With the widely use of cooling equipment such as air conditioner, the influence of cooling-load on security and economy operation of power system is increasingly obvious. According to the historical load data and temperature data, this paper analyses the correlation between cooling-load and temperature in summer in Guangzhou city. First, considering the daily increment of base load in summer, daily base load curve of power grid in summer is predicted by using GM(1,1) model, and then the cooling-load curve in summer is separated accurately, and the ‘W’ changing feature of daily cooling-load curve is also analyzed. Second, based on the correlation analysis between daily maximum cooling-load and daily maximum temperature, this paper establishes piecewise regression model between them, and predicts the daily maximum cooling-load. Finally, considering the influence of accumulative effect of temperature, this paper improves the piecewise regression model to further increase the prediction accuracy, which can provide basis to accurately predict the daily peak load in summer of power grid.