Abstract: In order to establish a basic meteorological database for energy saving of buildings in China, it is necessary to analyze the effect of different record periods on selection results of typical meteorological years. In this paper, 11 typical cities were selected which represended mateorolgical states of the 11 sub divisions of the five building thermal design divisions. The Sandia National Laboratory method was used to compare the differences of shortened record periods (3 to 15 years) with the long-term records of 30 years in typical meteorological years. The results showed that to meet the needs of the average fluctuation range of standard deviation for solar radiation within 1 MJ/m² and the average fluctuation range of standard deviation of temperature within 1 ℃, it was necessary to use data at least for 8 years. However, depending on the location of the city and climate, there would still be large fluctuations in standard deviations of temperatures or solar radiation in some cities