基于模糊聚类和BP神经网络的流域洪水分类预报研究  

Research on classified flood forecast based on fuzzy clustering and BP neural networks

作　　者：任明磊[1] 王本德[1] 

机构地区：[1]大连理工大学土木水利学院,辽宁大连116024

出　　处：《大连理工大学学报》2009年第1期121-127,共7页Journal of Dalian University of Technology

基　　金：国家自然科学基金资助项目(50479056)

摘　　要：传统的流域洪水预报大都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,其预报精度需要进一步提高.用模糊聚类ISODATA迭代模型将历史洪水分为若干类型,进行水文预报模型参数的分类调试;并建立BP神经网络分类模型判断实时洪水所属类别,选择其相应类别的模型参数实现流域洪水的分类预报.在辽宁省大伙房水库流域的实际应用表明:此方法不但可以实现洪水实时在线分类而且提高了流域整体洪水预报精度,是一种为水库实时调度提供可靠依据的有效洪水预报方法.Traditional flood forecast usually tries to find the generic or average disciplinarian of forming runoff in the basin by rating a set of hydrological model parameters, and its forecasting precision needs to be further improyed. Firstly, the historical floods were divided into several types by fuzzy clustering ISODATA iterative model, and several sets of hydrological model parameters were debugged separately. Secondly, BP neural networks classified model was established to judge the category of real-time flood, and the model parameters fit to the real-time flood were chosen to realize classified flood forecast. The factual application in Dahuofang reservoir basin shows that this method can realize the classification of real-time flood on-line, improve the forecasting precision integrally, and provide reliable information for real-time operation of the reservoir.

关 键 词：洪水预报 分类 BP神经网络 模糊聚类 

分 类 号：TV877[水利工程—水利水电工程]