DISTRIBUTED MODEL PREDICTIVE CONTROL FOR CONSTRAINED PIECEWISE AFFINE SYSTEM WITH CONTROL COUPLING AND ITS APPLICATION IN QUADRUPLE-TANK

Abstract

ABSTRACT: To improve the model predictive controller design of nonlinear system, a cooperative distributed model predictive controller (DMPC) for a system comprised of several constrained piecewise affine (PWA) subsystems with coupled control information is proposed in this paper. Distributed scheme is applied for saving time of solving optimization. To improve the optimal performance of entire system, cooperation was promoted with consideration to a greater portion of the system-wide objective. The terminal's positively invariant set and cost of the entire closed-loop system were added to guarantee stability. The optimal problem of each subsystem was converted to mixed-integer quadratic programming (MIQP) solving for obtaining control actions. This method decomposed the optimization problem of a large-scale nonlinear system into several lower-dimension MIQPs such that solving time could be reduced dramatically. Another advantage of the distributed model predictive control strategy is non-convex nonlinear optimization could be replaced by finite MIQPs, and an approximate global optimum could be obtained rather than a local optimum. Finally, a numerical example and a quadruple-tank simulation are presented in order to assess the proposed distributed model predictive control strategy. Keywords: Piecewise affine system, distributed control, model predictive control, mixed integer quadratic programming, quadruple-tank