基于双种群的多目标差分进化算法

针对约束多目标优化问题,为了提高约束多目标优化算法的收敛性和分布均匀性,本文提出一种基于双种群的多目标差分进化算法。该算法引进一种约束处理策略对非可行解进行处理的同时,也对可行解的目标函数值进行归一化处理,从而将种群所有个体目标函数向量进行修正,且新算法采用的双种群结构、基于部分计算拥挤距离的种群同步更新方式,对不同子种群采用不同变异策略,并运用变动的高斯分布对参数进行调整,最后采用一种类算术交叉操作加强扰动。为了验证算法的性能,本文选取了四个经典的约束多目标测试函数进行仿真实验,从收敛性与分布性上分析所提算法的有效性。

In order to solve the constrained multi-objective optimization problem and improve the convergence and distribution of the constrained multi-objective optimization algorithm,a multi-objective differential evolution algorithm based on dual population is proposed.The algorithm introduces a constraint strategy to deal with the non-feasible solution,and also normalizes the objective function value of the feasible solution to form a modified objective function value.Based on dual populations,the new algorithm,which implements synchronous update by partly calculating the crowded distance,adopt corresponding mutation strategies for different populations and controlling parameters by Gauss distribution.A variation of arithmetic crossover is incorporated to reinforce the effect of mutation.In order to verify the performance of the algorithm,we select four classical constrained multi-objective test functions for simulation experiments,and analyze the effectiveness of the proposed algorithm from the convergence and distribution.