Data‐driven optimal terminal iterative learning control with initial value dynamic compensation

Abstract

Iterative learning control is an effective control strategy for control of batch processes and initial condition is one of the most important factors affecting convergence of iterative learning batch process control. In this study, a novel initial value dynamic compensation-based data-driven optimal terminal iterative learning control (IDC-DDOTILC) approach is proposed for non-linear systems under random initial conditions. The unknown influence on the terminal output caused by the initial states is deduced by using a dynamical linearisation of the controlled non-linear system along the iteration direction, and then the unknown influence is estimated iteratively and incorporated into the learning control law. As a result, the proposed IDC-DDOTILC can drive the terminal output of the plant to attain the target value at the endpoint asymptotically under iteration-varying initial conditions. Two chemical engineering examples including a batch reactor and a fed-batch ethanol fermentation process are used to demonstrate effectiveness of the proposed control algorithm.