ECTI TRANSACTIONS ON COMPUTER INFORMATION TECHNOLOGYVolume 13, No. 02, Month NOVEMBER, Year 2019, Pages 170 - 177
Improving model using estimate error for daily inflow forecasting
Yaowalak Ngamsanroaj, Kreangsak Tamee
Abstract Download PDFInflow forecasting is one of the important components for reservoir operation and resource management. To obtain enhanced accuracy for forecasting reservoir inflow, this paper proposed an improved model for forecasting the inflow of Bhumibol reservoir. The 3,169 records of daily inflow data from June 1, 2008, to February 1, 2017, had been collected to calculate the inflow into the reservoir by using Artificial Neural Networks (ANN) Back-Propagation Learning Algorithm for forecasting the inflow of the reservoir in the main model and error prediction model. The performance of the model can be evaluated by four methods: the coefficient of determination (R2), the Nash-Sutcliffe efficiency (NSE), the Root Mean Square Error (RMSE) and the Mean Absolute Error (MAE). Thus, these proposed main and error prediction models were combined to support the forecast of reservoir inflow. The performance of the proposed model can be evaluated by following measured values: R2 was 0.927, NSE was 0.925, RMSE was 6.805 and MAE was 3.611. This indicates that the improved model provides more accurate value.
Articial Neural Network, Model Fore- casting, Reservoir In ow, Error Prediction