实时洪水预报中基于岭估计的AR修正模型研究

为提高实时洪水预报精度，经常将水文模型与误差修正模型相结合，AR模型因其结构简单广泛应用于实时洪水预报误差修正。然而，实际应用显示，AR模型时常出现修正结果不稳定现象，表现为流量修正幅度过大，甚至出现“震荡”现象，严重影响修正效果。鉴于此，本文从矩阵特征值角度解释了AR模型出现不稳定现象的原因，并引入岭估计方法选择性利用流量信息更新自回归系数，使其更满足真实流量的涨落过程，增强该模型的稳健性。将新方法应用于蔺河口流域，结果显示岭估计方法显著提高了AR模型的稳健性，从而改善了模型修正效果，进一步提高了洪水预报精度。

Research on ridge-estimation based autoregressive model in real-time flood forecasting

In order to improve the accuracy of real-time flood forecasting, hydrological model and error correction model are often combined. Autoregressive (AR) model is widely used in real-time flood forecasting to update forecasting flow because of its simple structure. However, some applications show that the results of AR model are often unstable, showing too large flow corrections or even oscillations, which seriously affects its performance. In view of this, this paper makes an explanation for the instability of AR model from the perspective of matrix eigenvalue, and then introduces ridge estimation method to update the autoregressive coefficient by using selective flow information to make it more meet the process of the real flow, and this enhances the robustness of AR model. The results show that the ridge estimation method significantly improves the robustness of AR model, and thus further improves the accuracy of flood forecasting.