Research on energy-saving scheduling of job shop based on deep reinforcement learning

Aiming at the job shop scheduling problem in the context of green manufacturing, a scheduling framework based on disjunctive graph is proposed, which can cope with the complex and changeable production scheduling environment and respond to the production state of the shop and the energy consumption of machine tools in real time. In transforming the scheduling problem into a Markov decision process, two static matrices and five dynamic matrices are defined as the state space, the combined scheduling rules about energy-saving strategies are designed, and the reward function is described by both global and local ways. Finally, the model is trained using a dueling deep Q-network. By testing and comparing with other optimization algorithms such as scheduling rules and genetic algorithms, it is proved that the method proposed in this paper can effectively reduce the completion time and the total energy consumption of the job shop.