資源描述：
《畢業(yè)設(shè)計（論文）-求解非線性規(guī)劃問題的遺傳算法設(shè)計與實現(xiàn)》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在學術(shù)論文-天天文庫。
1、摘要非線性規(guī)劃在工程、管理、經(jīng)濟、科研、軍事等方面都有廣泛的應用。傳統(tǒng)的解決非線性規(guī)劃問題的方法，如梯度法、罰函數(shù)法、拉格朗日乘子法等，穩(wěn)定性差，對函數(shù)初值和函數(shù)性態(tài)要求較高，且容易陷入局部最優(yōu)解。遺傳算法是模擬達爾文的遺傳選擇和自然淘汰的生物進化過程的計算模型。遺傳算法是一種全局搜索算法，簡單、通用、魯棒性強，對目標函數(shù)既不要求連續(xù)，也不要求可導，適用于并行分布處理，應用范圍廣。本文在分析傳統(tǒng)的非線性規(guī)劃算法的不足和遺傳算法的優(yōu)越性的基礎(chǔ)上，將遺傳算法應用于非線性規(guī)劃。算法引進懲罰函數(shù)的概念，構(gòu)造帶有懲罰項的適應度函數(shù)；通過
2、實數(shù)編碼，轉(zhuǎn)輪法選擇，雙點交叉，均勻變異，形成了求解非線性規(guī)劃問題的遺傳算法。與傳統(tǒng)的非線性規(guī)劃算法——外點罰函數(shù)法的比較結(jié)果表明該算法在一定程度上有效地克服了傳統(tǒng)的非線性規(guī)劃算法穩(wěn)定性差，對函數(shù)初值和函數(shù)性態(tài)要求較高，且容易陷入局部最優(yōu)解的缺陷，收斂更合理，性能更穩(wěn)定。關(guān)鍵詞：非線性規(guī)劃；遺傳算法；罰函數(shù)法ABSTRACTNon-linearprogramminghasawiderangeofapplicationsinengineering,management,economic,scientific,andmilitar
3、yaspects.Traditionalmethodstosolvethenon-linearprogrammingproblem,suchasthegradientmethod,penaltymethod,Lagrangemultipliermethod,havepoorstability.Theyaresensitivetothefunctioninitialvalueandrequesttheobjectivefunctiontobecontinuousanddifferential.Theresultsarealsoe
4、asilytrappedintolocaloptimalsolution.GeneticalgorithmisakindofcalculatemodelwhichsimulatesDarwin'sgeneticselectionandbiologicalevolutionofnaturalselection.Geneticalgorithmisaglobalsearchalgorithm.Ithassimple,universal,robustfeatures,anddoesnotrequesttheobjectivefunc
5、tiontobecontinuousanddifferential,andissuitableinparalleldistributionprocessing.Geneticalgorithmiswidelyappliedinmanyareas.Basedontheanalysisofthedisadvantageoftraditionalnon-linearprogrammingalgorithmandtheadvantageofgeneticalgorithm,geneticalgorithmisappliedtonon-
6、linearprogramminginthispaper.Theintroductionoftheconceptofpenaltyfunctionisusedtoconstructthefitnessfunctionwithpunishment.Byusingreal-coded,RouletteWheelselectionmethod,two-pointcrossover,uniformmutation,weformedageneticalgorithmtosolvethenon-linearprogrammingprobl
7、em.Comparedwiththemostclassicalandwidelyusedtraditionalnon-linearprogrammingproblemalgorithm–SUMTalgorithm,theresultsshowthatthenewalgorithmcouldeffectivelyovercomethedefectofthetraditionalalgorithminacertainextent.Thenewalgorithmismorestable,lesssensitivetothefunct
8、ioninitialvalueandconditions,andalwayscouldreceivetheoptimalsolutionorapproximateoptimalsolution.Itsconvergenceresultsaremorereasonable,th
