To obtain the reasonable irrigation strategies under deficit irrigation in arid regions, an irrigation regime optimization method (AquaCrop-OSPy-SIA) by using the irrigation interval or the single irrigation volume as the decision variables was proposed based on the crop growth model (AquaCrop-OSPy) and the swarm intelligence algorithm. In order to ensure the simulation precision of cotton yield, the field experimental data under the different irrigation regimes and environmental conditions in Korla, Xinjiang were used to calibrate the parameters of the AquaCrop-OSPy model. Moreover, the grey wolf optimization (GWO), particle swarm optimization (PSO), whale optimization (WO), and harris hawks optimization (HHO) algorithms were applied to combine with AquaCrop-OSPy model to discussed the effectiveness of irrigation regime optimization. Under the same weather conditions, the total irrigation amounts were assumed to six groups ranging from 200 mm to 450 mm with 50 mm intervals. The optimal irrigation regimes of different total irrigation amounts, including the single irrigation volumes with the fixed irrigation interval and the irrigation intervals with the fixed single irrigation volume, were obtained by using the maximum yield as the optimization objective. The simulation results showed that AquaCrop-OSPy and GWO algorithm have a good performance in simulating the crop growth and optimizing the irrigation regime. The cotton yields of optimized irrigation regimes were improved obviously with a universal increase of 0.63-1.79 t/hm2 compared with the traditional irrigation regimes with the fixed irrigation interval and volume. Furthermore, the regimes of optimized single irrigation volume and optimized irrigation interval increased the cotton yields by 10.6 %-17.8 % and 19.9 %-27.5 % and reduced the irrigation amounts by 45-50 mm, respectively. Therefore, the proposed optimization method provided a feasible way to obtain the more precise, water-saving, and real-time irrigation plan than the traditional irrigation regime.