Hybrid predictive control of nonlinear systems: method and applications to chemical processes

A hybrid control structure that unites bounded control with model predictive control (MPC) is proposed for the constrained stabilization of nonlinear systems. The structure consists of: (1) a finite-horizon model predictive controller, which can be linear or nonlinear, and with or without stability constraints, (2) a family of bounded nonlinear controllers for which the regions of constrained closed-loop stability are explicitly characterized and (3) a high-level supervisor that orchestrates switching between MPC and the bounded controllers. The central idea is to embed the implementation of MPC within the stability regions of the bounded controllers and employ these controllers as fall-back in the event that MPC is unable to achieve closed-loop stability (due, for example, to infeasibility of a given initial condition and/or horizon length, or due to computational difficulties in solving the nonlinear optimization). Switching laws, that monitor the closed-loop state evolution under MPC, are derived to orchestrate the transitions in a way that guarantees asymptotic closed-loop stability for all initial conditions within the union of the stability regions of the bounded controllers. The proposed hybrid control scheme is shown to provide a paradigm for the safe implementation of predictive control algorithms to nonlinear systems, with guaranteed stability regions. The efficacy of the proposed approach is demonstrated through applications to chemical reactor and crystallization process examples. Copyright # 2004 John Wiley & Sons, Ltd.