外部种群完全反馈的元胞差分算法设计及应用

摘要

To improve the diversity and convergence of traditional evolution algorithms on solving multi-objective optimization problems,a cellular differential algorithm was proposed based on complete feedback from the external population.The traditional cellular differential algorithm was improved,and the external population was truncated according to the rank and the k-nearest neighbor distance after each generation.Then it was assigned to two-dimensional grids randomly,and a new disturbance was introduced in original mutation to prevent the algorithm from trapping in local optimum.Through testing 6 benchmark functions,the result indicated that the new algorithm was superior to the other three typical algorithms concerning the coverage of Pareto fronts and that the new mutation operation could improve the algorithm ability to escape from local optimum.The feasibility and effectiveness of the algorithm was verified by an engineering example.%针对传统进化算法在求解多目标优化问题时存在多样性和收敛性不佳的问题,提出一种外部种群完全反馈的元胞差分算法.对标准元胞差分算法进行改进,在每一代进化之后,根据秩与k最近邻距离对外部种群进行修剪,并将修剪后的整个外部种群随机分配到二维网状结构,在原有变异操作中引入新的扰动来避免算法陷入局部最优.通过对6个基准函数进行测试表明,新算法相对于其他3种典型算法具有更好的前端覆盖性,新的变异方式能提高算法跳出局部最优解的能力.通过工程实例验证了所提算法的可行性与有效性.