ニューラルネットワークと一次元不定流を組合せた縦断的な河川水位予測手法

抄録

To prevent and mitigate the floods damage, it is important to observe and predict the river profile. In this study, we developed the combination model of artificial neural network model (ANN) and 1-D unsteady flow model for the real-time prediction of river profile. The upstream flow rate is predicted by deep neural network, which is trained with the past flood event. The input of ANN model is radar rainfall. The river profile are predicted by 1-D unsteady flow model by using the flow rate predicted by ANN as the boundary condition. The calculated river profile is corrected by using the real-time observation of the high-density river water observation station. The developed model is applied to the Jobaru River, and we confirmed a high accuracy.