Improved Iterative Learning Model Predictive Control for Nonlinear Batch Processes

Abstract

This paper once again focuses on the research of iterative learning model predictive control (ILMPC) in batch processes, which aims to ensure that the system has fast convergence speed and good non-repetitive disturbance suppression ability. Firstly, using the process input and output data, a nonlinear batch process composite model consisting of a nominal ARX model and a JITL model is established, where the former is used to describe the process dynamics and the latter to evaluate the modeling error caused by the process nonlinearity. Then, an improved ILMPC (IILMPC) method is proposed, which considers the current iteration input, the input increment along the iteration axis, and the input increment in the time axis in an integrated two-dimensional feedback design framework. Meanwhile, a slack variable is also taken into account in the IILMPC design algorithm to ensure that a feasible solution will always exist. These advantages drive the presented control strategy to give better tracking performance than existing ILMPC. The convergence of the IILMPC algorithm is analyzed under mild conditions. Finally, a simulation case is given to verify the effectiveness of the proposed control method.