融合聚类线性组合与优化状态自适应的差分进化算法

打开文本图片集

关键词：差分进化；聚类线性组合；状态自适应；参数自适应

中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2026）04-016-1098-14

doi：10.19734/j. issn.1001-3695.2025.08.0304

Differential evolution algorithm based on clustering linear combination and optimization state adaptation

Xiong Caiquanla，1b，LiHao1a，1b，Xia Dahai²+，Wu Xinyun1a，1b，Luo Maola，1b （1.a.Sholeliubelbelu HubeiUniversityofechnology，Wuhan43068，China；2.ShoolofComputerTechnology，ChangjangIstuteofTechnoogy，Wuan ，China）

Abstract：ToaddrestheissesoftheDEalgorithm，suchashighparametersensitivity，insuficientgobal explorationcapability，adimbalancebetweenexplorationandexploitationprocesssinhigh-dimensionalcomplexfunctionoptimization，this paperproposedanimprovedalgorithmnamedclustering linearcombinationandoptimizationstateadaptivediferentialevolution （CLOSADE），which integratedaclustering linear combinationapproach withanoptimization state adaptive mechanism.The research aimed to enhance the algorithm’s robustness and convergence performance when handling complex optimization problems.This methodfirstlydesigedaclustering strategybasedodualfactorsoffitnessanddistance togeneratemultipleclustersof inearcombinationvectorsandintroducedadynamicdistancethreshold toenhancepopulationdiversitySecondly，it constructedanindicatorofoptimizationstate （IOS）toquantifypopulationdistributioncharacteristics，driving theadaptiveadjustmentof mutationstrategiesandcontrolparameters.Experimentalresultsdemonstratethat，ontheCEC2Ol7and CEC2022 benchmark test functions，CLOSADE significantly outperformsadvanced algorithms suchas JSO，NL-SHADE-DP，and S-SHADE-DPin terms of bothconvergenceaccuracyandspeed.Particularlyonhigh-dimensionalhybridandcomposite functions，CLOSADE exhibits remarkable advantages，with an average improvement of 22% in convergence accuracy and approximately 40% inconvergence speed.Further population diversityanalysis reveals that the multi-subgroupstructure formed throughclusteringeffectivelymaintainsparalelsearchcapabilitiesithesolutionspace，whiletheoptimizationstateindicator ensuresadynamic balance between explorationand exploitation behaviors at diferent evolutionary stagesof thealgorithm.

Key Words：diffrential evolution（DE）；clustering linear combination；state adaptation；parameter adaptation

0引言

差分进化（DE）算法[是一种用于求解复杂优化问题的群体智能优化算法。（剩余21612字）