0% found this document useful (0 votes)
77 views

Short Term Load Forecasting

Uploaded by

Sunil Kumar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
77 views

Short Term Load Forecasting

Uploaded by

Sunil Kumar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 8

SHORT TERM LOAD FORECASTING IN POWER SYSTEM USING NEUROFUZZY

Sunil Kumar*
Dept. of ECE DGI, Gr. Noida, U.P., India 874480 !40 S"mee 07#$mail.%om

Kapil De& S'arma


Dept. (f Ele%troni%) En$$. DGI, Gr. Noida, U.P., India !808007888 Kd)'arma.en#$mail.%om

*a+ee& *an+an
Dept. of Ele%tri%al , Ele%troni%) En$$. DGI, Gr. Noida, U.P., India -el.No. E.mail of aut'or

ABSTRACT ABSTRACT
S'ort.term load fore%a)tin$ /S-012 i) a ne%e))ar3 dail3 ta)" for po4er di)pat%'. It) a%%ura%3 affe%t) t'e e%onomi% operation and relia5ilit3 of t'e )3)tem. -'i) paper pre)ent) a )olution met'odolo$3 u)in$ fu663 lo$i% approa%' and artifi%ial neural net4or" for )'ort term load fore%a)tin$ and i) implemented on 'i)tori%al 4eat'er )en)iti&e data i.e. temperature, 'umidit3, 4ind )peed and 'i)tori%al load data for fore%a)tin$ t'e load. -'e propo)ed fu663 lo$i% approa%' i) implemented on 4eat'er )en)iti&e data and t'e a%%ura%3 of t'e re)ult i) %ompared u)in$ t4o different mem5er)'ip fun%tion). 7rtifi%ial neural net4or" approa%' i) implemented on t'e propo)ed non.full3 %onne%ted neural net4or" %on)i)t) of fi&e full3 %onne%ted )upportin$ net4or") repre)entin$ 4eat'er &aria5le), da3 t3pe and load data a) input).
Keywords 0oad fore%a)tin$, Nueral model, 1u663 lo$i%, Neurofu663.

1. INTRODUCTION 1. INTRODUCTION
1u663 )3)tem) and neural net4or") 'a&e attra%ted t'e intere)t of re)ear%'er) in &ariou) )%ientifi% and en$ineerin$ area). -'e num5er and &ariet3 of appli%ation) of fu663 lo$i% and neural net4or") 'a&e 5een in%rea)in$, ran$in$ from %on)umer produ%t) and indu)trial pro%e)) %ontrol to medi%al in)trumentation , information )3)tem) and de%i)ion anal3)i) .-'e main idea of fu663 lo$i% %ontrol /10C2 i) to 5uild a model of a 'uman %ontrol e8pert 4'o i) %apa5le of %ontrollin$ t'e plant 4it'out t'in"in$ in term) of a mat'emati%al model. -'e %ontrol e8pert )pe%ifie) 'i) %ontrol a%tion) in t'e form of lin$ui)ti% rule). -'e)e %ontrol rule) are tran) lated into t'e frame4or" of fu663 )et t'eor3 pro&idin$ a %al%ulu) 4'i%' %an )imulate t'e 5e'a&ior of t'e %ontrol e8pert. -'e )pe%ifi%ation of $ood lin$ui)ti% rule) depend) on t'e "no4led$e of t'e %ontrol e8pert, 5ut t'e tran)lation of t'e)e rule) into fu663 )et t'eor3 frame4or" i) not formali6ed and ar5itrar3 %'oi%e) %on%ernin$, for e8ample, t'e )'ape of mem5er)'ip fun%tion) 'a&e to 5e made. -'e 9ualit3 of fu663 lo$i% %ontroller %an 5e dra)ti%all3 affe%ted 53 t'e %'oi%e of mem5er)'ip fun%tion). -'u), met'od) for tunin$ fu663 lo$i% %ontroller) are ne%e))ar3 neural net4or") offer t'e po))i5ilit3 of )ol&in$ t'e pro5lem of tunin$. 7lt'ou$' a neural net4or" i) a5le to learn from t'e $i&en data, t'e

