過程系統(tǒng)記憶增強型實時優(yōu)化方法.pdf

1、浙江大學信息學部博士學位論文過程系統(tǒng)記憶增強型實時優(yōu)化方法姓名：方學毅申請學位級別：博士專業(yè)：控制科學與工程指導教師：錢積新邵之江20091216AbstractRealtimeoptimization(RTO)forprocesssystemsiSimportanttolndustrialenergyeconomizationandprofitincreaseTherealtimeperformanceiSakeycriteriont

2、oevaluateRTOimplementationThus，theimprovementoftheperformanceattractsplentyofresearchinterestsTherearefruitfulproductionsconcerninglargescalerealtimeoptimizationalgorithmsHowever，asthesystemscalekeepsgrowingandtheprocess

3、modelstendstoincludemoremeso’scaleandmicroscaledetails，therealtimeperformanceremainsabottleneckofRTOThepurposeofthisworkaimsatimprovingthereal—timeperformancebasedontherepetitivefeatureofRTOproblemsequenceThisworkgivesad

4、etailedanalysisofRTOsystems，andproposesthemnemonicenhancementoptimization(MEO)methodOilthebasisofthesimilarityamongtheRTOproblemsThismethodviewstheoperationalconditionsandthevariousdisturbancesastheparametersofoptimizati

5、onproblemsItpreservespreviousSuccessfulsolutionsasexperience，andthenUSeStheexperiencetoapproximatetheoptimumtobesolvedTheapproximationisthensenttotheoptimizationalgorithmasthestartingpointtolocatetherealoptimumpreciselyT

6、heoreticalanalysisshowsthat，generallytheoptimumisacontinuousfunctionoftheparameters，basedonwhichtheMEOapproximationisprovedtoconvergetotherealoptimumwithprobability1asRTOcontinuestoIU兒Inthiswork，theMEOframeworkissetupItm

7、ainlyconsistsoftwoparts：themanagementoftheexperiencedatabaseandthemultivariateapproximationmethodsofMEOThethresholdoftheexperiencedatabaseisproposedtolimitthespacecomplexityAtthesametime，anincrementalDelaunaytriangulatio

8、nforarbitrarydimensionalspaceisconstructedSequentiallythenestednodeselectionmethodisintroduced，whichbuildsasolidfoundationforthenodeselectionforMEOapproximationTheMEOmethodwithzeroorderapproximationisproposedfirstBasedon