To study the dynamic changes of soil moisture, the gray correlation and multiple linear regression analysis were used based on hydrological meteorological and soil water measurements data from 1989 to 2015 at Wudaogou Hydrological Experiment Station, and Hydrological and meteorological factors and soil moisture prediction models in different soil layers at each growth stage of winter wheat were established. The results indicated that the correlation degree between soil moisture and meteorological factors in different soil layers at the same growth stage was the same. The correlation degree between soil moisture and temperature and groundwater depth was the strongest, with correlation coefficient 0.92 and 0.95, respectively. During the tillering-overwintering period, the correlations between soil moisture and groundwater depth, sunshine hours was the strongest, while the correlation degree between soil moisture and temperature, groundwater depth was the strongest in other growth stages. The fitting of soil moisture by hydrometeorological factors was relatively high (R2=0.94). The soil moisture calculation models have different prediction ability in different soil layers at different growth stages (R2= 0.80).This study provides a basis for the implementation of irrigation plans at different stages of crop growth.