Batch control improvement by model predictive control based on multiple reduced-models

In this research study, model predictive control (MPC) utilizing multiple reduced-models running in series is developed and studied to investigate an improved temperature-control performance of an exothermic batch reactor. Three steps of batch-model construction are presented, which involve (i) a reference-profile determination, (ii) an operating-condition selection and (iii) a model-reduction. Different pseudo steady-states conditions are properly selected along the closed-loop reference profiles with regards to overall closed-loop poles of the system. The models further individually determined their minimal-phases to attain only controllable and observable states. Consequently, different model-orders can be chosen corresponding to their controllability and observability. Simulation results have shown that, in a nominal case, the proposed controller provides control performances as good as a single-model based controller does. However, in presences of plant/model mismatches, the reduced-controller provides much better and more robust control performances.