基于动态种群多策略差分进化模型的多目标进化算法  

Multi-objective Evolutionary Algorithm Based on Dynamic Population Multi-strategy Differential Models

作　　者：王亚辉[1] 吴金妹[1] 贾晨辉[2] 

机构地区：[1]华北水利水电大学机械学院,河南郑州450011 [2]河南科技大学机电工程学院,河南洛阳471023

出　　处：《电子学报》2016年第6期1472-1480,共9页Acta Electronica Sinica

基　　金：国家自然科学基金(No.51475142)

摘　　要：针对复杂的多目标优化问题,根据不同差分进化策略的特点,提出一种基于动态种群多策略差分进化模型和分解机制的多目标进化算法(MOEA/D-DPMD).该算法将种群划分为3个子种群,每个子种群分配一种差分进化策略.为了提高算法的性能,依据每种差分进化策略的贡献度,动态的调整子种群的规模,各差分进化策略之间相互配合协同进化.采用具有复杂的PS的LZ09系列基准函数,测试新算法的性能,仿真结果表明邻域规模为25时性能最好.通过不同差分进化策略之间的对比分析,新算法也具有较强的优势.将其与MOEAD/DE和NSGA-II算法对比分析,结果显示该算法的收敛性和多样性均优于另外两种算法,是求解复杂多目标问题的有效方法.According to the characteristics of differential evolution,a multi-objective evolutionary algorithm based on dynamic population multi-strategy differential models and decomposition （MOEA/D-DPMD）is proposed to solve the ex-pensive problems.The algorithm divides the population into three sub-populations and each sub-population is corresponding to a differential evolution strategy.In order to improve the performance of the algorithm,the size of sub-population is adjus-ted dynamically on the basis of a differential evolution strategy contribution.Each strategy is adopted to participate in coordi-nation during the evolution process.Through the test simulation on the LZ09 benchmarks with complicated Pareto Set （PS）, MOEA/D-DPMD shows a best performance with a neighborhood size of 25 .Via the comparative analysis of different schemes of differential strategy,MOEA/D-DPMD also performs well.The experimental results indicate that MOEA/D-DPMD has a better performance in terms of convergence and diversity compared with MOEA/D and NSGA-II,which is an effective way for solving complex multi-objective optimization problems.

关 键 词：分解机制 多策略差分进化 动态种群 多目标优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]