EUVIMEDEUVIMED entdecken
Crossref

Day-Ahead Wind Power Ramp Events Prediction With Extreme-Value–Driven Learning and Confidence-Aware Detection

Chongyang Wan, Yuan Luo, Weimin Wu, Nikos D. Hatziargyriou, Shunbo Lei

IEEE Transactions on Sustainable Energy · 2026-10 · 17 (4) · 3296-3308

DOI: 10.1109/tste.2026.3669605 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Sustainable Energy, 2026-10. Verfasst von Chongyang Wan und 4 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 37 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Guangdong Basic and Applied Basic Research Foundation, Shenzhen Science and Technology Program.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Wind power ramp events (WPREs)—sudden, large-magnitude changes in wind generation—pose significant threats to power system stability. Accurate day-ahead prediction of WPREs is critical for proactive scheduling, reserve allocation, and grid security. However, existing methods often struggle with three key limitations: they underperform on rare and extreme events, lack sensitivity to multi-scale temporal dynamics, and provide little to no uncertainty quantification. To overcome these challenges, we propose a novel day-ahead WPREs prediction method that integrates three innovations. First, we design an Extreme-Value-Driven Learning model that decomposes wind power into multiple temporal frequency bands, enabling the simultaneous capture of long-term trends and short-term changes. Second, we introduce a tailored extreme-based loss function that rebalances model learning from average-case accuracy toward greater sensitivity to ramp events. Third, we develop a wind power–oriented conformal inference method that produces model-adaptive confidence intervals with formal statistical guarantees, enabling a confidence-aware ramp detection algorithm. Unlike conventional approaches that prioritize mean error metrics, our method is tailored to the operational significance of ramp events. Remarkably, experimental results on two real-world wind power datasets show that our method improves WPREs prediction accuracy by up to 29.6% in terms of F1-score with negligible added computational cost.

Förderung

  • National Natural Science Foundation of China
  • Guangdong Basic and Applied Basic Research Foundation
  • Shenzhen Science and Technology Program

Zitieren

APA

Wan, C., Luo, Y., Wu, W., Hatziargyriou, N. D., & Lei, S. (2026). Day-Ahead Wind Power Ramp Events Prediction With Extreme-Value–Driven Learning and Confidence-Aware Detection. IEEE Transactions on Sustainable Energy, 17(4), 3296-3308. https://doi.org/10.1109/tste.2026.3669605

Vancouver

Wan C, Luo Y, Wu W, Hatziargyriou ND, Lei S. Day-Ahead Wind Power Ramp Events Prediction With Extreme-Value–Driven Learning and Confidence-Aware Detection. IEEE Transactions on Sustainable Energy. 2026;17(4):3296-3308. doi:10.1109/tste.2026.3669605

BibTeX
@article{Wan2026,
  author = {Wan, Chongyang and Luo, Yuan and Wu, Weimin and Hatziargyriou, Nikos D. and Lei, Shunbo},
  title = {Day-Ahead Wind Power Ramp Events Prediction With Extreme-Value–Driven Learning and Confidence-Aware Detection},
  journal = {IEEE Transactions on Sustainable Energy},
  year = {2026},
  volume = {17},
  number = {4},
  pages = {3296--3308},
  doi = {10.1109/tste.2026.3669605}
}
Kostenlos starten →