Fault Equivalent Modeling of Doubly Fed Wind Farm Based on Improved K-means Algorithm

The methods for the fault modeling and analysis of the renewable energy stations are the basis of the fault characteristics and protection researches in the wind power and photovoltaic centralized grid connected areas. The multi-machine equivalent modeling in the stations mostly adopts the clustering algorithm. However, the traditional algorithm clusters based on the similarity of the operating states between the units, but classifying the units with high similarity into one class is not equivalent to higher accuracy of the model. Therefore, this paper first analyzes the influencing factors of the fault current of the doubly fed induction generator (DFIG) and takes it as the clustering index to propose an improved K-means algorithm. This algorithm takes the equivalent errors of the fault current of the units as the sample spacing, and the equivalent electrical quantity of the units divided into the same kind as the clustering center coordinate of the algorithm; Combined with the equivalent errors of the DFIG and the contour coefficient of the clustering algorithm, the number of the clusters of the algorithm is adjusted, and the multi-machine equivalent model of the doubly fed wind farm is constructed to improve the efficiency and accuracy of the network short-circuit calculation. The simulation results verify the rationality of the clustering criterion proposed in this paper. Compared with the single machine equivalent model and the multi- machine equivalent model obtained by the traditional K-means algorithm, the accuracy of the proposed method is improved to a certain extent. © 2023 Power System Technology Press. All rights reserved.