多目标协调进化算法研究

《多目标协调进化算法研究》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、维普资讯http://www.cqvip.com第24卷第9期计算机学报Vo1．24No．92001年9月CHINESEJCoMPUTERSSept．2001多目标协调进化算法研究崔逊学”李淼方廷健(中国科学技术大学自动化系台肥230627)z(中国科学院台肥智能机械研究所合肥230031)擒要进化算法适合解决多目标优化问题，但难以产生高维优化问题的最优解．文中针对此间题提出了一种求解高维多目标优化问题的新进化方法，即多目标协调进化算法，主要特点是进化群体按协调模型使用偏好信息进行偏好排序，而不是基于Pareto优于

2、关系进行个体排序．实验结果表明，所提出的算法是可行而有效的，且能在有限进化代数内收敛．关镀词进化算法，多目标优化，高维，协调模型中圈法分类号：TP18AMulti—ObjectiveConcordanceEvolutionaryAlgorithmCUIXunXueIAMiaoFANGTingJian。(DepartmentofAutomaO'~，UniversityScience＆TechnotogyofChina，Hefei230027(HefeiInstituteofInteltigentsl4achines，C

3、hineseAcadenn,ofScie~es．Hefei2300311AbstractEvolutionaryAlgorithms(EAS)arewelIsuitedforoptimizationproblemswithmulti—pieobjectives．Duetotheirinherentparallelismtheyareabletocaptureanumberofsolutionsconcurrentlyinasinglerun．AlthoughhAsbasedonParetorankingareinde

4、pendentofthecon—vexityorconcavityofatradeoffsurfaceanddonotrequireanypreferenceinformation，thedimensionalityofthesearchspaceinfluencestheirperformance．PureParetoEAscannotbeexpectedtOperformwellonproblemsinvolvingmanycompetingobjectivesandmaysimplyfailtOproduces

5、atisfactorysolutionsduetothelargedimensionalityandthesizeofthetradeoffsurface．InthispaperweproposeanewevolutionaryapproachformultiohjectiveoptimizationtOdealwiththisproblem，namelyMuhiObjectiveConcordanceEvolutionaryAlgorithm(MOCEA)．Itskevcharacterliesonthatevol

6、utionarypopulationisratherpreferencerankedbypreferenceinf0rmationbasedonconcordancemodelthannormallyrankedbyParetosuperiorrelationship．OptimizationprocessacceptsincomparabilityandUSeSanoutrankingrelationtOmodelthepreferencesofthedecision—maker，whichbelongstothe

7、familyofoutrankingapproaches(suchasELECTRE)1n—troducedbyBernardRoy．ThisapproachincludestWOphases：theconstructi。n。fan。utrankinrelationonthedifferentobjectives，andtheexploitationofthisrelationin。rdert。ziveananswertOthemuhiobjectiveoptimizationproblem．Evolutionary

8、processgeneratescumuIativePareto。p—timalsolutionsfromwhichELECTREselectsacertainsubsetbased。nthepreferencesofthedecision—maker·Thepreferencesareexpressedthroughpreferencerel