Distributed model predictive control synchronous and asynchronous

《Distributed model predictive control synchronous and asynchronous》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、732IEEETRANSACTIONSONSYSTEMS,MAN,ANDCYBERNETICSPARTA:SYSTEMSANDHUMANS,VOL.37,NO.5,SEPTEMBER2007DistributedModelPredictiveControl:SynchronousandAsynchronousComputationEduardoCamponogara,Member,IEEE,andSaroshN.Talukdar,Fellow,IEEEAbstractModelpredictivecontrol(MPC)hasbecomeoneofproblem(DP)ofop

2、eratingasystem,1theMPCalgorithmcantheleadingtechnologiestocontrolcomplexprocesses,toagreatbedividedinthreesteps.extent,asaresultofitsflexibilityandexplicithandlingofcon-1)UseadiscretizationtechniquetoconvertDPintoaseriesstraints.Givenadynamicproblem(DP),MPCconvertsDPintoaseriesofstaticoptimiz

3、ationproblems,therebyallowingtheuseofofstaticproblems{SPk}.standardoptimizationtechniquestocomputethecontrolsignals.2)SolveeachproblemSPkinturntoobtainasolutionXk.TherelianceofMPConcentralizedcomputations,however,stands3)CombinethesolutionsXkstothestaticproblemstoasabarriertoitsuseinthereal-

4、timeoperationoflargedynamicobtainasolutiontoDP.networks.Tothisend,thispaperproposesanextensiontoMPCbydecomposingDPintoanetworkofsmallbutcoupledsubproblemsOfprincipalimportanceistheeliminationofthetimevari-andsolvingthemwithanetworkofasynchronousagents.Thenetableinstep1):Typically,predictions

5、forthevariablesareresult,aftereachagentappliesMPCtoitsdynamicsubproblem,introducedoveratimehorizonand,accordingly,thesystemsisaseriesofsetsofstaticsubproblems.Ourfocusisonthesi-modelisapproximatedwithdifferenceequations.Then,stan-multaneoussolutionofthesesetsofstaticsubproblems.Thepaperdardo

6、ptimizationtechniquescanbeusedinstep2)tosolvethedeliversaframeworktocarryoutthedecompositionanddevelopsconditionsunderwhichtheiterativesynchronousprocessesofthestaticproblemsinplaceofanotherwiseintractableproblem.agentsconvergetosolutions.Furthermore,itproposesheuristicsInstep3),thecommonpra

7、cticeimplementsthecontrolinputsforasynchronousconvergenceandreportsexperimentalresultsgivenbyXkfromthebeginningofthetimehorizonofSPkfromprototypicaldynamicnetworks,demonstratingtheeffective-untilthatofSPk+1,atwhichinstantXk+1isimplemented.nessofthe