论文部分内容阅读
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。
In view of the fact that the model of Dahuofang reservoir flood forecasting is lumped model, its parameters not only need to be determined by optimization method or artificial trial and error method but also need to be corrected in real time. Therefore, according to the characteristics of Dahuofang basin, a semi-distributed BP neural network flood forecast Model, the auto-calibration of the parameters in the model is realized, and due to its semi-distributed characteristics, some disadvantages of the original lumped model are also avoided. That is, the DEM and ArcGIS are used to divide the watershed according to the hydrological station and the natural watershed to create a BP neural network, which is then applied to the cross-sections and entrance sections of the sub-river basins, and the flow values are forecasted. In each network, stepwise regression analysis The input method filters the input data to get the most significant factor. The model is applied to Dahuofang Reservoir, which has good forecasting accuracy and can be used for the formal forecast of Dahuofang Reservoir.