Wind speed prediction methods and applications: A review and roadmap

Authors

  • Yihe Zou Key Laboratory of Intelligent Control and Neural Information Processing, Ministry of Education, Yanshan University, Qinhuangdao 066000, Hebei Province, China
  • xuefang xu Yanshan University image/svg+xml https://orcid.org/0000-0002-3861-8733
  • Guoqian Jiang Key Laboratory of Intelligent Control and Neural Information Processing, Ministry of Education, Yanshan University, Qinhuangdao 066000, Hebei Province, China
  • Yan Wang School of Green Energy and Storage, Lanzhou University of Technology, Lanzhou, 730050, Gansu, China https://orcid.org/0009-0003-1149-9257
  • Wenxian Yang School of Computing and Engineering University of Huddersfield,Huddersfield HD1 3DH, UK
  • Hamid Reza Karimi Politecnico di Milano, Department of Mechanical Engineering, Milan, Italy https://orcid.org/0000-0001-7629-3266
  • Peiming Shi Key Laboratory of Intelligent Control and Neural Information Processing, Ministry of Education, Yanshan University, Qinhuangdao 066000, Hebei Province, China

Keywords:

Wind Speed Foecasting, Review And Roadmap, Physical And Statistical Methods, Machine Learning, Deep Learning

Abstract

Background and Purpose: Wind speed forecasting (WSF) is of paramount importance for the effective planning and operation of power and energy systems. However, the field currently lacks a comprehensive review article that systematically summarizes the state-of-the-art and provides guidance for future developments. To bridge this gap, this paper aims to present a holistic review and a strategic roadmap to analyze the developmental progress and outline future perspectives in wind speed prediction. Methods: This study systematically reviews the historical evolution of WSF methodologies. It traces the transition from early physical and statistical methods, which heavily relied on expert knowledge, through the era of traditional machine learning that began to reduce human intervention, to the current dominance of deep learning techniques. The analysis focuses on how these methods have progressively moved toward end-to-end prediction procedures, establishing a direct link between prediction models and large-scale wind speed data. Results: The review reveals a clear evolutionary trajectory in WSF. While early methods depended on manual feature extraction, the advent of deep learning since the late 2010s has significantly reformed the field by further releasing artificial assistance. The study identifies that future research will likely focus on lightweight real-time methods and ensemble methods driven by deep reinforcement learning to address engineering challenges. Furthermore, a detailed roadmap is illustrated to highlight potential research directions and address existing difficulties in the field. An appendix providing definitions of forecast time horizons, common evaluation metrics, and relevant publication titles is also structured for easy reference. Conclusion: The research concludes that wind speed prediction techniques will undoubtedly continue to improve. The proposed roadmap and comprehensive analysis serve as a valuable guide for researchers and engineers, offering a clear vision for the future implementation of WSF technologies in complex engineering settings.

Downloads

Published

2026-10-30