基于镜像判断和改进父代选择的多目标进化算法  

Multi-objective evolutionary algorithm based on image judgment and improved parent selection

作　　者：王嘉诚 邹雨恒 王珊珊 曾亮 WANG Jia-cheng;ZOU Yu-heng;WANG Shan-shan;ZENG Liang(School of Electrical and Electronic Engineering,Hubei Key Laboratory for High-Efficiency Utilization of Solar Energy and Operation Control of Energy Storage System,Hubei University of Technology,Wuhan 430068,China;School of Industrial design,Hubei University of Technology,Wuhan 430068,China)

机构地区：[1]湖北工业大学电气与电子工程学院太阳能高效利用及储能运行控制湖北省重点实验室,湖北武汉430068 [2]湖北工业大学工业设计学院,湖北武汉430068

出　　处：《陕西科技大学学报》2025年第2期215-225,234,共12页Journal of Shaanxi University of Science & Technology

基　　金：湖北省重点研发计划项目(2023BAB094)。

摘　　要：高维多目标进化算法在解决复杂帕累托前沿问题时,常面临收敛性和多样性难以平衡的问题.为解决这一问题,提出了一种基于镜像判断和改进父代选择的高维多目标进化算法.该算法首次结合成就标量函数和全局密度并应用在交配池中,使其在迭代过程中不仅关注当前最优解,还兼顾解在整个空间的分布情况,从而实现了收敛性和多样性的统一.此外,针对算法在迭代过程中可能出现镜像的问题,本文提出了解决方案.具体来说,算法首先采用非支配排序,将临界层个体与参考向量相关联,随后判断其是否满足镜像对称准则,若满足则通过全局密度选取个体,达成“内紧外松”的目的,最大限度保证候选解的分布性,从而有效解决了选择压力不均的问题.最后将本文算法与最新的五种多目标算法在4种不同维度的测试问题上进行对比实验,并应用在两个实际案例中.实验结果表明:所提算法不仅能高效解决高维多目标优化问题,且能有效平衡收敛性和多样性.When solving complex Pareto frontier problems,high-dimensional multi-objective evolutionary algorithms often face the difficulty of balancing convergence and diversity.In this paper,a multi-objective evolutionary algorithm based on image judgment and improved parent selection is proposed to solve this problem.This algorithm combines the achievement scalar function and global density for the first time and applies them to the mating pool,so that the algorithm not only pays attention to the current optimal solution in the iterative process,but also considers the distribution of the solution in the whole space,achieving the unity of convergence and diversity.Additionally,it was found that the algorithm may exhibit symmetrical phenomena during the iterative process.This paper identifies the drawbacks of this issue and proposes a solution.Specifically,The algorithm first employs non-dominated sorting to associate the critical layer individuals with reference vectors.It then evaluates whether they meet the symmetry criterion.If they do,individuals are selected based on global density to achieve the goal of"tight inside,loose outside",maximizing the distribution of candidate solutions.This effectively resolves the issue of uneven selection pressure.Finally,the proposed algorithm is compared with the latest five multi-objective algorithms on four different dimension test problems,and is applied in two practical cases.The experimental results show that the proposed algorithm can efficiently solve the high-dimensional multi-objective optimization problem and has good stability.

关 键 词：多目标进化算法 交配选择 聚合距离 收敛性 分布性 

分 类 号：TM614[电气工程—电力系统及自动化] TP18[自动化与计算机技术—控制理论与控制工程]