In the tea industry, achieving a high winnowing accuracy to produce high-quality tea is a complex challenge. The complex shape of the tea leaves and the uncertainty of the flow field lead to the difficulty in determining the wind selection parameters. The purpose of this paper was to determine the accurate wind selection parameters of tea through simulation and improve the precision of tea wind selection. This study used three-dimensional modeling to establish a high-precision simulation of dry tea sorting. The simulation environment of the tea material, flow field, and wind field wall were defined using a fluid–solid interaction method. The validity of the simulation was verified via experiments. The actual test found that the velocity and trajectory of tea particles in the actual and simulated environments were consistent. The numerical simulations identified wind speed, wind speed distribution, and wind direction as the main factors affecting the winnowing efficacy. The weight-to-area ratio was used to define the characteristics of different types of tea materials. The indices of discrete degree, drift limiting velocity, stratification height, and drag force were employed to evaluate the winnowing results. The separation of tea leaves and stems is best in the range of the wind angle of 5–25 degrees under the same wind speed. Orthogonal and single-factor experiments were conducted to analyze the influence of wind speed, wind speed distribution, and wind direction on wind sorting. The results of these experiments identified the optimal wind-sorting parameters: a wind speed of 12 m s−1, wind speed distribution of 45%, and wind direction angle of 10°. The larger the difference between the weight-to-area ratios of the tea leaves and stems, the more optimized the wind sorting. The proposed model provides a theoretical basis for the design of wind-based tea-sorting structures.