Time-Varying Setpoint Tracking for Batch Process Control Using Reinforcement Learning

Batch processes are indispensable for the production of low-volume and high-value products. However, control of a batch process is challenging due to the inherent nonlinearity, and time-varying characteristics. In this work, we aim to provide a comparative study of reinforcement learning approaches to perform tracking of a time-varying setpoint trajectory in a batch process. This can be accomplished by defining suitable state vectors, actions, and rewards for batch process control. Upon discretizing the state and action space, we utilize two model-free reinforcement learning methods. The efficacy of these two reinforcement learning methods is demonstrated in a non-isothermal batch reaction system. Further, we compare the performance of these reinforcement learning controllers with the nonlinear model predictive controller.