基于CPSO参数辨识的支持向量机增泄水量计算模型研究  

Support vector machine parameter identification method for calculating the increased water yield based on chaos particle swarm optimization

作　　者：张范平[1] 唐德善[1] 戴会超[1] 孙意翔[1] 

机构地区：[1]河海大学水利水电学院,南京210098

出　　处：《干旱区资源与环境》2014年第12期117-121,共5页Journal of Arid Land Resources and Environment

基　　金：国家重点基础研究发展计划(973)项目(2012CB417006);国家杰出青年科学基金(50925932)资助

摘　　要：水利工程实施的效果可以用增泄水量的多少来评价,文中构建了一种计算增泄水量的计算模型。以径向基函数（RBF）作为核函数,建立了以上游来水量、中游GDP增长、人口增长、降水量为输入,下游下泄水量为输出的支持向量机计算模型,为了提高支持向量机的预测精度,利用混沌粒子群算法（CPSO）的全局寻优特性进行支持向量机（SVM）的参数辨识,克服了人工选取的不足。以黑河流域（1990~2007年）18年的数据样本集作为训练样本,将后5年（2008~2012年）的数据样本集作为检验样本,选择参数如下：C=100,ε=0.001,σ=14。通过支持向量机模型计算的最大相对误差为8.01%,平均相对误差为6.50%。结果表明：文中建立的基于CPSO-SVM的增泄水量计算模型具有很好的效果,可以用于对水利工程实施效果的评价。We constructed a calculating model for calculating the increased water yield which can be used to evaluated the effect of water conservancy projects. The calculating model was established based on support vector machine（ SVM）,which considers the upstream water,the midstream GDP growth,the population growth and the precipitation as input indexes,and takes the discharged water of downstream as output index. In order to increase the calculating accuracy of SVM,the chaos particle swarm optimization（ CPSO） was used to optimize the SVM parameters,which can overcome the shortage for arbitrary selection of SVM parameters. Taking the data sets of 1990 ~ 2007（ 18 years） as training samples,the data sets of 2008 ~ 2012（ 5 years） as testing samples,the SVM parameters were selected as C = 100,ε = 0. 001,σ = 14. Through calculating the increased water yield by CPSO- SVM model,the maximum relative error is 8. 01% and the average relative error is 6. 50%. The result demonstrates that the calculating model of the increased water yield based on CPSO- SVM has a good effect and can be used to evaluate the effect of the water conservancy projects.

关 键 词：混沌粒子群算法 支持向量机 参数辨识 增泄水量 

分 类 号：TV214[水利工程—水文学及水资源]