基于神经网络模型的输沙过程推估  

Estimation of Suspended Sediment Loads by Artificial Neural Network Methods

作　　者：全胜蓝 江衍铭 QUAN Sheng-lan;CHIANG Yen-ming(Institute of Hydrology and Water Resources Engineering, Zhejiang University, Hangzhou 310058, China)

机构地区：[1]浙江大学建筑工程学院水文与水资源工程研究所,杭州310058

出　　处：《中国农村水利水电》2020年第9期112-116,共5页China Rural Water and Hydropower

基　　金：中央高校基本科研业务费专项资金资助(2018QNA4022);国家重点研发计划(2016YFC0402406)。

摘　　要：以皇甫川流域为研究对象,在分析区域降雨及来水量对含沙量影响的基础上,分别构建静态倒传递神经网络及动态反馈式神经网络模型推估汛期输沙过程,并对模型推估结果与误差进行对比分析。研究结果表明:动静态模型对于流域输沙过程的推估均具有良好的模拟效果,效率系数均为0.82以上,动态反馈式神经网络模型的精度略微优于静态倒传递神经网络模型;动态网络模型推估沙峰值的效果更好,更加接近于实际观测沙峰值,推估沙峰误差最小达到了2%;静态网络模型在拟合退沙阶段的效果比动态网络模型的效果好。The back propagation neural network(BPNN)and recurrent neural network(RNN)based sediment load estimation are constructed in the Huangfuchuan Basin according to the influence of precipitation and streamflow on suspended sediment load.Moreover,the performance of model estimations and the relative errors are further analyzed.The results indicate that:The RNN model is slightly better than that of BPNN model.Both BPNN and RNN models perform well in sediment load estimations and the CE values are higher than 0.82.The accuracy of peak sediment discharge simulation obtained from RNN model is closer to actual observations and the relative error of peak sediment discharge estimation is within 2%.The performance of BPNN model is significantly better than that of RNN model during recession periods.

关 键 词：输沙量 倒传递神经网络 反馈式神经网络 输沙过程推估 

分 类 号：TV213[水利工程—水文学及水资源]