trained neural net4or" i) $enerall3 under)tood a) a 5la%" 5o8. Neit'er it i) po))i5le to e8tra%t )tru%tural information from t'e trained neural net4or" nor %an 4e inte$rate )pe%ial information into t'e neural net4or" in order to )implif3 t'e learnin$ pro%edure. (n t'e ot'er 'and, a fu663 lo$i% %ontroller i) de)i$ned to 4or" 4it' t'e )tru%tured "no4led$e in t'e form of rule) and nearl3 e&er3t'in$ in t'e fu663 )3)tem remain) 'i$'l3 tran)parent and ea)il3 interpreta5le. :o4e&er, t'ere e8i)t) no formal frame4or" for t'e %'oi%e of &ariou) de)i$n parameter) and optimi6ation of t'e)e parameter) $enerall3 i) done 53 trial and error. Com5ination of neural net4or") and fu663 lo$i% offer) t'e po))i5ilit3 of )ol&in$ tunin$ pro5lem) and de)i$n diffi%ultie) of fu663 lo$i%. -'e re)ultin$ net4or" 4ill 5e more tran)parent and %an 5e ea)il3 re%o$ni6ed in t'e form of fu663 lo$i% %ontrol rule) or )emanti%). -'i) ne4 approa%' %om5ine) t'e 4ell e)ta5li)'ed ad&anta$e) of 5ot' t'e met'od) and a&oid) t'e dra45a%") of 5ot'. In t'i) paper, neuro.fu663 %ontroller ar%'ite%ture i) propo)ed, 4'i%' i) an impro&ement o&er t'e e8i)tin$ neuro fu663 %ontroller). It o&er%ome) t'e ma+or dra45a%") of t'e e8i)tin$ neuro.fu663 approa%'e); of eit'er "eepin$ neural net4or") and fu663 lo$i% a) )eparate entitie) /%o.operati&e model)2 4or"in$ to4ard) a %ommon $oal or in mo)t of t'e e8i)tin$ neuro.fu663 approa%'e), t'e trained %ontroller no lon$er %an 5e interpreted a) fu663 lo$i% %ontroller.

