Reservoir inflow forecasting by artificial neural networks

This study applied the artificial neural networks(ANNs) for daily reservoir inflow forecasting of Lam Takong and Lam Phra Ploeng reservoirs in the Upper Mun basin, Nakhon Ratchasima province. The daily meteo-hydroiogical data including the reservoir inflow, runoff from the upstream gages, rainfall, temperature, relative humidity and air pressure in the vicinity area of the reservoirs were used to develop the ANNs models for daily reservoir inflow forecasting. The daily data between 1987-2000 and 1992-2000 were used in the study for Lam Takong and Lam Phra Ploeng Reservoirs respectively. The result showed that the models that used both the meteorological data and the hydrological data as the input variables gave better forecast than the models that used only rainfall-runoff data. The best model for Lam Takong and Lam Phra Ploeng Reservoirs were 10-63-1 and 15-17-17-17-1 respectively. The R ² for training of both reservoirs were 0.90. The R ² for testing using the data of 1999-2000 were 0.55 and 0.72 for Lam Takong and Lam Phra Ploeng Reservoirs respectively.