Abstract: Based on daily cloud cover data with 3 hours interval and corresponding NCEP data with 6 hours interval from July of 2001 to July of 2011 in Yuzhong of Gansu province, relationships between forecast factors built by the NCEP data and forecast object such as the total cloud cover and low cloud cover were analyzed. A series of monthly forecast equations of cloud cover with daily 8 times was established by a stepwise regression analysis method and were tested by back substitution, and then prediction effect was checked using data of 2012. The results show that cloud cover is mainly affected by the whole layer humidity, vertical velocity, instability energy, trough intensity index and divergence of moisture flux in 700 hPa, especially the first two elements. Forecast effect of total cloud cover is better than that of low cloud cover; average errors of total cloud cover and low cloud cover are about 20% and about 30%, respectively. Tendency of predicted values can partly reflect that of observed values.