求解众目标优化问题的演化算法研究

《求解众目标优化问题的演化算法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、万方数据ABSTRACTEvolutionaryalgorithmisaheuristicmethodwithgroupsearchstrategy,andisap—pliedinscienceandindustrysuccessfully．Inrecentyears，moreandmoreresearchersfocusedtheirattentiononmany-objectiveoptimizationproblems，andmoreandmorealgorithmsbasedonevolutionaryalgorithmshavebeenproposedtOsolveth

2、oseprob-lems．Wehaveproposedtwoalgorithmstodealwithmanyobjectiveproblemswithnoconstraintsandwithconstraintsrespectivelyinthispaper．1．Many—objectiveoptimizationproblemshavemanyobjectiveswhomaybecon·flictwitheachother,SOoptimizingallobjectivessimultaneouslyfacesmanydif-ficulties，forexample，wecan

3、’tcomparesolutions．Weproposedanewmany—objectiveevolutionaryalgorithm(MOEA)whosenameisParetopartialdomi-nanceontwoselectedobjectiVesmany—objectiveevolutionaryalgorithm(PPDSO-MOEA)．InPPDSO-MOEA，weproposeamethodofchoosingobjectives，andonlycalculatethedominancerelationshipbetweensolutionsonthetwo

4、selectedobjectivesandthenreflectthisworkonthechoosingofparentpopulation．Thetwoobjectivesusedincalculatingdominancerelationshipbetweensolutionsarechosenthroughcalculatingtheaverageobjectivedistancetothebestpoint局历优andweselectthetwoobjectiveswiththebiggestaveragedistances．Rp。in￡storesthebestval

5、ueandisupdatedwhennewsolutionscome．Weswitchthetwoob—jectivesineverysomeevaluationstooptimizeallobjectivesandkeepanarchivepopulationtostoretheParetooptimalsolutions．WeverifieditsperformanceonDTLZ2andDTLZ4．MOEA／D，SPEA2+SDE，MyODEMR，PPD—MOEA，andanalgorithmselectingobjectiveswithrandommethodaresel

6、ectedasrivalalgo—rithms．Theexperimentresultsshowthatouralgorithm，PPDSO—MOEA，outper-formsotheralgorithmsonmanycases．2．ManyproblemsintherealworldhaveconstraintsbutPPDSO-MOEAisdevel。opedtosolvethemany—objectiveproblemswithnoconstraints．So，weimprovetheabovealgorithmtogetanewalgorithm，namedPPDS02一

7、MOEA，whoisusedtoproblemswithconstraints．Therearemanydifferencesbetweenthosetwoalgo—rithms．Here，welistthedifferencesinPPDS02一MOEA．PPDS02-MOEAusesthegreedyrepairmethodtorepairinfeasiblesolutions．Solutionsinthearchivepopulationarefeasiblesolutio