摘要
应用模糊C均值(FCM)和自组织映射网络(SOM)两种方法将洪水流量过程线进行分解,并聚成不同的类别,结合多层前馈神经网络(MFN)建立了两个综合神经网络模型(FCMMFN和SOMMFN),进行洪水预报。在王家厂水库流域洪水预报的应用结果表明,两种聚类方法能够将流量过程分解为具有不同内在规律的若干过程,两种综合神经网络模型预报精度均优于单一的多层前馈网络模型,而且FCMMFN的精度高于SOMMFN。
Fuzzy C Means （FCM） clustering method and Self-Organizing Feature Map （SOM） clustering method were both employed to decomposed the flow hydrograph to several segments, and the situation of rain and runoff was analyzed in each segment, then two hybrid artificial neural networks （FCMMFN ＆ SOMMFN）, based on Fuzzy C Means clustering method and Serf-Organizing Feature Map clustering method separately, were applied to simulate the rainfall-runoff relationship. The case study in Wangjiachang Reservoir indicated that two clustering methods have the ability to decomposed the flow hydrograph to several segments in which the under-lying mechanisms of streamflow generation appear different. Besides, the two classification-based artificial neural networks are both superior to the single multi-layer feedforward network, furthermore, the performance of FCMMFN is better than the one of SOMMFN.
出处
《四川大学学报（工程科学版）》
EI
CAS
CSCD
北大核心
2007年第3期34-40,共7页
Journal of Sichuan University (Engineering Science Edition)
基金
国家自然科学基金资助项目(50579053
50239050)
湖北省自然科学基金资助项目(2005ABA288)
关键词
模糊C均值
自组织映射网络
洪水预报
聚类分析
人工神经网络
Fuzzy C Means（FCM）
Serf-Organizing Feature Map（SOM）
flood forecasting
clustering analysis
artificial neural networks
