The application of fuzzy reasoning techniques and neural network structures to model-based predictive control (MPC) is studied. First, basic structures of MPC algorithms are reviewed. Then, applications of fuzzy systems of the Takagi-Sugeno type in explici

SOFTCOMPUTINGINMODEL–BASEDPREDICTIVECONTROL
´PIOTRTATJEWSKI,MACIEJ AWRYNCZUK
InstituteofControlandComputationEngineering
FacultyofElectronicsandInformationTechnology
WarsawUniversityofTechnology
ul.Nowowiejska15/19,00–665Warsaw,Poland
e-mail:{P.Tatjewski,M.Lawrynczuk}@ia.pw.edu.pl
Theapplicationoffuzzyreasoningtechniquesandneuralnetworkstructurestomodel-basedpredictivecontrol(MPC)isstudied.First,basicstructuresofMPCalgorithmsarereviewed.Then,applicationsoffuzzysystemsoftheTakagi-SugenotypeinexplicitandnumericalnonlinearMPCalgorithmsarepresented.Next,manytechniquesusingneuralnetworkmodelingtoimprovestructuralorcomputationalpropertiesofMPCalgorithmsarepresentedanddiscussed,fromaneuralnetworkmodelofaprocessinstandardMPCstructurestomodelingpartsorentireMPCcontrollerswithneuralnetworks.Finally,asimulationexampleandconclusionsaregiven.
Keywords:processcontrol,modelpredictivecontrol,nonlinearsystems,fuzzysystems,neuralnetworks
1.Introduction
TheModel-BasedPredictiveControl(MPC)techniqueistheonlyoneamongadvancedcontroltechniques–usu-allyde nedastechniquesmoreadvancedthanthewell-knownPIDcontrol–whichwastremendouslysuccessfulinpracticalapplications,exertinggreatin uenceonthedirectionsofthedevelopmentofindustrialcontrolsys-temsaswellasresearchconductedinthisarea(Allgöweretal.,1999;Brdy´sandTatjewski,2005;Eder,1999;Ma-ciejowski,2002;Mayneetal.,2000;MorariandLee,1999;QinandBadgwell,2003;Rossiter,2003;Tatjew-ski,2002).Thereareseveralreasonsforthissuccess.First,predictivecontrolalgorithmscantakeintoaccountinanaturalwayconstraintsonbothprocessinputs(con-trolsignals)andprocessoutputvalues(controlledvari-ables),whichoftendecideonthequality,effectivenessandsafetyofproduction.Theygeneratecontrolsignalstakingon-lineintoaccounttheseconstraintsand,owingtoadirectuseofamodel,alsointernalinteractionswithintheprocess.Thus,theycannaturallybeappliedtomul-tivariableprocesscontrol,alsowhenthenumbersofthecontrolinputsandthecontrolledvariablesdiffer.Third,theprincipleofoperationofthesealgorithmsiscompre-hensibleandrelativelyeasytoexplaintoengineeringandoperatorstaff,whichisaveryimportantaspectwhenin-troducingnewtechniquesintoindustrialpractice.
2.Model-BasedPredictiveControl
2.1.PredictiveControlPrinciple
Thegeneralprincipleofpredictivecontrolisthefollow-ing:Ateachconsecutivesamplinginstantk,thecontrolinputsu(k)=u(k|k),u(k+1|k),...,u(k+Nu 1|k)arecalculated,assumingu(k+p|k)=u(k+Nu 1|k)forp≥Nu,whereNuisthecontrolhorizon.Theap-pliednotation‘u(k+p|k)’meansthepredictionofthecontrolinputvalueforthefuturetimek+p,performedatthetimek.Thecontrolinputsarecalculatedinsuchawayastominimizedifferencesbetweenthepredictedcontrolledoutputsy(k+p|k)andtheforeseensetpointsfortheseoutputsysp(k+p|k)overthepredictionhorizonN(p=1,2,...,N).Then,onlythe rstelementu(k|k)ofthecalculatedsequenceisappliedtotheprocess,i.e.,u(k)=u(k|k).Atthenextsample(k+1),thereoc-cursanewmeasurementoftheprocessoutputsandthewholeprocedureisrepeated,withthepredictionhorizonofthesamelengthN,butshiftedbyonestepforward(theprincipleofarecedinghorizon,alsocalledtherepetitivecontrolprinciple(e.g.,Findeisen,1997;Findeisenetal.,1980)).ThisprincipleispresentedinFig.1,foraSISO(single-inputsingle-output)process.
AmodeloftheprocessusedintheMPCalgorithmisusuallyonlyanapproximationofreality.Moreover,thereissomeuncertaintyintheuncontrolledinputs,whichcanbeinaccuratelymeasuredornotmeasuredatall.There-fore,theoutputpredictionsusuallydifferfromthe(later)measuredvalues.ThisfactisdepictedinFig.1asadis-
Theworkwassupportedbyastatutoryresearchproject.
SOFT COMPUTING IN MODEL–... 20页 免费 Nonlinear Model Predicti... 12... Model Predictive Control 模型预测控制—应用方法 Model-based Predictive Control...
model-based predictive c... 暂无评价 21页 免费 SOFT COMPUTING IN MODEL... Multistep Model Predictive Control Based on Artificial Neural Networks 基于...
model-based predictive c... 暂无评价 21页 免费 SOFT COMPUTING IN MODEL... Model predictive control of CSTR based on local model networks Abstract-- ...
model-based predictive control_专业资料。A long-range predictive control ... SOFT COMPUTING IN M... 20页 免费 Fuzzy model-based pred... 28页 ...
Application of Nonlinear Predictive Control Based on RBF Network Predictive Model in MCFC Plant_专业资料。维普资讯 http://www.cqvip.com Junlo hnhiJatn ...
LS-SVM model based nonlinear predictive control for MCFC system_专业资料。维普资讯 http://www.cqvip.com 784 Ce tldhjn n cA07()4-5 hne /ZeigUiSi 2...
暂无评价 21页 免费 SOFT COMPUTING IN MODEL–... 20页 免费喜欢此文档的... Active SLAM using Model Predictive Control and Attractor based Exploration ...
model-based predictive... 暂无评价 21页 免费 SOFT COMPUTING IN M... ... Fuzzy Model Based Predictive Control by Instantaneous Linearization János ...
GA-Based Model Predictive Control of Semi-Active Landing Gear_专业资料 暂无评价
homepage: www.elsevier.com/locate/renene Technical Note Fuzzy model based multivariable predictive control of a variable speed wind turbine: LMI approach S....
热门文档