The Multi-objective Evolutionary Algorithm Based on Decomposition (MOEA/D) overcomes the limitation of evolutionary algorithm based on a Pareto dominant relationship in dealing with the problem of super multi-objective optimization and has wide application prospects, but there are also some problems, such as the lack of diversity and slow convergence speed in the later-stage evolution species. This article specifically conducts a systematic study on the population diversity of the MOEA/D algorithm and proposes three improvements: firstly, the evolutionary strategy of competition between SBX and DE operator is adopted to overcome the problem of the species diversity degradation of a single operator; secondly, an adaptive adjusting strategy of modulation probability is introduced to promote the variability of later-stage evolution species; finally, a method of double-faced mirror theory boundary optimization is used to prevent species aggregating at the boundary. The research shows that the above three improvement measures can effectively improve the population diversity of the MOEA/D algorithm. On the basis of this research, an improved MOEA/D algorithm with adaptive evolution strategy (AES-MOEA/D) is proposed. Simulation experiment indicators show that the convergence and comprehensive performance of the AES-MOEA/D algorithm are better than that of the basic MOEA/D algorithm and the other four comparison algorithms, which shows that the research on the maintenance of the diversity of the MOEA/D algorithm population helps improve the comprehensive performance of the MOEA/D algorithm.

基于分解的多目标进化算法（MOEA / D）克服了基于帕累托优势关系的进化算法在解决超多目标优化问题上的应用前景广阔，但也存在一些问题，例如后期进化物种缺乏多样性和收敛速度慢。本文专门对MOEA / D算法的种群多样性进行了系统的研究，并提出了三点改进：首先，采用SBX与DE算子竞争的进化策略来克服单个算子的物种多样性退化的问题。其次，引入调制概率的自适应调整策略，以促进后期进化物种的变异性。最后，为了防止物种在边界处聚集，采用了双面镜像理论边界优化方法。研究表明，以上三种改进措施可以有效地提高MOEA / D算法的种群多样性。在此基础上，提出了一种具有自适应进化策略的改进型MOEA / D算法（AES-MOEA / D）。仿真实验指标表明，AES-MOEA / D算法的收敛性和综合性能均优于基本的MOEA / D算法和其他四种比较算法，这表明在维持MOEA多样性方面的研究/ D算法填充有助于提高MOEA / D算法的综合性能。