The daily power peak loads are highly correlated with weather conditions. Base on the analysis of daily meteorological data and power peak load data in the summer (from 1 May to 30 September) of 2002 to 2004 in the Beijing area, the temporal characteristics of the power peak loads are studied. The power peak loads in the Beijing area have obvious seasonal and weekly variation trends. By using the regression and artificial neural network methods, the power peak loads are simulated and forecasted with weather data. The results show that these methods can be used to forecast the daily power peak loads, and the artificial neural networks method is better.