Abstract：A hybrid energy storage frequency regulation instruction prediction method is proposed to address the problem of profit loss caused by instruction transmission delay in the joint frequency regulation of thermal-energy storage. The method combines area determination method with dynamic gated scaling gated recurrent unit (DGS-GRU). The traditional variational mode decomposition (VMD) method has false components, which affect the prediction accuracy. This method introduces the area determination method: firstly, the enclosed area between each subsequence after VMD decomposition and the time axis is calculated. Combined with the prediction error of the experimental group, a determination function is constructed to screen for false components. A grouping optimization strategy is adopted for false components: generating random replacement sequences from multiple false sequences, and implementing mean-based symmetry axis mirror reconstruction for single false sequences. The reconstructed subsequence is input into a DGS-GRU network for prediction, and finally overlaid to output the complete instruction. The experiment shows that compared with the traditional VMD-gated recurrent unit (GRU) model, this method reduces the mean absolute percentage error (MAPE) from 9.20% to 5.16% and effectively improves the frequency regulation revenue of power plants.
李志鹏, 姜楠, 兀鹏越, 孙国辉, 董树鑫. 基于面积判定法与动态门控网络的混合储能调频指令预测方法[J]. 电气技术, 2025, 26(11): 11-17.
LI Zhipeng, JIANG Nan, WU Pengyue, SUN Guohui, DONG Shuxin. A hybrid energy storage frequency regulation instruction prediction method based on area determination method and dynamic gated network. Electrical Engineering, 2025, 26(11): 11-17.
2025, Vol. 26 
