Int ern at i onal  Journ al of  P ower E le ctr on i cs a n d  Drive  S ystem   (I J PE D S )   Vo l.   11 ,  No.   4 ,  Decem be r   2020 , p p.   209 9 ~ 2106   IS S N:  20 88 - 8694 ,  DOI: 10 .11 591/ ij peds . v 1 1 .i 4 . pp 209 9 - 2106        2099       Journ al h om e page :  http: // ij pe ds .i aescore.c om   Islandin g   detecti on   in   a   distri bution   net work   wit h   distri buted   generat ors   using   signal   pr ocessin g   techni ques       Seong - Ch e ol   Kim 1 ,   P ap ia   R ay 2 ,   Sur ender   Redd y   Sa lk ut i 3   1,3   Depa rtment   of   Railroad   and   Elec tr ic a l   Engi ne er ing,   Woosong   U nive rsity ,   Da ej e on,   Repub li c   of   Korea .   2   Depa rtment   of   El e ct ri ca l   Eng in ee ring ,   Ve er   Sur endr a   Sa i   Univ er sity   of   Technol o gy,   Burl a,   Indi a.       Art ic le   In f o     ABSTR A CT   Art ic le   hist or y:   Re cei ved   A pr   1 4 ,   20 20   Re vised   M a y   1 ,   20 20   Accepte d   J un   17 ,   20 20       Thi s   pap er   prop oses   quic k   &   a cc ura te   isla ndin g   detec t ion   t ec h nique   for   a   distri buti on   sys t em   with   distr ibu te d   generat ors   ( DG s).   Here   two   sche me s   of   isla nding   d et e ct i on   base d   on   sig nal   proc essing   is   proposed   of   which   one   is   base d   on   disc rete   wave le t   tra nsf orm   (DWT )   wit h   art if icial   neur al   ne twork   (AN N),   and   ano the r   one   is   base d   on   S - tra nsfor m   wi th   ANN.   T he   n ega t ive   seque nce   cur ren t /vol ta g e   signal s   are   re tri ev ed   at   t arg eted   DG   loca ti on   which   are   used   fo r   isl a nding   d et e ction   in   the   distr ibut io n   sys te m .   Here ,   the   wave le t   and   S - tra nsform s   are   used   fo r   f aul t   locati on   an d   cl assifi cation   appl i ca t ions.   Further,   the   fe ature   ex tracti on   is   used   for   red u cing   the   siz e   of   da ta   ma tr ix   by   tra nsforming   it   i nto   set   of   fe at ur es.   In   thi s   work,   par ticle   sw arm   o pti mizatio n   (PS O)   base d   f eature   sel ec t ion   sc hem e   is   appl i ed.   Simu la t ion   resu lt s   on   te st   sys te m   ind ic a te   t he   eff icac y   of   pr oposed   isla nd ing   detec t ion   te chn i ques.   Ke yw or d s :   Ar ti fici al   ne ur a l   netw orks   Distrib uted   ge ner at ion   Feat ur e   sel ect ion   Islan ding   detec ti on   Sign al   proces sing   te ch niques   Stoch a sti c   opti miza ti on   This   is   an   open   acc ess   arti cl e   un der   the   CC   BY - SA   l ic ense .     Corres pond in g   Aut h or :   Su r en der   Re dd y   Sal ku ti ,     Dep a rtme nt   of   Ra il ro ad   a nd   E le ct rical   Eng i ne erin g,   Woos ong   U nive rsity,   17 - 2,   Ja ya ng - Don g,   D ong - G u,   Daejeo n   -   3460 6,   Re public   of   K orea.   Emai l:   su re nde r@wsu.ac . kr       1.   INTROD U CTION   Pr otect io n   e ng ineers   face   c halle ng e s   in   certai n   a pp li c at ion s   s uc h   as   isl and i ng   de te ct ion   in   a   distrib ution   ne twork .   Isla nd   is   a   sit uation   w her e   a   pa rt   of   a   util it y   gr i d   is   energize d   by   DGs   an d   el ect rical ly   separ at e d   from   rest   of   t he   netw ork.   It   is   very   m uc h   im portant   to   det ect   isl and in g   conditi on   co rrec tl y   &   qu ic kly ,   ot herwise   ris k   of   da mage   to   the   powe r   s ys te m   rises   &   sa fety   hazar d   for   the   perso nnel   bec om es   a   matt er   of   c oncern.   Se ver al   powe r   s ys te m s   face   t he   iss ue   of   l ow   f re qu e nc y   os ci ll at ion   due   hea vy   loa d   conditi on   or   to   sy ste m   int ercon necti on.   The   sta bili ty   of   a   pow er   s ys te m   is   a   non - li near,   dyna mic   ph e nome non,   and   it   de pends   on   poorl y   da mp e d   low   fr e quenc y   os ci ll at ion s .   T hese   os c il la ti on s   play   vi ta l   ro le   in   the   a nalysi s   of   sta bili ty   of   t he   s ys te m.   If   t he se   os ci l la ti on s   are   not   dam pe d   s uffici ently,   an   unsta ble   op e rati on   may   oc cu r   an d   it   may   le ad   to   a   netw ork   colla pse .   T he refor e ,   it   is   esse ntial   to   m on it or   the   modal   par a mete rs   of   t he   low   fr e quen cy   os ci ll at ory   s ign al s   for   dyna mic   sy ste m   sec ur it y   [ 1].     Re fer e nce   [2]   pr e sents   sev eral   isl and i ng   detect ion   te chn i qu e s   s uch   as   act ive   in ve rter - reside nt   te chn iq ues ,   pa ssive   in ver te r - reside nt   te chn i qu e s,   co m munica ti on s - ba sed   te ch niqu es,   an d   util it y   le vel   te chn iq ues   for   distrib uted   power   gen e rati on   sy ste ms.   In   Re fer e nce   [ 3],   an   isl and i ng   sch eme   base d   on   wav el et   singular   e ntr opy   is   im plement ed   in   a   micr ogrid   with   DG.   A   hy br i d   a naly zi ng   meth od   w hich   a nalyze s   the   d - axis   volt age   c omp on e nt   in   2   way s   f or   detec ti ng   the   isl an di ng   preci sel y   is   pr ese nted   in   [ 4].   A   hy br id   isl a nd i ng   detect ion   meth od   f or   micr ogr ids   ( MGs)   wit h   var io us   c onne ct ion   po i nts   to   sma rt   gri ds   ( SG s )   is   pr opose d   in   [5],   a nd   it   ba se d   on   pr ob a bili ty   of   isl and i ng   cal culat ed   at   the   SG   side   an d   sent   to   the   ce nt ral   co ntro l   f or   MG.   Re fer e nce   [6]   analyzes   the   se ns it ivit y   of   16   powe r   s ys te m   par a mete rs   w hi ch   are   us e d   in   passi ve   met hods   to   Evaluation Warning : The document was created with Spire.PDF for Python.
