| Batch reactors perform an important role in the production of low volume, high quality chemicals. The advantages of batch reactors lie in the fact that they can be adapted to small production of various product types, they present less scale-up problems as compared to their continuous counterparts and they are better suited to carry out reactions that require stringent sterile conditions.; Batch reactors pose challenging control problems due to their inherently non-stationary and nonlinear nature. Effective control of batch reactors requires controllers that must be able to provide both good regulation and setpoint control over the entire operating range and in the presence of constraints.; This work proposes two interlinked model-based control paradigms to control batch processes. The first methodology is a multiple-model predictive control (MMPC) strategy that attempts to determine the present and future optimal control actions based on a set of models. Critical issues relating to the closed-loop stability of the resulting transient response are investigated. A theorem that establishes a criterion for local stability of the nonlinear system as well as a corollary that places an upper bound on the closed-loop response of the system are presented and proven. The performance of the MMPC control strategy is demonstrated on two example batch reactors: an electrochemical batch reactor and a chemical batch reactor.; Whereas MMPC and other conventional model predictive control (MPC) paradigms use essentially global time-scale information to determine the optimal control action, having local multi-scale information can lead to more selective disturbance compensation. This is accomplished using multi-scale models defined on trees that not only represent the dynamic nature of the process but also of the input disturbance structure. The criterion for the controllability of downward multi-scale systems is established. With this tree (hierarchical) structure, it is necessary to re-define the optimization strategy. To wit, a new (top-down) multi-scale optimization algorithm is developed. A multi-scale disturbance estimation algorithm that makes use of the Rauch-Tung-Streibel smoothing algorithm is developed and integrated into the multi-scale optimization methodology. The control strategy is implemented on both a continuous and a batch process and its performance compared to that of conventional model predictive control. |
