Building automatic artificial neural network program to forecast reservoir inflows in a river basin – Case studies in Viet Nam

The comments of scientists and managers of water resources management have recently emphasized that water is very precious to serve and develop socio-economic. Therefore, the use of Artificial Intelligence (AI) technology, especially Artificial Neural Network (ANN) to manage water resources and forecast the hydrological variables for river basins is essential. Hereinafter, the ANN-river basin program system has been introduced to allow data management, computation to predict the inflow into reservoirs, which introduces convenient mechanisms for creating input data, automatic solution finding for ANN program. The program has been applied to forecast the inflows into Tri An reservoir in different time steps and daily inflow into Thac Mo reservoir. The results show that this ANN program system is rather convenient, and consumes rather less time to obtain the solutions with high accuracy. For larger Dong Nai river basin, a diagram showing the relationship between the accuracy and time step to forecast the inflow into Tri An reservoir was obtained. It showed that, to ensure the forecasting result about 80% of accuracy (EI), the forecast time should not exceed 10 days. While, for smaller Be river basin, the accuracy to forecast daily inflow into Thac Mo reservoir was very high, 99.93%. However, for longer time steps (7-days, 10-days), the result for forecasting was almost unattainable.