Day-Ahead Electricity Prices Prediction Method Based on Spatial-temporal Transformer with Variable-Specific Modeling

Electricity price predictions occupy an important position in the electricity market and are a key to the healthy operation. in order to solve the problem that the past prediction models can only capture the average characteristic due to the variables sharing the same parameters, this paper proposes a spatial-temporal Transformer prediction model based on variable-specific modeling, by designing specific temporal projection matrices and spatial projection matrices that enable specific modeling of each variable and correlations between variables so that this prediction model captures the unique spatial-temporal characteristic of day-ahead electricity prices. Taking electricity price data from Australia as an example, verified the effectiveness of the prediction model by comparing it with six mainstream prediction models.