                   IS S N :   2088 - 8 694   In t J   P ow  Ele c  &  Dr i   S ys t ,   V ol .  1 1 , N o.   4 ,  D ecembe r   2020   :   209 9   –   210 6   2100   detect   non - isl a nd i ng   an d   isl a nd i ng   eve nts.   An   isl an ding   de te ct ion   met hod   base d   on   isl and i ng   disc rimi nation   factor   is   pro pose d   in   [7],   a nd   it   is   de rive d   from   the   s uperim posed   c omp on e nts   of   v oltages.   An   isl and i ng   detect ion   meth od   ba sed   on   a ve rag e   abs olu te   fr e qu e nc y   dev i at ion   value   is   presente d   in   Re f eren ce   [ 8].     An   isl an ding   de te ct ion   met hod   for   M G s   c onsiderin g   small   scal e   synch r onou s   ge ne rato rs   is   pr e sente d   in   [ 9].   T un e d   f il te rs   connecte d   at   the   DG   te rmin al s   a re   util iz ed   f or   isl a nding   detect ion   in   MG   is   pr opose d   in   [10].   An   isl an di ng   detect ion   method   base d   on   modal   c ompone nts   of   volt ages   is   disc us s ed   in   [11].   An   act ive   isl and in g   detect ion   meth od   for   an   in ve rter - ba sed   DG   is   pr opose d   in   [ 12].   A   pas sive   loca l   mu lt i - crit eria   base d   isl and in g   detec ti on   te ch nique   with   fa ult   dete ct ion   as   well   as   isl and i ng   ve r ific at ion   lo gic   is   pro po se d   in   [13].   Re fer e nce   [14 ]   prese nts   an   e f fici ent   a ppro ac h   for   buil ding   decisi on   tree s - base d   intel li ge nt   rela y.   A   ne w   la r g e   performa nce   isl and in g   searc h   seq uen ce   met hod   im plemente d   to   4   isl a nd i ng   detect ion   a ppr oac hes   is   pr opose d   in   [ 15].   A   ne w   passive - based   anti - isl and i ng   te chn i qu e   f or   both   s ynch ron ous   mac hin e   &   inv e rter   base d   DGs   is   pro po se d   in   [16].     It   can   be   noti c ed   from   the   li te ratur e   s urve y   that   there   is   a   pr essi ng   re quir ement   for   fast   and   acc ur at e   al gorithm   f or   isl and i ng   detect ion .   T her e fore,   in   t his   pap e r,   two   sig nal   pr oc essing   ba sed   isl and in g   detec ti on   methods ,   i.e. ,   DWTwit h   A N N,   a nd   S - tra nsfo rm   with   ANN   a re   propose d.   He re,   PS O   base d   feat ur e   s el ect ion   (F S )   al gorith m   is   us e d.   T he   s uitabil it y   an d   eff ect ive ness   of   pro pose d   isl and i ng   detect io n   met hods   has   been   examine d   on   di stribu ti on   net work   wit h   DGs.         2.   PROP OSE D   SIGNAL   P R OCESS IN G   T ECHNIQ UES   FO R   ISL ANDING   DETE CTIO N   In   t his   pap e r,   DWT   a nd   S - tra nsfo rms   are   a pp li e d   for   detect in g   the   isl and i ng   in   t he     distrib ution   ne twork .       2.1.   I slan ding   detecti on   usin g   discr e te   wav el et   tr an s fo r m   (DWT )   wit h   ANN   Wav el et   t ran s f orm   ( WT )   is   an   a da ptive   si gnal   processi ng   t echn i qu e   for   non - sta ti on a r y   s ign al s   a nd   it   extractsi nfo rm at ion   from   the   data   se ries.   Ch oice   of   desire d   m oth e r   wa velet   is   vital   f or   good   pe rfo rma nc e,   a nd   Daubec hies   is   consi der e d   as   the   a pprop riat e   on e   f or   a nal yzing   the   tra ns ie nt   even t   [17 ].   H ere,   WT   is   a pp li ed   for   isl an ding   de te ct ion .   Sig na l   is   colle ct ed   f rom   the   rela yi ng   e nd   of   dist rib ution   li ne   is   dec omp os ed   i nto   8   le vel   by   WT,   and   it   is   de picte d   in   Fig ure   1.   Sam pling   f requen c y   is   co ns i der e d   as   30   k Hz,   t her e fore,   8   le vel   decomp os it io n   can   be   pe rformed   [ 18].   For   the   acc ur ac y   of   the   m odel ,   samplin g   f requen c y   of   30   kHz   is   require d.   Dec omp os it ion   proc ess   de pe nds   on   sam pling   f requen c y   (F s )   a nd   num ber   of   sa mp le s   Decomp os it io n   process   with   dig it al   filt erin g   ap proac h   is   dep ic te d   in   Fi g.1.   In   t his   fig ur e ,   X [n]   is   sign al ,   g[ n]   is   hi gh   pa ss   filt er,   a nd   h[n ]   is   low   pass   filt er   [19 - 20] .   ‘d1’ a nd   ‘a1’   a re   1 st   le vel   deta il   and   appr ox imat io n   coeffic ie nts,   re sp ect ively .           Figure   1 .   Di gital   filt ering   te ch nique   of   wa vel et   trans form   ( WT).       ANN   is   a   c omp utati on al   m od el   t hat   sim ul at es   the   str uct ur al   a nd   funct ion al   as pects   of   bi ologica l   neural   net work   [21].   