基于BP神经网络的大伙房水库洪水预报  

Dahuofang Reservoir Flood Forecasting Based on BP Neural Network

作　　者：王竹[1] 杨旭 张静[1] 何俊仕[1] 

机构地区：[1]沈阳农业大学水利学院,辽宁沈阳110161 [2]辽宁省大伙房水库管理局,辽宁抚顺113007

出　　处：《水电能源科学》2016年第4期31-34,共4页Water Resources and Power

基　　金：高等学校博士学科点专项科研基金新教师类资助项目(20112103120003)

摘　　要：鉴于大伙房水库洪水预报模型为集总式模型,其参数不仅需要优选法选定或人工试错法确定,还需要实时校正,因此根据大伙房流域特点提出了一种半分布式BP神经网络洪水预报模型,实现了模型中参数的自动率定,且由于其半分布式的特点还规避了原集总式模型的部分劣势。即采用DEM和ArcGIS根据水文站及自然流域分水线划分流域,创建BP神经网络,然后应用于各子流域断面及入库断面,预报其流量值,并在每个网络中均运用逐步回归分析法对输入层数据进行筛选,以得到影响最显著因子。将所建模型应用于大伙房水库,预报精度较好,可用于大伙房水库的正式预报。The Dahuofang flood forecasting model is a lumped model and its parameters not only require to be selected by optimum seeking method or manual trial,but also need real-time correction.Based on Dahuofang basin characteristics,this paper proposed a semi-distributed BP neural network model for flood forecasting,which can calibrate the model parameters automatically.Due to its semi-distributed characteristics,it also can avoid some disadvantages of the original lumped model.According to the hydrological station and natural river watershed,it divided the basin by using the digital elevation model（DEM）and ArcGIS and established BP neural network,which was applied to each sub section and reservoir section,to forecast the flow value.Each network input data were screened by stepwise regression analysis method in order to get the most significant impact factors.The model was applied in Dahuofang reservoir.Its prediction accuracy was satisfactory,and it can be used for the official forecasting of Dahuofang reservoir.

关 键 词：BP神经网络 半分布式水文模型 洪水预报 逐步回归分析 

分 类 号：TV124[水利工程—水文学及水资源]