Marco comparativo para el pronóstico de demanda eléctrica con machine learning y validación temporal rodante

Contenido principal del artículo

Juan Carlos Castillo
Jessica N.Castillo
Gabriel Pesantez
Wilian Guaman

Resumen

La precisión en el pronóstico de la demanda eléctrica es un elemento central para la planificación y operación de los sistemas de potencia, en particular ante la variabilidad temporal de la carga y la presencia de deriva temporal. En este trabajo se desarrolla un marco comparativo reproducible de modelos de machine learning con validación temporal rodante (rolling-origin expanding), análisis multihorizonte y una métrica operativa de tolerancia relativa (Tol). Se evalúan cuatro modelos representativos: EvoXGB (ensamble secuencial de XGBoost sobre residuales), XGB, TabNet y FT-Transformer, aplicados al pronóstico horario de potencia activa en subestaciones de distribución de un sistema eléctrico ecuatoriano. Para asegurar la comparabilidad cuando existen diferencias de cobertura o desalineación temporal entre predicciones, se incorpora una auditoría explícita basada en alineación y un conjunto común de evaluación (COMMONMASK), complementada con un bloque contiguo común para la figura de zoom. En la subestación representativa (con métricas sobre el conjunto común), XGB logra el mejor desempeño, con R2= 0.993 (corto) y 0.983 (mediano), y un RMSE de 21.16 y 30.84 kW, respectivamente. EvoXGB se mantiene competitivo, mientras que TabNet y FT-Transformer muestran mayor degradación en el horizonte mediano. En la verificación de holdout (90/10) se observa la caída esperada por deriva temporal, preservándose el orden comparativo. El marco propuesto entrega una base trazable para comparar modelos en series reales de subestaciones y para extender el análisis hacia esquemas híbridos y adaptativos.

Detalles del artículo

Sección
Artículo Científico

Referencias

[1] T. Hong and S. Fan, “Probabilistic electric load forecasting: A tutorial review,” International Journal of Forecasting, vol. 32, no. 3, pp. 914–938, 2016. [Online]. Available: https://doi.org/10.1016/j.ijforecast.2015.11.011

[2] IEA, “Renewables 2023: Analysis and forecast to 2028,” International Energy Agency, Paris, France, Tech. Rep., 2023, accessed: May 15, 2026. [Online]. Available: https://upsalesiana.ec/ing36ar2r2

[3] M. G. Pinheiro, S. C. Madeira, and A. P. Francisco, “Short-term electricity load forecasting—a systematic approach from system level to secondary substations,” Applied Energy, vol. 332, p. 120493, Feb. 2023. [Online]. Available: https://doi.org/10.1016/j.apenergy.2022.120493

[4] S. Akhtar, S. Shahzad, A. Zaheer, H. S. Ullah, H. Kilic, R. Gono, M. Jasiński, and Z. Leonowicz, “Short-term load forecasting models: A review of challenges, progress, and the road ahead,” Energies, vol. 16, no. 10, p. 4060, May 2023. [Online]. Available: https://doi.org/10.3390/en16104060

[5] F. Rodrigues, C. Cardeira, J. M. F. Calado, and R. Melicio, “Short-term load forecasting of electricity demand for the residential sector based on modelling techniques: A systematic review,” Energies, vol. 16, no. 10, p. 4098, May 2023. [Online]. Available: https://doi.org/10.3390/en16104098

[6] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, ser. KDD ’16. ACM, Aug. 2016, pp. 785–794. [Online]. Available: https://doi.org/10.1145/2939672.2939785

[7] G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T.-Y. Liu, “Lightgbm: a highly efficient gradient boosting decision tree,” in Proceedings of the 31st International Conference on Neural Information Processing Systems, ser. NIPS’17. Red Hook, NY, USA: Curran Associates Inc., 2017, p. 3149–3157. [Online]. Available: https://upsalesiana.ec/ing36ar2r4

[8] L. Prokhorenkova, G. Gusev, A. Vorobev, A. V. Dorogush, and A. Gulin, “Catboost: unbiased boosting with categorical features,” in NIPS’18: Proceedings of the 32nd International Conference on Neural Information Processing Systems. arXiv, 2017, pp. 6639–6649. [Online]. Available: https://doi.org/10.48550/arXiv.1706.09516

[9] C. Bentéjac, A. Csörgő, and G. Martínez-Muñoz, “A comparative analysis of gradient boosting algorithms,” Artificial Intelligence Review, vol. 54, no. 3, pp. 1937–1967, Aug. 2020. [Online]. Available: https://doi.org/10.1007/s10462-020-09896-5

