Abstract: Based on meteorological and electricload data of Suining from 2021 to 2023, the meteorological load component was extracted from the daily maximum electrical load, establishing the correlation between meteorological load ratio and meteorological factors in Suining, the graded forecasting models of meteorological load rate were established by stepwise regression, multiple linear regression and BP neural network. The results show that several meteorological factors, including temperature and atmospheric pressure, significantly affect electricity load across different seasons. especially in summer and winter, the relationship between the meteorological load rate and temperature shows obviously nonlinear characteristics. The multiple linear regression method provides the best predictive performance in most temperature ranges, better meeting the forecasting needs of the power department. In certain temperature ranges, the BP neural network method exhibits superior prediction accuracy, however, it suffers from issues of prediction instability.