Dynamic equvalent model of the wind farm based on the k nearest neighbor

A clustering method based on k nearest neighbor is proposed for the dynamic equivalent of wind farm in this paper. After dimension reduction, this method can detect boundary points (BPs) of clustering parameters of wind turbines along the border of it. So the training samples and store space are reduced significantly. Initial clustering result is obtained according to the sequence of BPs. Final clustering result with pre-defined number of clusters or accuracy can be obtained by simply modifying the initial result. Consequently, the efficiency of clustering is enhanced greatly. The method is demonstrated on a wind farm consisting of 133 wind turbines. Dynamic responses of equivalent model of the wind farm are compared against the responses of the detail model. Simulation results demonstrate the efficiency of the method.