies [22] and [24]. A u he ad an age o his
me hod is ha iden i ica ion o he aul and classi i-
ca ion o he aul ype a e comple ely independen o
line and aul pa ame e s.
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 156
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
2. Wa ele T ans o m Analysis
The inpu s o he designed algo i hms we e ea u es
ex ac ed om he main eede cu en , which we e de-
i ed om wa ele ans o ms. Wa ele ans o ms can
be conside ed as an ex ension o Fou ie ans o ms,
bu ins ead o wo king on one scale ( equency o ime),
hey wo k on mul iple scales. This mul i-scale ea u e
o wa ele ans o ms leads o he decomposi ion o
a signal in o se e al scales, wi h each scale ep esen -
ing a pa icula ea u e o he signal unde s udy [25].
Wa ele ans o ms di ide he signal in o di e en le -
els, each le el con aining equency- ime in o ma ion
o he signal. In his s udy, ea u es o he p o ile o
ansien s a e aken o he 2000–4000 Hz ange. The
p ocess o Mul i-Resolu ion Analysis (MRA) o he
inpu signal is shown in Fig. 1.
HPFLPF
HPFLPF
HPF
LPF
HPFLPF
HPFLPF
HPFLPF
Inpu signal
2000-4000 Hz
0-2000 Hz
1000-2000 Hz
500-1000 Hz
250-500 Hz
125-250
Hz
62.5-125
Hz
0-1000 Hz
0-500 Hz
0-250 Hz
0-125 Hz
0-62.5 Hz
500-4000 Hz62.5-500 Hz0-62.5 Hz
T ansien
Cha ac e is ics
Ha monic
Cha ac e is ics
Main equency
Cha ac e is ics
Fig. 1: F equency di ision mul i- esolu ion le els up o 6.
3. Adap i e Neu o-Fuzzy
In e ence Sys ems (ANFIS)
ANFIS is one o he models o neu o- uzzy sys ems.
Neu al ne wo ks and uzzy sys ems a e bo h indepen-
den sys ems. Inc easing aining p ocesses, inc easing
membe ship unc ions, and independen uzzy ules a e
ac o s in complexizing p oblem sol ing. This led o
he de elopmen o he ANFIS me hod, which com-
bines he bene i s o bo h neu al ne wo ks and uzzy
logic. ANFIS aims o elimina e he disad an ages o
each o hese sys ems while e aining hei complemen-
a y bene i s [26]. Fuzzy logic in his sys em is used as
a con ibu o o he aining algo i hm and can adjus
he pa ame e s o he uzzy sys em.
4. Es ima ing Faul Time
Algo i hm
Fi s ly, he main eede cu en s in each cycle a e e-
cei ed and hei essen ial cha ac e is ics a e ob ained
om hei wa ele ans o ms. Based on a wa e o m
analysis o he aul signal a di e en imes, i was
ound ha in all cases, he wa e o m ob ained om
he wa ele ans o m o he main eede cu en , a
he ime o ansien aul , possessed he highes jump.
By compa ing momen a y a ia ions in a sample wi h
he p e ious sample in each cycle, he maximum a ia-
ion o he wa ele ans o m can be de ec ed. I he e
we e no aul in he selec ed cycle, he sum o a ia-
ions would be equal o ze o. I a aul akes place,
he ime o maximum change is conside ed as he ime
o aul occu ence. Fo example, wa ele ans o m o
phase-A du ing aul occu ence in node 6 o he 25 kV
powe dis ibu ion sys em o Canada [27] wi h a esis-
ance o 40 ohms and a aul incep ion o 10 deg ees,
is shown in Fig. 2. Figu e 3 shows changes om mo-
men o momen . I can be seen ha he momen wi h
he highes change in alue was conside ed as he aul
occu ence ime.
As shown in he lowcha in Fig. 4, o a oid in e -
e ence in de ec ing he ime o he ansien aul wi h
en e o exi loads, a e de e mining he s a ime o
dis u bance in he wa ele ans o m cu en signal, he
du a ion o he dis u bance is calcula ed using a an-
sien de ec ion lag. I his ime is less han 3 cycles,
he dis u bance in he cu en signal is conside ed as
a ansien aul and he aul ime is es ima ed. Us-
ing he algo i hm in Fig. 4, he aul de ec ion ime
and aul occu ence we e 4 mic oseconds, he eby in-
dica ing high p ecision in a cycle o 16.6 milliseconds.
Dis u bance du a ion
Fig. 2: Cu en wa ele ans o m o phase A (Du ing aul ).
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 157
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
Dis u bance du a ion
Fig. 3: Ins an aneous changes o he cu en wa ele ans o m
o phase A.
