Reinforcement learning for energy-efficient control of multi-stage production lines with parallel machine workstations

An effective approach to enhancing the sustainability of production systems is to use energy-efficient control (EEC) policies for optimal balancing of production rate and energy demand. Reinforcement learning (RL) algorithms can be employed to successfully control production systems, even when there is a lack of prior knowledge about system parameters. Furthermore, recent research demonstrated that RL can be also applied for the optimal EEC of a single manufacturing workstation with parallel machines. The purpose of this study is to apply an RL for EEC approach to more workstations belonging to the same industrial production system from the automotive sector, without relying on full knowledge of system dynamics. This work aims to show how the RL for EEC of more workstations affects the overall production system in terms of throughput and energy consumption. Numerical results demonstrate the benefits of the proposed model.