000888993 001__ 888993
000888993 005__ 20240712112901.0
000888993 0247_ $$2doi$$a10.1016/j.jprocont.2020.06.012
000888993 0247_ $$2ISSN$$a0959-1524
000888993 0247_ $$2ISSN$$a1873-2771
000888993 0247_ $$2Handle$$a2128/26599
000888993 0247_ $$2WOS$$aWOS:000567786500021
000888993 037__ $$aFZJ-2020-05380
000888993 082__ $$a004
000888993 1001_ $$0P:(DE-HGF)0$$aVaupel, Yannic$$b0
000888993 245__ $$aAccelerating nonlinear model predictive control through machine learning
000888993 260__ $$aAmsterdam [u.a.]$$bElsevier Science$$c2020
000888993 3367_ $$2DRIVER$$aarticle
000888993 3367_ $$2DataCite$$aOutput Types/Journal article
000888993 3367_ $$0PUB:(DE-HGF)16$$2PUB:(DE-HGF)$$aJournal Article$$bjournal$$mjournal$$s1609245469_13522
000888993 3367_ $$2BibTeX$$aARTICLE
000888993 3367_ $$2ORCID$$aJOURNAL_ARTICLE
000888993 3367_ $$00$$2EndNote$$aJournal Article
000888993 520__ $$aThe high computational requirements of nonlinear model predictive control (NMPC) are a long-standing issue and, among other methods, learning the control policy with machine learning (ML) methods has been proposed in order to improve computational tractability. However, these methods typically do not explicitly consider constraint satisfaction. We propose two methods based on learning the optimal control policy by an artificial neural network (ANN) and using this for initialization to accelerate computations while meeting constraints and achieving good objective function value. In the first, the ANN prediction serves as the initial guess for the solution of the optimal control problem (OCP) solved in NMPC. In the second, the ANN prediction is improved by solving a single quadratic program (QP). We compare the performance of the two proposed strategies against two benchmarks representing the extreme cases of (i) solving the NMPC problem to convergence using the shift-initialization strategy and (ii) implementing the controls predicted by the ANN prediction without further correction to reduce the computational delay. We find that the proposed ANN initialization strategy mostly results in the same control policy as the shift-initialization strategy. The computational times are on average 45% longer but the maximum time is42% smaller and the distribution is tighter, thus more predictable. The proposed QP-based method yields a good compromise between finding the optimal control policy and solution time. Closed-loop infeasibilities are negligible and the objective function is typically greatly improved as compared to benchmark (ii). The computational time required for the necessary second-order sensitivity integration is typically an order of magnitude smaller than for solving the NMPC problem to convergence.    Previous article in issue
000888993 536__ $$0G:(DE-HGF)POF3-899$$a899 - ohne Topic (POF3-899)$$cPOF3-899$$fPOF III$$x0
000888993 588__ $$aDataset connected to CrossRef
000888993 7001_ $$0P:(DE-HGF)0$$aHamacher, Nils C.$$b1
000888993 7001_ $$0P:(DE-HGF)0$$aCaspari, Adrian$$b2
000888993 7001_ $$0P:(DE-HGF)0$$aMhamdi, Adel$$b3
000888993 7001_ $$0P:(DE-HGF)0$$aKevrekidis, Ioannis G.$$b4
000888993 7001_ $$0P:(DE-Juel1)172025$$aMitsos, Alexander$$b5$$eCorresponding author$$ufzj
000888993 773__ $$0PERI:(DE-600)2000438-2$$a10.1016/j.jprocont.2020.06.012$$gVol. 92, p. 261 - 270$$p261 - 270$$tJournal of process control$$v92$$x0959-1524$$y2020
000888993 8564_ $$uhttps://juser.fz-juelich.de/record/888993/files/JPROCONT_2019_618_Rev2.pdf$$yOpenAccess
000888993 909CO $$ooai:juser.fz-juelich.de:888993$$pdnbdelivery$$pdriver$$pVDB$$popen_access$$popenaire
000888993 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-HGF)0$$aRWTH Aachen$$b0$$kRWTH
000888993 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-HGF)0$$aRWTH Aachen$$b1$$kRWTH
000888993 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-HGF)0$$aRWTH Aachen$$b2$$kRWTH
000888993 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-HGF)0$$aRWTH Aachen$$b3$$kRWTH
000888993 9101_ $$0I:(DE-588b)5008462-8$$6P:(DE-Juel1)172025$$aForschungszentrum Jülich$$b5$$kFZJ
000888993 9101_ $$0I:(DE-588b)36225-6$$6P:(DE-Juel1)172025$$aRWTH Aachen$$b5$$kRWTH
000888993 9131_ $$0G:(DE-HGF)POF3-899$$1G:(DE-HGF)POF3-890$$2G:(DE-HGF)POF3-800$$3G:(DE-HGF)POF3$$4G:(DE-HGF)POF$$aDE-HGF$$bProgrammungebundene Forschung$$lohne Programm$$vohne Topic$$x0
000888993 9141_ $$y2020
000888993 915__ $$0StatID:(DE-HGF)0200$$2StatID$$aDBCoverage$$bSCOPUS$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0160$$2StatID$$aDBCoverage$$bEssential Science Indicators$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)1160$$2StatID$$aDBCoverage$$bCurrent Contents - Engineering, Computing and Technology$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0600$$2StatID$$aDBCoverage$$bEbsco Academic Search$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0100$$2StatID$$aJCR$$bJ PROCESS CONTR : 2018$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0113$$2StatID$$aWoS$$bScience Citation Index Expanded$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0150$$2StatID$$aDBCoverage$$bWeb of Science Core Collection$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)9900$$2StatID$$aIF < 5$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0510$$2StatID$$aOpenAccess
000888993 915__ $$0StatID:(DE-HGF)0030$$2StatID$$aPeer Review$$bASC$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0300$$2StatID$$aDBCoverage$$bMedline$$d2020-09-03
000888993 915__ $$0StatID:(DE-HGF)0199$$2StatID$$aDBCoverage$$bClarivate Analytics Master Journal List$$d2020-09-03
000888993 920__ $$lyes
000888993 9201_ $$0I:(DE-Juel1)IEK-10-20170217$$kIEK-10$$lModellierung von Energiesystemen$$x0
000888993 9801_ $$aFullTexts
000888993 980__ $$ajournal
000888993 980__ $$aVDB
000888993 980__ $$aUNRESTRICTED
000888993 980__ $$aI:(DE-Juel1)IEK-10-20170217
000888993 981__ $$aI:(DE-Juel1)ICE-1-20170217