We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
An energy prediction approach using bi-directional long short-term memory for a hydropower plant in Laos.
- Authors
Kaewarsa, Suriya; Kongpaseuth, Vanhkham
- Abstract
Hydropower remains the largest source of renewable electricity while most hydropower plants, especially commercial hydropower plants, require accurate future energy or reservoir incoming flow ahead prediction for ensuring effective generation planning and revenue maximization as well as for flood control, and environmental protection purposes. This study develops effective daily and monthly prediction models using a Bi-Directional Long Short-Term Memory (Bi-LSTM) network-based meteorological and hydrological data, properly pre-processed by cross-correlation function (CCF), and normalization techniques for predicting the future energy portion corresponding to the natural incoming flow expected flowing into the reservoir (reservoir inflow) in the near future (days and months ahead) for Nam Theun 2 Hydropower Plant (NT2HPP) in Laos. The proposed models were tested with a separated dataset excluded from the training dataset and then evaluated using measurement or statistical indices including the root mean square error (RMSE), relative error (RE), and efficiency index (EI) in couple with Taylor diagrams. For comparison purposes, the two widely used prediction approaches namely, Feedforward Backpropagation Neural Network (FFBPNN), and Long Short-Term Memory (LSTM) based on the same datasets were also adopted and analyzed. The study results revealed that the proposed daily prediction model, with the statistical indices of EI = 74.872%, RMSE = 11.313, and RE = 4.560% for 2020; EI = 90.910%, RMSE = 7.237, and RE = 1.896% for 2021; and EI = 82.807%, RMSE = 4.918, and RE = 2.763% for 2022, was superior to the other models. Similarly, the proposed monthly prediction model, with the statistical indices of EI = 98.133%, RMSE = 55.213, and RE = 8.407% for 2020; EI = 99.819%, RMSE = 19.314, and RE = 0.611% for 2021; and EI = 97.316%, RMSE = 44.780, and RE = 6.878% for 2022, was also superior to the other models. These proved that the proposed models were more robust and efficient for future energy prediction.
- Subjects
LAOS; HYDROELECTRIC power plants; WATER power; FEEDFORWARD neural networks; STANDARD deviations; STATISTICAL measurement; RESERVOIRS; ENERGY futures
- Publication
Electrical Engineering, 2024, Vol 106, Issue 3, p2609
- ISSN
0948-7921
- Publication type
Article
- DOI
10.1007/s00202-023-02096-8