基于产流机制的中长期径流预测RBF神经网络模型研究

Research on RBF Neural Network Model for Medium and Long Term Runoff Forecasting Based on Runoff Generation Mechanism

从产流的物理过程出发,提出了影响产流的主要因子集,利用逐步回归分析法筛选影响各月径流的主要因子;利用筛选的主要因子建立了逐月径流预报RBF神经网络模型,并进行了实证研究。结果表明,不同月份的径流量主要影响因子不完全相同,存在明显的季节性差异;基于产流机制的RBF神经网络模型对于月尺度的径流过程,具有较好的模拟与预测能力。

The main factor set that affects runoff generation is firstly established based on the physical process of runoff, and the main factors influencing monthly runoff are selected by using stepwise regression analysis. Then a monthly runoff prediction model based on Radial Basis Function（ RBF） neural network is set up. A case study is carried out. The results show that the main factors influencing monthly runoff are not identical in different months which show obvious seasonal difference, and the RBF neural network model based on runoff generation mechanism has better simulation and forecast ability for monthly scale runoff process.