In   this   work,   a   m ulti la yer   feed - f orw ard   ne ur al   net work   (FFN N)   with   bac k   pro pa gati on   trai ning   al gorit hm   is   a pp li ed .   Flow   c ha rt   of   i sla nd i ng   detect ion   te ch nique   in   a   distri bu ti on   netw ork   usi ng   WT   and   ANN   is   de picte d   in   Fig ure   2.   At   the   ta r ge te d   distri bu ti on   gen e rati on   ( DG)   locat i on,   the   neg at ive   se qu e nc e   current/v oltage   sign al s   are   ret ri eve d.   T hese   s ign al s   a re   decomp os e d   us in g   DWT   [ 22] .   T he   obta ined   sta ti sti cal   featur e s   f rom   the   sig nals   are   reconstr ucted   f rom   detai le d   c oeffici ents.   Fe at ur e   set   with   best   predict ive   power   is   cho s en   by   us in g   P SO   al gorith m.   T her e a fter,   t he   trai n   &   te st   data   i s   generate d   f rom   wi de   va riat ion   of   Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  P ow Elec   & Dri S ys t   IS S N:  20 88 - 8 694       Islandi ng d et ec ti on  in  a distri bu ti on  network  wi th d ist rib ute d gen e ra t or s  usi ng … ( Se ong - Che ol Kim )   2101   loading   c ondit ion   a nd   it   is   nor mali zed   [ 23] .   A NN   is   trai ne d   with   t he   se le ct ed   featu res   as   input.   T he n,   the   te sti ng   is   performe d   with   the   trai ne d   neural   ne twork .       A t   t h e   t a r g e t e d   D G   l o c a t i o n ,  r e t r i e v e     n e g a t i v e   s e q u e n c e   v o l t a g e / c u r r e n t   s i g n a l s S i g n a l s   a r e   d e c o m p o s e d   u s i n g   D W T S t a t i s t i c a l   f e a t u r e s   o b t a i n e d   f r o m   t h e   s i g n a l s   a r e   r e c o n s t r u c t e d   f r o m   d e t a i l F e a t u r e   s e t   w i t h   g o o d   p r e d i c t i v e   p o w e r   i s   c h o s e n   b y   a p p l y i n g   P S O T r a i n   a n d   t e s t   d a t a   i s   g e n e r a t e d   f r o m   w i d e   v a r i a t i o n   o f   l o a d i n g   c o n d i t i o n   a n d   i t   i s   n o r m a l i z e d A N N   i s   t r a i n e d   w i t h   t h e   e x t r a c t e d   f e a t u r e T e s t i n g   w i t h     t r a i n e d   a r t i f i c i a l   n e u r a l   n e t w o r k     Figure   2 .   Isla ndin g   detect ion   te chn iq ue   us i ng   DWT   a nd   A NN.       2.2.   I slan ding   detecti on   usin g   S - Tr ansf or m   wit h   ANN   S - tra ns f orm   re pr ese nts   the   ti me - f re qu e nc y   relat ion s hip   with   insta ntane ous   val ues   of   fr e qu e nc y,   ph a se   an d   am plit ud e   [24 ].   It   giv es   the   a bsolutel y   ref e re nced   phase   in formati on   &   f reque ncy   i nv a rian t   amplit ude   res ponse   &   pro vide s   bette r   si gn al   cl arit y   f or   tra ns ie nt   si gn al .   Gen e rates   a   c onto ur   w hich   is   simple   to   vis ualiz e,   w her eas   WT   re quires   sta nd a r d   mu lt i - res olu ti on   anal ys is.   H oweve r,   it   re quires   hi gh e r   c omplex   com pu ta ti on   [ 25].   S - tra ns f or m   co nf in es   the   real   &   imagina ry   c omp onents ,   phase   as   well   as   amplit ud e   sp ect r um s   i ndepende ntly.   It   is   te rmed   as   a bs ol utely   ref e r enced   phase   i nformat ion.   S - trans form   is   use d   for   sever al   a ppli cat ion s   s uc h   as   fa ult   cl assifi cat ion ,   l ocati on   a nd   moda l   a nalysi s   of   sig nal.     Figure   3   de picts   the   flo w   cha rt   of   pro posed   isl and in g   detect ion   sch eme   us ing   S - tra ns f orm   &   A NN .   Neg at ive   se quence   vo lt age/c urren t   si gnal s   are   ac qu ire d   at   the   ta r geted   DG   locat io n.   S - t ran s f or m   is   ap plied   to   these   sig nals,   and   e ne rgy   i s   determi ned   [ 26 - 27].   T he   c umulat ive   sum   (CU M S U M)   of   ene rgy   si gn al   is   check e d   f or   t he   co nv e r gen ce   conditi on,   a nd   this   CU M S U M   is   us e d   f or   isl a nd i ng   detect io n.         A t   t h e   t a r g e t e d   D G   l o c a t i o n ,   c o l l e c t   n e g a t i v e   s e q u e n c e   v o l t a g e A p p l y   S - t r a n s f o r m   f o r   t h i s   s i g n a l C a l c u l a t e   t h e   E n e r g y C u m u l a t i v e   s u m   ( C U M S U M )   o f   e n e r g y   i s   d e t e r m i n e d N o n - i s l a n d i n g I s   ( C U M S U M )   ³   T h r e s h o l d   ? N o Y e s I s l a n d i n g     Figure   3 .   Isla ndin g   detect ion   te chn iq ue   us i ng   S - tra ns f orm   and   A N N.       2.3.   Fea tu re   e xtracti on   It   is   us e d   for   re du ci ng   data   m at rix   siz e   by   tr ansfo rming   int o   feat ur es .   F rom   dec omp os ed   coeffic ie nt s   of   D WT,   si x   s ta ti sti ca l   feature s   are   e xtracte d.   F or   W T,   t he   num ber   of   re const ru ct e d   c oe ff ic ie nts   a re   8   (i.e.,   d1 - d8)   a nd   t he   total   featu re   matri x   is   96   (i. e.,   8 th dec omp osi ti on   le ve l ×   2nos   of   sig nals   ×   6   nos   of   featu res ).   Total   featu re   set   for   each   si gn al   for   S - tra ns f orms   is   18   featur e s   [ 28].   In   S - tra ns f or m,   3   pa ramete rs,   i.e. ,   fr e qu e nc y,   pha se   an d   a mp li tu de   a re   c on si dered.     