Combined Iterative Learning and Model Predictive Control Scheme for Nonlinear Batch Processes

Abstract

Iterative learning control (ILC) and model predic-tive control (MPC) are both effective control methods for batch processes. Using ILC and MPC together, we propose a combined design scheme for nonlinear constrained batch processes. This scheme utilizes the historical batch data, as well as the current measurements about the process through a two-dimensional (2D) framework. In our combined 2D design scheme, the ILC part is designed using optimal run-to-run feedback with the historical batch data, while the MPC part is designed using real-time feed-back with the current sampled measurements within the batch. By combining the run-to-run ILC and the real-time feedback-based MPC, the current control inputs are optimized based on historical batch data and real-time measurements, resulting in enhanced control performance in both the batch and time directions, as well as the ability to deal with enforced constraints in the time direction. Our design allows control objectives to be attained in several successive batches, not necessarily in a single batch. Finally, a rigorous theoretical analysis has been presented to demonstrate the perfect tracking stability of the combined scheme.