基于神經(jīng)網(wǎng)絡(luò)的感應(yīng)電機(jī)定子電阻參數(shù)辨識(shí).pdf

1、Y1404875分類(lèi)號(hào)：TPl83UDC：——密級(jí)：學(xué)校代號(hào)：10150學(xué)號(hào)：20052160峽建交積摯碩士學(xué)位論文基于神經(jīng)網(wǎng)絡(luò)的感應(yīng)電機(jī)定子電阻參數(shù)辨識(shí)IdentifieationofInductionmotorStatorResistanceBasedonNeutralNetwork學(xué)生姓名：導(dǎo)師及職稱(chēng)：學(xué)科門(mén)類(lèi)：專(zhuān)業(yè)名稱(chēng)：研究方向：申請(qǐng)學(xué)位級(jí)別：論文答辯日期：學(xué)位授予單位：李曉芳張繼和教授工學(xué)控制理淪與控制上程智能控制碩士2007

2、qi12月20日大連交通大學(xué)人_迕交通人學(xué)l：’Z碩‘1?！酆握撐腁bstractGreatdevelopmentshavetakenplacesinceDirectTorqueControlsystem(DTC)wasbroughtforward1thasbeenafocusamongdomesticandforeignscholarsbecauseithascharacterwitllsimplestructurenoveltyth

3、oughtandwonderfulpropertyDTCsystemisdifferentfromvectorcontrolinthatitiSlocatedinstatorcoordinateAsaresultfewparametershaveeffectonDTCsystemsandithaslittlercalculationThispapermainlyhasobtainedagreatdealofdataaboutthesta

4、torresistancewhichchangeswithdifferentvoltagefrequencies，thestatorcurrents，thestatorendtemperatethroughtheexperiment，andobtainsstatorresistanceandvoltagefrequency，statorcurrentaswellasthestatorendwarmbetweenchangerelatio

5、nsfromthedatathenusestheneuralnetworkalgorithmtoconstructthemodelofstatorresistancechangingwithvarietyvoltagefrequencies，statorcurrentsandthetemperatureandhascardedonthesimulationusingtheMATLABsimulationt001Thestatorresi

6、stanceisaffectedbymanyfactors，suchasthedriver’Sruntime，currentmagnitude，frequency，andtemperatureoftheenvironmentetcItisextremelydifficulttofindtherelationshipthroughthetraditionmethodsbecauseofcomplicatedvarietyofresista

7、nceThereforethemainpointofthispaperistoapplygeneralizedregressionneuralnetworktothedetectionofstatorresistance；tocombinemanyfactorsintothreeparameters：frequency，electriccurrentandtemperatureUsingthegeneralizedregressionn

8、euralnetwork，thispaperstruoturesthemodelinordertoconfirmstatorresistanceInfullspeedarea，thispapercarriesontheidentificationtothestatorresistanceBecauseofinthelowfrequencyareathestatorresistanceparameterchangeinfluencesth

9、efluxlinkagecharacteristicThekeypointwhichimproveslowfrequencycharacteristicofthenon—velocitygeneratordirecttorquecontrolsystemishowtOenhancetheprecisionofthestatorresistanceobservationInordertOenhancetheneuralnetworkthe

10、trainingspeed，andreducethemutualinfluenceofthenetworkweightbetweenmiddle&highfrequencyandlowfrequencysampleintrainingprocess，thispaperseparatelyestablishesrespectiveneuralnetworktOmiddle&high—frequencyandlowfrequencystat

11、orresistancemodelComparingwiththeidentificationresultofthestatorresistancebytheBPneuralnetwork，thegeneralizedregressionneuralnetworkwhichthispaperdesignshasfollowmerits：1TheabilityofapproachisverystrongTheconcealmentleve

12、lnodefunction(primaryfunction)ofthegeneralizedregressionneuralnetworkusestheGaussfunctionAsanonnegativenonlinearfunction，ittakesonekindofpartialdistributionwhichweakensthecenterradialsymmetry，theGaussfunctionwillhavether