洪水聚类有效性分析  

Cluster Validity Analysis for Flood Clustering

作　　者：卢正波[1] 侯召成[2] 

机构地区：[1]青岛市水利局 [2]南水北调中线干线工程建设管理局,北京100053

出　　处：《南水北调与水利科技》2007年第2期87-90,共4页South-to-North Water Transfers and Water Science & Technology

摘　　要：在聚类分析中,聚类数是一个非常重要的参数,最佳聚类数的确定问题是聚类分析研究的热点之一。在模糊聚类迭代模型的基础上,首先提出了基于类间相关系数的聚类有效性指标确定最佳聚类数,并给出了最佳聚类数的确定步骤。随后以IRIS和Ruspini数据集作为分析样本验证了所提聚类有效性指标的有效性。以双牌水库的45场典型洪水过程和碧流河水库的13场典型洪水过程为例,重点进行了洪水聚类分析,进一步验证了所提公式的有效性。碧流河水库洪水聚类,遵循了成因分析和聚类分析相结合的原则,其目的是确定各类主要天气系统的典型代表过程,以便进行水库防洪分类预报调度设计。In clustering analysis, the number of clusters is a very important parameter, and the determination of the optimal cluster number is one of the key points. On the basis of the Fuzzy Clustering Iterative Algorithm, the correlation coefficient-based cluster validity criterion is first proposed to specify the optimal cluster number, and the flow for specifying the optimal cluster number is given as well. IRIS and Ruspini datasets are used as standards for testing the validity of the proposed criterion. Taken 45 typical hydrographs of the Shuangpai reservoir and 13 typical hydrographs of the Biliuhe reservoir as examples, the clustering analysis is carried out, and the validity of the proposed criterion is further testified. In flood clustering analysis of the Biliuhe reservoir, the cause analysis and the clustering analysis are combined to determine the typical floods of the different weather systems and carry out the reservoir flood operation design with classified forecasts.

关 键 词：模糊聚类 洪水聚类 聚类有效性分析 最佳聚类数 

分 类 号：TV213[水利工程—水文学及水资源]