The Limits of Point Forecasting for Electricity Price Spikes: Evidence for Probabilistic Methods
ACM SIGEnergy Energy Informatics Review · 01.09.2026 · 6 (3) · 282-288
DOI: 10.1145/3861130.3861152 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM SIGEnergy Energy Informatics Review, online erschienen am 11.11.2026. Verfasst von Maria Margarida Mascarenhas und 1 weiteren Autor:innen. Die Arbeit nennt 18 Literaturangaben.
Abstract (Originalsprache)
Accurate forecasting of extreme electricity prices is critical for risk management in day-ahead energy markets, yet most models are optimised for average accuracy rather than tail performance. This paper presents a study on the Belgian day-ahead market evaluating how covariate selection and loss function choice affect spike forecasting, and whether probabilistic forecasting can address the limitations of deterministic models. Using the LEAR model, we find that improving the covariate set with renewable generation, load, and cross-border market signals improves overall MAE by 26% relative to a price-only baseline, with the largest gains at the price tails. A DNN extends the overall reduction to 32%, but its gains are confined to the central deciles. At both spike deciles, no DNN variant improves on the best LEAR configuration, regardless of loss function. A quantile LightGBM model (QLGBM) captures 80% of spike hours within the 90% prediction interval, making it a valuable tool for risk-aware downstream optimisation.
Zitieren
APA
Mascarenhas, M. M., & Kazmi, H. (2026). The Limits of Point Forecasting for Electricity Price Spikes: Evidence for Probabilistic Methods. ACM SIGEnergy Energy Informatics Review, 6(3), 282-288. https://doi.org/10.1145/3861130.3861152
Vancouver
Mascarenhas MM, Kazmi H. The Limits of Point Forecasting for Electricity Price Spikes: Evidence for Probabilistic Methods. ACM SIGEnergy Energy Informatics Review. 2026;6(3):282-288. doi:10.1145/3861130.3861152
BibTeX
@article{Mascarenhas2026,
author = {Mascarenhas, Maria Margarida and Kazmi, Hussain},
title = {The Limits of Point Forecasting for Electricity Price Spikes: Evidence for Probabilistic Methods},
journal = {ACM SIGEnergy Energy Informatics Review},
year = {2026},
volume = {6},
number = {3},
pages = {282--288},
doi = {10.1145/3861130.3861152}
}
EUVIMED