The   tra ns ie nt   e nerg y   sig nal   ha s   la rg e   val ue   as   com pa red   to   normal   si gn a l.   Stan d ar d   de viati on   of   an   undisto rted   si gnal   is   eq ual   to   on e ,   howe ver,   f or   tra ns ie nt   sign al ,   t his   val ue   dev ia te s   f r om   one.   Durin g   the   isl and in g   eve nt ,   the   sig nal   m ean   bec om es   non - zer o,   wh e re as   its   value   is   zero   for   undist or te d   si gnal   [29 - 30].   Fo r   an   undist ort e d   sig nal,   t he   val ue   of   kurto sis   is   3,   w he re as   its   value   is   gr eat er   tha n   3   for   a   tra ns ie nt   sign al .   Fo r   an   undisto r te d   sig nal,   the   value   of   s ke wness   is   ze ro   &   non - zer o   f or   dis torted   case.   T he   val ue   of   e ntr opy   is   la rg e   value   f or   transient   sig nal,   an d   it   is   le ss   f or   an   un distor t ed   si gn al .       Evaluation Warning : The document was created with Spire.PDF for Python.
                   IS S N :   2088 - 8 694   In t J   P ow  Ele c  &  Dr i   S ys t ,   V ol .  1 1 , N o.   4 ,  D ecembe r   2020   :   209 9   –   210 6   2102   3.   FEATU RE   S EL ECTION   Feat ur e   sel ect ion   ( FS)   is   sel ect ing   s mall   s et s   of   feature s   w hich   dete r mine’s   ta r get   pro per ly .   It   reduces   the   c omplexit y   of   predict ion   &   le a rn i ng   te ch niqu e   &   ma kes   pr e dicti on   acc ur a cy   to   rise.   In   va rio us   su pe r vised   le a rn i ng   act ivit ie s,   m os t   of   the   featur e s   sho ws   re dunda ncy   [ 31].   T her e f or e ,   FS   is   a   sig ni ficant   appr oach   to wa rd s   le ar n ing   of   la rg e   mu lt i - fea ture d   data   matr ix.       3.1.   Fea tu re   s el ection   using   PSO   algorith m   PSO   is   stoc ha sti c   opti miza ti on   ap proac h   m otivate d   by   the   beh a vior   of   bi rd   floc king   or   fis h   sch oo li ng.   In   P SO ,   the   pote nti al   so luti on   is   c al le d   pa rtic le s,   w hich   interact s   am ong   thems el ves   to   fi nd   glo ba l   op ti mal   s olu ti on.   In   e very   it er at ion ,   eac h   pa r ti cl e   has   its   own   posit ion   &   ve locit y,   an d   is   updated   by   two   best   values   cal le d   pb e st&g best.   Fo r   furthe r   det ai ls   of   P SO   al gorithm ,   t he   r eader   ma y   re f er   ref e ren ce   [32].   T he   flo w   c har t   of   P SO   base d   FS   is   de picte d   in   Fi g.   4.       I n i t i a l i z e     t r a i n i n g   d a t a   s e t C a l c u l a t e   p o s i t i o n   &   v e l o c i t y   f o r   e a c h   p a r t i c l e E v a l u a t e   f i t n e s s I s   p a r t i c l e   f i t n e s s   >   p b e s t   ? I s   p a r t i c l e   f i t n e s s   >   g b e s t   ? R e a c h e d   m a x i m u m   n u m b e r   o f   i t e r a t i o n s ? U p d a t e   p b e s t U p d a t e   g b e s t U p d a t e   p a r t i c l e   p o s i t i o n   a n d   v e l o c i t y Y e s Y e s Y e s O p t i m i z e d   p a r a m e t e r s   a n d   f e a t u r e   s u b s e t No     Figure   4:   Fl ow   cha rt   of   P SO   base d   feat ur e   s el ect ion .       3.2.   Par ame te r   set ting   f or   ge nera ting   tr ain   &   test   d ata   mat ri x   The   t rain   &   te st   data   matri x   is   no rmali zed   betwee n   [ 0,1].   F or   eac h   s ign al ,   48   feat ures   (8   le vel   decomp os it io n   ×   6   feat ur es )   f or   D WT,   a nd   18   featu res   (3   modal   pa ramet ers   ×   6   featu res )   f or   S - tra nsfo rms   are   c onside red.   O ptimal   featu res   a re   obta ine d   from   featur e   sel ect ion   met hod,   a nd   the y   are   8   f or   DWT   an d   4   for   S - tra nsf or m.   In   this   pa pe r,   t wo   ty pes   of   trai ni ng   data   s et s   are   co ns ide red.   T he   first   da ta   set   is   form ed   by   the   c ombinati on   of   7   re sist an ces   ( 0Ω,   1Ω ,   5Ω ,   10 Ω ,   25Ω,   35 Ω ,   45Ω )   an d   6   i ncep ti on   ang le s   ( 10 ° ,   20 ° ,   30 ° ,   40 ° ,   50 ° ,   85 ° ).   Her e ,   te n   form s   of   eve nt   are   consi der e d   with   300   locat io ns   [33 ].   T he refo re,   the   total   of   126000   (i.e.,   10   ×   300   ×   42)   trai ning   da ta   matri x   is   f ormed.   T he   sec ond   data   set   is   f ormed   by   the   com bin at io n   of   9   resist ances   (2 Ω ,   4 Ω ,   6 Ω ,   9 Ω ,   12 Ω ,   20 Ω ,   30 Ω ,   40 Ω ,   50 Ω )   an d   8   ince ption   an gles   (5 ° ,   11 ° ,   17 ° ,   24 ° ,   32 ° ,   45 ° ,   60 ° ,   90 ° ).   The r efore,   the   total   of   216000   (i.e. ,   10   ×   300   ×   72)   trai ning   data   se t   is   gen e rated .   In   this   pa pe r,   the   accurac y   (in   %)   ca n   be   cal cul at ed   by   us i ng   [ 34],     %   Accu racy =   | Total   num be r   of   samp l es − Mi sc lassi fie d | Tot al   numb er   of   sample s × 100                                       ( 1 )       4.   SY STE M   DESCRIPTIO N   The   s ys te m   un der   stu dy   f or   is la nd in g   detect ion   in   a   distrib ut ion   netw ork   is   de picte d   in   Figure   5.   