基于知识发现的电力需求复合预测研究

《基于知识发现的电力需求复合预测研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、华北电力大学（保定）博士学位论文基于知识发现的电力需求复合预测研究姓名：李春祥申请学位级别：博士专业：技术经济及管理指导教师：乞建勋;牛东晓20090501华北电力大学博士学位论文摘要AbstractPowerdemandforecastingisanimportantfoundationworkforpowersystemplanningandoperation．Itprovidesfoundationforpowercompaniestosetoutpurchaseofelectricityandpowerproductionplansandalsodispla

2、ysallimportantguaranteeforgridsecureandeconomicoperation．Powerdemandindicatorswillbesubjecttoallkindsoffactors．TheoriesandmethodsofknowledgediscoveryCanbeusedtominetheintrinsiclawofindicatorsvaryingandmutualrelationwithinfluencefactors．Theprimaryworkdoneinpowerdemandforecastinginthispa

3、perisasfollows．Firstly，acompoundforecastingmodelbasedonthreeindexquantitiesisresearchedandputforward．Threeindexquantitiesaretotalindexquantity，increasingindexquantityandindexgrowthrate．Convertingpredictionindicatorsequenceintothreeindexquantitiessequencesandproceedinganalysisandforecas

4、tingrespectivelyforthem，afterthatdosynthesistogetfinalforecastingresult，whichiscalledcompoundforecasting．Inthispaper，powerdemandisdividedintoquantityofelectricityforecastingandelectricloadprediction,andgreyassociationanalysisisappliedtoanalyzethemutualrelationbetweenquantityofelectrici

5、tyindicatorandinfluencefactors．Anachievementmodelofcompoundforecastingispresentedon，whichiscomprehensivemodelofquantityofelectricitycompoundforecasting．Itbenefitsfromcombinationforecastingidea．Atfirst，ananalytichierarchyprocessmodelisconstructedtoanalyzeandestimatethethreeindexquantiti

6、esrespectively，thenselectingoutoptimalforecastingmodelforeachindexquantity，inwhichevaluationcriterioninvolvemodelforecastingerror,fittingdegreeofmodel，experttrustdegreeofthemodelandconfidencelevelofthetrendofforecastingresults．Andthen，twosynthesismethodsareresearched，whicharesynthesism

7、ethodbasingforecastingefficiencydegreeandradialbasicfunctionneuralnetworksynthesizemodel，andcomparerelativemeritsofthem．Finally，throughinstanceanalysistocontrastdominantofthecomprehensivemodeltotraditionalmodels．Compoundforecastingmethodologyinapositiontoviatoanalyzeforecastedindices