Faulty feeder detection of single phase-to-ground fault for distribution networks based on improved K-means power angle clustering analysis

Abstract

High-resistance ground faults often occur in distribution networks, and the fault current can be as low as 0.1A, making it extremely difficult to realize faulty feeder detection. The application of traditional faulty feeder detection methods based on single fault characteristic is limited. Thus, a method of single-phase grounding fault feeder detection based on improved K-means power angle clustering analysis was proposed. First, collect various fault characteristic quantities of each feeder under different operation and fault conditions, and establish a historical sample database. Then, the improved K-means clustering algorithm is used to quickly classify the historical sample database, and the cluster center is optimized through the spatial density and maximum neighbor distance principle. Finally, the power angle similarity is calculated according to the relative position of the real-time feature sample and the cluster center in the fault space, so as to realize the sensitive detection of the fault feeder. The PSCAD simulation and the 10 kV distribution network test field showed that the proposed method effectively eliminates the misjudgment of a single fault characteristic caused by external interference. And without a fixed setting value, the accurate fault feeder detection of intermittent arc grounding, high resistance grounding and other fault conditions can be realized.