T he   gr i d   data   c onsidere d   in   t his   w ork   has   a   f requ ency   of   50   Hz,   vo lt age   of   120   kV,   zer o   se quence   pa ramete r s:   R 0   is   1.7 28   Ω/ km   and   L 0   is   0.0 55   H/km,   an d   t he   posit ive   se que nce   par a mete rs :   R 1   is   0.5 76   Ω /km   a nd   L 1   is   0.018   H/km.   Distrib ut ion   li ne   data   c on si der e d   in   this   w ork   has   the   li ne   le ng t h   of   30   km ,   zer o   se qu e nce   pa ra me te rs:   Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  P ow Elec   & Dri S ys t   IS S N:  20 88 - 8 694       Islandi ng d et ec ti on  in  a distri bu ti on  network  wi th d ist rib ute d gen e ra t or s  usi ng … ( Se ong - Che ol Kim )   2103   R 0   is   0.826   Ω/ km ,   L 0   is   6.6 4   mH/km,   C 0   is   0.100 2   nF / km,   posit ive   se qu e nce   pa rameters :   R 1   is   0.2 306   Ω/ km ,   L 1   is   2.1   m H/km   a nd   C 1   is   0.226 6   nF/ km.   The   DG   par a mete rs   ha s   ge ner at or   data:   wind   s peed is 10   m/s,   numb e r   of   wind   tu rb i nes   a re   6,   nominal   po wer   is   9   MW   (i .e.,   6   ×   1.5   MW) ,   volt age   is   400   V,   fr e quenc y   is   50   Hz;   DC   bus   volt age   re gula tor   gain   ha s     is   8,     is   400;   gri d - side   c onve rter   c urren t   re gula to r   gain   ha s     is   0.83,     is   5;   spe ed   re gu la to r   ga in   has     is   3,     is   0.6;   r otor   side   co nverter   c urr ent   regulat or   ga in   has     is   0.6,     is   8;   reac ti ve   powe r   &   vo lt age   re gu la t or   gain   a re   0.0 5,   20;   pitch   c ontr oller   gai n   is   150;   co nverte r   data   ha s   nomin al   DC   bus   volt age   has   1150   V   a nd   DC   bus   capaci tor   is   10000   F.             Figure   5:   S ys te m   un der   stu dy   for   isl an ding   de te ct ion .       Fr om   fi gure   5,   it   can   be   obse rv e d   that   the   s ys te m   un der   st udy   has   two   9   MW   wind   f ar ms   dri ve n   by   wind   t urbine.   Each   9   MW   w ind   fa rm   co ns i sts   of   si x   wind   tur bin e s   of   1.5MW   capa ci ty   co nnect ed   to   120   kV   gr i d   t hroug h   25kV,   30   km   fe eder.   T he   sam pling   fr e qu e nc y   c onside red   is   200   kHz   &s yst em   f re qu e nc y   is   50   Hz.   T he refor e ,   there   is   4000   samples/c ycle.   Loa d   dema nds   are   va ried   at   DG   e nd   as   well   as   at   po int   of   commo n   c oupl ing .   Cu rr e nt   sa mp le s   a re   retrieved   at   DG - 1   and   DG - 2   e nd s.   In   t his   w ork ,   the   wind   s pe ed   is   consi der e d   as   10   m/s .   Her e ,   t wo   c ycles   of   c urren t   sig nal   is   co ns i der e d,   one   j us t   befo re   i sla nd i ng   a nd   a no t her   after   isl an ding.       5.   RESU LT S   A ND   DI SCUS S ION   In   t his   pap e r,   18   dif fer e nt   c ombinati ons   of   loading   c onditi on s   are   util iz ed   to   ge ner at e   t rain   &   te st   data.   F or   each   sign al ,   48   featur es   (8   le vel   de com posit ion   ×   6   feat ur e s)   a re   consi der e d   f or   DWT.   Te n   opti mal   featur e s   a re   se le ct ed   f r om   the   PS O   based   op ti mal   feat ure   sel ect ion   me thod,   a nd   the y   are   fed   to   A NN   for   trai ning.   T his   t raine d   ne ur al   ne twork   is   use d   f or   te sti ng   pu r po s e.   To   a void   ove r   fitt ing   of   da ta ,   te n   f old   cro ss   validat io n   is   c arr ie d   out   [35 ] - [ 36].7 0%   trai ning   data   a nd   30%   te sti ng   da ta   will   be   sel e ct ed   ra ndoml y   f or   t he   te st   set .   Be cau se   of   this,   s ome   obse rv at io ns   ma y   not   be   sel ect ed   in   t he   te st   set ,   w her eas   ot her s   may   be   sel ect ed   more   t han   once.   T his   resu lt s   in   te st   s et   ov e rlap ping   [37].   To   overc om e   this   sit uation,   cr os s - valid at ion   is   performe d.   In   this   pap e r,   t he   cr os s   valida ti on   is   pe rfo rm ed,   a nd   chec ke d   that   for   al l   the   te st   cases   e rror   is   within   the   li m it ed   range.   Also ,   the   pro pos ed   meth od   in   the   pa per   is   more   r obus t   than   c ross - validat ion,   because   in   cr oss - validat io n   the   op e rati ng   co nd it io n   is   al rea dy   bee n   seen   by   the   neural   ne twork ,   w her eas   in   the   pro po se d   te c h nique,   operati ng   co ndit ion   f or   te st   data   is   al te red   from   t rain ed   one   [38 ].   As   mentio ne d   ea rlie r,   in   this   w ork,   t wo   isl an ding   de te ct ion   a ppro a ches   a re   us e d,   and   the y   a re   •   Metho d   1:   Isla nd i ng   detect io n   us in g   D WT   with   ANN.   •   Metho d   2:   Isla nd i ng   detect io n   us in g   S - T ransform   with   A N N.     Evaluation Warning : The document was created with Spire.PDF for Python.