Recei e cu en o main
eede
Ia,Ib,Ic
Wa ele T ans o m
SWa=WT(Ia)
SWb=WT(Ib)
SWc=WT(Ic)
ΔSWa= SWa(i)-SWa(i-1)
ΔSWb= SWb(i)-SWb(i-1)
ΔSWc= SWc(i)-SWc(i-1)
(ΔSWa) 0
and
(ΔSWb) 0
and
(ΔSWc) 0
Yes
No
Wha ime is maximum o ΔSWa?
o
Wha ime is maximum o ΔSWb?
o
Wha ime is maximum o ΔSWc?
Nex cycle
Maximum o (ΔSWa, ΔSWb,
ΔSWc) is
Conside ed as Time o aul
Calcula e Dis u bance Du a ion using
T ansien De ec ion lag
>3 cycle
T ansien Faul has no
occu ed
<3 cycle
Maximum o (ΔSWa, ΔSWb,
ΔSWc) is
Time T ansien aul occu ed
Fig. 4: Flowcha o he algo i hm o es ima ing aul ime (iis
he numbe o samples aken in one cycle).
5. Faul Classi ica ion
Algo i hm
In mos aul -loca ing algo i hms, aul classi ica ion
is one o he mos impo an pa s o he p ocess. In
his s udy, a new algo i hm o classi y he ype o aul
based on ea u es ex ac ed om he main eede cu -
en is p esen ed. Analysis o he wa ele o he aul
signals e ealed ha signals ex ac ed om wa ele
ans o ms displayed speci ic beha io o each ype o
aul .
Fo example, in he case o phase- o-phase aul s,
he sum o he wa ele ans o ms o he phases in-
ol ed om he main eede cu en was almos equal
o ze o (Fig. 5), while in he case o single-phase- o-
g ound aul s, wa ele ans o ms o he wo phases
wi hou aul s we e almos equal. Figu e 6 shows he
wa ele ans o ms o he h ee-phases when a C-phase-
o-g ound aul ook place. I can be seen ha he
Phase-C wa ele ans o m wi h he aul is dis inc ,
bu wa ele ans o ms o he o he wo phases dis-
play simila beha io .
Fig. 5: Wa ele ans o m o h ee-phase cu en du ing AC
phase- o-phase aul (du ing aul ).
Fig. 6: Wa ele ans o m o h ee-phase cu en du ing Cg
single-phase- o-g ound aul (du ing aul ).
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 158
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
I he abo e condi ions a e no es ablished o de e -
mine he ype o aul s, he absolu e alue o he di e -
ence be ween he maximum cu en peak o each phase
du ing and be o e he aul is compu ed. Then, mag-
ni udes ob ained om each phase a e compa ed wi h
each o he . I alues ob ained o one phase a e negligi-
ble compa ed o o he phases, he phase is conside ed
o be wi hou aul and he numbe o aul y phases
can be es ima ed (Fig. 7, Fig. 8, Fig. 9 and Fig. 10).
Fo example, in Fig. 8 a double-phase- o-g ound
aul occu ed in he AB phases. I can be seen ha
he di e ence in cu en peak in phase C be o e and
du ing he aul is insigni ican , bu in phases A and
B, he e a e signi ican di e ences (app oxima ely 2
imes g ea e han be o e he aul o phase A and
app oxima ely 6 imes g ea e o phase B). Simila ly,
o h ee-phase- o-g ound aul s, shown in Fig. 7, he
di e ence be ween cu en peak be o e and du ing he
Fig. 7: Wa e o m o h ee-phase cu en du ing ABCg h ee-
phase- o-g ound aul (du ing aul ).
Fig. 8: Wa e o m o h ee-phase cu en du ing ABg double-
phase- o-g ound aul (du ing aul ).
aul o all h ee phases is high. Fo phase- o-g ound
aul s, only he di e ence be ween he cu en peak
be o e and du ing he aul in he aul y phase is
high (Fig. 10). Simila ly, o phase- o-phase aul s
(Fig. 9) gi en he di e ence be ween he cu en peak
be o e and du ing he aul , he same conclusion can
be eached o each phase. Figu e 11 shows a lowcha
o he aul classi ica ion algo i hm. This algo i hm is
independen o he esis ance, loca ion and aul incep-
ion.
The algo i hm o he classi ica ion o aul ype is
ini ially implemen ed wi h espec o he de ec ion al-
go i hm, p o ided ha a aul has aken place in he
ne wo k.
In he i s s ep, he wa ele ans o ms o he h ee-
phases o he main eede cu en a e summed up pai -
wise and p o ided ha he es ima ed sum o a leas
one o he wa ele ans o ms is ze o i is conside ed
Fig. 9: Wa e o m o h ee-phase cu en du ing AC phase-
phase aul (du ing aul ).
Fig. 10: Wa e o m o h ee-phase cu en du ing Cg single-
phase- o-g ound aul (du ing aul ).
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 159

