Journals Proceedings

Abstract

Artificial Neural Network (ANN) model has been successfully being used to predict the river flow for rainfall-runoff process. The important question including how long the forecasting time step should be, which rainfall stations should be selected, which discharge and rainfall data at different time steps should be used for the input layer most suitably to forecast reservoir inflow with highest accuracy. Focusing these questions, hereinafter, ANN model was used to forecast inflow into Tri An reservoir in Dong Nai River Basin, Vietnam. Research results showed that six main rainfall stations including Bao Loc, Da Lat, Da Te, Di Linh, Dai Nga, Lam Dong; and the inflows into Tri An reservoir at current and previous time steps, directly affected forecasting inflow; and the appropriate time step to forecast the inflow into Tri An reservoir is from 7 days to 10 days. This is shown by the accuracy, Efficiency Index (EI) in network training period obtaining 90.83% and 89.65%, respectively; and in network testing phase obtaining 92.15% and 77.43%, respectively