基于K均值聚类分析的流域洪水实时分类修正  

Classified Correction of Real-time Flood Forecasting Based on K-Cluster Analysis

作　　者：李月玉 周建奕 蒋汝成 周密 

机构地区：[1]云南省水利水电勘测设计研究院,昆明650021 [2]云南农业大学建筑工程学院,昆明650201

出　　处：《中国农村水利水电》2016年第12期160-162,167,共4页China Rural Water and Hydropower

基　　金：国家水体污染控制与治理科技重大专项(2013ZX07102-006);云南省技术创新人才计划(2011CI092)

摘　　要：针对传统的流域实时洪水预报通过率定一组水文模型参数来寻求一个流域径流形成的一般性或平均化规律的方法在特殊情况下误差较大,以及在实时洪水预报误差修正中历史信息未得到充分利用的问题,利用洪水的相似性扩大实时修正信息量,结合K均值聚类分析方法,提出了洪水实时分类误差修正方法。并在沿渡河流域用30场洪水进行分类参数率定,采用分类修正的方法对4场洪水进行实时修正,分析结果表明:实时分类修正方法能大幅提高预报精度尤其是洪峰流量的预报精度。Targeted at the problem of traditional flood forecast which usually tries the genetic or average disciplinarian of forming runoff in thebasin by supposing a set of hydrological model parameters will produce bigger errors in special case, and the problem of not considering fullythe history information in flood forecasting error correction, by using the K-Cluster analysis and similarity of the flood that can enlarge the re-al-time information, a classified correction of real-time flood forecasting method is presented. Hydrological model parameters of 30 floods areclassified and the classified correction of real-time flood forecasting method are made for real-time correction of 4 floods in the Yandu RiverBasin. The results show that the classified correction of real-time flood forecasting method can improve forecasting precision, especially theaccuracy of peak discharge to a very great extent.

关 键 词：实时洪水预报 误差 相似性 分类误差修正 

分 类 号：TV122[水利工程—水文学及水资源]