Scientific irrigation scheduling is essential for agricultural production in arid and semi-arid regions. Located in the northwest arid region, the Heihe River is the second largest inland (terminal lake) river in China and faces increasing competition for water between the middle reach agricultural irrigation and lower reach ecosystem services. In this study, we use the Decision Support for Agro-technology Transfer (DSSAT) model to simulate the growth of four main crops:maize, wheat, rape seed and potato in the middle reaches of the Heihe River Basin. We first compared the temporal differences between the water demands of the four crops and the precipitation during the growing season as well as the differences between the crop water demands and the current irrigation scheduling. Subsequently, we explored multiple irrigation scheduling combinations during the growth period to optimize irrigation scheduling of the four crops. Finally, we calculated the water saving potential under the optimal irrigation scheduling. Results show after calibration and validation with in situ observations, the DSSAT model has better simulation performance for the four crops in the study region. The standardized root mean square error (nRMSE) of the crop yields is less than 15.0%, and the coefficient of determination (R²) is above 0.65. The annual average water deficit of the four crops ranged from 122.5 to 367.0 mm during the growth season. By adopting the optimal irrigation scheduling, the water use efficiency of maize, wheat, rape seed, and potato could be improved by 54.8%, 25.0%, 18.3% and 51.3%, respectively, and the variation of the simulated crop yield was all lower than 5.0%, achieving both high yield and water conservation. If the optimal irrigation scheduling is applied in the study area, potential water saving could reach to 8.1×10⁸ m³ annually in the middle reaches, which could be used to support downstream ecological protection.