摘要
提出一种求解动态优化问题的改进差分进化算法.新算法将种群分为跟踪和搜索两个种群.通过监测跟踪种群的当前最优解和次优解来判断环境是否发生变化.发现环境变化时,重新计算种群适应值,分别找出变化后两个种群新的最优解.最优解好的种群,变为跟踪种群,保持不变,采用DE/best/1变异策略,在其最优解附近进行局部搜索;最优解差的种群,变为搜索种群,重新初始化,采用DE/rand/1变异策略全局搜索,扩大搜索范围,寻找新的最优解.搜索过程中,跟踪种群和搜索种群各负其责,相互配合提高了算法的搜索效率.比较跟踪和搜索种群的最优解,好的最优解作为动态优化问题的解.最后,用Dynamic Function1(DF1)函数对算法进行了验证,实验结果表明该算法可行有效.
Modified differential evolution（MDE） is proposed for dynamic optimization problems. MDE divides the population into tracking group and searching group and judge whether there is a change in the enviroment by monitoring the extrema and hypo-extre- ma of the tracking group. When a change occurs, MDE revaluates the two groups, then finds their new extrema after change. The group with the better extremum will become the tracking group and keeps invariant, then adopts DE/best/1 to search locally around the better extremum. The other group will become the searching group and is re-initialized at random, then adopts DE/rand/1 to ex- tend the searching area and search globally. During the evolution, tracking and searching group take on different tasks and cooperate with each other. The searching efficiency of MDE is improved greatly. The extrema of the two groups are compared and the better one will be the solution of the dynamic optimization problem. The performance of MDE is tested using Dynamic Function 1. The re- sults show that the MDE is feasible and effective.
出处
《小型微型计算机系统》
CSCD
北大核心
2013年第12期2837-2840,共4页
Journal of Chinese Computer Systems
基金
国家″九七三″重点基础研究发展计划项目(973-61338)资助
关键词
差分进化算法
动态优化
多种群
differential evolution
dynamic optimization
multi-population
