Article Details

Abstract

Wind speed prediction is a core aspect of meteorological disaster prevention and mitigation, playing a critical role in safeguarding public safety and supporting government decision-making. To improve the accuracy of nowcasting for extreme wind speed, this paper constructs an extreme wind speed predictive model based on the Long Short-Term Memory (LSTM) network. Using regional minute-level observational data from the China Meteorological Administration's Tianqing Big Data Platform, wind speed parameters are identified as core features through a combination of statistical analysis and Machine Learning methods. By optimizing the input and prediction time windows, a 15-minute historical sequence combined with a 15-minute prediction lead time is determined as the optimal configuration. During testing, the model achieves a Mean Absolute Error of 1.387 m/s and a Relative Error of 7.96% for wind speed prediction. Evaluation on real gust events shows a goodness of fit (R2) of 0.893 1, with the model effectively identifying wind events exceeding Level 7 on the Beaufort scale. Combined with confidence level calculation, the model can provide quantitative support for extreme wind nowcasting, demonstrating practical value for operational applications.

Keywords

meteorological nowcasting; extreme wind; wind speed prediction; Long Short-Term Memory network; feature selection

References

1. [1] ZHOU K H，ZHENG Y G，YANG B，et al. Objective Nowcasting of Severe Convective Weather: Technological Progress and Outlook [J].Journal of Meteorological Research，2025，39（3）：724-740.
2. [2] Dong Fu, Zhang Ling, Zhang Haipeng, etc A review of ensemble forecasting of strong weather processes based on WRF model [J]. Arid Meteorology, 2020, 38 (5): 699-708
3. [3] Yin Chuanhao Research on Short term Precipitation Forecasting Method Based on Convolutional Neural Network [D]. Nanjing: Nanjing University of Information Science and Technology, 2024
4. [4] Wang Qinglin Research on Short term Thunderstorm and Gale Prediction Based on Machine Learning [D]. Shenyang: Shenyang University of Technology, 2022
5. [5] INNOCENT J J，JACQUELINE B T，ABRAHAM O，et al. Precipitation Nowcasting in Dar es Salaam: Comparative Analysis of LSTM and Bidirectional LSTM for Enhancing Early Warning Systems [J].Journal of Geoscience and Environment Protection，2025，13（4）：327-342.
6. [6] HUANG C，KARIMI H R，MEI P，et al. Evolving Long Short-Term Memory Neural Network for Wind Speed Forecasting [J].Information Sciences，2023，632：390-410.
7. [7] XIAO H X，WANG Y Q，ZHENG Y，et al. Convective-gust Nowcasting Based on Radar Reflectivity and a Deep Learning Algorithm [J].Geoscientific Model Development，2023，16（12）：3611-3628.
8. [8] Ma Zhifeng, Zhang Hao, Liu Jie A review of short-term precipitation forecasting based on deep learning [J]. Computer Engineering and Science, 2023, 45 (10): 1731-1753
9. [9] Zhuang Xiaoran, Zheng Yu, Wang Yaqiang, etc Research on Fusion Precipitation Near Range Forecasting Method Based on Deep Learning and Its Application in Eastern China [J]. Journal of Meteorology, 2023, 81 (2): 286-303
10. [10] Cao Weihua, Nan Gangqiang, Chen Mingxuan, etc Research on Fine Scale Precipitation Near Range Forecasting in the Beijing Tianjin Hebei Region Based on Deep Learning [J]. Journal of Meteorology, 2022, 80 (4): 546-564
11. [11] Zeng Kang, Min Jinzhong, Zhuang Xiaoran, etc A Strong Convection Near Range Forecasting Method Based on Generative Adversarial Networks and Its Application Evaluation in Eastern China [J]. Atmospheric Science, 2024, 48 (6): 2316-2328
12. [12] Zhang Haiyan, Wu Qishu, Wei Guofei, etc Short term heavy precipitation short-term close range forecasting technology based on multi-source data fusion integration [J]. Atmospheric Science, 2025, 49 (6): 1711-1726
13. [13] Jin Haoyu, Ye Xiaoling, Xiong Xiong, etc Research on Strong Wind Prediction Technology Along High Speed Rail [J]. Journal of Railway Science and Engineering, 2021, 18 (3): 615-622
14. [14] Li Long, Wang Rui, Zhang Weijiao Research on LSTM based High Speed Rail Gale Prediction Model and Algorithm [J]. Railway Computer Applications, 2021, 30 (2): 18-21

Funding

Zhejiang Provincial Meteorological Bureau Science and Technology Plan Project (2024YB35); Shaoxing Meteorological Bureau Science and Technology Plan Project (2025KY02)

About the authors

Yu Jinhe (1997-), male, Han ethnicity, native of Shaoxing, Zhejiang, assistant engineer, master's student, research direction: integration application of artificial intelligence and meteorological elements.