An Improved Aggregated Equivalent Modeling of DFIG Wind Farm Based on Dynamic Clustering Strategy for Post-fault Analysis

Abstract

Research communities have made significant efforts to derive dynamic equivalent models (DEM) for wind farms due to the high penetration of wind energy. The whole development can be divided into two categorical approaches: aggregated DEM and multi-machine pre-clustered DEM. However, with the stochastic characteristics of wind speeds and different crowbar activation patterns inside the wind farm resulted from different fault types, conventional DEMs may not be accurate for all kinds of faults, therefor are not suitable for post-fault analysis. This paper focuses on developing a DEM for LVRT studies. Firstly, a two-level clustering algorithm was proposed to build dynamic clusters based on the crowbar triggering waveforms and the grouping indicators' data, where Gaussian density distance clustering algorithm was applied. Then the established aggregated DEM was updated according to the previous clustering results. Different fault scenarios were tested. Compared with conventional methods, the simulation results showed that the proposed method could be more accurate to represent the dynamic response of the wind farm when there were several different triggering patterns from crowbar systems for post fault analysis.