This research presents a new classified real-time flood forecasting framework. In this framework, historical floods are classified by a K-means cluster according to the spatial and temporal distribution of precipitation, the time variance of precipitation intensity and other hydrological factors. Based on the classified results, a rough set is used to extract the identification rules for real-time flood forecasting. Then, the parameters of different categories within the conceptual hydrological model are calibrated using a genetic algorithm. In real-time forecasting, the corresponding category of parameters is selected for flood forecasting according to the obtained flood information. This research tests the new classified framework on Guanyinge Reservoir and compares the framework with the traditional flood forecasting method. It finds that the performance of the new classified framework is significantly better in terms of accuracy. Furthermore, the framework can be considered in a catchment with fewer historical floods.
Skip Nav Destination
Research Article| March 20 2015
Research on classified real-time flood forecasting framework based on K-means cluster and rough set
1College of River and Ocean Engineering, Chongqing Jiaotong University, Chongqing 400074, China; National Engineering Research Center for Inland Waterway Regulation, Chongqing Jiaotong University, Chongqing 400074, China; and State Key Laboratory of Hydraulics and Mountain River Engineering, Sichuan University, Sichuan 610065, China
Search for other works by this author on:
Wei Xu, Yong Peng; Research on classified real-time flood forecasting framework based on K-means cluster and rough set. Water Sci Technol 1 May 2015; 71 (10): 1507–1515. doi: https://doi.org/10.2166/wst.2015.128
Download citation file: