基于多归档协同进化的约束多目标优化算法研究

Research on Constrained Multi-Objective Optimization Algorithm Based on Multi-Archiving Collaborative Evolution

摘要: 近些年以来，约束多目标优化问题在真实的能源系统以及工程问题中广泛应用，但是这类问题在求解过程中可能遇到可行域狭窄不连续，收敛速度慢，解集分布不均衡等问题使得求解这类问题变得困难。为解决求解中所遇到的问题，本文提出一种增强型多归档协同进化算法Enhanced Multi-Archive Cooperative Evolutionary Algorithm (EMFEACD)，该算法利用收敛、多样性以及可行性归档，并引入技能因子进行跨归档知识迁移。最后通过自适应调整算子动态平衡搜索的探索与开发强度。并在MW与LIR-CMOP总计28个典型测试问题中的大多数问题测试结果上优于现有主流算法。

Abstract: In recent years, constrained multi-objective optimization problems have been widely applied in real-world energy systems and engineering challenges. However, solving these problems often encounters difficulties such as narrow and discontinuous feasible regions, slow convergence rates, and uneven solution distribution. To address these issues, this paper proposes an enhanced multi-archive cooperative evolutionary algorithm (EMFEACD). The algorithm utilizes convergence, diversity, and feasibility archiving, while introducing a skill factor to facilitate knowledge transfer across archives. Additionally, it dynamically adjusts the operator to balance exploration and development intensity. The proposed algorithm outperforms existing mainstream algorithms in most of the 28 typical test problems, including MW and LIR-CMOP.