2. S'()T T*)+ L(A, F()*-AST.N/ 2. S'()T T*)+ L(A, F()*-AST.N/


Ele%tri%al load fore%a)tin$ i) an e))ential tool u)ed to en)ure t'at ener$3 )upplied 53 utilitie) meet t'e load plu) t'e ener$3 lo)t in t'e )3)tem. It i) a %entral and inte$ral pro%e)) in t'e plannin$ and operation of ele%tri% utilitie). 0oad fore%a)tin$ 'elp) an ele%tri% utilit3 to ma"e important de%i)ion) on pur%'a)in$ and $eneratin$ ele%tri% po4er. Factors Affecting Accurate Load Forecast /i2 <eat'er Influen%e, /ii2 -ime fa%tor),/iii2 E%onom3,/i&2 Cu)tomer %la)) 2.1 Short Term Load Forecasting Using Artificial Neural Networ 2.1.1 Neural Networ 7 Neural net4or" i) an artifi%ial repre)entation of t'e 'uman 5rain t'at trie) to )imulate t'e learnin$ pro%e)). It re)em5le) t'e 5rain in t4o re)pe%t)= >. Kno4led$e i) a%9uired 53 t'e net4or" from it) en&ironment t'rou$' a learnin$ pro%e)). . Interneuron %onne%tion )tren$t'), "no4n a) )3napti% 4ei$'t), are u)ed to )tore t'e a%9uired "no4led$e. -'e pro%edure u)ed to perform t'e learnin$ pro%e)) i) %alled a learnin$ al$orit'm, t'e fun%tion of 4'i%' i) to modif3 t'e )3napti% 4ei$'t) of t'e net4or" in an orderl3 fa)'ion to attain a de)ired de)i$n o5+e%ti&e. 2.1.2 Artificial Neuron Neuron i) t'e 5a)i% 5uildin$ 5lo%" of an artifi%ial neural net4or". In %al%ulatin$ t'e output of t'e neuron, t'e a%ti&ation fun%tion ma3 5e in t'e form of a t're)'old fun%tion, if a t're)'old le&el i) rea%'ed, t'e output of t'e neuron i) ?> and 0 ot'er4i)e. -'e 5a)i% %omponent) of artifi%ial neuron are )ummari6ed a)=. @ <ei$'tin$ 1a%tor), @ Summation 1un%tion, @ -ran)fer 1un%tion A 7%ti&ation 1un%tion @ S%alin$ and 0imitin$, @ (utput 1un%tion, @ Error 1un%tion and Ba%".propa$ated Calue @ 0earnin$ 1un%tion. 2.2 ANN Techni!ue for Short Term Load Forecasting -'e artifi%ial neural net4or" /7NN2 te%'ni9ue to )'ort.term load fore%a)tin$ 'a) $ained a $reat deal of intere)t and )e&eral re)ear%'er) 'a&e reported t'e effe%ti&ene)) of t'e 7NN approa%'. -'e 7NN te%'ni9ue) 'a&e 5een applied to %orrelate t'e 4eat'er information and t'e load &ariation) to fore%a)t t'e dail3 pea" load and 'ourl3 load. 2.2.1"ro#lem Formulation -'e propo)ed de&elopment and implementation of ne4 7NN 5a)ed te%'ni9ue for t'e )'ort.term load fore%a)tin$ of 4eat'er )en)iti&e load) %on)i)t of four )ta$e). -'e 7NN model u)ed a non full3.%onne%ted neural net4or" %on)i)tin$ of one main and four )upportin$ neural net4or". Data )ele%tion and anal3)i) for )ele%tion of num5er of input &aria5le) t'ere i) no 5a)i% rule. It i) %arried out almo)t entirel3 on 'it and trial 5a)i). 2.$ Short Term Load Forecasting Using Fu%%y Logic 2.$.1 Fu%%y Logic & Fu%%y Sets 7 fu663 )et, a) t'e name implie), i) a )et 4it'out a %ri)p 5oundar3. -'at i), t'e tran)ition from D5elon$ to a )etE to Dnot 5elon$ to a )etE i) $radual, and t'i) )moot' tran)ition i) %'ara%teri6ed 53 mem5er)'ip fun%tion) t'at $i&e fu663 )et) fle8i5ilit3 in modelin$ %ommonl3 u)ed lin$ui)ti% e8pre))ion), )u%' a) D4ater i) 'otE or Dt'e temperature i) 'i$'E. 7 fu663 )et e8pre))e) t'e de$ree to 4'i%' an element 5elon$) to a )et. :en%e t'e %'ara%teri)ti% fun%tion of a

fu663 )et i) allo4ed to 'a&e &alue) 5et4een 0 and >, 4'i%' denote) t'e de$ree of mem5er)'ip of an element in a $i&en )et. If F i) a %olle%tion of o5+e%t) denoted $enerall3 53 8, t'en a fu663 )et 7 in F i) defined a) a )et of ordered pair) <'ere G7/82 i) %alled t'e mem5er)'ip fun%tion /H12 for t'e fu663 )et 7. -'e H1 map) ea%' element of F to a mem5er)'ip $rade 5et4een 0 and >. 1u663 I1 -:EN *ule) 7 fu663 if.t'en rule /fu663 rule, fu663 impli%ation, or fu663 %onditional )tatement2 i) e8pre))ed a) follo4= I1 8 i) 7 -:EN 3 i) B 4'ere 7 and B lin$ui)ti% &alue) defined 53 fu663 )et). D8 i) 7E i) %alled Dante%edentE or Dpremi)eE, 4'ile D3 i) BE i) %alled t'e D%on)e9uen%eE or D%on%lu)ion. 2.$.2 Fu%%y inference Systems 0F.S1 -'e fu663 inferen%e )3)tem i) a popular %omputin$ frame4or" 5a)ed on t'e %on%ept of fu663 )et t'eor3, fu663 if.t'en rule), and fu663 rea)onin$.

