Short-Term Power Load Forecasting Based on VMD-Pyraformer-Adan

Abstract

For the characteristics of fluctuation, periodicity and nonlinearity of power load data, this paper proposes a short-term power load forecasting model based on VMD-Pyraformer-Adan. Firstly, the variational modal decomposition (VMD) algorithm is used to modally decompose the electric load data, the over-zero rate and Pearson correlation coefficient are introduced to divide the modal components to obtain the low-frequency, mid-frequency and high-frequency parts, and the reconstructed data are formed with the original load data respectively. Secondly, the reconstructed data are input to the Pyraformer prediction network containing pyramidal attention module (PAM) and coarse-scale construction module (CSCM). Then a new momentum optimizer Adan is used to optimize the parameters of the prediction network. The final output prediction results. The experimental results show that the proposed model in the paper exhibits higher prediction accuracy compared with other models.