基于多策略排序变异的多目标差分进化算法  

Multi-objective differential evolution algorithm with multi-strategy and ranking-based mutation

作　　者：艾兵[1] 董明刚[1] 敬超[1] Ai Bing;Dong Minggang;Jing Chao(School of Information Science ＆ Engineering,Guilin University of Technology,Guilin Guangxi 541004,China)

机构地区：[1]桂林理工大学信息科学与工程学院,广西桂林541004

出　　处：《计算机应用研究》2018年第7期1950-1954,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61563012;61203109);广西自然科学基金资助项目(2014GXNSFAA118371;2015GXNSFBA139260);广西研究生教育创新计划资助项目(YCSZ2015165)

摘　　要：针对多目标差分进化算法求解多目标优化问题时收敛慢和均匀性欠佳等不足,提出了一种基于多策略排序变异的多目标差分进化算法。该算法利用基于排序变异算子来快速接近真实的Pareto最优解,同时引入多策略差分进化算子以保持种群的多样性;通过自适应策略动态调整控制参数以提高算法的鲁棒性,并且从理论证明的角度分析了所提算法的收敛性。仿真实验结果表明,该算法相对于近期相关文献中的改进算法具有更好的收敛性与多样性,从而表明了所提算法的有效性。Focused on slow convergence and poor uniformity of multi-objective differential evolution algorithm in solving multiobjective optimization problems,this paper put forward a multi-objective differential evolution algorithm with multi-strategy and ranking-based mutation. This algorithm took full advantage of ranking-based mutation operator to approximate the true Pareto optimal solutions quickly,and introduced the multi-strategy differential evolution operator to maintain the diversity of population. It adjusted the control parameters dynamically through the adaptive parameter adjustment to enhance the robustness. This paper analyzed the convergence of the proposed algorithm from the point of view of theoretical proof. Simulation results indicate that,compared with some recently proposed improved algorithms,the proposed algorithm has better convergence and diversity,which demonstrates the effectiveness of the proposed algorithm.

关 键 词：多目标优化 多策略差分进化 排序变异算子 自适应参数调整 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]