Infinite Horizon MPC applied to batch processes. Part II

In the first stage of this work (part I), an infinite horizon model predictive controller (IHMPC) using the closed-loop paradigm, was adapted to be applied to batch processes. Batch processes fall into the category of finite-duration systems, where the control objective is to track a given output profile (reference trajectory) during a finite time period, and so the use of an infinite horizon controller is not intuitive a priori. However, based on previous IHMPC results, this work shows that prediction along a fictitious horizon beyond the actual final batch time allows good performance and, more important, the stability (in the sense of the finite-duration systems) can be guaranteed. Simulation results show the closed-loop performance of the proposed strategy when a linear system is controlled. In addition, the proposed methodology allows a direct extension to a new MPC with learning properties - i.e. a MPC strategy that "learns" from previous batches - which accounts for the second important property required to batch processes: the repetitiveness. The way the proposed strategy is extended to the MPC with learning properties is presented in a second part of this work (Gonzalez et al., 2009).