Elevating Model Predictive Control Using Feedforward Artificial Neural Networks A Review.pdf

《Elevating Model Predictive Control Using Feedforward Artificial Neural Networks A Review.pdf》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、ChemicalProductandProcessModelingVolume4,Issue12009Article45ElevatingModelPredictiveControlUsingFeedforwardArtificialNeuralNetworks:AReviewSenthilKumarArumugasamy,UniversitiSainsMalaysiaZainalAhmad,UniversitiSainsMalaysiaRecommendedCitation:Arumugasamy,SenthilKumarandAhmad,Zainal(20

2、09)"ElevatingModelPredictiveControlUsingFeedforwardArtificialNeuralNetworks:AReview,"ChemicalProductandProcessModeling:Vol.4:Iss.1,Article45.DOI:10.2202/1934-2659.1424Availableat:http://www.bepress.com/cppm/vol4/iss1/45©2009BerkeleyElectronicPress.Allrightsreserved.ElevatingModelPre

3、dictiveControlUsingFeedforwardArtificialNeuralNetworks:AReviewSenthilKumarArumugasamyandZainalAhmadAbstractProcesscontrolinthefieldofchemicalengineeringhasalwaysbeenachallengingtaskforthechemicalengineers.Hence,themajorityofprocessesfoundinthechemicalindustriesarenon-linearandinthes

4、ecasestheperformanceofthelinearmodelscanbeinadequate.Recentlyapromisingalternativemodellingtechnique,artificialneuralnetworks(ANNs),hasfoundnumerousapplicationsinrepresentingnon-linearfunctionalrelationshipsbetweenvariables.Afeedforwardmulti-layeredneuralnetworkisahighlyconnectedset

5、ofelementarynon-linearneurons.Model-basedcontroltechniquesweredevelopedtoobtaintightercontrol.Manymodel-basedcontrolschemeshavebeenproposedtoincorporateaprocessmodelintoacontrolsystem.Amongthem,modelpredictivecontrol(MPC)isthemostcommonscheme.MPCisageneralandmathematicallyfeasiblesc

6、hemetointegrateourknowledgeaboutatarget,processcontrollerdesignandoperation,whichallowsflexibleandefficientexploitationofourunderstandingofatarget,andthusproducesoptimalperformanceofasystemundervariousconstraints.TheneedtohandlesomedifficultcontrolproblemshasledustouseANNinMPCandhas

7、recentlyattractedagreatdealofattention.Theefficacyoftheneuralpredictivecontrolwiththeabilitytoperformcomparablytothenonlinearneuralnetworkstrategyinbothsetpointtrackinganddisturbancerejectionprovestohavelesscomputationexpensefortheneuralpredictivecontrol.Theneuralnetworkmodelpredict

8、ivecontrol(NNMPC)me