洪水预报自回归实时校正多步外延方法研究

Research on Multi-step Epitaxy Method for Auto-regressive Real-time Correction of Flood Forecasting

自回归模型在洪水预报实时校正中应用广泛。针对自回归模型进行连续多时段校正时中间误差系列缺失问题,提出一种基于历史洪水预报误差系列的样本重组自回归外延方法,以淮河流域王家坝断面为背景,选用洪量相对误差、洪峰相对误差、峰滞时间和确定性系数四个指标开展校正效果评估,并与时程递推外延方法对比。结果表明:样本重组外延方法可以提升洪水预报精度,延长洪水预报有效预见期,特别在降低洪量误差和提高洪水过程的拟合精度上优势更为显著。同时,该方法泛化能力较强,具有实用价值。

Auto-regressive model is widely used in real-time correction of flood forecasting. In order to solve the problem of missing intermediate error series in continuous multi-step epitaxy correction of auto-regressive correction model, this paper proposed a sample recombination auto-regressive epitaxy method based on historical flood forecasting error series. Taking the cross-section of Wangjiaba section in the Huaihe River basin as the study area, relative flood volume error, relative flood peak error, peak lag time and deterministic coefficient were applied to evaluate the correction effect, and compared with the time history recursive epitaxy method. The results show that the sample recombination epitaxy method can improve the accuracy of flood forecasting and extend the effective forecasting period, especially in reducing the flood volume error and improving the fitting precision of flood process.