| This work describes the application of model predictive control to hybrid systems. Hybrid systems are modeled with linear-time time-invariant state space equations. The nonlinear variables present in this model are linearized with the use of a so-called Glover's transformation. With a linear hybrid model, model predictive control is implemented. The resulting control law is represented as a mixed integer quadratic programming problem with several sets of linear constraints which arise from linearization, safety, and performance issues. Several simple hybrid systems are simulated to illustrate this model predictive control technique. However, the main focus of this dissertation is to develop and apply this technique to large-scale chemical engineering systems including batch processes. Through the simulation of a batch chemical plant, it is demonstrated how this methodology can be applied to batch process supervision. Simulation results reveal that this method controls batch processes in a safe and efficient manner. Consequently, this control methodology would be quite useful during the design and operational phases of the plant. During the design phase, startup/shutdown and emergency operation procedures for the batch plant could be synthesized which conform with necessary safety and operating constraints. During operation of the plant, this synthesis package could be used to aid the plant operator when unexpected plant states arise by finding a control sequence to move the plant within nominal operating states. |
