Forecast of Daily Inflow Series in Dry Season by Recurrent Neural Network Model

Abstract

This study presents an application of neural network model to forecast of daily inflow series in a dry season with hydrological data at Makio dam basin area. The daily inflow and daily precipitation of a period before the day of forecast are used for the input to neural network. The lead time are one day and five days. In the case of one day, the average of relative error to data is about 20%. And in the case of five days, it is shown that the study of recurrent neural network with feed back is effective to progress of precision of forecast.