Article Details

Abstract

Affected by the randomness and volatility of wind energy, the integration of large-scale wind power into the grid poses significant challenges to stable grid operation. Accurate wind power prediction is a crucial basis for formulating power generation plans and promoting the consumption of wind power. Traditional single model methods often fail to fully account for the differences between samples, leading to large prediction deviations and limited model performance. To address these issues, this paper proposes an ensemble Just-In-Time Learning wind power forecasting method based on Gaussian Process Regression (GPR). The method first selects modeling samples that match the current wind power process state using different similarity functions, then constructs diverse local models combined with GPR modeling techniques, and finally employs a stacking strategy to combine these models and produce the final forecasting output. Experimental results on real data from a wind farm in Yunnan Province demonstrate that the proposed method effectively improves prediction accuracy and stability, reducing the Root Mean Square Error to 15.228 and improving the coefficient of determination to 0.478.

Keywords

wind power forecasting; Gaussian Process Regression; Ensemble Learning; Just-In-Time Learning

References

1. [1] Ding Yiting The cumulative installed capacity of power generation in China increased by 17.5% year-on-year [N]. People's Daily, October 27, 2025 (1)
2. [2] Lu Xueping, Dong Cun, Wang Zheng, etc Research on Short term Wind Power Prediction Technology under Low Temperature Cold Wave Weather [J]. Grid Technology, 2024, 48 (12): 4833-4843
3. [3] Li Te, Huang Ziying A review of the application of machine learning in wind power prediction [J]. Shandong Electric Power Technology, 2025, 52 (8): 45-55
4. [4] Yang Ming, Li Menglin, Wang Bo, etc Overview of Numerical Weather Forecasting Techniques and Applications for New Power Systems [J]. Power System Automation, 2025, 49 (17): 1-20
5. [5] Wang Lei, Teng Wei, Wu Xin, etc A review of key technologies and research applications for wind power prediction [J]. Journal of Power Engineering, 2026, 46 (2): 98-112
6. [6] Xu Weijun, Pei Huan, Wei Yong Ultra short term prediction of SVR wind power based on particle swarm optimization [J]. Journal of Northeast Normal University (Natural Science Edition), 2017, 49 (1): 73-77
7. [7] Mustaffa Z，Sulaiman M H. Random Forest Based Wind Power Prediction Method for Sustainable Energy System [J/OL].Cleaner Energy Systems，2025，12：100210.https://doi.org/10.1016/j.cles.2025.100210
8. [8] Yang Qiang, Yan Zonghui, Du Xiuju, etc Research on New Energy Generation Prediction Model Based on ANN Architecture [J]. Electrical Applications, 2024, 43 (2): 76-82
9. [9] Wu Chen, Cui Qiushi, Xie Yigong, etc LSTM-AM Short term Wind Power Prediction Based on Feature Optimization and Similar Sample Fusion [J]. Southern Power Grid Technology, 2025, 19 (6): 162-172
10. [10] Li Yunlong, Jin Huaiping, Fan Shouyuan, etc Online Selective Integration Real time Learning Wind Power Adaptive Prediction [J]. Journal of Solar Energy, 2024, 45 (10): 487-496
11. [11] Jin Huaiping, Li Jiangang, Qian Bin, etc Integrated real-time learning soft sensing modeling based on multimodal disturbances [J]. Information and Control, 2020, 49 (3): 257-266

Funding

Science Research Fund of Yunnan Provincial Department of Education (2024J1689)

About the authors

Li Jiangang (1994-), male, Han ethnicity, born in Qujing, Yunnan. He is a teaching assistant and a master's student. His research interests include industrial data analysis and predictive modeling, as well as wind power prediction.