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
P oposed algo i hm o
aul ime iden i ica ion
Yes
NSWa+NSWb 0
O
NSWb+NSWc 0
O
NSWa+NSWc 0
YES
A1=max(Ia)
B1=max(Ib)
C1=max(Ic)
A2=max(-Ia)
B2=max(-Ib)
C2=max(-Ic)
A= A1-A2
B= B1-B2
C= C1-C2
Which one is
smalle ?
i
A<<
i
C<<
i
B<<
0<A/B<e
&
0<A/C<e
0<B/A<e
&
0<B/C<e
0<C/A<e
&
0<C/B<e
yes
Phase o phase aul
BC Phase o phase aul
AC Phase o phase aul
AB
yes yes
Th ee phase aul
ABCg/ABC
NO
NO
NO
NO
NSWa NSWb
O
NSWb NSWc
O
NSWa NSWc
YES
YES
A1=max(Ia)
B1=max(Ib)
C1=max(Ic)
A2=max(-Ia)
B2=max(-Ib)
C2=max(-Ic)
A= A1-A2
B= B1-B2
C= C1-C2
i
A>> i
C>>
i
B>>
0<B/A<e
&
0<C/A<e
0<A/B<e
&
0<C/B<e
0<A/C<e
&
0<B/C<e
YES
Single phase o g ound aul
Ag Single phase o g ound aul
Bg
Single phase o g ound aul
Cg
YES YES
Which one is
bigge ?
NO NO
A2=max(-Ia)
B2=max(-Ib)
C2=max(-Ic)
A1=max(Ia)
B1=max(Ib)
C1=max(Ic)
A= A1-A2
B= B1-B2
C= C1-C2
NO
NO
NO
go nex s ep
No Nex cycle
Recei e cu en o main
eede
Ia,Ib,Ic
No malized Wa ele T ans o m
NSWa=NWT(Ia)
NSWb=NWT(Ib)
NSWc=NWT(Ic)
YES
Which one is
smalle ?
i
A<< i
C<<
i
B<<
0<A/B<e
&
0<A/C<e
0<B/A<e
&
0<B/C<e
0<C/A<e
&
0<C/B<e
yes
Double phase o g ound
BCg Double phase o g ound
ACg Double phase o g ound
ABg
yes yes
Th ee phase aul
ABCg/ABC
NO
NO
NO
0<e<0.5
Fig. 11: Flowcha o he aul classi ica ion algo i hm.
as a phase- o-phase aul . In he nex s ep, o ensu e
ha he aul is phase o phase, he cu en peak di -
e ence be o e and du ing he aul is checked o each
phase. I he di e ence be ween he cu en peak be-
o e and du ing he aul is high o all h ee phases,
he aul ype is de e mined as h ee-phase. I he sum
o wa ele ans o ms o pai ed phases is no app oxi-
ma ely equal o ze o, i p oceeds o he nex s ep. A
his poin , i he wa ele ans o m o one phase is sim-
ila o ano he phase, a single-phase- o-g ound aul is
conside ed. Then, o ensu e ha he aul is single-
phase, he di e ence be ween he cu en peak be o e
and du ing he aul is calcula ed. I he di e ence
be ween he cu en peak be o e and du ing he aul
is signi ican in mo e han one phase a single-phase-
o-g ound is ejec ed and he nex s ep is conside ed.
The lowcha in Fig. 11 shows ha in classi ying phase-
phase- o-g ound and h ee-phase aul ypes only he
absolu e alue o he di e ence be ween he maximum
cu en peak o each phase du ing and be o e he aul
is used.
6. Faul A ea Loca ion
Algo i hm
To a oid he ailu e o all lines du ing a aul and o
con inue powe supply o he sys ems, dis ibu ed sys-
ems we e di ided in o sepa a e egions [28]. The di-
ision o egions in he dis ibu ion sys em was based
on sys em opology, he p esence o p o ec i e de ices,
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 160
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
load load
load
load load
load
load
load load
load
1 2 3 4 5 6 8 9 10 11 12
22 24 25
15
19 18
20 16
load
23
13
14 21
17
7
R
zone1
zone2
zone3
zone4
zone5
52 R
F1 F2
F3
Fig. 12: Diag am o he 25 kV powe dis ibu ion sys em o Canada.
Faul classi ica ion
