A grouping method of wind turbines for the power system analysis is presented， based on the coherency analysis of the input wind speeds for individual wind turbines， and the influence of wind direction random change on the wake effect. A 3-dimension of coherency coefficient matrix of wind turbines is used to group turbines in a large wind farm. The research shows that the wind direction is a dominating factor in the grouping method of wind turbines. The wind turbine groups on 3-dimension of coherency coefficient matrix can be easily acquired with the wind speed and direction random fluctuation， which lays the foundation for building stochastic model of wind farms and studying the characteristics of wind farms.