Energy-Efficient Optimization Of Flexible Job-Shop Scheduling Based On DQN Co-Evolutionary Algorithm

As fossil fuels such as coal and oil become increasingly depleted,the energy problem has become more prominent,and strengthening energy-saving awareness and improving energy efficiency levels have become the necessary way for sustainable development in the country.At the same time,flexible job shop scheduling,as an important part of manufacturing systems,has received increasing attention.Efficient and reasonable workshop scheduling can not only achieve cost reduction and efficiency improvement,but also effectively reduce environmental pollution.Therefore,this paper conducts in-depth research on energy-efficient scheduling problems in flexible job shops,with the following main research contents:To achieve energy-saving in workshop operations,three strategies,including machine selection,speed adjustment,and timely on/off,are simultaneously implemented as energy-saving measures.A mixed integer linear programming model is established with the objective of minimizing comprehensive energy consumption,and a cooperative coevolution algorithm based on DQN is proposed to solve it.The algorithm incorporates cooperative evolution ideas,allowing the three decisions of processing sequence,machine selection,and speed level selection to cooperate and compete for common evolution.Multiple local search operators are designed for each decision,and DQN reinforcement learning is used to recommend the operators that are more consistent with the current workshop operating state and more conducive to energy saving and consumption reduction.In addition,the algorithm inherits the strong pertinence and fast convergence speed of local search algorithms,and designs a restart strategy based on the archive set and cross operation to help the algorithm jump out of the local optimum.To reduce energy consumption and improve production efficiency,a multiobjective mixed integer programming model is constructed,and the above algorithm is improved to adapt to multi-objective problem solving.The model includes four objectives,namely,processing energy consumption,transportation energy consumption,standby energy consumption,and manufacturing cycle,to achieve energy-saving in the manufacturing/transportation process,reduce invalid energy consumption,and improve production efficiency.Under the guidance of objectives,multiple heuristic initialization rules are constructed,and the solution with the maximum crowding distance is selected using the tournament selection rule to prompt the algorithm to start from the most promising solution space.Local search operators are designed based on objectives to achieve objective-based cooperative evolution.An adaptive dynamic restart threshold is added to the restart mechanism to ensure stable operation in the early stage of the algorithm and rapid jumping out of the local optimum in the later stage.Experimental results show that the cooperative optimization algorithm based on multi-decision and the other five comparison algorithms have significant advantages in energy consumption indicators and stability when solving the studied problems,and the improvement in energy saving is up to 13.8%.The multi-objective cooperative optimization method and the other five comparison algorithms can obtain a Pareto front solution set with better convergence and diversity when solving the studied problems.