One-day ahead forecasting of energy production from run-of-river hydroelectric power plants with a deep learning approach
Mehmet Bilgili, Sinan Keiyinci, Fırat Ekinci
Abstract
Accurate energy production forecasting is critical when planning energy for the economic development of a country. A deep learning approach based on Long Short-Term Memory (LSTM) to forecast one-day-ahead energy production from the run-of-river hydroelectric power plants in Turkey was introduced in the present study. In addition to the LSTM network, three different data-driven methods, namely, adaptive neuro-fuzzy inference system (ANFIS) with fuzzy c-means (FCM), ANFIS with subtractive clustering (SC), and ANFIS with grid partition (GP) were applied. The correlation coefficient (R), mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE) were used as quality metrics for prediction. Predicted values of the LSTM, ANFIS-FCM, ANFIS-SC, and ANFIS-GP models were compared with observed values by evaluating their errors. MAPE values in the testing process are 5.98%, 6.14%, 6.16%, and 6.40% for the LSTM neural network, ANFIS-FCM, ANFIS-SC, and ANFIS-GP models, respectively. The comparison revealed that the LSTM neural network provided high accuracy results in one day-ahead short-term energy production prediction and gave higher performance than other ANFIS models used in the study.