A novel protection method for a wind farm collector line based on FCM clustering analysis

Abstract

To solve the problem of incorrect collector line protection in large-scale wind farms, we propose a protection method based on fuzzy c-means (FCM) clustering analysis. The intrinsic relationship between the fault current and the voltage sag of a doubly-fed induction generator (DFIG) was analysed under fault conditions. According to the component characteristics contained in the DFIG fault current, the DC, AC power frequency, and rotation frequency under different operation conditions were extracted to form the historical data sample set. The historical feature samples were divided into either the non-fault or fault category by the FCM algorithm in the multi-dimensional fault space, thereby obtaining the clustering centres. The distance similarity criterion was used to determine the attributes of the real-time feature sample to evaluate its protection. A centralized large-scale wind farm model was established in DIgSILENT to simulate various fault and operation conditions of the collector line. The simulation results showed that the proposed protection method could trip the fault collector line accurately by considering DFIG feeding current under different fault conditions. The proposed protection method improved the relay adaptability under variable wind farm operation conditions.