A novel minute-scale prediction method of incoming wind conditions with limited LiDAR data

Han Wang, Yunzhou Li, Jie Yan, Wuyang Xiao, Shuang Han and Yongqian Liu

Renewable Energy, 2025, vol. 240, issue C

Abstract: Accurate incoming wind conditions in future are essential inputs for wind farm wake optimization control. Nacelle SCADA data is commonly used for wind prediction at present. However, the influence of wind-blocking and nacelle vibration introduces significant disparities between SCADA data and incoming wind conditions, leading to a serious reduction in prediction accuracy. Furthermore, the existing prediction technologies often offer limited resolution, typically exceeding 15 min, making it challenging to meet the timeliness requirements of wake optimization control. Wind LiDAR can provide accurate incoming wind conditions, but due to its expensive cost, wind farms usually install it at specific wind turbines by leasing, which poses significant challenges in acquiring long-term incoming wind conditions at multiple wind turbines. Therefore, a novel minute-scale prediction method of incoming wind conditions with limited LiDAR data for wind farm wake optimization control is proposed in this paper. For wind turbines equipped with leased LiDAR, a temporal soft-sensing model based on Sequence to Sequence is established by considering the autocorrelation in wind speed and wind direction time series. For wind turbines without LiDAR equipment, a spatial soft-sensing model based on Domain Adversarial Neural Network is established, the unsupervised transfer from wind turbines with LiDAR to those without is realized for the first time. On the above basis, aiming at the characteristics of faster frequency fluctuation and larger amplitude of minute-level wind conditions, a joint wind speed and wind direction prediction model based on multi-task learning is established to achieve the accurate prediction of future incoming wind conditions. Four wind turbines are taken as examples to validate the effectiveness and robustness of the proposed method. The results show that the proposed method has better performance than traditional methods. When RMSE is used as the evaluation index, the deviation of LiDAR temporal soft-sensing can be reduced by 3.8%–59.7% (wind speed) and 17.2%–61.0% (wind direction), spatial soft-sensing can be reduced by 2.2%–50.0% (wind speed) and 10.8%–53.0% (wind direction). The prediction accuracy of incoming wind conditions can be improved by 1.7%–7.4% (wind speed) and 2.4%–7.6% (wind direction), and the wind farm power generation can be improved by 0.15%–4.09%.

Keywords: Wind LIDAR; Temporal soft-sensing; Spatial soft-sensing; Unsupervised transfer learning; Joint prediction (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0960148124023036
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:renene:v:240:y:2025:i:c:s0960148124023036

DOI: 10.1016/j.renene.2024.122235

Access Statistics for this article

Renewable Energy is currently edited by Soteris A. Kalogirou and Paul Christodoulides

More articles in Renewable Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().