Algo i hm 2Algo i hm 1 Algo i hm 3 Algo i hm 4
Loca e aul ed zone
Single phase o
g ound aul Phase o phase aul Double phase o
g ound aul Th ee phase aul
Fig. 13: Flowcha o loca e he aul y zone.
he leng h o eede s, e c. [29]. As an example, he
25 kV powe dis ibu ion sys em o Canada [27] was
s udied as a p ac ical sys em (Fig. 12). This sys em
was di ided in o i e egions wi h espec o p o ec-
i e de ices, and in case o a aul in each egion, he
aul y zone was de ec ed and only he speci ied a ea
was disconnec ed om he ci cui .
A e comple ing he p ocess o es ima ing aul ime
and classi ica ion o aul ype, he aul zone loca ion
was de e mined using he ANFIS sys em. To loca e he
aul zone, da a om he h ee-phase wa ele ans o m
cu en was di ec ed o he app op ia e algo i hm, one
o each ype o aul (Fig. 13). In his algo i hm, o
educe da a in e e ences and diminish e o s in he
loca ion o aul zone using he ANFIS sys em, da a
we e analysed in smalle ba ches. Acco ding o he
algo i hm lowcha shown in Fig. 14, a e de e mining
he ype o aul , he ained in e ence sys em o his
aul ype was used.
7. E alua ion o he
Pe o mance o he
P oposed Algo i hms
7.1. Es ima ing he Faul Time
To e alua e he pe o mance o he algo i hm in es i-
ma ing aul ime, i was implemen ed o en ypes o
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 161
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
Wha is he ype o
aul ?
Faul y zone
Ag
Cg
Bg AB
AC
BC ABg
ACg
BCg ABC ABCg
Single phase
aul Double phase
o g ound aul Th ee phase
aul
Double phase
aul
T ained
ne wo k
(ANFIS1)
T ained
ne wo k
(ANFIS2)
T ained
ne wo k
(ANFIS3)
T ained
ne wo k
(ANFIS4)
T ained
ne wo k
(ANFIS5)
T ained
ne wo k
(ANFIS6)
T ained
ne wo k
(ANFIS7)
T ained
ne wo k
(ANFIS8)
T ained
ne wo k
(ANFIS9)
T ained
ne wo k
(ANFIS10)
T ained
ne wo k
(ANFIS11)
Fig. 14: Flowcha o he aul a ea loca ion algo i hm.
aul s wi h a ied s a ing angles a 20 andom loca-
ions in he 25 kV powe dis ibu ion sys em o Canada
[27]. The esul s a e shown in Tab. 1. The accu acy o
he algo i hm was compu ed using Eq. (1):
pe cen age accu acy =
eal ime o aul −es ima ed aul ime
du a ion o one cycle ·100.(1)
In Tab. 1, i can be seen ha he la ges e o in es-
ima ing he aul ime was 1 %, which indica es a high
accu acy o he algo i hm.
7.2. E alua ing he Pe o mance o
he Faul -Type Classi ica ion
Algo i hm
To e alua e he pe o mance o he aul - ype classi i-
ca ion algo i hm, he algo i hm was un o en ypes
o aul in one loca ion in he 25 kV powe dis ibu ion
sys em o Canada (node 6, aul esis ance 40 ohms,
aul incep ion 8 deg ees) [27]. The esul s a e p e-
sen ed in Tab. 2. Fo example, in he i s ow, a single-
phase- o-g ound aul Ag ook place. As discussed in
Sec. 4. , in single-phase aul mode, he wa ele
ans o ms o he aul - ee phases a e almos iden i-
cal (SWb =SWc) and by compa ing he di e ence
be ween he peak cu en o he phase, bo h du ing
and be o e he aul , wi h he o he phases, i can be
de e mined whe he a aul has occu ed in phase A

