电子学报 ›› 2021, Vol. 49 ›› Issue (3): 578-585.DOI: 10.12263/DZXB.20200228

• 学术论文 • 上一篇    下一篇

基于多邻居结构的自适应元胞差分算法

王亚良, 倪晨迪, 金寿松   

  1. 浙江工业大学机械工程学院, 浙江杭州 310023
  • 收稿日期:2020-03-02 修回日期:2020-09-01 出版日期:2021-03-25 发布日期:2021-03-25
  • 通讯作者: 金寿松
  • 作者简介:王亚良 男,1977年生于浙江绍兴.现为浙江工业大学机械工程学院高级实验师.主要研究方向为生产系统优化和智能制造.E-mail:wangyaliang@zjut.edu.cn;倪晨迪 男,1995年生于浙江绍兴.现为浙江工业大学机械工程学院硕士研究生.主要研究方向为智能制造和优化算法.E-mail:2060662618@qq.com
  • 基金资助:
    国家重点研发计划(No.2018YFB1308100);浙江省自然科学基金资助项目(No.LY16G010013);国家863计划资助项目(No.2015AA0430020)

Adaptive Cellular Differential Evolutionary Algorithm Based on Multi-neighborhood Structure

WANG Ya-liang, NI Chen-di, JIN Shou-song   

  1. College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou, Zhejiang 310023, China
  • Received:2020-03-02 Revised:2020-09-01 Online:2021-03-25 Published:2021-03-25

摘要: 针对传统多目标优化算法在求解Pareto解集时存在全局搜索能力与局部寻优能力无法得到有效平衡的问题,提出了一种基于多邻居结构的自适应元胞差分算法.该算法在保留传统元胞差分算法进化特点的基础上,使用更加丰富的多邻居结构替换原有的单一邻居结构,并且依据相应元胞个体的性能优劣来对其邻居结构进行选择分配.同时,面对进化过程中的复杂性能需求,算法定义了一种周期性变化的变异策略来实现不同进化阶段的自适应调节.最后,利用DTLZ系列测试函数对算法性能进行测试,并通过与四种经典的多目标优化算法相比较,证明了改进后的算法拥有更好的收敛性与分布性.

 

关键词: 多目标优化, 多邻居结构, 随机扰动, 自适应变异, 元胞自动机, 进化策略

Abstract: To solve the problem that the global search ability and local search ability of traditional multi-objective evolutionary algorithm cannot be effectively balanced when solving the Pareto solution set,an adaptive cellular differential evolutionary algorithm based on multi-neighborhood structure is proposed.Based on the characteristics of the traditional cellular differential evolutionary algorithm,the improved algorithm uses a richer multi-neighbor structure to replace the original single neighbor structure,and the neighbor structure is adjusted reasonably according to the performance of the corresponding individual.At the same time,in the face of the complex requirements in the whole evolution process,the algorithm defines a mutation strategy with periodic variation to realize the adaptive adjustment in different evolution stages.Finally,the DTLZ series of test functions are used to test the performance of the algorithm.Compared with four classical multi-objective optimization algorithms,it is proved that the improved algorithm has better convergence performance and diversity of solution set.

Key words: multi-objective optimization, multi-neighborhood structure, random disturbance, adaptive mutation, cellular automata, evolutionary strategy

中图分类号: