基于機器學(xué)習(xí)和智能算法的鍋爐燃燒優(yōu)化研究.pdf

1、東南大學(xué)博士學(xué)位論文基于機器學(xué)習(xí)和智能算法的鍋爐燃燒優(yōu)化研究姓名：周建新申請學(xué)位級別：博士專業(yè)：動力機械及工程指導(dǎo)教師：徐治皋20090201AbstraetStudyontheBoilerCombustionOptimizationviatheMethodsofMachineLearningandIntelligentizedAlgorithmsAbstractHighenergyconsumptionandseriouspollut

2、ionarethetwoprimaryproblemsinthedevelopmentofthepowerplantAsoneofthethreemajorequipmentsinthermalpowerplant，boiler’sefficiencywillimpactonthewholee伍ciencyoftheunitMeanwhiletheproductofthecombustionemissionsfromboileristh

3、esourceoftheatmospherepollutionThereforethehighe伍ciencyandlowpollutionoftheboileroperationisattractingmoreandmoreattentionInalltheworktheoptimizationadjustingofboilercombustionisconsideredtobetheeffectivemethodtesolvethe

4、problemThereforemoreandmoreelectricallytechnicalworkersfocustheirresearchontheoptimizationtechnologyofboilercombustionOnthebasisofsummarizingthedevelopmentcourseandthepresentstateoftheboilercombustionoptimization，thepape

5、remphaticallystudiestheboilercombustionadjustingbasedonmachinelearningandinteUigentizedoptimizationTheresearchhasimportantpracticalsignificanceformodelingandoptimizationofboilercombustionanddevelopmentofmachinelearningan

6、dintelligentoptimizationalgorithminthefieldofthermalengineeringapplicationsAndithasachievedsomeusefulconculsionstoprovidereferencefortheoperationofboilerThemaincontentscanbedescribedasfollow：l、Thecombustionofflinemodelin

7、gbasedonmachinelearningisemployedTheparametersselectionofSVRmodelisinvestigatedmainlyinthedissertationThegeneticalgorithm(GA、isalsousedtosolvetheproblemofparametersselectioninSVRmodelThismethodhasachievedgoodsimulationre

8、sultsandcanbeusedtoaccuratelyidentifytheproblemquickly2、AsupportvectorregressionmodeloftheNOxemissionpropertyisbuiltwiththehelpoftheexperimentdataforboilercombustionCombinedwiththepredictionmodel，thequantitativerelations

9、hipbetweenNOIemissionandoperationalparametersareobtainedCombinedwithGAthefiduciallyoxygencontentisreceivedaccordingtotheoptimalobjectoftheunit’spowersupplycoalconsumptionrateItalsooffersanewthoughtandmethodfordetermining

10、theoperatingoxygencontentduringactualvariantconditions3、AnnewalgorithmbasedonKarushKuhuTUcker(KKT)conditioilsandnearestneighbor仆n呵)methodandallimprovedadaptivemodelingmethodbasedonamodifiedaccurateon1inesupportvectorregr

11、ession(AOSVR)areadvisedtevisetheparametersofsoftsensormodelonlinetoimprovetheprecisionofthemodelAndkemelprincipalcomponentanalysis(KPCAlisusedtodealwiththenonlinearrelationshipbetweenoriginaldatasetsandextractsecondaryva

12、riablesforsoftsensorThenanewapproachforthechoice0fKPCAparametersandprincipalcomponentisproposed4、Acomprehensiveevaluationgoalcombiningwitheconomy，environmentandenergy—savingisproposedforthemodeloftheboilercombustionUsing

13、theimprovedmultiobjectiveparticleswarmoptimizationalgorithm，theoptimizationofboilerefficiencyNOxemissionsandauxiliaryelectricalconsumptionissettledAiudgementontheresultofmulti—objectiveoptimizationbasedonthebenefitisprop

14、osed，andallcomposesthecompletesolutionoftheboilercombustionperformanceoptimization5、Acompletedevelopmentprogramoftheboilercombustionoptimizationsystem(BCOS)isdesignedwhichincludessystemtheoryarchitecturedesignandsystemfu

15、nctionAndtheintegrationBCOSintoSISisalsointroducedKeywords：Thermalpowerplant，Boilercombustion，Machinelearning，Intelligentizedoptimization，Supportvectorregression，Geneticalgorithms，Kernelprincipalcomponentanalysis，Multiob