Abstract: In order to meet the development requirements of wind resource assessment of typical terrains, wind farm micro-siting and wind power prediction, two wind profile models are proposed based on measured wind field data: one is the modified Gryning wind profile model; the other is the fitted wind profile model by using the piecewise function and theoretical formula. Considering the atmospheric thermal and geostrophic effects, and combined with the modified k \text- \varepsilon model for Reynolds-Averaged Navier-Stokes (RANS), an atmospheric CFD model is proposed to simulate the wind field of complex terrains under different atmospheric stable states, thus a CFD software of LSWS (large scale wind simulation) for the wind field simulation is developed. The simulation accuracy of the self-developed LSWS software is compared with that of the commercial software Fluent, based on the wind test data from the Xilinhot flat terrain and the Shanxi hills. The results show that the simulation data of the two softwares are in good agreement with the measured data of the Xilinhot flat terrain. The computational accuracy of the self-developed LSWS software is 6.33% higher than that of the Fluent software under stable atmospheric states. The computational accuracy of the self-developed LSWS software is significantly higher than that of the Fluent software, especially in simulating the wind field behind the Shanxi hills, and the computational accuracy is at most 40% higher than that of the Fluent software.