基于云差分进化算法的约束多目标优化实现

针对现有约束多目标算法存在收敛性、分布性不高等问题,提出一种基于云差分进化算法的约束多目标优化方法,通过云模型对差分进化算法的参数进行自适应处理;采用建立外部种群分别存储可行解和不可行解的方式处理约束条件,并对已有可行解集的更新方法进行改进,有效提高解集的分布性.提出新的变异策略,利用优秀可行解和不可行解的方向信息增强算法对解的探索能力.通过对CTP类标准问题的求解表明,与另外2种较为优秀的约束多目标算法相比,本算法显著提高了Pareto解集的分布性,且更接近于真实的Pareto前沿,有效地解决了约束多目标问题.

A cloud differential evolutionary algorithm for constrained multi-objective optimization

Considering that the diversity and convergence of a constrained multi-objective optimization algorithm is not sufficient,a constrained multi-objective optimization algorithm based on a cloud differential evolutionary algorithm was proposed in this paper.First,parameter CR was adjusted adaptively based on a cloud model.Secondly,constraint conditions were considered by utilizing external population to store feasible and unfeasible solutions,and the updating method of the feasible solution set was improved to increase distribution.Finally,a novel mutation strategy was proposed which enhanced the exploration ability of the algorithm by utilizing direction information of excellent feasible and unfeasible solutions.The simulation results of CTP standard test functions show that the constrained multi-objective optimization algorithm in this paper outperforms the other two state-of-the-art constraint multi-objective evolutionary algorithms in terms of diversity metrics and is closer to the true front of Pareto,which proves the superiority of this algorithm in solving constraint multi-objective optimal problems.