Variable rate irrigating hill-seeder can automatically adjust the amount of the irrigating water according to the variation in soil moisture content in the field to alleviate drought and save water. Mathematical models for minimizing water application and determining the control time of the electromagnetic valve are developed. Static experiment is conducted with the developed variable rate irrigating hill seeder. Changing water amount in tank and irrigating time, the water amount ejected from the electromagnetic valve varies. The mathematical models for the control time of the electromagnetic valve are established based on the relations among water amount in tank, irrigating time and water amount supplied using linear regression and artificial neural networks. Regression results show that the adjusted coefficients of determination for the linear regression models are 0.992, 0.991 and 0.998 when the tank is 100% full, 1/2 full and 1/8 full, respectively. A 2-5-1 artificial neural network is used to predict irrigating time. The network provides high prediction accuracy with a maximum deviation of 3.95ms and an average deviation of 1.46ms.