1i$.4.> 1u663 rule

S'ort term

1i$.4. 1u663 rule me%'ani)m

1i$.4.I 1u663 mem5er)'ip fun%tion input plot -'e)e fi&e )tep), )upplemented 53 t'e re)olution of %onfli%t) and in%orporation of e8pert opinion, %on)titute t'e %on)tru%tion of t'e fu663 rule 5a)e for trape6oidal mem5er)'ip fun%tion.

1i$.4.4 1u663 mem5er)'ip fun%tion input plot

1i$.4.J 1u663 mem5er)'ip fun%tion input plot

1i$.4.K 1u663 mem5er)'ip fun%tion input plot

1i$.4.7 1u663 mem5er)'ip fun%tion oAp plot

2.2 Neuro3Fu%%y +odels 7 neuro.fu663 /N12 model i) a fu663 )3)tem dra4n in a neural net4or" )tru%ture, %om5inin$ t'e learnin$, parallel pro%e))in$ and $enerali6ation %apa5ilitie) of neural net4or") and lo$i%alit3, tran)paren%3 and u)e of a priori "no4led$e in fu663 )3)tem). 7 lo%al linear neuro.fu663 )3)tem, a) a nota5le %ate$or3 of N1 model), de%ompo)e) an a priori un"no4n $lo5al nonlinear )3)tem into a )erie) of lo%al linear model) and t'en trie) to %arr3 out t'e )impler ta)" of identif3in$ parameter) of linear )u5.model).

3. S.+ULAT.(N AN, )*SULT ANAL4S.S 3. S.+ULAT.(N AN, )*SULT ANAL4S.S


-'e di))ertation 4or" 'a) 5een fo%u)ed on t'e preliminar3 in&e)ti$ation of fea)i5ilit3 and performan%e of fu663 lo$i% model and artifi%ial neural net4or" model to %arr3 out )'ort term load fore%a)tin$ on 'ourl3 5a)i) and pea" load demand predi%tion for 4ee"da3) and 4ee"end da3). *eal time data t'at in%lude) 'i)tori%al 'ourl3 load demand o&er a 4ee", and 4eat'er data in term) of temperature, 'umidit3, 4ind )peed, i) %olle%ted from, Ha'ara)'tra Cid3ut Para)aran Ni$am /HCN2. -'e fore%a)ted load i) %ompared 4it' t'e a%tual load and a&era$e per%enta$e error i) %al%ulated.

1i$.K.> -'e a5o&e fi$ i) )'o4in$ t'e re)pon)e of fore%a)tin$ for different )e))ion, in rainin$ )ea)on load i) re9uired more.

-ran)for med real load

-ran)for med )%aled load

-ime/Da32 1i$.K. -'e a5o&e fi$ure )'o4 t'e load fore%a)tin$ of *eal load &) )%aled load

-ime/Hont'2 1i$.K.I -'e a5o&e fi$. i) )'o4in$ t'e load fore%a)tin$ of different time

1i$.K.4 -'e a5o&e fi$ure i) )'o4 t'e Surfa%e rule of fore%a)tin$

Defle%ted

*eal load fore%a)tin$


1i$.K.J -ime/Da32

3.CONCLUSION 3.CONCLUSION

-'i) paper 'a) propo)ed a )op'i)ti%ated )'ort.term ener$3 demand fore%a)tin$ approa%' 5a)ed on t'e :P filter, HI. 5a)ed input )ele%tion al$orit'm and S0N1 model. E8tra%tion of t'e trend and %3%li% %omponent) of t'e demand )erie) and effe%ti&e input )ele%tion 53 t'e HI te%'ni9ue, %on)idera5l3 impro&ed t'e modelin$ performan%e of t'e S0N1 net4or". Bot' lo%al and $lo5al e)timation of rule %on)e9uent parameter) of t'e S0N1 model 4ere %on)idered RE re)ult) in 'i$'er modelin$ performan%e. and it 4a) demon)trated t'at t'e $lo5al e)timation te%'ni9ue RE

FE FE RE RE L>M. U.S. Ener$3 Information 7dmini)tration /EI72.NC 7&aila5le online= 'ttp=AA444.eia.$o& /a%%e))ed on J Nune NC ES 0>>2. ES L M. :arri), N.0.; 0iu, 0.H. D3nami% )tru%tural anal3)i) and fore%a)tin$ of re)idential ele%tri%it3 %on)umption. Int. J.

