Research On Multi-product Batch Scheduling Problem With Limited Buffers Based On Improved PSO Algorithm

Production scheduling is the core of the enterprise in organizing and managing production process. A good scheduling scheme plays a key role in improving production efficiency and equipment utilization, reducing energy and cost for enterprise. As the increasingly diversified and personalized market demand, multi-product batch plant scheduling problem gets people’s attention as one of he classic modes of production. This paper investigated the multi-product batch plant scheduling problem under different storage policy through designing and improving Particle Swarm Optimization algorithm. Experiments results showed that the effectiveness and superiority of the proposed algorithms.To minimize the total flow time of multi-product batch plant scheduling problem with limited buffers, the multi-swarm particle swarm optimization algorithm was proposed. The algorithm used multiple swarms to increase the diversity of initial particles and selected several good particles of each sub-swarm as the immigrant particles in the process of parallel evolution of sub-swarms to make the sub-swarms affect and promote each other, which prevented the result running into the local optimum prematurely and enhanced the global research ability. It utilized the variable neighborhood search on the elite swarm consisting of each sub-swarm’s best particle to further improve its convergence accuracy.To minimize the total flow time of multi-product batch scheduling under uncertain processing time with zero wait, a batch scheduling algorithm based on the differential evolution particle swarm optimization was proposed. With basic particle swarm optimization as the algorithm’s framework, the algorithm used the opposition-based learning to initialize population to improve the quality of the initial population, utilized permutation-based differential evolution to optimize individual extremum, then through the variable neighborhood search to further improve its convergence accuracy.In order to solve the multi-product batch plant schedule with multi-objective, the differential evolution particle swarm optimization was improved. It updated pareto solutions of particles based on the Pareto dominance relations. Generational distance was calculated to decide when using differential evolution to change particles search area, which prevented the results running into local optimum. Variable neighbor search was imported to increate the local research ability.Different scale instances were simulated and simulation results demonstrated that the improved algorithms have more remarkable performance.