摘要
由于水沙作用机理和演进规律以及河道形态变化的复杂性,泥沙预报一直是洪水预报的难点.本文将径向基函数神经网络方法应用于复杂河网洪水水沙预报中,并采用k-均值聚类算法确定径向基函数的中心,采用最小二乘法求解权值.在探讨建立预报模型基本方法的基础上,分别建立了具有2个预报期和3个预报期的珠江三角洲河网洪水水沙预报模型.计算结果表明,该方法能够较好地识别洪水水沙的演进规律,预测结果与实测结果吻合较好,运算速度快、简便易行且预报精度较高.
Sediment and flood forecast is a very difficult problem due to the complicated mechanism of water and sediment movement and channel bed variation. This paper discusses the fundamental issues regarding how to establish an intelligent forecast method. The radial basis function is used to construct models for sediment and flood forecast in a complicated river network. The Gaussian kernel function is selected as the transform function in the hidden layer and the k- mean clustering algorithm is proposed for the parametric estimation, then the least square estimation algorithm is used to produce the weights. The proposed methodology is finally applied to the Pearl River Delta river network to forecast the sediment and flood. The result of the simulation indicates that the RBF network can be applied successfully and provide high accuracy and reliability of sediment and flood forecast in a complicated river network.
出处
《水利水运工程学报》
CSCD
北大核心
2008年第1期47-52,共6页
Hydro-Science and Engineering
基金
水文水资源与水利工程科学国家重点实验室开放基金项目(2007490411)
关键词
径向基函数
神经网络
洪水水沙预报
最小二乘法
radial basis function
neural network
sediment and flood forecast
the least square estimation algorithm