max 
id
a
−max 
ib
a

.
Simila ly, o he ou h ow whe e an AB phase-
o-phase aul ook place, by compa ing he wa ele
ans o m o he cu en , i was obse ed ha he
wa ele ans o ms o he A and B phases we e sym-
me ical (SWa +SWb = 0). Also, by compa -
ing he di e ence o he cu en peaks, bo h be-
o e and du ing he aul , wi h he o he phases, i
could be de e mined whe he a aul had aken place
in phases A o B 
max 
id
a

−
max 
ib
a

and


max 

id
b




−

max 

ib
b



. I should be no ed
ha he p oposed me hod had 100 % accu acy in clas-
si ying he aul ype, his was no obse ed in p e ious
s udies [22].
7.3. E alua ing he Faul A ea
Loca ion Algo i hm
As men ioned p e iously, he main pu pose o he p e-
sen ed pape is o de e mine he aul egion in he dis-
ibu ion sys em, and he p ecise loca ion o he aul is
no conside ed. On he o he hand, o ob ain su icien
da a o aining he ANFIS ne wo k, di e en ypes o
aul s in a ying nodes and unde di e en aul esis-
ance, aul incep ion and aul ypes ( o 1000 epochs)
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 162
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 2 |2018 |JUNE
Tab. 1: E alua ion o he pe o mance o he es ima ing aul ime algo i hm.
Type o Faul loca ion Faul impedance Real ime Es ima ed aul E o o es ima ed
aul (node) Ωo aul (ms) ime (ms) aul ime
ABg 1 10 15 14.8293 1 %
BCg 6 20 11 10.9283 0.4 %
Bg 12 40 7 6.9601 0.2 %
Ag 8 80 4 4.0156 0.09 %
ACg 9 10 1 1.0984 0.5 %
BCg 11 20 15 14.9519 0.2 %
Bg 13 40 11 10.9323 0.4 %
Bg 14 80 7 7.0155 0.09 %
Bg 15 10 4 4.0354 0.2 %
Bg 16 20 1 1.075 0.4 %
Bg 17 40 15 14.9804 0.1 %
Cg 18 80 11 10.9876 0.07 %
Cg 19 10 7 6.9325 0.4 %
ABC 6 0 11 10.9402 0.3 %
AC 8 20 11 10.9679 0.1 %
ABg 11 80 4 4.0275 0.1 %
Bg 14 80 11 10.9679 0.1 %
Bg 16 40 7 7.0036 0.02 %
Cg 21 20 4 4.0354 0.2 %
Cg 18 20 11 10.9718 0.01 %
Tab. 2: E alua ion o he pe o mance o he es ima ing aul ime algo i hm.
Type o SWa SWb SWc max ib
amax ib
bmax ib
cmax id
amax id
bmax id
c
aul
Ag 5.703 -3.24 -3.24 144.1712 149.0179 152.1604 943.6845 146.4857 143.3865
Bg 4.556 -2.566 -2.566 180.7221 149.0179 144.0064 171.35 7109.5 143.3865
