Deep application of nitrogen (N) fertilizer helps to increase crop yield in dryland area, but the effect of depth of N fertilizer on wheat yield in hilly and gully regions of the Loess Plateau under different precipitation types is still unclear. This study used the meteorological data from 1990 to 2020, to explore the effect of N application depth on wheat yield in different precipitation types based on APSIM model and mathematical statistical methods. The correlation coefficient (R²) of the simulated and observed value of wheat yield and biomass was greater than 0.9, the model validity index (M_E) was greater than 0.8, and the normalized root mean square error (NRMSE) was less than 0.2, indicating that the model had good applicability in this region. The NRMSE of simulated and observed soil moisture dynamics during the growth period and soil profile moisture content at the harvest stage of the wheat field were 0.05 ~ 0.07 and 0.13 ~ 0.29, respectively, demonstrating that the model could accurately simulate the wheat field soil moisture changes. The yield potential of deep application of N fertilizer was the highest in wet years, followed by normal years. Compared with drought years, increasing the depth of N application in wet and normal years under the same nitrogen application rate could improve the stability and sustainability of farmland productivity. When the nitrogen application level was 150 kg/hm², increasing the depth of N application in wet and normal years had no obvious effect on the stability of farmland productivity, but it helped to improve the sustainability of farmland productivity. The interaction relationship between N application depth and amount and yield under different precipitation types showed that the relationship between wet years and normal years was better, with R² of 0.76 (P < 0.05) and 0.61 (P < 0.05), respectively. Increasing the depth of N application to 20 ~ 23 cm could obtain the highest potential yield. In a drought year, R² was 0.29 (P > 0.05), increasing the depth of N application had no significant effect on wheat yield. The results of the study are expected to provide theoretical guidance for optimizing fertilization measures at the regional scale for wheat in the hilly and gully regions of the Loess Plateau.