Research On Fast Model Predictive Control Algorithms And The Applications Based On Embedded Systems

In the last twenty years,the theory and technology of predictive control have been gradually mature.Due to the high demands for the computing resources,the algorithms of predictive control have been widely used in long control rate process,such as petrochemical industry and power sys-tem.In recent years,with the rapid development of embedded systems,the control requirements of high sampling rate process such as aerospace and automobile have been continuously improved so that fast model predictive control(MPC)based on embedded systems becomes hot gradually.FPGA(Field Programmable Gate Array),ARM(Advanced RISC Machine)are two well-known embedded platforms,which have large numbers of programmable logic resources so that they can fulfill the complex computation such as matrix calculation.This thesis copes with the design of the fast model predictive control algorithms and the applications on embedded systems related to the control of high sampling time system like Buck DC-DC converters.The following issues are addressed in this paper:·The issue of predictive functional control algorithm for second-order system such as Buck DC-DC converter which has complex poles is addressed here.To avoid the limitation of the computation burden from the receding optimization in conventional model predictive control,this paper proposes the algorithm of predictive functional control(PFC),which can fast reach the set point to achieve the control of the forward DC-DC converter in fuel cells through its characteristic of one-step prediction method.Finally,by using the implementa-tion of the predictive functional control on General Purpose Inverter Controller(GPIC)to control the DC-DC circuit,the experimental results show that the DC-DC converter output voltage can track the design voltage quickly,and has a good response of step change and load variation.·A fast model predictive control(MPC)is addressed,which uses closed-loop dual mode structure to guarantee the stability for the Buck DC-DC converter and utilizes its characteris-tic of terminal optimal to simplify the computation of converting MPC problem to quadratic programming problem(QP).Meanwhile based on the measured data through simulation this algorithm employs singular value decomposition(SVD)to reduce the computation com-plexity of the QP solver and finally chooses dual fast gradient-projection(GPAD)method for the optimization.It is clearly shown from the simulation experiments that output voltage of the DC-DC converter can track the design point quickly,and has a good response to a wide range change in load variation.·A novel fast model predictive control(MPC)methodology based on linear parameter vary-ing(LPV)systems is addressed here.The proposed approach can deal with large-scale problems better than conventional fast MPC methods.First,we do not eliminate the equal-ity constraints given by the model equations to get a condensed quadratic programming(QP)problem,as the model of the LPV system changes and it will be time-consuming to refor-mulate the QP problem at each sampling time.Instead,the proposed approach constructs a sparse QP problem by keeping the equality constraints.Although the resulting QP problem has larger dimension than the condensed one,it can be reformulated and solved as a system of piecewise affine(PWA)equations given by the Karush-Kuhn-Tucker(KKT)conditions of optimality and solved through a Newton-method and exact line search in a fast way.The performance is tested and compared with off-the-shelf QP solvers on the conventional buck DC-DC converter control problem in simulations.The proposed methodology works well for the controller and significantly increase the optimization speed.·Fast model predictive controllers based on embedded systems are developed here.To verify the feasibility of fast MPC controller based on chips,a motor tracking system and a wind turbine in low wind speed are successfully controlled by all three algorithms of fast solver in this thesis.