The summer maximum power load variation and its correlation with meteorological factors were analyzed by using the 15 min power load and the daily meteorological data in Wuhan from 2016 to 2018. Based on stepwise regression and double hidden layer BP neural network algorithm, the prediction model of summer maximum power load was established. The results show that there is a significant positive correlation between the mean temperature, average maximum temperature and average minimum temperature and the meteorological load. The correlation between the maximum load on the day and the load on the previous day is the best. The load on the day is most sensitive to the average temperature and comfort index in the previous two days. With historical load and meteorological data as joint forecasting factors, the stepwise regression and BP neural network algorithm have a good simulation effect on the maximum summer power load in Wuhan, especially on the maximum load of high-level operation caused by continuous high temperature in 2018. When the sensitivity is lower than 10%, the positive contribution of the meteoro-logical factor in the stepwise regression algorithm is less than the negative contribution, and the positive contribution in the BP neural network algorithm is higher than the negative contribution. But when the sensitivity is higher than 10%, the meteorological factors in both algorithms contribute positively.