資源描述:
《基于自適應(yīng)進(jìn)化模型的粒子群優(yōu)化算法.pdf》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫(kù)。
1、2014年8月計(jì)算機(jī)工程與設(shè)計(jì)Aug.2014第35卷第8期COMPUTERENGINEERINGANDDESIGNVoL35No.8基于自適應(yīng)進(jìn)化模型的粒子群優(yōu)化算法王雪瑞,宋全有。(1.河南工程學(xué)院計(jì)算機(jī)學(xué)院,河南新鄭451191;2.河南交通職業(yè)技術(shù)學(xué)院交通信息工程系,河南鄭州450052)摘要:針對(duì)標(biāo)準(zhǔn)粒子群算法在處理復(fù)雜優(yōu)化問(wèn)題時(shí)易出現(xiàn)收斂速度慢和陷入局部最優(yōu)的問(wèn)題,提出了一種自適應(yīng)進(jìn)化模型的粒子群優(yōu)化算法。通過(guò)設(shè)定的閾值limit將種群進(jìn)化狀態(tài)劃分為正常狀態(tài)和“早熟”狀態(tài),當(dāng)種群全局最優(yōu)位置信息連續(xù)超過(guò)limit次沒(méi)有更新時(shí),認(rèn)為算法處于“早熟”狀態(tài),此時(shí)對(duì)種群的個(gè)體最優(yōu)位置
2、進(jìn)行反向?qū)W習(xí),幫助算法逃離局部最優(yōu),并采用新的進(jìn)化模型;否則視為正常進(jìn)化狀態(tài),并采用標(biāo)準(zhǔn)粒子群進(jìn)化模型。8個(gè)基準(zhǔn)測(cè)試函數(shù)的仿真結(jié)果表明,該算法與一些其它改進(jìn)粒子群算法如FIPS、CLPSO、MPSO-SFLA算法相比,在全局尋優(yōu)能力、收斂速度和收斂精度方面都具有明顯的優(yōu)勢(shì)。關(guān)鍵詞:粒子群優(yōu)化算法;自適應(yīng)進(jìn)化;反向?qū)W習(xí);快速收斂;局部最優(yōu)中圖法分類號(hào):TP301.6文獻(xiàn)標(biāo)識(shí)號(hào):A文章編號(hào):1000-7024(2014)08—2901-06ParticleswarmoptimizationbasedonadaptiveevolutionmodelWANGXue—rui。SONGQuan-yo
3、u。(1.CollegeofComputerScienceandTechnology,HenanInstituteofEngineering,Xinzheng451191,China;2.DepartmentofInformationandCommunicationEngineering,HenanVocationalandTechnicalCollegeofCommunications,Zhengzhou450052,China)Abstract:Asstandardparticleswarmoptimizationalgorithmhadsomeshortcomings,suchas
4、convergingslowlyandgettingtrappedinthelocalminima,3newimprovedPSOalgorithmbasedonadaptiveevolutionwasproposed.Thealgorithm’Spopula—tionevolutionstatewasdividedinto“normalandpremature”bysettingthethresholdvaluelimit.Whenpopulation’Sglobalopti—malpositionwasnotupdatedcontinuouslyformorethanlimittim
5、es.thealgorithmwasconsideredtObeinthe“premature’’state.Atthispoint,theindividuals’optimalpositionadoptedopposition-basedlearningstrategywhichhelpedthealgorithmtoescapefromlocaloptimum,andthealgorithmadoptedtheimprovedevolutionmode1.Otherwise,itwasconsideredtobein“nor—mal”evolutionstate,andthestan
6、dardmodelwasadopted.Thesimulationexperimentresultsoneightclassicalbenchmarkfunc—tionsshowedthatthecapacityofsearchingoptimalsolution,convergencespeedandconvergenceaccuracyoftheAEMPSOwasbetterthansomeofthecommonimprovedparticleswarmoptimizations,suchasFIPS,CLPSOandMPSO-SFLA.Keywords:particleswarmo
7、ptimization;adaptiveevolution;opposition-basedlearning;fastconvergence;localoptima題時(shí),在進(jìn)化后期存在種群多樣性喪失嚴(yán)重和易陷入局部0引言最優(yōu)等缺點(diǎn),嚴(yán)重制約了它的發(fā)展。后期眾多學(xué)者對(duì)其進(jìn)粒子群優(yōu)化算法是一種基于種群搜索的群智能優(yōu)化算行了相關(guān)改進(jìn)。周龍甫等人將粒子群搜索分成探索性和開(kāi)法,它是模擬自然界鳥(niǎo)類群體覓食行為的一種隨機(jī)搜索算發(fā)性搜索兩種狀態(tài),提