PatchTST-based power market price prediction method and system

Abstract

The application relates to the technical field of power market prediction, and provides a power market price prediction method and system based on PatchTST, which comprises the following steps: acquiring power market historical operation multi-source data; performing multi-dimensional feature extraction on the power market historical operation multi-source data in terms of time characteristics, lag characteristics, rolling statistical characteristics and energy supply characteristics; pre-establishing a time sequence feature extraction network based on PatchTST; adopting the time sequence feature extraction network to perform Patch segmentation on the multi-dimensional characteristics, and mapping the multi-dimensional characteristics segmented through Patch to an embedding space through a linear layer, so that the time sequence information between Patches is reserved, an embedding sequence is formed, the embedding sequence is encoded through a Transformer, and a future prediction value with a preset time length is output. Through the PatchTST model, the multi-source data is subjected to feature extraction and time sequence modeling, the calculation complexity is effectively reduced, the problem of difficult long-distance dependence capture is solved, and the precision and efficiency of power market price prediction are significantly improved.

Images