多极小值粒子群优化算法.docx

多极小值粒子群优化算法
多极小值粒子群优化算法
Abstract
Particleswarmoptimization(PSO)isawell-knownoptimizationalgorithm.However,itmaysufferfromprematureconvergencewhendealingwithmulti-modalormulti-peakfunctionoptimizationproblems.Toovercomethisproblem,anovelPSOalgorithmcalledmulti-localminimaparticleswarmoptimization(MLMPSO)algorithmisproposed.Inthisalgorithm,twostrategies,namedvelocitydampingandparticlebehaviordiversity,areintroducedtoenhancetheabilityoflocalsearch.Inaddition,adisturbanceoperatorisusedtomovethesearchtrajectoriesfromonelocalminimumtoanother.Theperformanceoftheproposedapproachisvalidatedbyusingasetofbenchmarkfunctions,andtheresultsshowthattheMLMPSOalgorithmoutperformstraditionalPSOalgorithms.
Introduction
Particleswarmoptimization(PSO)isawell-establishedoptimizationalgorithminspiredbythesocialbehaviorofbirdsflockingorfishschooling.InPSO,apopulationofparticlesisusedtosearchtheoptimalsolutionbyupdatingtheirpositionsandvelocitiesaccordingtoasetofrules.PSOhasbeenwidelyusedinvariousfields,suchasengineering,finance,andcomputerscience.
However,theclassicalPSOalgorithmhassomelimitations,particularlywhendealingwithcomplexoptimizationproblems.OneofthemajorchallengesforPSOistohandlemulti-modalormulti-peakfunctionoptimizationproblems,whichiscommoninreal-worldapplications.Insuchproblems,thesearchspaceusuallycontainsmanylocalminima,andtraditionalPSOalgorithmsmayfallintolocalminimaprematurely,resultinginpoorsearchperformanceandprematureconvergence.
Toenhancethelocalsearchcapabilityandavoidprematureconvergence,variousmodificationshavebeenproposed.Thesemodificationscanbeclassifiedintotwocategories:diversity-enhancingstrategiesandlocalsearchstrategies.Diversity-enhancingstrategiesaimtomaintaindiversityamongtheparticlesandexplorethesearchspaceefficiently.Localsearchstrategiesaimtoimprovetheabilityoftheparticlestoexploitlocalinformation.
Inthispaper,weproposeanovelPSOalgorithm,calledmulti-localminimaparticleswarmoptimization(MLMPSO)algorithm,whichusesbothdiversity-enhancingandlocalsearchstra