支持向量機(jī)在電力系統(tǒng)短期負(fù)荷預(yù)測中的應(yīng)用及改進(jìn)

1、西南交通大學(xué)碩士學(xué)位論文支持向量機(jī)在電力系統(tǒng)短期負(fù)荷預(yù)測中的應(yīng)用及改進(jìn)姓名：李云飛申請學(xué)位級別：碩士專業(yè)：電力系統(tǒng)及其自動(dòng)化指導(dǎo)教師：黃彥全20061201西南交通大學(xué)碩士研究生學(xué)位論文第1I頁AbstractPowersystemloadforecastingisallimportantfactorofenergymanagementsystemAccordingtothepredictedloads，electricitypurch

2、aseandtheoperationmodecouldbearrangedreasonablyShorttermloadforeoastingwhichaimsatpredictingelectricloadsforaperiodofminutes，hours，days，orweeks，playsasignificantroleinthepowernetwork’SeconomicandstablerunningWiththeriseo

3、fderegulationandfreecompetitionoftheelectricpowerindustryallaroundtheworld，shorttermloadforecastingbecomesmoreimportantthaneverbeforeGenerallytherearetwokindsofmethodologiesforloadforecastingOneisinatraditionalwayreprese

4、ntedbytimeseries，anotheroneistermedasnewartificialintelligencemethod，representedbytheartificialneuralnetworkBasicallythetraditionalwaymayinvolvemethodoftimeseries，multilinearregressionandtheFourierexpansionete，whilethear

5、tificialinttelligencemetIlodmainlyinvolvesexpertsystemmethod、fuzzylo#cmethod、f舊neuralmodelandartificialneuralnetworkmethodetcneforecastingeffectofnon1inearregressionandtimeseriesisineffectwhenelectricnetworkworkswellandl

6、ittlechangeoccurredintheproductionandweatherhoweverwithomconsideringsomefactorsaffectingtheloadsuchasholidaysandweatherHence，theforecastingprecisionwillbeaffectedbysuddenchangeoftheseabovef_actorsneuralnetworkandfuzzythe

7、orytakeintoaccountoftheindefinitefactorsaffectingtheload，butstillhavenotsolvethedesigningproblemofnetworkstructurecompletelyyetandneedalongtimeoftrainingaswellSupportvectormachineisamachinelearningalgorithmofthenewe|abas

8、edonstatisticstheorywhichequalstosolvingaquadraticprogrammingproblemintheprincipleofmininlunlstructuralriskTllisalgorithmfeaturesstrongforecastingabilityglobaloptimizationandfastspeedofapproaching，eteAftertheintroduction

9、ofsupportvectormachinealgorithmsimulationswascarriedoutcomparedwithshortterraloadforecastingbasedonBPneuralnetworkConsideringtheeffectofparametersonforecastprecisionandthegeneralizationabilityGridsearchmethodwasproposedf