16477

1741 دكتري

دكتري

كنترل

اصفهان : دانشگاه صنعتي اصفهان

1399

نه، 75، 14ص.: مصور، جدول، نمودار

محسن مجيري

جواد عسكري

فريد شيخ الاسلام، مرضيه كمالي

1400/03/10

كتابنامه

برق

مهندسي برق و كامپيوتر

1400/03/11

2689031

1 Fast Model Predictive Control Based on Sensitivity Analysis for Linear Systems Hamid Kalantari Dehaghi hamid kalantari@ec iut ac ir January 2021 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 IranDr Mohsen Mojiri mohsen mojiri@cc iut ac irDr Javad Askari j askari@iut ac irDr Stevan Dubljevic Stevan Dubljevic@ualberta ca Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Department of Chemical and Materials Engineering University of Alberta Edmonton Alberta T6G 2V4 Canada AbstractIn this thesis we have proposed a new strategy based on sensitivity analysis and Wolfe method to solve a sequenceof the linear programmings LPs such as those arise in l1 MPC and the parametric quadratic programmings QPs such as that arise in quadratic model predictive control QMPC The main idea is to nd a relationship betweeneach two successive parametric LPs by using sensitivity analysis strategy For l1 MPC TA l1 MPC and SA l1 MPCare introduced TA l1 MPC takes O N n2 operations per step time where N and n are the prediction horizon andthe number of states respectively SA l1 MPC has not any limitation in usage and it reduces the runtime signi cantlycompared with common solvers For QMPC the Wolfe method based on Karush Kuhn Tucker conditions has beenused to convert parametric QPs to parametric Linear Programmings LPs and then the sensitivity analysis wouldbe applied to solve the sequence of the parametric LPs TA QMPC and SA QMPC are two obtained algorithmsby this strategy It is proved that the computational complexity of TA QMPC is O N n2 for a region of the initialconditions and the computational complexity of SA QMPC is O N 2 n2 for suf ciently small sampling time andfor all initial conditions By these algorithms the runtime is reduced signi cantly compared with usual methodssuch as interior point and active set Also SA l1 MPC has been used for urban traf c network managment problemwith modi ed Store and forward model KeywordsFast Linear MPC Sensitivity Analysis Wolfe method Linear and Quadratic Programming ComputationalComplexityIntroduction Model Predictive Control MPC is used in industry to solve a very wide spectrum of controlproblems by optimizing the closed loop response of multivariable systems under constraintson command inputs internal states and outputs Objective function is usually expressed asquadratic or linear form which leads to quadratic programming QP or linear programming LP respectively To date many effective algorithms for LP and QP has been developed to

محسن مجيري

جواد عسكري

فريد شيخ الاسلام، مرضيه كمالي