Iterative Learning Control Guided Reinforcement Learning Control Scheme
for Batch Processes
Abstract
Iterative learning control (ILC) offers an effective learning control
scheme to solve the control problems of the batch processes. Although
the control performances of ILC systems can be improved batch-by-batch,
the convergence still strongly depends on the repeatability of the
process and thus lack of robustness. Meanwhile, the data-driven-based
deep reinforcement learning (DRL) algorithms have good robustness due to
the generalization of the neural network, but it has lower data
efficiency in training. In this paper, we propose a complementary
control scheme for the batch processes by employing a DRL guided by a
classical ILC, termed as the IL-RLC scheme. This scheme has higher data
efficiency than the DRL without guidance and better robustness than the
ILC, which are demonstrated by the numerical simulations on a linear
batch process and a nonlinear batch reactor. This work provides a way
for the application of DRL algorithm in the batch process control.