Daily Load Forecasting of Electric Power Based on Casual Relationship Analysis and Graph Convolutional Network

The prediction of daily average load has important practical significance for early distribution or load dispatch, but currently there is a lack of attention in the field of short-term load forecasting. The existing daily power load forecasting methods only rely on time data and autoregressive processes to predict loads, lacking exploration of the impact of meteorological factors on forecasting accuracy. In order to fill the research gap mentioned above, this paper proposes a graph convolutional neural network daily load forecasting method based on causal correlation analysis. In detail, firstly we analyze the main meteorological influencing factors of daily electricity load through causal relationship analysis; Then, the causal relationship and daily load data are input into the graph neural network for training. The results indicate that the proposed model outperforms other reference models with smaller differences and higher prediction accuracy.