Abstract: The aim of this project is to study and explore the correlation between the load and meteorological factors in Quzhou City, and to develop corresponding digital management methods to improve the safety and economic operation of the power system. The research background is based on the frequent occurrence of extreme weather events caused by global climate change, which have had a significant impact on the power system, especially in the Quzhou area. The research method involves multi-source data fusion, including electricity data and meteorological data, and handles outliers and missing values through data preprocessing. In terms of algorithm and model design, the LightGBM algorithm was adopted, and the model was trained using 5-fold cross validation and early stop methods, combined with Bayesian methods for optimization. The research results show that there is a saddle shaped correlation between the load and temperature in the substation area, that is, the load reaches its peak in the temperature range of too high or too low. The impact of wind speed on load is mainly significant in winter, and there is also a certain correlation between weather type and load distribution. Based on these findings, a big data model was established, and through model prediction and early warning monitoring, the number of heavy overloads in the platform area during the Spring Festival in 2024 is effectively reduced.