基于近邻估计的年径流预测动态联系数回归模型  

Nearest neighbor estimate based dynamic connection number regression model for predicting annual runoff

作　　者：蒋尚明[1] 金菊良[2,3] 袁先江[1] 汤广民[1] 于凤存[1] 

机构地区：[1]安徽省水利部淮河水利委员会水利科学研究院安徽省水利水资源重点实验室,安徽蚌埠233000 [2]合肥工业大学土木与水利工程学院,安徽合肥230009 [3]合肥工业大学水资源与环境系统工程研究所,安徽合肥230009

出　　处：《水利水电技术》2013年第7期5-9,共5页Water Resources and Hydropower Engineering

基　　金：水利部公益性行业科研专项(200901077,200901026);国家自然科学基金项目(51079037,51209001)

摘　　要：集对分析理论为处理确定、不确定系统提供了新的途径,根据集对分析理论建立起来的预测联系数回归模型可以明显改善回归模型的预测精度。对于预测因子结构具有的动态性,文中将利用近邻估计,通过计算各个预测因子的变异系数,来判断预测因子在某次预测中处于强势或者弱势,进而动态地选择预报功能大的强势因子,消除对预报起负面作用的弱势因子的作用,这样很好地体现了预测因子结构中具有的动态性。基于此建立了基于近邻估计的年径流预测动态联系数回归模型(NNEDCNR)。结果说明:用NNE-DCNR去预测年径流量,预测精度比常用预测方法有显著提高,在水文水资源的预测中具有推广应用价值。The set pair analysis theory provides a new way to identify the uncertain system. The connection number regres- sion model for prediction established in accordance with the set pair analysis theory can significantly improve the prediction accuracy of the regression model. For the dynamics of the structure of the predictive factor, the nearest neighbor estimate is to be adopted for estimating that the predictive factors within a prediction are to be strong or weak through the calculation of the variation coefficients of all the predictive factors, moreover, the strong factors with large predicting function are dynami- cally selected to eliminate the negative effect from the weak factors on the prediction, thus the dynamics of the structure of the predictive factor is better reflected. On the basis of this, the nearest neighbor estimate based dynamic connection num- ber regression model for predicting annual runoff（NNE-DCNR） is established. The result shows that the predicting accuracy is to be significantly enhanced in comparison with the conventional predicting method, if NNE-DCNR is utilized for predic- ting the annual runoff, and then it has a high value of popularization and application in the prediction of hydrology and water resources.

关 键 词：年径流预测 近邻估计 回归模型 集对分析 联系数 变异系数 

分 类 号：P332[天文地球—水文科学]