広域雨量データを用いたニューラルネットワークによるダム流入量予測

概要

論文の詳細を見る
At a multipurpose dam, it is necessary to forecast inflow to control flood safely and to operate hydraulic power plant efficiently. In this paper, we propose a method of forecasting the inflow of several hours later by neural network. The correlation is high about the inflow and rain which fell in the dam basin, but it is difficult to forecast by mathematical methods, because the relation is non-linear model. The neural network system can forecast the inflow by learning the past data of inflow and rain in the basin. This system can forecast inflow well after 1 hour or so. However, this system becomes inaccurate rapidly when it tries to forecast inflow at 3 or more hours later, because we use the rain data of the dam basin. Therefore, we also use rain data which is out of the dam basin and in the direction of the windward. The rain data contains information of the rain which will fall at the dam in future. Then, forecast results show that our system is effective.
社団法人電気学会の論文
2004-04-01