Water Volumes Estimation of Dams and Flood Forecasting System of the Sagami River Using Artificial Neural Networks

Abstract

With designating the first class Sagami river that is under the control of Kanagawa prefecture as a river flood control implemented, a flood forecasting system to predict the water level during flood from 1 to 3 hours in advance was studied. It is necessary to ensure the water volumes estimation accuracy of dams for keeping the water level accuracy of each prediction point. In this study, artificial neural networks model that is one of the correlation analysis methods was employed to estimate water volumes of dams such as water volume discharge and flowing. As a result, following conclusions were obtained. First, application of artificial neural networks indicated the possibility of estimating water volumes of dams to 3 hours in advance by using only actual measured values. Second, the application of both correlation analysis and runoff analysis at the same time indicated that it would ensure the stable estimation accuracy and perform as a backup system. It successfully predicted reaching to the warning water level about 3 hours in advance during the tentative operation in the fiscal 2004 year.