中文摘要：

径流预报在水资源综合开发利用、科学管理、优化调度等方面起着至关重要的作用。针对不同预报方法预报精度存在差异的问题，选用多元线性回归模型（MVLR）、灰色动态GM（1，N）模型、人工神经网络模型（ANN）分别对宜昌水文站日、月、汛期、非汛期流量进行模拟与比较分析，并在此基础上建立组合预报模型（CP）。结果表明：ANN模型预报精度相对较高，预报效果总体优于MVLR、GM（1，N）2种模型；CP模型能够降低预报的风险，但预报合格率略有下降；各模型日尺度流量预报精度高于月尺度，非汛期预报精度高于汛期。

英文摘要：

Runoff forecasting plays an important role in the development and utilization of water resources, scientific management and optimal scheduling. In order to select a suitable model for the stream flow predication at Yiehang Station, the four models are select ed for comparison, including multiple linear regression （MVLR） model, gray dynamic GM （1, N） model, artificial neural network （ANN） model and combination model （CP）. The runoff at different time scales is simulated. Furthermore, forecasting accuracy is analyzed from different aspects. The results show that the prediction accuracy of ANN model is relatively high, and the simulation results are better than MVLR model and GM（1, N） model; The CP model can improve the prediction of uncertainty, but the forecast pass rate declined slightly; Every model has good ability for daily runoff forecasting and non-flood season forecasting, but unsatisfied for monthly runoff forecasting and flood season forecasting.