Generation Prediction in a Mini Hydroelectric Power Plant Using Machine Learning and Open-Source Software
DOI:
https://doi.org/10.56294/saludcyt20252244Keywords:
machine learning, generation, Prediction, forecasting modelsAbstract
Introduction: Ecuadorian electric companies that own run-of-the-river hydroelectric plants with regulating reservoirs equal to or less than one day must submit the hourly generation curve planned for the following day to the National Electricity Operator of Ecuador (CENACE) before 10:00 am daily, in addition to having long-term estimates that allow for optimizing their operational planning.
Objective: To predict the behavior of electrical power generation at the Illuchi 1 run-of-river plant by applying machine learning methods, and then determine the most efficient method for each time scenario.
Method: For the development of this study, a historical database of electrical power generation from the Illuchi 1 mini hydroelectric plant was compiled, corresponding to a period of 3 years, 7 months, which was ordered chronologically and subsequently preprocessed. The open-source software Python was used, applying a methodology based on the model and evaluation of machine learning techniques such as Linear Regression, LSTM, GRU and XGBoost.
Results: The XGBoost algorithm showed better prediction performance for one and seven days, obtaining mean absolute error MAE values of 39.26 [W] and 25.60 [W] respectively and the coefficient of determination R2 of 0.44 and 0.79. On the other hand, the GRU model showed greater prediction accuracy in the two-day horizon, reaching a mean absolute error MAE of 36.03 [W] and its coefficient of determination R2 of 0.61.
Conclusions: XGBoost and GRU stand out from other prediction methods due to their ability to identify non-linear models, in order to optimize their forecast accuracy at different time intervals.
References
1. Electricity Master Plan – Ministry of Energy and Mines [Internet]. [citado 20 de julio de 2025]. Disponible en: https://www.recursosyenergia.gob.ec/plan-maestro-de-electricidad/
2. Suntaxi J, Salazar Y, Loor R. Forecast of energy and power demand of the Quito Electric System. Rev Téc Energ. junio de 2019;15(2):12-21.
3. Nadeem M, Oroszlanyova M, Farag W. Effect of Digital Game-Based Learning on Student Engagement and Motivation. Computers. septiembre de 2023;12(9):177.
4. Aljohani A. Predictive Analytics and Machine Learning for Real-Time Supply Chain Risk Mitigation and Agility. Sustainability. enero de 2023;15(20):15088.
5. Gallo A, Pérez F, Salinas D, Gallo A, Pérez F, Salinas D. Data Mining and Short-Term Projection of Power Demand in the Ecuadorian Electric System. Rev Téc Energ. diciembre de 2021;18(1):72-85.
6. Mhlanga D. Artificial Intelligence and Machine Learning for Energy Consumption and Production in Emerging Markets: A Review. Energies. enero de 2023;16(2):745.
7. Gökçe MM, Duman E. Performance Comparison of Simple Regression, Random Forest and XGBoost Algorithms for Forecasting Electricity Demand. En: 2022 3rd International Informatics and Software Engineering Conference (IISEC) [Internet]. 2022 [citado 12 de mayo de 2025]. p. 1-6. Disponible en: https://ieeexplore.ieee.org/document/9998213
8. Zor K, Buluş K. A benchmark of GRU and LSTM networks for short-term electric load forecasting. En: 2021 International Conference on Innovation and Intelligence for Informatics, Computing, and Technologies (3ICT) [Internet]. 2021 [citado 12 de mayo de 2025]. p. 598-602. Disponible en: https://ieeexplore.ieee.org/document/9581373
9. Segovia JA, Toaquiza JF, Llanos JR, Rivas DR. Meteorological Variables Forecasting System Using Machine Learning and Open-Source Software. Electronics. enero de 2023;12(4):1007.
10. Laurencio RM, Breff OT, Ramirez SM, Jiménez DJ. Predicción de Consumo y Demanda de Electricidad Mediante Redes Neuronales Artificiales y Algoritmo Iterativo. Rev Politécnica. 31 de diciembre de 2024;54(3):45-58.
11. Tucumbi L, Guano J, Salazar-Achig R, Jiménez J. DL. Solar Radiation Prediction Using Decision Tree and Random Forest Models in Open-Source Software. Benhala B, Lachhab A, Raihani A, Qbadou M, Sallem A, editores. E3S Web Conf. 2025;601:00051.
12. Achig RS, Gonzales JA, Hidalgo CA. Flow Prediction for Hydropower Generation using LMST Neural Networks. J Phys Conf Ser. octubre de 2023;2609(1):012006.
13. Sohrabpour V, Oghazi P, Toorajipour R, Nazarpour A. Export sales forecasting using artificial intelligence. Technol Forecast Soc Change. 1 de febrero de 2021;163:120480.
14. Bayram F, Aupke P, Ahmed BS, Kassler A, Theocharis A, Forsman J. DA-LSTM: A dynamic drift-adaptive learning framework for interval load forecasting with LSTM networks. Eng Appl Artif Intell. 1 de agosto de 2023;123:106480.
15. Hua Q, Fan Z, Mu W, Cui J, Xing R, Liu H, et al. A Short-Term Power Load Forecasting Method Using CNN-GRU with an Attention Mechanism. Energies. enero de 2025;18(1):106.
16. Dispatch Procedures.pdf [Internet]. [citado 21 de junio de 2025]. Disponible en: https://arconel.gob.ec/wp-content/uploads/downloads/2025/01/ProcedimientosDespacho.pdf
17. Freire FSB, Fiallos JNC, Caiza CIQ, Peñaherrera HRR. Unidades recurrentes cerradas (GRU) vs redes neuronales artificiales en la predicción de la generación eléctrica de la Central Hidroeléctrica Illuchi. AlfaPublicaciones. 30 de agosto de 2023;5(3):150-66.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Lenin Ulloa-Chipantiza , Freddy Pilataxi-Molina , Roberto Salazar-Achig , Diego L. Jiménez J (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
The article is distributed under the Creative Commons Attribution 4.0 License. Unless otherwise stated, associated published material is distributed under the same licence.