Cg 13.78 1.758 1.758 144.1712 149.0179 143.3865 149.0797 197.2185 4566.4
AB 0.6724 -0.6752 0.0009 144.1714 149.0179 143.3915 6400.5 6401.8 143.3865
BC 0.0006 -7.908 7.906 144.1712 149.0179 143.3865 153.959 7395.7 7287.2
AC -3.794 0.1803 3.807 144.1712 149.0179 143.3865 2334.2 149.029 2446.8
ABg -1.497 -0.5625 4.46 182.5789 149.0179 159.6409 885.6958 5454.3 143.3865
BCg 1.714 -8.864 6.0343 144.1712 149.0179 143.3865 144.174 6700.6 3773
ACg 7.531 2.004 -5.567 144.1712 149.0179 143.3865 636.7509 156.3637 3719
ABCg 3.241 -0.629 -2.207 144.1712 149.0179 143.3865 949.447 6627.6 3902.8
we e simula ed and he collec ed da a used o aining
he ANFIS ne wo k (Tab. 3).
The da a necessa y o es ing he ained ne wo k
a e shown in Tab. 4. Faul s iden i ied ou side he aul
zone by he ANFIS ne wo k we e conside ed as inco -
ec . The accu acy o he p oposed me hod is calcu-
la ed om Eq. (2):
pe cen age e o =inco ec answe
o al o answe ·100.(2)
Tab. 3: T aining da a.
Faul y node 1,9,12,13,16,19,20,23,25
Faul esis ance 0,10,20,40,80
Faul incep ion 11,43,75,118,161
To al o da a 1125
Tab. 4: Tes ing da a.
Faul y node 6,10,15,18,22
Faul esis ance 5,15,30,50,70
Faul incep ion 8,54,96,108,144
To al o da a 575
Table 5 shows he da a used o each aul , based on
he aul phase as well as he sha e o each one in he
p ocess o aining and es ing he ANFIS ne wo k. In
Tab. 6, he e alua ion o he aul loca ion algo i hm
and he accu acy o he algo i hm o each aul phase
based on he numbe o da a is shown.
Tab. 5: Da a used o ain and es he aul a ea loca ion al-
go i hm.
Type o aul Faul y phase P ocess
Single phase Ag Bg Cg T ain Tes
o g ound 225 425 225 275 150
Double phase ABg ACg BCg T ain Tes
o g ound 150 150 150 300 150
Double AB AC BC T ain Tes
phase 150 150 150 300 150
Th ee ABCg/ABC T ain Tes
phase 150 100 50
The p oposed me hod o aul iden i ica ion was
assessed o 10 ypes o aul s in di e en loca ions
and unde di e en condi ions and compa ed wi h he
me hod de eloped in [24]. The esul s a e shown in
Tab. 7.
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 163