ニューラルネットワークと回帰式を適用した連接水系ダム残流予測システムの開発

抄録

This paper describes a water flow forecasting system for dams using neural networks and regression models. Water flow forecasting task is very important for reliable and economic operation. Many conventional methods have been used. They take much time to develop an accurate forecasting system, because it is difficult to adjust parameters.
Water flow forecasting system for dams, which have much flood data, can be developed by neural networks. On the contrary, that system for dams, which have few flood data, must be developed by regression models. The system has been used for three years at Tadami/Agano river basin to assure forecasting performance by the proposed method.