基于微粒群算法的多目标优化问题研究

《基于微粒群算法的多目标优化问题研究》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、中南民族大学硕士学位论文基于微粒群算法的多目标优化问题研究姓名：杨勋申请学位级别：硕士专业：计算机应用技术指导教师：王江晴20080528基于微粒群算法的多目标优化问题研究策方案；在实验和理论分析的基础上我们给出了Pareto-ε动态策略的评价。4、基于PSO算法和Pareto-ε优胜关系提出了一种新的PεPSO算法框架；基于面向对象理论提出了一种相应的数据结构，提高了算法实现的通用性、复用性以及兼容性；选用了经典测试函数集中的部分ZDT系列函数进行了测试，实验显示PεPSO算法是有效的。5、在采用动

2、态调整ε的策略后，通过动态调整ε的值，使算法开始时快速向Pareto真实前沿逼近，最终让ε在算法运行过程中逐步回归为0，从而更好的逼近真实Pareto前沿，不受ε的影响。既可以提高算法的搜索和收敛速度，又可以消除ε值对最终解的质量的影响。关键词：计算智能；进化算法；微粒群算法；多目标优化；Pareto-εII中南民族大学硕士学位论文AbstractItisanindispensablecapabilityforthemoderndecisionandassistancesystemtoprovided

3、ecision-makerswithscientific,properandtimelydicisionschemes.Sincemostoftheactualdecision-targetsaremulti-objective,theresearchonMulti-objectiveOptimizationProblem(MOP)hasgainedmoreandmoreattention.Forthefactthatthesub-objectivesinMOParecontradictorytoea

4、chotherandtheyhavenounifiedmeasurestandards,ithasbecomeaprincipalconcerntodefinitetheoptimalsolutionofMOPwhensolutingMOP.Thedefinitionofnon-dominancedsolution,basedontheideaofchoosingPareto,isbecomingincreasinglypeople’sgeneralconsensus.TheoptimalPareto

5、solutionofMOP,however,isnotanexclusivesolution.Sometimesthereareevennumerousones.Inthiscase,theobtainedsolutionsdonotfacilitatebuttroublethedicision-makers.Furthermore,ittakesalongtimetoobtaintheoptimalsolutions.Soitisratherimportanttoprovideproperandfe

6、asiblesolutionschemesrapidlyforthedecisionmakers.ParticleSwarmOptimization(PSO)isakindofswarmintelligentalgorithmwhichhasbeendevelopedinrecentyears.Itisanewintelligentsearchalgorithm.Thealgorithmutilizestheeffectiveinformation,whichissharedbyeverypartic

7、leinthepopulationfromitspastexperiencesandotherparticles’experiences,tosearchtheoptimalsolutionssynergically.Atpresent,thealgorithmresearchbasedonPSOhasbeenattachedmoreandmoreimportanceinthefieldofmulti-objectiveoptimizationanditisevenahottopicintherese

8、arch.TherelevantPareto-εconceptsareproposedinthepapergroundedonthedefinitionofParetooptimalsolutions.Bymeansofanalysisandexperiments,itcanbeprovedthattheprocessofoptimizationinsolutiingMOPisimprovedfortheusageofPareto-εconcepts,t