改进变异算子的差分进化算法并行性分析  

Parallel Analysis of Improving Mutation Operator Differential Evolution Algorithm

作　　者：孙康[1] 潘广贞[1] 

机构地区：[1]中北大学计算机与控制工程学院,山西太原030051

出　　处：《中北大学学报（自然科学版）》2014年第6期704-708,713,共6页Journal of North University of China(Natural Science Edition)

摘　　要：针对并行差分进化算法的全局搜索能力和寻优的稳定性弱的现状,基于DE/best/1变异算子提出了一种改进的差分进化算法变异算子.该算子前期采用DE/best/1变异方法,当进化代数超过设定的进化代数值时,采用改进的变异算子.通过拓宽变异算子的搜索域来提高种群的多样性,提高了差分进化算法的寻优能力.对改进变异算子的并行差分进化算法进行了函数测试,实验结果表明:相比普通的变异算子,在相同种群规模的前提下,改进的差分变异算子拓宽了遗传算法的搜索域,提高了算法的全局搜索能力;在不同的种群规模下,改进的变异算子增强了算法的寻优稳定性.In order to improve weak situation of the global search ability and the stablity of optimization in parallel differential evolution algorithm, an improved differential evolution algorithm＇s mutation operator was proposed based on DE/best/1 mutation operator. The operator used DE/best/1 in the early, and the improved mutation operator was used if the generations were greater than set generations. Through broadening the field of searching to improve the diversity of populations, the new operator enhanced the optimization capability of differential evolution algorithm. Several functions are tested and the results show that the proposed algorithm can enhance the global convergence ability remarkably in the same population sizes and has more stability in the different population sizes.

关 键 词：差分进化 变异算子 并行化 寻优质量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]