The analysis and classification are conducted on the error sources in data assimilation, and according to the categories and the demands of data assimilation, the matching quality control method is developed under the strict strategy For one week data selected from the area of Beijing, Tianjin and Hebei Province of new type Automatic Weather Stations (AWS), including pressure, temperature, humidity, and wind, numerical experiments are carried outThe results show that the method is applicable to reduce rough errors and representative errors effectively, keep the normal distribution characteristic of random errors, and lower the root mean square error between observation and background significantly