基于动态多策略差分进化模型的MOEA/D算法*

在基于分解技术的多目标进化算法的框架中,引入一种动态多策略差分进化模型。该模型在分析不同差分进化策略的特点基础上,选择了三种差分进化策略,并对每种策略分配一子种群。在进化过程中,依据每种策略对邻域更新的贡献度,动态的调整其子种群的大小。对比分析采用不同差分进化算法的性能,结果表明运用多个策略之间相互协同进化,有利于提高算法性能。将新算法同NSG-II和MOEA/D算法在LZ09系列基准函数上进行性能对比,实验结果显示该算法的收敛性和多样性均优于对比算法。将新应用于I型梁多目标优化设计问题中,获得的Pareto前沿均匀,且解集域较宽广,对比分析表明算法的工程实用性。

MOEA/D Based on Dynamic Multi-strategy Differential Evolution Model

In the framework of multi-objective evolutionary algorithm based on decomposition (MOEA/D), a dynamic multi-strategy differential evolution model was introduced to it (MOEA/D-DMDE). The model chooses three differential evolution strategies and each sub-population is corresponding to a differential evolution strategy based on the analysis of the characteristics of different strategies. In order to improve the performance of the algorithm, the size of sub-population is adjusted dynamically on the basis of a differential evolution strategy contribution for updated of neighborhood. Each strategy was adopted to participate in coordination during the evolution process. Via the comparative analysis of different schemes of differential strategy, MOEA/D-DMDE also performs well. Comparing with NSGA-II and MOEA/D on the LZ09 benchmarks, the experimental results indicate that MOEA/D-DMDE has a better performance in terms of convergence and diversity. To validate its performance on constraint multi-objective optimization problems, the proposed MOEA/D-DMDE is applied for solving the I-Beam. The uniformly distributed Pareto sets obtained by MOEA/D-DMDE show its practicability for engineering problems.