Abstract: During the fault process, crowbar activated and deactivated DFIG have a significantly different transient response characteristics. The crowbar status of units is a good indicator for cluster division in the wind farm equivalent. However, the fault characteristics between the doubly-fed induction generator(DFIG) based wind farm and a single unit are clearly different, leading to inaccurate identification in existing literature methods by using a single-unit model to determine the crowbar status of the units in the farm, thereby reducing the accuracy of wind farm equivalent modeling. Therefore, this paper proposed an equivalent modeling method of DFIG-based wind farm considering improved identification of crowbar status. By analyzing the fault characteristics of wind farm, the crowbar status feature vector was constructed. And the sample data of wind farm under various conditions were collected to establish the identification model based on support vector machine(SVM). For new conditions, the identification results of crowbar status and the input wind speed were used as the clustering index to divide the units in the field in turn, so as to establish the equivalent model of wind farm. Finally, the simulation results showed that the SVM-based crowbar status judgment method proposed in this paper has higher identification results than the traditional method in various fault scenarios; the established equivalent model and detailed model fault transient characteristics are highly consistent; and the equivalent method is reasonable and effective.