Short-term power load forecasting based on VMD-Pyraformer-Adan

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