Volume 10,Issue 8
Research on Medium- and Long-Term Wind Power Forecasting Technology Based on Ensemble Forecasting
With the continuous growth of grid-connected wind power capacity and the full participation of new energy in power market transactions, medium and long-term power forecasting has become a key supporting technology to guarantee the safe and economic operation of power systems. The forecasting ability of deterministic numerical weather prediction (NWP) decays sharply for lead times longer than 4 days. Ensemble forecasting provides a probabilistic information foundation to break this bottleneck, yet its application at wind farm scale is confronted with critical problems including coarse spatial resolution, prominent systematic bias, and low utilization rate of multi-member information. Taking Yuanqu Wind Farm in Shaanxi Province as the research object, this paper systematically carries out research on multi-source ensemble forecast data adaptation processing, hybrid forecasting modeling driven by random forest and physical models, as well as probabilistic forecasting methods focusing on medium- and long-term wind power forecasting based on ensemble forecasts. In terms of data adaptation processing, four global ensemble forecasting systems including ECMWF-ENS, UKMO-EGRR, CMA-BABJ and NCEP-GEFS are compared and evaluated for their applicability in complex mountain wind farms. The results show that ECMWF-ENS has the lowest forecasting error and optimal seasonal stability. On this basis, a complete processing workflow consisting of spatial downscaling, quantile mapping bias correction, and ensemble spread consistency reconstruction is established, which reduces the root mean square error (RMSE) of wind speed forecasting by 15–25%. In terms of forecasting modeling, a hybrid driving framework combining random forest wind speed correction and equivalent power curve physical conversion is proposed. The random forest is adopted to learn local nonlinear residuals, while physical models provide hard constraints for energy conversion. The model is trained with data from January 2024 to February 2025 and verified in the independent test period from January to April 2026. The power generation accuracy of ensemble mean point forecasting reaches 83.26%, with an RMSE of 42.73 MW, meeting the requirements of power industry standards. This study constructs a complete technical chain from raw ensemble forecast data to wind farm-level probabilistic power forecasting, and verifies the effectiveness and engineering practicability of the hybrid driving method for medium and long lead times.
[1] National Development and Reform Commission, 2025, Circular on Deepening the Market-Oriented Reform of New Energy On-Grid Electricity Prices and Promoting High-Quality Development of New Energy: FGJG (2025) 36.
[2] National Energy Administration, 2026, 2025 National Electric Power Industry Statistical Data. Beijing.
[3] Shu Y, Chen G, He J, et al., 2022, Research on Framework of New Power System Dominated by New Energy Sources. Proceedings of the Chinese Society for Electrical Engineering, 42(12): 4233–4245.
[4] Jin L, Wang Y, Chang J, 2023, Ultra-Short-Term Wind Power Forecasting Based on Parallel Machine Learning. Journal of Hydroelectric Engineering, 42(1): 40–51.
[5] Yuan C, Tang Y, Mei R, et al., 2021, A PSO-LSTM Model of Offshore Wind Power Forecast Considering the Variation of Wind Speed in Second-Level Time Scale. Mathematical Problems in Engineering, 2021: 1–9.
[6] Li M, 2021, Wind Power Forecasting Based on Wind Farm Meteorological Data, thesis, Inner Mongolia University of Science and Technology.
[7] Mu M, Wang Q, 2016, Research on Predictability in Numerical Weather Prediction. Meteorology, 42(7): 781–790.
[8] Bao M, Ni Y, 2003, Research on Error Growth in Medium-Range Numerical Weather Prediction. Chinese Journal of Atmospheric Sciences, 27(5): 889–900.
[9] Wang Z, Zhang Y, et al., 2014, Analysis on Application of ECMWF Ensemble Forecast Products in Medium-Range Weather Forecasting. Meteorology, 40(11): 1361–1368.
[10] Liu Y, Zhao D, Song L, 2020, Research on Wind Speed Correction Method Based on ECMWF Ensemble Forecast. Meteorology, 46(8): 1069–1078.
[11] Wang B, Wang N, Liu C, 2014, Wind Power Forecasting Technology Based on Numerical Weather Prediction. Power System Technology, 38(10): 2741–2747.
[12] Jin R, Ma J, Li Y, 2015, Research on Application of ECMWF Ensemble Forecast for Medium-Range Weather Forecasting in China. Meteorology, 41(5): 573–582.
[13] Du J, Chen J, 2021, Review and Prospect of Ensemble Forecasting Methods. Meteorology, 47(12): 1433–1447.
[14] Feng S, Wang B, Wang Z, et al., 2021, Research Progress and Prospect of Wind Power Forecasting Technology in China. Power Generation Technology, 42(1): 80–87.
[15] Ma J, Wang B, Liu C, 2019, Research on Photovoltaic Power Forecasting Method Based on Ensemble Forecasting. Renewable Energy Resources, 37(11): 1601–1607.
[16] Zhan W, Che J, Wang B, et al., 2024, Regional Multi-Output Photovoltaic Forecasting Method Based on Gridded Numerical Weather Prediction. Electric Power, 57(3): 144–151.
[17] Zhu R, Zhang H, Li W, 2025, Application of Numerical Weather Prediction Technology in Renewable Energy Generation. Hydropower and New Energy, 39(9): 69–72.
