A Spring Festival Load Forecasting Model Considering Annual Growth and Lunar Characteristics

Aiming at the phenomenon of low accuracy in load forecasting before and after the Spring Festival, a Spring Festival load forecasting model based on annual growth coefficient and lunar indicators is proposed. Firstly, the average load of the month before the Spring Festival effect is used to calculate the annual growth coefficient over the years, flattening the load during the Spring Festival period to the same range. Based on the absolute value difference between the number of days on the day of the Spring Festival, a lunar indicator is constructed, and LightGBM is used for prediction. The prediction results show that the proposed method reduces $M_{\text {APE }}$ by 1.66 percentage points compared to the traditional algorithm. R2 increased by 11.5%.