Two-Stage Equivalent Dynamic Modeling of DFIG with Crowbar-Based LVRT

With the popularity of wind power generation, the accuracy and speed of equivalent modeling for wind farms are challenged. In this paper, a two-stage equivalent method for wind farm considering the action state of Crowbar circuit is proposed. Firstly, the transient characteristics of Crowbar circuits of different units during the fault period are analyzed, and their action states are used as the preliminary clustering basis. An ISODATA clustering method for wind farms with small data sets is proposed. Secondly, it aims at the interval range of incomplete action of Crowbar circuit, and the adaptive clustering method based on Gaussian mixture model clustering algorithm optimized by Bayesian information is proposed. Finally, the equivalent model of wind farm under different voltage dip is obtained. Simulation examples show that the proposed method has high precision and effective clustering.