[10] Z. Mustaffa and M. H. Sulaiman, “Advanced forecasting of building energy loads with XGBoost and metaheuristic algorithms integration,” Energy Storage and Saving, vol. 4, no. 4, pp. 421–438, Dec. 2025. [Online]. Available: https://doi.org/10.1016/j.enss.2025.03.005

[11] T.-N. Tran and Q.-D. Nguyen, “Research on the influence of genetic algorithm parameters on XGBoost in load forecasting,” Engineering, Technology & Applied Science Research, vol. 14, no. 6, pp. 18 849–18 854, Dec. 2024. [Online]. Available: https://doi.org/10.48084/etasr.8863

[12] B. Liang, W. Qin, and Z. Liao, “A differential evolutionary-based xgboost for solving classification of physical fitness test data of college students,” Mathematics, vol. 13, no. 9, p. 1405, Apr. 2025. [Online]. Available: https://doi.org/10.3390/math13091405

[13] S. Arik and T. Pfister, “TabNet: Attentive interpretable tabular learning,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 8, pp. 6679–6687, May 2021. [Online]. Available: https://doi.org/10.1609/aaai.v35i8.16826

[14] Y. Gorishniy, I. Rubachev, V. Khrulkov, and A. Babenko, “Revisiting deep learning models for tabular data,” in NIPS’21: Proceedings of the 35th International Conference on Neural Information Processing Systems. arXiv, 2021, pp. 18 932–18 943. [Online]. Available: https://doi.org/10.48550/arXiv.2106.11959

[15] V. Borisov, T. Leemann, K. Seßler, J. Haug, M. Pawelczyk, and G. Kasneci, “Deep neural networks and tabular data: A survey,” IEEE Transactions on Neural Networks and Learning Systems, vol. 35, no. 6, pp. 7499–7519, 2024. [Online]. Available: https://doi.org/10.1109/TNNLS.2022.3229161

[16] L. Grinsztajn, E. Oyallon, and G. Varoquaux, “Why do tree-based models still outperform deep learning on tabular data?” in NIPS’22: Proceedings of the 36th International Conference on Neural Information Processing Systems. arXiv, 2022, pp. 507–520. [Online]. Available: https://doi.org/10.48550/arXiv.2207.08815

[17] R. Shwartz-Ziv and A. Armon, “Tabular data: Deep learning is not all you need,” Information Fusion, vol. 81, pp. 84–90, May 2022. [Online]. Available: https://doi.org/10.1016/j.inffus.2021.11.011

[18] V. Cerqueira, L. Torgo, and I. Mozetič, “Evaluating time series forecasting models: an empirical study on performance estimation methods,” Machine Learning, vol. 109, no. 11, pp. 1997–2028, Oct. 2020. [Online]. Available: https://doi.org/10.1007/s10994-020-05910-7

[19] R. J. Hyndman and G. Athanasopoulos, Forecasting: Principles and Practice, 3rd ed. Melbourne, Australia: OTexts, 2021, accessed: May 15, 2026. [Online]. Available: https://upsalesiana.ec/ing36ar2r16

[20] C. Bergmeir and J. M. Benítez, “On the use of cross-validation for time series predictor evaluation,” Information Sciences, vol. 191, pp. 192–213, May 2012. [Online]. Available: https://doi.org/10.1016/j.ins.2011.12.028

[21] M. Q. Raza and A. Khosravi, “A review on artificial intelligence based load demand forecasting techniques for smart grid and buildings,” Renewable and Sustainable Energy Reviews, vol. 50, pp. 1352–1372, Oct. 2015. [Online]. Available: https://doi.org/10.1016/j.rser.2015.04.065

[22] C. Borges, Y. Penya, I. Fernández, J. Prieto, and O. Bretos, “Assessing tolerance-based robust short-term load forecasting in buildings,” Energies, vol. 6, no. 4, pp. 2110–2129, Apr. 2013. [Online]. Available: https://doi.org/10.3390/en6042110

[23] W. Guamán, P. Benalcázar, J. Córdova-García, and M. Torres, Machine Learning-Based Projections of Long-Term Electricity Consumption: The Case Study of Ecuador. Springer Nature Switzerland, 2025, pp. 174–187. [Online]. Available: https://doi.org/10.1007/978-3-031-83432-5_12

[24] G. Pesántez, W. Guamán, J. Córdova, M. Torres, and P. Benalcázar, “Reinforcement learning for efficient power systems planning: A review of operational and expansion strategies,” Energies, vol. 17, no. 9, p. 2167, May 2024. [Online]. Available: https://doi.org/10.3390/en17092167

[25] J. C. Castillo. (2026) Forecasting-rollingenergy. GitHub repository. [Online]. Available: https://upsalesiana.ec/ing36ar2r27