基于粒子群最小二乘支持向量机的水文预测.doc

基于粒子群最小二乘支持向量机的水文预测
摘要:支持向量机理论为研究中长期水文预测提供了新的方法。针对最小二乘支持向量机模型参数选择费时且效果差这一问题，给出基于粒子群算法的最小二乘支持向量机水文预测模型(pso.lssvm)。该模型运用最小二乘支持向量机回归原理建立，参数选取采用具有全局搜索能力的粒子群算法进行寻优。用此模型对南桠河冶勒水电站月径流进行预测，仿真计算结果表明，该算法可提高预测效率与预测精度。
关键词:最小二乘支持向量机；粒子群算法；水文预测；参数优化；回归


leastsquaresupportvectormachinesmodelbasedon
particleswarmoptimizationforhydrologicalforecasting

liwen.li1,2*,liyu.xia1
1.schoolofwaterresourcesandhydro.electricengineering,xi’anuniversityoftechnology,xi’anshaanxi710048,china;
2.schoolofcomputerscience,shaanxinormaluniversity,xi’anshaanxi710062,china
abstract:
supportvectormachine(svm)algorithmprovidesanewwayforthestudyofmid-and-longtermhydrologicalforecastingthatneedsalearningoffinitesamples.avoidingwasteoftimeandunsatisfactoryinconventionalparameterchoosingmethod,aleastsquaresupportvectormachine(ls.svm)modelbasedonparticleswarmoptimizer(pso)isgiveninthispaper.themodelisbuiltbyusingtheregressionprincipleofleastsquaresupportvectormachine,thekeyparametersinthismodelareoptimizedbypsoalgorithmwithrandomseekingstrategy.monthlyrunoffforecastinginyelehydropowerstationonnanyariverindicatesthatthealgorithmisabletopromoteefficiencyandaccuracy.
supportvectormachine(svm)algorithmprovidesanewwayforthestudyofmid.and.longtermhydrologicalforecastingthatneedsalearningoffinitesamples.concerningthetime.consumptionandunsatisfactoryperformanceintheconventionalparameterchoosingmethod,aleastsquaresupportvectormachine(ls.svm)modelbasedonparticleswarmoptimization(pso)wasgiveninthispaper.themodelwasbuiltbyusingtheregressionprincipleofleastsquaresupportvectormachine,thekeyparametersinthismodelwereoptimizedbypsoalgorithmwithrandomseekingstrategy.monthlyrunoffforecastinginyelehydropowerstationonnanyariverindicatesthatthealgorithmisabletopromoteefficiencyandaccuracy.keywords:
leastsquare.supportvectormachines(ls.svm)；particleswarmoptimization(pso)；hydrologicalforecasting；parameteroptimization；regression

0引言
我国是水资源严重缺乏的国家，可靠的河川径流预测对水资源的合理利用具有重要价值。然而径流的形成过程涉及气候、流域自然地理以及人类活动等多种因素