基于改进NSGA-Ⅲ的高维多目标柔性调度研究

In order to better simulate real production environments, this paper proposes a high-dimensional multi-objective flexible job shop scheduling problem that considers five types of indicators, including completion time, carbon emissions, machine load, delivery time, and production costs. Although the classical NSGA-III algorithm is very suitable for solving multi-objective optimization problems, it suffers from slow convergence speed and poor diversity in high-dimensional objective spaces. Therefore, this paper proposes an improved NSGA-III algorithm. Firstly, the traditional operation of associating reference points is improved by introducing a penalty-based boundary intersection method to associate the nearest reference point, and reinforcement learning is introduced to adaptively adjust the penalty parameter based on the current population space distribution, dynamically balancing the convergence and diversity of the algorithm. Secondly, to address the problem of overlapping solutions and poor diversity in high-dimensional spaces, a niche elimination method is introduced to equally eliminate individuals with poor diversity in different non-dominated layer. Finally, to adapt to the flexible job shop scheduling problem, the original crossover and mutation operator are improved. By conducting comparative experiments on 11 standard instances of flexible job shop scheduling problems, it was verified that the proposed improved algorithm exhibits better convergence and diversity in solving high-dimensional multi-objective flexible job shop scheduling problems, and has higher solution quality.