Wind gusts prediction using multivariate time series classification

Authors

David Guijo-Rubio

Antonio Manuel Gómez-Orellana

Víctor Manuel Vargas-Yun

Jorge Pérez-Aracil

Marta Vega-Bayo

Luis Prieto-Godino

Antonio Manuel Durán-Rosal

César Hervás-Martínez

Sancho Salcedo-Sanz

Pedro Antonio Gutiérrez

Published in Journal

Energy and AI

COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Impact FactorJCR 2025

9.7

JCR Ranking

25/ 210

Position

ISSN2666-5468

Pages100924

Abstract

Wind speed intermittency is a key challenge for integrating wind energy into the energy mix. Accurate wind power forecasting in wind farms is crucial for mitigating this issue. A major source of uncertainty in nowcasting and short-term wind power prediction arises from wind gusts. At very short timescales (seconds), gusts impact turbine stability and loads, while at longer scales (hours), they significantly affect wind speed and power prediction accuracy. This paper proposes Multivariate Time Series Classification (MTSC) approaches specifically designed to capture temporal dependencies in wind speed time series, which are crucial for identifying and predicting wind gusts. We introduce a novel definition of short-term wind gusts affecting power production, based on detecting consecutive hourly increases and decreases (or vice versa) above a given threshold. This setting leads to highly imbalanced datasets, where only a small fraction of samples qualify as wind gusts. To tackle this challenge, we evaluate multiple MTSC models and resampling strategies and demonstrate their effectiveness in improving predictive performance. Our results highlight the potential of these approaches for enhancing short-term wind power forecasting, ultimately contributing to more reliable and efficient wind energy integration.

Keywords

imbalanced datasets time series classification wind farms wind gusts wind speed prediction

BibTex Citation

@article{Guijo2026Wind,
	author = {Guijo-Rubio, David and G{\' o}mez-Orellana, Antonio Manuel and Vargas-Yun, V{\' i}ctor Manuel and P{\' e}rez-Aracil, Jorge and Vega-Bayo, Marta and Prieto-Godino, Luis and Dur{\' a}n-Rosal, Antonio Manuel and Herv{\' a}s-Mart{\' i}nez, C{\' e}sar and Salcedo-Sanz, Sancho and Guti{\' e}rrez, Pedro Antonio},
	journal = {Energy and AI},
	doi = {10.1016/j.egyai.2026.100924},
	year = {2026},
	pages = {100924},
	title = {Wind gusts prediction using multivariate time series classification},
}

BibTex Unicode Citation

@article{Guijo2026Wind,
	author = {Guijo-Rubio, David and Gómez-Orellana, Antonio Manuel and Vargas-Yun, Víctor Manuel and Pérez-Aracil, Jorge and Vega-Bayo, Marta and Prieto-Godino, Luis and Durán-Rosal, Antonio Manuel and Hervás-Martínez, César and Salcedo-Sanz, Sancho and Gutiérrez, Pedro Antonio},
	journal = {Energy and AI},
	doi = {10.1016/j.egyai.2026.100924},
	year = {2026},
	pages = {100924},
	title = {Wind gusts prediction using multivariate time series classification},
}

APA Citation

Guijo-Rubio, D., Gómez-Orellana, A. M., Vargas-Yun, V. M., Pérez-Aracil, J., Vega-Bayo, M., Prieto-Godino, L., Durán-Rosal, A. M., Hervás-Martínez, C., Salcedo-Sanz, S., & Gutiérrez, P. A. (2026). Wind gusts prediction using multivariate time series classification. Energy and AI, 100924. https://doi.org/10.1016/j.egyai.2026.100924

RIS Citation

TY  - JOUR
AU  - Guijo-Rubio, David
AU  - Gómez-Orellana, Antonio Manuel
AU  - Vargas-Yun, Víctor Manuel
AU  - Pérez-Aracil, Jorge
AU  - Vega-Bayo, Marta
AU  - Prieto-Godino, Luis
AU  - Durán-Rosal, Antonio Manuel
AU  - Hervás-Martínez, César
AU  - Salcedo-Sanz, Sancho
AU  - Gutiérrez, Pedro Antonio
DA  - 2026///
PY  - 2026
DO  - 10.1016/j.egyai.2026.100924
ID  - temp_id_950165934080
SP  - 100924
T2  - Energy and AI
TI  - Wind gusts prediction using multivariate time series classification
ER  -

CV Citation

D. Guijo-Rubio, A.M. Gómez-Orellana, V.M. Vargas-Yun (CA), J. Pérez-Aracil, M. Vega-Bayo, L. Prieto-Godino, A.M. Durán-Rosal, C. Hervás-Martínez, S. Salcedo-Sanz, P.A. Gutiérrez (3/10). "Wind gusts prediction using multivariate time series classification". Energy and AI, pp. 100924, 2026. (Q1, IF: 9.7).