                   IS S N :   2088 - 8 694   In t J   P ow  Ele c  &  Dr i   S ys t ,   V ol .  1 1 , N o.   4 ,  D ecembe r   2020   :   209 9   –   210 6   2104   5.1.   I slan ding   detecti on   by   u sing   me thod   1   As   me ntio ned   earli er,   in   this   method,   D WT   with   ANN   is   use d   for   isl an din g   detect io n   in   distrib utio n   netw ork.   Table   1   presents   pro po s ed   meth ods   accu racy   f or   isl and in g   &   non - isl and in g   c ondi ti on s.         Table   1.   Acc uracy   of   pro pose d   a ppr oach es   f or   isl and i ng   &   non - isl and i ng   conditi ons   Co n d itio n   No .   of   sa m p les   Co rr ect   id en tific ati o n   Mis - id en tification   Accuracy   ( %)   Metho d   1   Metho d   2   Metho d   1   Metho d   2   Metho d   1   Metho d   2   Metho d   1   Metho d   2   Islan d in g   18   18   18   17   0   1   100   9 4 .4   No n - islan d in g   54   54   54   49   0   5   100   91       The   ob ta i ned   ou t pu t   a fter   usi ng   discrete   wav el et   tra nsf orm   ( D WT)   is   fed   to   an   A NN,   a nd   it   is   dep ic te d   in   Fig ur e   6.           Figure   6.   Isla ndin g   detect ion   by   us i ng   DWT   an d   ANN.       Figure   6   s how s   '0'   for   fau lt   free   co nd it io n   in   the   s ys te m,   i.e .,   under   no rma l   op e r at in g   c onditi on s ,   a nd   it   sh ows   ' 1'   w he n   the   fa ult   occ ur s   un der   isl an ding   c onditi on.     5.2.   I slan ding   detecti on   by   u sing   me thod   2   Figure   7   de pic ts   the   plo t   for   retrieve d   sig na l   samples   ta ke n   at   the   sta nda rd   fr e quenc y   of   50   Hz ,   an d   the   fa ult   is   occ urred   after   4000   sa mp le s.           Figure   7.   Isla ndin g   detect ion   us in g   S - tra nsfo rm.       Fr om   fi gure   7,   it   can   be   observ e d   that   a fter   40 00   sa mp l es,   there   is   t he   detect ion   of   isl and in g   conditi on,   a nd   hen ce   the re   is   a   sud den   sur ge   in   the   s ys te m   f reque ncy.       Evaluation Warning : The document was created with Spire.PDF for Python.
In t J  P ow Elec   & Dri S ys t   IS S N:  20 88 - 8 694       Islandi ng d et ec ti on  in  a distri bu ti on  network  wi th d ist rib ute d gen e ra t or s  usi ng … ( Se ong - Che ol Kim )   2105   6.   CONCL US I O NS   This   pa pe r   pr opos es   an   a ppro ac h   for   dete ct ing   isl an ding   in   distrib ution   s ys te m .   T wo   isl a nd i ng   detect ion   te ch ni qu es   a re   pro pose d   in   t his   pa per   base d   on   DWTwit h   ANN,   an d   S - tra nsfo rm   wit h   A N N.   F rom   the   simulat io n   resu lt s   on   c ons idere d   syst em   for   isl an ding   de te ct ion   in   a   di stribu ti on   network   sho ws   th at   the   DWT   in   c omb inati on   with   A NN   ga ve   100%   accu racy,   a nd   it   is   r obust   and   m or e   acc ur at e   tha n   t he   oth e r   te chn iq ues   pr e sented   in   the   l it eratur e.   In   this   pa per,   featu re   sel ect io n   is   pro po se d   b y   usi ng   P SO   al go rithm.   Feat ur e   sel ect ion   ma kes   pro po s ed   ap proac h   m ore   s uperi or   tha n   oth e r   methods   re por te d   in   the   li te ratur e .   Exten ding   the   pr ese nt   w ork   to   meet   the   pro blem   cau sed   due   to   the   sud de n   cha nge   in   l oad   w hich   ma y   create   false   al arm   in   t he   isl an di ng   de te ct ion   pr ocess   is   a   f uture   rese arch   sco pe.       ACKN OWLE DGE MENTS   This   resea rc h   work   has   bee n   carried   out   ba sed   on   t he   sup port   of   “ W oos ong   Un i ver sit y' s   Aca demic   Re search   F undi ng   -   (20 19 - 2020) ” .       REFERE NCE S   [1]    A.   Pouryekt a,   V.   K.   R am a chandara mur thy,   N.   Mithu l ana n tha n ,   A.   Arula mp alam,   "Is la nd ing   Dete c ti on   and   Enha nc em en t   of   Microgr id   Perfor ma nc e, "   I EEE   S yste ms   Journal ,   vol.   12 ,   no .   4,   pp .   3131 - 3141 ,   De c.   2018 .   [2]    I. J.B.   Álvar ez,   E. I. O .   R ive r a,   " Survey   of   Distr ibut ed   Gene r ati on   Island ing   De te c ti on   Methods , "   I EE E   Latin   Ame rica   Tr ansacti ons ,   vol .   8,   no .   5,   pp .   565 - 570 ,   Sept.   2010.   [3]    A.   Samui   and   S.   R.   Sam antara y,   "Wa v elet   Singula r   Ent r opy - Based   Isla nding   Det ec t ion   in   Di str ibut ed   Gene ration, "   in   I EE E   Tr ansacti o ns   on   Powe r   Del iv ery ,   vo l.   28,   no .   1,   pp .   411 - 418 ,   Jan.   2013.   [4]    S.   Murugesan,   V.   Mural i,   S.A.   D ani e l,   "H ybrid   Analyzing   T ec hn i que   for   A ct iv e   Is la nding   Detect io n   Based   ond - Axis   Curre nt   Inj ec t ion, "   I EE E   Sy stems   Journal ,   v ol.   12 ,   no .   4,   pp .   3608 - 3617,   De c.   2018.   [5]    S.   D.   Kerm any ,   M.   Joorabia n ,   S.   Deil a mi   and   M.   A.   S.   Masoum,   "H ybrid   Islandi n g   Dete c ti on   in   Microgr idWith   Multi ple   Conn e ct ion   Points   to   Smar t   Grids   Us ing   Fuzzy - N eur al   Network , "   IE EE   Tr ansacti ons   on   Powe r   Syste ms ,   vol .   32 ,   no.   4,   pp.   2640 - 2651,   Jul.   2017.   [6]    S.   Ra za,   H.   Mokhlis,   H.   Arof,   J . A.   La gh ari,   H.   Mohama d,   "A   Sensiti vi ty   Analy sis   of   Diffe r ent   Pow er   Sys te m   Para meter s   on   Islandi ng   Det ectio n, "   I EE E   Tr ansacti ons   on   Sustai nable   En ergy ,   v ol.   7,   no .   