Abstract: In response to the difficulties of low accuracy of equivalence modeling and insufficient clustering basis under dynamic operating conditions of wind farms,a wind farm equivalence modeling method based on the idea of Gaussian mixture model clustering is proposed. First,the dynamic response characteristics of a single doubly-fed induction wind turbine during LVRT are analyzed,and the clustering indexes are constructed based on the clustering characteristics of the response characteristics. Then,a two-stage equivalence modeling method based on Gaussian mixture model with dynamic preliminary clustering and optimized number of clusters is proposed,and an optimization search algorithm for the number of clusters under the criteria of red pool information and Bayesian information is derived. Simulation tests with different fault ride-through conditions are carried out in Matlab/Simulink platform for a typical mediumscale wind farm,and the results show that the proposed equivalence modeling method for wind farms is effective in clustering with high accuracy.