TITLE:
A Wind Speed Prediction Model Based on Machine Learning in Guyuan Area
AUTHORS:
Shiyun Mu, Yuming Zhai, Hongxia Shi, Shujie Yuan, Lin Han, Lixin Su, Hailing Shi, Juan Gu
KEYWORDS:
Guoyuan Strong Wind, BP Neural Network, Support Vector Machine, Random Forest, Wind Speed Prediction
JOURNAL NAME:
Journal of Geoscience and Environment Protection,
Vol.13 No.11,
November
21,
2025
ABSTRACT: Under the context of global climate change, the frequent occurrence of strong winds in Guyuan has significantly hindered the development of local facility agriculture. Using hourly meteorological data from the Sanying National Station and the Guyuan Greenhouse Station between April 2024 and April 2025, this study employed machine learning methods to develop wind speed prediction models based on BP neural network, support vector machine, and random forest (referred to as BP, SVM, and RF models), aiming to provide references for local disaster prevention and mitigation. The results indicate that: 1) Wind speed at the Guyuan Greenhouse Station exhibits the strongest correlation with that at the National Station (0.489 - 0.595), followed by temperature and 24-hour precipitation (0.116 - 0.336). 2) The mean absolute error (MAE) of the BP, RF, and SVM models at all heights is below 1.5 m/s, the root mean square error (RMSE) is under 2.1 m/s, and the forecast accuracy (FA) exceeds 75%, indicating satisfactory model performance. Compared to 3 m, the MAE and RMSE of 0.5 m are larger, while the FA is smaller. This indicates that the wind speed of 0.5 m is close to the ground, and is more affected by surface roughness and turbulence effects, resulting in greater randomness and making the model more difficult. 3) Based on case analyses of May 10 and May 1, 2024, the overall simulation performance ranks as “RF model > SVM model > BP model”; however, the SVM model demonstrates higher accuracy in simulating strong wind events.