基于鲁棒PCA聚类分析的径流周期特性研究  

Runoff periodicity analysis based on robust PCA and cluster

作　　者：邓红霞[1] 李存军[2] 朱兵[1] 丁晶[1] 

机构地区：[1]四川大学水电工程学院,四川成都610065 [2]四川大学建筑与环境学院,四川成都610065

出　　处：《武汉大学学报（工学版）》2006年第1期10-14,共5页Engineering Journal of Wuhan University

基　　金：国家自然科学基金项目(编号:50279023编号:40271024)

摘　　要：为了研究多年来河流天然径流的变化周期，引入鲁棒PCA（Principal Component Analysis）方法实现大量非线性流量数据的压缩和特征提取，得到了更具表征性的主成分，在此基础上提出了基于加权主成分距离和密度的径流聚类分析方法和径流周期特性分析的模糊模式识别算法，通过嘉陵江支流白龙江下游的三磊坝水文站1964～2002年周平均流量数据的实验和仿真，得出了该站径流具有7年的显著周期和相应的隶属度，从而证明了本文提出的基于鲁棒PCA聚类分析的径流周期特性研究方法具有较好的适用性。An effective analysis method of fuzzy pattern recognition algorithm based on cluster analysis by distance and density, with the more representative weighted principal component what is reduced and distilled from large numbers of nonlinear original data by robust PCA, is proposed to study periodicity of nature runoff in many years; accordingly weekly average runoff of Sanleiba Hydrology Station in lower Bailong River, a branch of Jialingjiang River, in the period of 1964-2002 year, is tested by this method; resulted in prominent 7-year period with remarkable subjective. Experimental results have shown that the method of runoff periodicity analysis based on robust PCA and cluster runs approving adaptability.

关 键 词：径流周期特性分析 鲁棒PCA 聚类分析 模式识别 

分 类 号：TV121.4[水利工程—水文学及水资源]