Multi-timescale Load Forecasting Based on Deep Diffusion Modeling

Accurate load forecasting can provide guidance for power grid planning and dispatch operation. In order to further improve the accuracy of power load forecasting, this paper proposes a multi-timescale power load forecasting method based on a deep diffusion model, which combines the variational modal decomposition (VMD) and the diffusion model of U-network to construct a multi-timescale power load forecasting model. First, the original power load sequence is decomposed by VMD to reduce its non-stationarity and volatility and extract the more regular modal components; second, a deep diffusion model suitable for the time series is constructed using the UNet network; then, the deep diffusion model is utilized to carry out the forward noise addition and backward denoising processes from the monthly, weekly, and daily time scales, to train the multi-timescale temporal diffusion model; and finally, the Finally, an example analysis is carried out based on the load data of a domestic network province to verify that the method is better than the traditional model in terms of prediction accuracy, and has better prediction accuracy and universality.