双向随机多策略变异的自适应差分进化算法  

Adaptive differential evolution algorithm with bidirectional randomly multi-mutation strategy

作　　者：孔祥勇[1] 高立群[1] 欧阳海滨[1] 邵煜博 

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110004

出　　处：《计算机集成制造系统》2014年第8期1948-1958,共11页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(61273155)~~

摘　　要：针对差分进化算法中局部搜索和全局搜索之间的均衡难题,设计了一个基于符号函数的多策略变异算子,进而提出一种改进的自适应差分进化算法。新算法为提高跳出局部最优和搜索到全局最优解的可能性,用正负随机数代替了原有的变异率F,实现了两个方向上的随机搜索。同时为进一步简化参数选择过程,提高算法的寻优性能和通用性,新算法还设计了交叉率CR的两区间选择策略,在进化过程中通过学习以往的成功经验,实现自适应调整。对比实验结果表明,该算法具有更快的精确寻优和跳出局部最优的能力。Aiming at the balance problem between global search and local search in the adaptive differential evolution algorithm, a multi-strategy mutation operator based on symbolic function was designed and an improved Adaptive Differential Evolution algorithm with Bidirectional Randomly Multi-mutation （ADE-BRM） algorithm was proposed further. To increase the possibility of escaping from the local optimums and finding the global optimal solution, the original mutation rate F was replaced by a random number which leaded the population to search randomly in two directions. Two interval selection strategy of crossover rate CR was designed to further simplify the complex parameter selecting and improve the versatility of the algorithm. Comparison tests indicated that ADE-BRM algorithm had faster convergence, higher precision and stronger ability of jumping out of the local optimums.

关 键 词：差分进化算法 多策略变异 双向随机搜索 两区间选择策略 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]