有序聚类分析法的改进及其在水文序列突变点识别中的应用  

Improvement of Sequential Clustering Method and Its Application to Diagnose Jump Points of Hydrological Series

作　　者：袁满[1] 王文圣[1] 叶濒璘 

机构地区：[1]四川大学水利水电学院,四川成都610065

出　　处：《水文》2017年第5期8-11,共4页Journal of China Hydrology

基　　金：国家自然科学基金项目(51679155)

摘　　要：有序聚类分析法是水文学中识别突变点的有效方法,但该法只考虑了同类之间的离差较小原则,忽略了类与类之间的离差较大原则。基于此,提出了改进的有序聚类分析法,改进法同时考虑了同类之间的离差较小和类类间的离差较大原则。将改进的有序聚类分析法应用于年平均流量序列突变点识别中,并与传统有序聚类分析法进行对比分析,研究结果表明,改进的有序聚类分析法原理明确,识别突变点更加有效。The sequential clustering method is an effective way to diagnosejump points in hydrologic research. However, the method only considers smaller deviations between the same classes, but ignores the larger deviations between different classes. For that reason, this paper proposedthe improved sequential clusteringmethod considering both smaller deviations between same classes and larger deviations between different classes. The improved sequential clustering methodhas been applied to identify jump points ofsome average annual flow seriesand comparedwith traditional sequential clustering method. The results show that theimproved sequential clusteringmethod is clear and effective to identify jump points.

关 键 词：突变点识别 改进的有序聚类分析法 水文序列 显著性检验 

分 类 号：P333[天文地球—水文科学]