2,   pp .   461 - 470,   Apr .   2016.   [7]    Y.M.   Makwana,   B. R.   Bh al j a,   "Is la nding   de tection   te chn ique   b a sed   on   superim p osed   com ponen t s   of   volt ag e, "   IET   Re n ewabl e   Powe r   Gene rat i on ,   vol .   11 ,   no .   1 1,   pp .   1371 - 138 1,   Sept .   2017 .   [8]    P.   Gupta,   R. S.   Bhat ia,   D.K .   Ja in,   "A ver age   A bsolute   Frequ en cy   Devi at ion   Value   Based   Ac t ive   Isl andi ng   Dete c ti on   Techn ique , "   IEEE   Tr ansacti ons   on   Sm art   Gr id ,   vol .   6,   no.   1,   pp .   26 - 35 ,   Jan.   2015.   [9]    R.   B ekhr adian,   M.   Dava rp ana h ,   M.S.   Pasand,   " Novel   Approa ch   for   Secur e   Isl an ding   De te c ti on   in   Synchronous   Gene rat or   Based   Microgr ids,"   IE EE   Tr ansacti ons   on   Pow er   Del ivery ,   vo l.   34,   no.   2,   pp .   457 - 466 ,   Apr.   2019.   [10]    R.   Dhua ,   D.   Chat t erj e e,   S.K.   Gos wami ,   "H ar moni c   fi lt er - b ase d   im prove d   isl andi ng   detec t io n   t ec hniqu e   fo r   micr ogr id, "   I ET   Re newab le   Pow er   Gene ration ,   v ol.   13 ,   no .   13 ,   pp .   2443 - 2450 ,   Oc t .   2019 .   [11]    Y.M.   Makwana,   B. R.   Bha lj a ,   " Expe ri me nt al   Pe rform ance   of   an   Islandi ng   Det ection   Sch em e   Ba sed   on   Modal   Compone nts,"   I EE E   Tr ansacti o ns   on   Smar t   Gr id ,   vol .   10 ,   no .   1,   pp.   1025 - 1 035 ,   J an.   2019 .   [12]    H.H.   Z ei ne ldi n ,   E. F.   El - Sa ada ny ,   M.M.A.   Sa lama ,   "Is la nd ing   de tecti on   of   inv erter - base d   distr ibut ed   gene r at i on, "   IEE   Proceedi ngs   -   Gen erati on,   Tr ansm ission   and   Distributi on ,   vo l.   153 ,   no .   6,   pp .   644 - 652,   Nov.   2 006.   [13]    H.   L aa ksonen ,   " Advanc ed   Is la nd ing   De te c ti on   Functi on al it y   for   F uture   Elec t ricity   Distribut ion   Net works , "   IEEE   Tr ansacti ons   on   Powe r   Del iv er y ,   vol.   28 ,   no .   4,   pp .   2056 - 2064 ,   Oc t .   2013 .   [14]    Q.   Cui ,   K. E.   Ar roudi,   G.   Joós ,   "Is la nding   De tecti on   of   Hybrid   Distributed   Ge ner ation   Under   Reduc ed   Non - Dete c ti on   Zone, "   IEE E   Tr ansact ions   on   Smar t   G rid ,   vol .   9,   no .   5,   pp.   5027 - 5037,   Sept.   2018 .   [15]    E.   Serba n ,   C.   P ondic he ,   M.   Ord onez ,   "Is la nd ing   Det ec t ion   Se arch   Sequen ce   for   Distribut ed   Pow er   Gen era to rs   Under   AC   Grid   Fault s,"   IE EE   Tr ansacti ons   on   P ower   Ele ct roni cs ,   vol .   30 ,   no .   6,   p p.   3106 - 3121 ,   Ju n.   2015 .   [16]    M.   Bakhshi,   R.   Noroozi an,   G.B.   Ghare hpe ti an ,   " Islandi ng   de tection   sche me   base d   on   ada p ti ve   id ent ifier   sign al   esti mation   meth od",   ISA   Tr ansacti ons ,   vol .   71 ,   n o.   2,   pp .   328 - 34 0,   Nov.   2017.   [17]    P.P.   Mishra,   C.   N .   Bhende,   "Is l andi ng   de tecti on   using   sparse   S - tra nsform   in   dis tri bute d   gen era t i on   sys te ms",   El e ct rica l   Eng in ee ring ,   vol .   100 ,   no.   4,   pp.   2397 - 2406,   De c.   2018 .   [18]    M.   Ahmadi pour,   H.   Hiza m ,   M.L .   Othma n,   M.A .   Radz i ,   "Is la ndin g   det e ct ion   me th od   using   ridge l et   prob abi l isti c   neur al   net work   in   distri but ed   g en era t ion",   N euroc omputing ,   vo l.   3 29,   pp .   188 - 209 ,   Feb.   2019.   [19]    B.   M.   B ia z ,   V.   H.   Ferr ei ra ,   M.   Z.   Fortes,   T.   T.   Lope s   and   G.   B.   A.   L im a ,   "Is la n ding   De tecti on   in   Distribu te d   Gene ration   usin g   Uns uper vised   Le arn ing   T ec hn i ques, "   IEEE   Lat in   Ame ri ca   Tr ansacti ons ,   vo l.   16,   no.   1,   pp.   118 - 125,   Jan .   20 18.   [20]    G.   Bay rak ,   E.   Kaba lci,   "Imp lementati on   of   a   new   re mote   islanding   d et e ction   me thod   for   win d – solar   hybr i d   power   pla n ts",   R ene wabl e   and   Su stainabl e   Ene rgy   Revie ws ,   vol .   58 ,   pp .   1 - 15 ,   May   2016.   [21]    R.   Sirja n i,   C . F.   Okw ose,   "Combini ng   two   technique s   to   de vel op   a   novel   i slandi ng   de tection   me thod   for   distri bute d   gen er at ion   unit s" ,   Me asur eme nt ,   vol .   81,   pp .   66 - 79 ,   M ar.   2016 .   Evaluation Warning : The document was created with Spire.PDF for Python.
