Two-stage forecasting of TCN-GRU short-term load considering error compensation and real-time decomposition

Abstract With the continuous development of power system and the growth of load demand, efficient and accurate short-term load forecasting (SLTF) provides reliable guidance for power system operation and scheduling. Therefore, this paper proposes a two-stage short-term load forecasting method based on temporal convolutional network and gated recurrent unit (TCN-GRU) considering error compensation and real-time decomposition. In the first stage, the original sequence is processed by improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), the time series characteristics of the data are extracted by TCN, and the initial load forecasting is realized based on GRU. At the same time, in order to overcome the problem that the prediction model established by the original subsequence has insufficient adaptability in the newly decomposed subsequence, the real-time decomposition strategy is adopted to improve the generalization ability of the model. In the second stage, the error sequence is constructed by the difference between the original sequence and the prediction sequence. The unpredictability of the error sequence is reduced by adaptive variational mode decomposition (AVMD), and the initial prediction result is corrected by TCN-GRU error compensator. Taking the real load as an example, the analysis results show that the proposed method can better capture the nonlinear and unstable characteristics in the load data, and the average absolute percentage error of prediction is 0.819%, which has high accuracy in SLTF.