神经模糊系统研究及其在电厂协调系统中的应用

《神经模糊系统研究及其在电厂协调系统中的应用》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）博士学位论文神经模糊系统研究及其在电厂协调系统中的应用姓名：林碧华申请学位级别：博士专业：热能工程指导教师：韩璞20050401英文摘要AbstractNeuralFuzzySystemfNFS)isanewfuzzyinferencesystemstructurcthatcombinedfuzzylogicsystemwi也neuralnetworks．ItistheneuralnetworkrealizationforTakagiandSugenofuzzyinferencesystem．Itcallapproachanylinearandnonli

2、nearfunction试thanyprecision,butalsoquickenconvergingspeed，decreasepreciseerrors。andlessertrainingsamplesthatareneeded．Thispaperresearchedthelearningability,generalizationability,methodofinputselection，overfittingproblemanddimensiondisasterofNFS．Intheend，NFSWasappliedtothemodelingandcontrol

3、ofCoordinatedControlSystem(ccs)thatiscomplexsystemofpowerplanttovalidatethevalidityofNFS．1、Theeffectthatthecombinationofinput,thentunbcrofmembershipfunctionofinput，trainingtimesandtrainingsampleaffe：．etthelearningabilityandgeneralizationabilityofNFSwasresearchedrespectively．Therelationbetw

4、eenthelearningabilityandgeneralizationabilityWasstudied．2、Aperformanceindexoferrorwaspresented．ThisindexisakindofevaluationofneuralfuzzymodelperformanceandsyntheticallyconsideredtrainingerrorandcheckingerrorofNFS．3、AnewmethodofinputselectionforNFSWasproposed．Thismethodiseasy,fast，andcalldi

5、rectlyobtainn，mandtdoftheobjectsystem．Anotheradvantageisthatprecedenceorderofinputselectionindicatestheimportancegradeofinputinfluencingoutput．4、AsfortheoverfittingproblemofNFS．thispaperproposedamethodofselectingtrainingsampleandcheckingsampleandamethodofselectingthemostbefittingtrainingti

6、mes．5、AsforthecomplexsystemmodelingbasedonNeuralFuzzySystem，Thispaper(1)researchedthestructureandlearningalgorithmfortheMultistageNeuralFuzzySystem(MNFS)；(2)presentedamethodofinputselectionbasedonMNFS；(3)presentedamethodofbuildingthemodelofthemulti·outputsystemsbasedonMNFS．6、Becauseofthela

7、rgetime-delayofCCSofBoiler-turbineunit，thispaperdesignedtwo-stageneuralfuzzysystemtObuildthemodelsofCCS．NFSwasrespectivelyusedtobuildthenonlinearmodel，thelinearmodelandtheNFSmodelbuihn华北电力大学博士学位论文accordingtothefielddataoftheCCS．Thesimulationresultsshowthatthem