年径流预测的最小二乘支持向量机-马尔可夫链组合模型  

Annual runoff prediction based on least square support vector machines-Markov chain combined model

作　　者：周秀平[1] 李天翔 王文圣[3] 

机构地区：[1]中国水电顾问集团成都勘测设计研究院,成都610072 [2]四川省清源工程咨询有限公司,成都610072 [3]四川大学水利水电工程学院,成都610065

出　　处：《水力发电学报》2013年第4期16-19,共4页Journal of Hydroelectric Engineering

基　　金：国家自然科学基金项目(51179110;50739002);科技基础性工作专项项目(2011IM011000)

摘　　要：基于最小二乘支持向量机(LLSVM)与马尔可夫链(MC)提出了一种新的年径流预测模型——最小二乘支持向量机-马尔可夫链组合模型(LLSVM-MC)。首先利用LLSVM模型对年径流进行第一步预测;其次将预测误差看作是马尔可夫过程,用MC对预测误差进行估计,作为第二步预测;最后,将两步预测结果叠加得到年径流预测值。实例预测表明,建议模型有效地提高了预测精度。A new annual runoff prediction model, LSSVM-MC, was proposed by combining least squares support vector machines with markov chains. This model uses LSSYM to predict annual runoff as the first step, then its prediction errors are taken as a MC process and estimated by MC method for the second step prediction. The final prediction values are calculated by adding up the two predictions. A case study shows that the combined model can improves prediction accuracy.

关 键 词：水文学 最小二乘支持向量机 马尔可夫链 组合模型 年径流预测 

分 类 号：TV121[水利工程—水文学及水资源]