A Multi-label Classification Model for Type Recognition of Single-Phase-to-Ground Fault Based on KNN-Bayesian Method

In non-solidly earthed distribution networks, single-phase-to-ground fault (SPGF) greatly threats the safety of person and equipment. Though the existing fault line selection and location techniques have made great contributions in reducing the harm of SPGF, certain amount of power loss still exists in SPGF because of low efficiency in detecting and repairing faults of current techniques. Accurate type classification of SPGF could help reveal the fault nature from different views and further improve the fault repair efficiency. In this paper, a multi-label classification model for recognizing types of SPGF is proposed, In the proposed model, SPGF are classified respectively, and corresponding symptom features are determined according to the following five aspects: time-domain continuity, time-domain stability, volt-ampere characteristics of transition impedance, transition impedance size and fault point medium. In addition, a multi-label classification architecture for SPGF is constructed with an 8-dimension feature space and a 14-label fault type space. Finally, a KNN-Bayesian method is designed to solve the multi-label classification problem. The feasibility and advantages of the proposed model and methods are verified by the field data and the comparison with KNN method.