Forecast. 2556, 9, 4I7O4JJ. LIM. Edi$era, C.; 7"ar5, S. 7*IH7 fore%a)tin$ of primar3 ener$3 demand 53 fuel in -ur"e3. Energy Policy 2556, 35, >70>O>708. L4M. Bian%oa, C.; Han%a, (.; Nardinia, S. Ele%tri%it3 %on)umption fore%a)tin$ in Ital3 u)in$ linear re$re))ion model). Energy 2557, 34, >4>IO>4 >. LJM. 76ade', 7.; K'a"e)tani, H.; Sa5eri, K., D7 fle8i5le fu663 re$re))ion al$orit'm for fore%a)tin$ oil %on)umption e)timationE. Energy Policy 2557, 37, JJK7OJJ7!. LKM. S'a"ouri, :.; Nadimia, G.; G'aderia, 1. D7 '35rid -SK.1* model to )tud3 )'ort.term &ariation) of t'e ele%tri%it3 demand versus t'e temperature %'an$e)E. Expert Syst. Appl. 2558, 36, >7KJO>77 . L7M. 76ade', 7.; 7ra5, *.; Be'fard, S., D 7n adapti&e intelli$ent al$orit'm for fore%a)tin$lon$term$a)oline demand e)timation= -'e %a)e) of US7, Canada, Napan, Ku4ait and IranE. Expert Syst. Appl. 2515, 37, 74 7O74I7. L8M. Po"o3ama, *.; <a"ui, -.; Sata"e, *., D Predi%tion of ener$3 demand) u)in$ neural net4or" 4it' model identifi%ation 53 $lo5al optimi6ationE. Energy onvers. !anag. "ol. 2557, 5#, I>!OI 7. L>0M. :a'n, :.; He3er.Nie5er$, S.; Pi%"l, S., D Ele%tri% load fore%a)tin$ met'od)= -ool) for de%i)ion ma"in$E. Eur. J. $per. %es. 2557, &99, !0 O!07. L>>M. Su$ant'i, 0.; Samuel, 7. DEner$3 model) for demand fore%a)tin$= 7 re&ie4E. %ene'. Sustain. Energy %ev. 2511, &#, 8>OI>>. L> M. C'en, -., D 7 %olla5orati&e fu663.neural approa%' for lon$.term load fore%a)tin$ in -ai4anE. o(put. In). Eng.* 2511, doi=>0.>0>KA+.%ie. 0>>.0K.00I. L>IM. 76ade', 7.; 7)ad6ade', S.H.; G'an5ari, 7. D7n adapti&e net4or".5a)ed fu663 inferen%e )3)tem for )'ort. term natural $a) demand e)timation= Un%ertain and %omple8 en&ironment)E. Energy Policy 2515, 3+, >J !O>JIK. L>4M N.P. *ot'e, 7.K. <ad'4ani, S. <ad'4ani, D:35rid and inte$rated approa%' to )'ort term load fore%a)tin$E, IEEE -ran)a%tion) on Po4er S3)tem), &ol. , i))ue > , pp. 7> 7.7>I , 0>0. L>JM Qu'aim3I)mailaand *o)naliniHan)or5, D1u663 0o$i% 7pproa%' for 1ore%a)tin$ :alf'ourl3 Hala3)ia Ele%tri%it3 0oad DemandE, International S3mpo)ium on 1ore%a)tin$ O Pra$ue pro%eedin$), pp.>.>K, Nul3 0>>. A99endi: 0ma8RBma8 *eSCma8 , 0minR Bma8* eSCmin ,B/"2ma8 R B/"2minR H7PERH7F/ , R 2*>00 ,HSPER , R */i2/ . 2?

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy