Research on short-term wind speed prediction model based on physical feature-guided gate recurrent unit
LI Zhipeng
JIANG Nan
GUO Xinyu
ZHAO Junbo
XIE Jia
GAO Huanhuan
Abstract：To improve the prediction accuracy of the energy storage frequency regulation instructions for wind farms,the global curvature-stability analysis method is adopted to calculate the non-linear degree(curvature scaling factor)S1 of the deviation from the linear trend of the original wind speed sequence and the stability(change suppression index)S2 of the sequence value changes.The update gate parameters of the gated recurrent unit(GRU)are dynamically adjusted using the non-linear degree S1 and stability S2,and the original wind speed sequence is placed in the improved GRU network for prediction,obtaining the final prediction result.The results show that in terms of prediction error,the method proposed in this study reduces the MAE of the GRU network by an average of 83.34%,the SSE by an average of 98.24%,the RMSE by an average of 74.21%,and the MAPE by an average of 86.10%compared to the GRU network.
Keywords：wind speed predictionglobal curvature-stability analysis methodimproved GRUnew energywind-storage coordinated frequency regulation
Publication Date：2026-01-31
Online Publishing Date：2026-03-11(First online date of this platform, not the publication date of the document)
Pages：4( 54-57 )