资源描述:
《求解非线性规划问题的遗传算法设计实现分析范文》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、WORD格式可编辑摘要非线性规划在工程、管理、经济、科研、军事等方面都有广泛的应用。传统的解决非线性规划问题的方法,如梯度法、罚函数法、拉格朗日乘子法等,稳定性差,对函数初值和函数性态要求较高,且容易陷入局部最优解。遗传算法是模拟达尔文的遗传选择和自然淘汰的生物进化过程的计算模型。遗传算法是一种全局搜索算法,简单、通用、鲁棒性强,对目标函数既不要求连续,也不要求可导,适用于并行分布处理,应用范围广。本文在分析传统的非线性规划算法的不足和遗传算法的优越性的基础上,将遗传算法应用于非线性规划。算法引进惩罚函数的概念,构造带有惩罚项的适应度函数;通过实数编码,转轮法选择,双点交叉,均匀变异,形成了求
2、解非线性规划问题的遗传算法。与传统的非线性规划算法——外点罚函数法的比较结果表明该算法在一定程度上有效地克服了传统的非线性规划算法稳定性差,对函数初值和函数性态要求较高,且容易陷入局部最优解的缺陷,收敛更合理,性能更稳定。关键词:非线性规划;遗传算法;罚函数法专业知识分享WORD格式可编辑ABSTRACTNon-linearprogramminghasawiderangeofapplicationsinengineering,management,economic,scientific,andmilitaryaspects.Traditionalmethodstosolvethenon-lin
3、earprogrammingproblem,suchasthegradientmethod,penaltymethod,Lagrangemultipliermethod,havepoorstability.Theyaresensitivetothefunctioninitialvalueandrequesttheobjectivefunctiontobecontinuousanddifferential.Theresultsarealsoeasilytrappedintolocaloptimalsolution.Geneticalgorithmisakindofcalculatemodelwh
4、ichsimulatesDarwin'sgeneticselectionandbiologicalevolutionofnaturalselection.Geneticalgorithmisaglobalsearchalgorithm.Ithassimple,universal,robustfeatures,anddoesnotrequesttheobjectivefunctiontobecontinuousanddifferential,andissuitableinparalleldistributionprocessing.Geneticalgorithmiswidelyappliedi
5、nmanyareas.Basedontheanalysisofthedisadvantageoftraditionalnon-linearprogrammingalgorithmandtheadvantageofgeneticalgorithm,geneticalgorithmisappliedtonon-linearprogramminginthispaper.Theintroductionoftheconceptofpenaltyfunctionisusedtoconstructthefitnessfunctionwithpunishment.Byusingreal-coded,Roule
6、tteWheelselectionmethod,two-pointcrossover,uniformmutation,weformedageneticalgorithmtosolvethenon-linearprogrammingproblem.Comparedwiththemostclassicalandwidelyusedtraditionalnon-linearprogrammingproblemalgorithm–SUMTalgorithm,theresultsshowthatthenewalgorithmcouldeffectivelyovercomethedefectofthetr
7、aditionalalgorithminacertainextent.Thenewalgorithmismorestable,lesssensitivetothefunctioninitialvalueandconditions,andalwayscouldreceivetheoptimalsolutionorapproximateoptimalsolution.Itsconvergenceres