Collaborative Decomposition Multi-Objective Improved Elephant Clan Optimization Based on Penalty-Based and Normal Boundary Intersection  

作　　者：Mengjiao Wei Wenyu Liu 

机构地区：[1]School of Electrical Engineering,Northeast Electric Power University,Jilin,132012,China [2]Northeast Electric Power Design Institute,Changchun,130000,China

出　　处：《Computers, Materials & Continua》2025年第5期2505-2523,共19页计算机、材料和连续体(英文)

摘　　要：In recent years,decomposition-based evolutionary algorithms have become popular algorithms for solving multi-objective problems in real-life scenarios.In these algorithms,the reference vectors of the Penalty-Based boundary intersection(PBI)are distributed parallelly while those based on the normal boundary intersection(NBI)are distributed radially in a conical shape in the objective space.To improve the problem-solving effectiveness of multi-objective optimization algorithms in engineering applications,this paper addresses the improvement of the Collaborative Decomposition(CoD)method,a multi-objective decomposition technique that integrates PBI and NBI,and combines it with the Elephant Clan Optimization Algorithm,introducing the Collaborative Decomposition Multi-objective Improved Elephant Clan Optimization Algorithm(CoDMOIECO).Specifically,a novel subpopulation construction method with adaptive changes following the number of iterations and a novel individual merit ranking based onNBI and angle are proposed.,enabling the creation of subpopulations closely linked to weight vectors and the identification of diverse individuals within them.Additionally,new update strategies for the clan leader,male elephants,and juvenile elephants are introduced to boost individual exploitation capabilities and further enhance the algorithm’s convergence.Finally,a new CoD-based environmental selection method is proposed,introducing adaptive dynamically adjusted angle coefficients and individual angles on corresponding weight vectors,significantly improving both the convergence and distribution of the algorithm.Experimental comparisons on the ZDT,DTLZ,and WFG function sets with four benchmark multi-objective algorithms—MOEA/D,CAMOEA,VaEA,and MOEA/D-UR—demonstrate that CoDMOIECO achieves superior performance in both convergence and distribution.

关 键 词：Multi-objective optimization elephant clan optimization algorithm collaborative decomposition new individual selection mechanism diversity preservation 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]