水文时间序列的相似性搜索研究

Similarity search in hydrological time series

将时间序列相似性搜索的数据挖掘方法应用于水文时间序列数据中,挖掘相似的水文过程.在分析欧氏距离和动态时间扭曲距离两种相似性距离度量方法特点的基础上,采用对时间轴的伸缩和弯曲具有较好适应性的动态时间扭曲距离法对塔里木河流域源流区出山口水文站沙里桂兰克站1961—2000年共220场洪水流量过程进行相似性搜索,基于相似性距离度量矩阵,挖掘出相似的洪水流量过程.结果表明,沙里桂兰克站洪水过程虽形态多样,但也表现出一定的相似性,基于动态时间扭曲法的相似性搜索能有效挖掘出相似的水文过程.

Data mining in the time series similarity search was applied to hydrological time series data for mining similar hydrological processes. The characteristics of two kinds of similarity distance measuring methods,Euclidean distance and dynamic time warping distance,were analyzed. Owing to its satisfactory adaptability in terms of stretching and warping to the time axis,the dynamic time warping distance method was employed to perform a similarity search of 220 floods at Shaliguilanke Station in the Tarim River Basin in China from 1961 to 2000. Based on the similarity matrices,the similar flood discharge processes were mined. The results show that although the flood discharge processes at Shaliguilanke Station are diverse,they exhibit a certain similarity. The similarity search based on the dynamic time warping can be employed for effective mining of similar hydrological processes.