神经网络理论在河道洪水预报中的应用  

Application of Neural Network in Channel Flow Forecasting

作　　者：王光生[1] 苏佳林[1] 沈必成 刘超 

机构地区：[1]水利部水文局,北京100053 [2]黑龙江省水文局,黑龙江哈尔滨150001 [3]北京市水文总站,北京100089

出　　处：《水文》2009年第5期55-58,共4页Journal of China Hydrology

摘　　要：本文将神经网络用于松花江干流河道汇流计算和河道洪峰水位的预报。对各种转移函数的效果进行了比较,线性函数和双正切函的精度较好超过传统的马斯京根法,其中线性转移函数最好,说明对于大江大河线性转移函数最好。由上游断面洪峰水位预报下游断面洪峰水位也取得了良好的效果。The application of BP neural network in channel routing and flood peak stage predicting in the Songhuajiang River Basin was introduced, what this study focused on was the effect of different activation function and the comparing with tradition hydrological models. In the channel routing results, all the activation functions are perfect, among which results from hyperbolic function and linear function are better than the Muskingum model, the result of linear function is the best. In the flood peak stage prediction, good results were obtained from sigmoid and linear function.

关 键 词：水文预报 河道汇流计算 洪峰水位预报 神经网络 应用 

分 类 号：P338.1[天文地球—水文科学]