河网洪水预报径向基函数人工神经网络方法

River system flood forecasting based on artificial neural network of radial basis function

讨论了神经网络在河网水流数值模拟中的运用现状,并基于河网水流数值计算模拟的特点,将径向基函数神经网络方法应用于复杂河网水流.模型采用混合学习算法,选用高斯核函数作为隐藏层基函数,充分发挥其表示形式简单、径向对称、光滑性好和解析性好的优势,并采用k-均值聚类算法来确定径向基函数的参数,运用最小二乘法求解权值.建立了珠江三角洲河网的洪水预报模型,计算表明,预测结果与实测数据吻合较好,该模型具有运算速度快、简便易用且预报精度较高等特点.

The neural network of radial basis function （RBF） is used to construct a model for flood forecasting in a complicated river network based on its characteristic. The Gaussian kernel function is selected as the transform function in the hidden layer exerting its good properties as simplicity, radial symmetry and smoothness and accurate analyses. And the self-organizing learning method and the k-means clustering algorithm are proposed for the parametric estimation of the network, and then the least square estimation algorithm is used to produce a weighted sum of the output from the hidden layer. The proposed methodology is finally applied to the Pearl River Delta river network to forecast the flood discharge and water level. The result of the simulation indicates that the RBF network can be applied successfully and high accuracy and reliability of flood forecasting in the complicated river network can be achieved.