Abstract: Based on the monthly mean temperature data and electric power consumptions in Guangzhou, Hong Kong and Macao from 2000 to 2005, the monthly mean weather components of 3 cities are divided by linear regression method and a model between temperature and weather components is build up. It shows that the percentages of the monthly mean weather components of the cities are 12.9%,17.1% and 18.5%, respectively.When the monthly mean temperature is up to 24.43,24.80 and 24.52℃, the weather components is more sensitive to the change of mean monthly temperature. The 1℃ effect and correlation coefficients successively decease in hot, transitional and cold seasons in the cities.