Research On Hybrid Particle Swarm Optimization For Distributed Flow Shop Scheduling

Production scheduling plays a crucial role in many manufacturing systems,and effective production planning is a key element to improve industrial production efficiency and resource utilization.Meanwhile,with the development of economic globalization and green manufacturing,it is of great practical importance to study the green scheduling of distributed flow shop.Most of the existing distributed flow shop scheduling problems lack the consideration of heterogeneous factories,and the effectiveness of existing algorithms cannot meet the production requirements due to the complexity of this kind of problems.Therefore,this thesis proposes the hybrid particle swarm optimization algorithm to solve the above two types of distributed flow shop scheduling problems considering energy consumption.First,a Q-learning-based multiobjective particle swarm optimization with local search within factories is proposed for a distributed permutation flow shop scheduling problem with makespan and total energy consumption.The algorithm uses an improved particle swarm optimization algorithm as a global search strategy to enable the population to quickly converge on multiple regions of the Pareto front.Meanwhile,the algorithm employs Q-learning guided variable neighborhood search as a local search strategy to balance the exploration and exploitation capabilities of the algorithm,thus further improving the quality of the particles.The proposed algorithm and comparison algorithms are tested on the benchmark problems,and the experimental results prove that the proposed algorithm has better convergence performance and diversity performance.Second,a multiobjective memetic algorithm with particle swarm optimization and Qlearning-based local search is proposed to solve the distributed heterogeneous hybrid flow shop scheduling problem.Multigroup particle swarm optimization is used as a global search strategy,aiming to improve the fast convergence of the solution set at the Pareto front.Meanwhile,two local search strategies are combined with multigroup particle swarm optimization to improve problem-specific neighborhood search for specific problems.Among them,the inter-factory local search can fine-tune the sequence of jobs between critical factories to balance the capacity of factories to process jobs,while the Q-learning intra-factory-based local search mainly guides the variable neighborhood search to better balance the exploration and exploitation of the algorithm,thus improving the solutions.Comparative experiments are conducted on benchmark problems with energy consumption,and the results verify that the proposed algorithm has superior convergence and distribution performance.The hybrid particle swarm optimization algorithm proposed in this thesis combines the global search strategy of the improved particle swarm optimization algorithm and the local search strategy based on Q-learning for problem-specific knowledge search,which ensures fast convergence and distribution diversity of the algorithm.The validation of the experimental part shows that the proposed algorithm has higher solution efficiency and also provides useful references for solving other complex distributed shop scheduling problems.