A Dynamic Chemical Production Scheduling Method based on Reinforcement Learning

The capacity and demands of the modern chemical industry are increasing day by day. Dynamic chemical production scheduling refers to the processing strategy implemented in response to changing order demands. This problem is difficult to solve by traditional methods due to the uncertainty of task objectives. The proximal policy optimization algorithms, which is a kind of reinforcement learning methods, is introduced into the dynamic chemical production scheduling problem in this paper. And an improved state function considering the difference between short-term and long-term orders is proposed. It effectively solved the chemical production scheduling problem with uncertainty. In addition, experiments are also carried out on the dynamic chemical production scheduling model, and compared with the policy gradient algorithm. The results show that the method proposed in this paper obtains more rewards in scheduling, faster convergence, and less performance fluctuation. Which proved that the methods presented in this paper can stably deal with the complexity and uncertainty of chemical production scheduling.