                   IS S N :   2088 - 8 694   In t J   P ow  Ele c  &  Dr i   S ys t ,   V ol .  1 1 , N o.   4 ,  D ecembe r   2020   :   209 9   –   210 6   2106   [22]    B.   Anude ep,   P.K.   Naya k ,   "S eque nce   Compon e nt - Based   Impro ved   Pass iv e   Isl andi ng   D et e ct io n   Method   for   Distribut ion   Sys te m   with   Distrib ute d   Gen erati on s" ,   Inte rnat ional   Journal   of   Eme rging   Elec tri c   P ower   Syste ms ,   vol.   20 ,   no .   2,   A pr.   2019 .   [23]    K.   Sare en ,   B . R.   Bhal ja,   S.   Srivas ta va ,   Y.   Sw arn kar ,   R . P.   Mah eshwari ,   "Is l andi ng   Dete c ti on   Techn ique   b ase d   on   Karl   Pear son’s   Coeff icient   of   Corre lation   for   Distribut ion   N e twork   with   Hig h   Penet r at ion   of   Distributed   Gene rations" ,   In te rnational   Jour nal   of   Eme rging   El e ct ric   Pow er   S yste ms ,   vol .   19 ,   no.   3,   Jun.   2018.   [24]    K.   Sar ee n ,   B. R.   Bha lj a ,   R. P.   M ahe shw ari ,   "A   Hybrid   Mult i - fe at ure   b ase d   Isla nding   De tecti on   T ec hn ique   for   Grid   Connec t ed   Distribut ed   Gen era t ion",   Int ernati onal   Journal   of   Eme rging   Elec t ric   Powe r   Syst e ms ,   vol.   18,   no.   1,   Feb .   2017 .   [25]    J.R.   Reddy ,   A.   Pandi an,   R.   Dh ana sekha r an,   Ch . R.   R eddy,   B. P.   La kshmi ,   B . N.   Devi,   "Is la ndin g   detec t ion   of   int egr at ed   distributed   g ene ra ti on   with   adv anced   cont roller",   Indo nesian   Journal   of   Elec tric al   En gine ering   and   Computer   Scien ce ,   vol .   17 ,   no .   3,   pp .   1626 - 1631 ,   Mar.   2020.   [26]    S. S.   Reddy ,   You ng   Hawn   Lho ,   “ Optim um   Locat i on   of   Volt age   R egul a tors   in   the   Radi al   Distribu tion   Sys te ms”,   Inte rnational   Jo urnal   of   Eme rgi ng   Elec tri c   Pow er   Syste ms ,   vol .   17,   no .   3,   pp .   35 1 - 361,   Jun.   2016 .   [27]    K.   Suresh,   P.   A nusha,   Sk.   Na jma,   B . I.   R aj kum a r,   Ch. R .   R eddy ,   B. P.   L akshmi,   "A   passive   isla n ding   detec t ion   me thod   for   hybr id   distri bu te d   ge ner ation   sys te m   under   b alanc ed   isla nd ing",   Ind onesian   Journal   of   El e ct ri ca l   Engi ne ering   and   Computer   Sc ie n ce ,   vol .   14 ,   no .   1,   pp .   9 - 19 ,   Apr.   2019.   [28]    T.   Sinha ,   P.   R ay ,   S.S.   Red dy ,   “Prote c ti on   Coordi nat ion   in   Micro grid   using   Fault   Curre nt   L im i te rs ”,   Journal   of   Gr ee n   Engi n ee ri ng,   vol .   8,   no .   2,   pp.   125 - 150 ,   20 18.   [29]    S.S.   Reddy,   “ Op ti mal   r ea c ti v e   p ower   sizi ng   in   a   distri but ed   n etw ork ”,   Int ernational   Journal   of   Engi ne ering   &   Technol ogy ,   vo l.   7,   no.   4,   pp.   331 2 - 3316,   2018 .   [30]    S.C.   Shek ar,   G. R.   Kumar ,   S.V. N.L .   Lalit h a,   "A   tra nsien t   cur r ent   b ase d   mi cro - grid   conn ecte d   power   sys tem   prote c ti on   sch e me   using   wav el e t   appr oa ch",   In ter nati onal   Journ al   of   El e ct rica l   and   Computer   E ngine ering ,   vol .   9,   n o.   1,   pp .   14 - 22,   Feb .   2019 .   [31]    M.A.   Abido,   " Optim al   power   flow   using   p art i cl e   sw arm   optimizati on",   Int ernati onal   Journa l   of   E le c tric al   Powe r   &   En ergy   Syste ms ,   vol.   24 ,   no .   7,   pp .   563 - 571,   Oct .   2002 .   [32]    Ch. R .   Reddy ,   K. H.   Reddy,   "A   Pa ss ive   Islandi ng   Dete c ti on   Metho d   for   Neutr a l   poi nt   clampe d   Mult il ev el   Inve rte r   base d   Distribute d   Gene ration   usi ng   Rat e   of   Ch an ge   of   Freque n cy   Analysis",   Inte r nati onal   Journal   of   Elec tri cal   and   Computer   E ngine ering ,   vol .   8,   no .   4,   pp .   196 7 - 1976,   Aug.   20 18.   [33]    S.   Raz a ,   H.   Am ja d,   M.   Umer ,   A.   Khali d ,   "A nal ysis   of   power   sys te m   par am e te r s   for   isla nding   d et e ct ion   using   wave let   tra nsfor m",   Indon esian   Journal   of   El e ctr ic al   Engi n ee rin g   and   Computer   Sc ie nc e ,   vo l.   1 7,   no .   3,   pp.   1184 - 1193,   Mar .   2 020.   [34]    P.   Ray,   B . K.   Pa nigra hi ,   N.   Senr oy,   "H ybrid   m ethodology   for   fa ult   dist anc e   estimat ion   in   ser ie s   com pensa te d   tra nsmiss ion   li ne , "   IET   Gen erati o n,   Tr ansm ission   &   Distributi on ,   vol.   7,   no .   5,   pp .   431 - 439,   May   2 013.   [35]    D.   Pal,   B . K.   Pa nigra hi ,   S.   Kew at ,   "A   Hybr id   M et hod   for   Isl anding   Dete c ti on   of   Inve rte r   Inte rf aced   Distributed   Gene rat ors   Utili zi ng   Super im po sed   Compone n t   of   d - axi s   Volt age , "   IEEE   Ener gy   Conve rs ion   Congress   and   Ex positi on ,   B al t i more ,   MD,   USA,   2019 ,   pp .   1020 - 1025.   [36]    H.   R.   B aghaee,   D.   Mlakić,   S.   N ikol ovski,   T.   Dragi č evi ć ,   "S upp ort   Vec tor   Ma c hine - base d   Islan ding   and   Grid   Fault   Det ection   in   Acti v e   Distri buti on   Network s,"   IEE E   Journ al   of   Eme rging   and   Sel e cted   Topics   in   Powe r   El e ct ronics ,   201 9.   [37]    S.S.   Reddy ,   C . M.   Jung,   “Sho rt - Te r m   Loa d   Forec asti ng   Us i ng   Artifici al   Neura l   Networks   and   W ave l et   Tra nsform” ,   In ter nati onal   Journal   of   Applied   Enginee ring   Re searc h ,   vol .   11 ,   no .   19 ,   pp .   9831 - 9836 ,   Oct.   2016.   [38]    R.   Nal e,   M.   Bis wal,   N.   Kishor,   "A   Tra nsien t   Co mpone nt   Based   Approac h   for   Isl andi ng   De tecti o n   in   Distr ibute d   Gene ration, "   I E EE   Tr ansacti ons   on   Sustainab le   Ene rgy ,   vol .   10 ,   no.   3,   pp .   1129 - 1138,   July   2019 .     Evaluation Warning : The document was created with Spire.PDF for Python.