基于BP神经网络的河道断面变形预测模型

Model for predicting cross section deformation of river channel based on BP neural networks

采用“试错法” ,以及通过建立网络训练学习过程与网络特征参数之间的反馈机制 ,对BP神经网络隐含层单元数和特征参数进行优化选择。在此基础上 ,以河段水沙条件、水流主流位置及河道边界条件为输入向量 ,河道断面高程或冲淤变形为输出向量 ,建立了基于BP神经网络的河道断面变形预测模型。经长江中游马家咀河段实测资料验证 。

The hidden layers and characteristic parameters of back propagation(BP) neural networks are optimized by means of try and error method and the feedback mechanism between training and determination of characteristic parameters. On this basis, by using flow and sediment conditions and position of main flow and river boundary as the input vectors, using cross section elevation and variation of silting as the output vectors, the model for predicting cross section deformation of river channel is established. The validity of this model is verified by observation data.