Model predictive control and optimal operation of continuous and batch nonlinear processes

o account for operating constraints, modeling errors and nonlinear characteristics, a significant number of Model Predictive Control (MPC) algorithms that utilize nonlinear process models in the on-line optimization have appeared in the literature. Almost all of the nonlinear MPC algorithms require solving a nonlinear program on-line which is computationally expensive and therefore not feasible for practical on-line implementation. This dissertation formulates an observer based nonlinear QDMC algorithm for open-loop stable and unstable processes. The algorithm can handle nonlinear models represented in state space and/or input-output form and has provisions for various kinds of disturbance models. Although a nonlinear model is used, only a single Quadratic Program (QP) is solved on-line. The proposed algorithm generalizes the previously published nonlinear QDMC algorithm for open-loop stable processes. There are very few results rigorously addressing the robust stability and performance aspects of MPC algorithms either for linear or nonlinear processes. In this dissertation, an on-line objective is proposed for a model predictive control algorithm for linear processes, which allows tuning for guaranteed robustness with respect to modeling errors, utilizing the concepts of...