资源描述：
《disturbance models for offset free model prodictive control》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、DisturbanceModelsforOffset-FreeModel-PredictiveControlGabrielePannocchiaDept.ofChemicalEngineering,UniversityofPisa,56126Pisa,ItalyJamesB.RawlingsDept.ofChemicalEngineering,UniversityofWisconsin,Madison,WI53706Modelpredicti®econtrolalgorithmsachie®eoffset-freecontrol
2、objecti®esbyaddingintegratingdisturbancestotheprocessmodel.Thepurposeoftheseadditionaldistur-bancesistolumptheplant-modelmismatchandrorunmodeleddisturbances.Itseffec-ti®enesshasbeenpro®enforparticularsquarecasesonly.Forsystemswithanumberofmeasured®ariablespgreatertha
3、nthenumberofmanipulated()®ariablesm,itisclear()thatanycontrollercantrackwithoutoffsetatmostmcontrolled®ariables.Onemaythinkthatmintegratingdisturbancesaresufficienttoguaranteeoffset-freecontrolinthemcontrolled®ariables.Weshowthisideaisincorrectandpresentgeneralcondit
4、ionsthatallowzerosteady-stateoffset.Inparticular,anumberofintegratingdisturbancesequaltothenumberofmeasured®ariablesareshowntobesufficienttoguaranteezerooffsetinthecontrolled®ariables.Theseresultsapplytosquareandnonsquare,open-loopstable,integratingandunstablesystems
5、.IntroductionModelpredictivecontrolMPCarosefromtheindustrialŽ.stableandunstablesystems.AsshowninLeeetal.1994,Ž.applicationscalledIdentificationandCommandIDCOMŽ.steporimpulseresponsemodelsareparticularcasesoftheŽ.Richaletetal.,1978,andDynamicMatrixControlDMCŽ.state-sp
6、acemodels.Ž.CutlerandRamaker,1979.ThesecontrolalgorithmsusedIntheseformulations,offset-freeobjectivesareobtainedfinite-impulseorstep-responsemodelstopredictthefuturebyaugmentingthesystemstatewithintegratingdisturbances.processbehavior.Inordertoobtainoffset-freecontro
7、l,theIntheoriginalformulationsofMPCnoanalysiswasgivenmodelisupdatedwithfeedbackinformation.Comparingtheregardingsteady-stateoffset.InRawlingsetal.1994itwasŽ.currentmeasuredprocessoutputandthecurrentpredictedshownthataconstantoutputdisturbanceguaranteesoffset-output,a
8、constantbiastermisaddedtothefuturemodelfreeperformanceforsquaresystemswithoutintegratingforecasts.Theseconvolutionmodelscannotbeuse
