Decision tree-based fault line selection method for mine power systems using transient polarity and 5th harmonic direction dual criteria

Abstract

To address the difficulty of single-phase grounding fault line selection in mine resonant grounding systems, this paper proposes a decision tree-based fault line selection method integrating transient polarity and 5th harmonic direction dual criteria. The transient polarity characteristics of zero-sequence currents and the steady-state directional characteristics of the 5th harmonic are analyzed. The polarity correlation coefficients and 5th harmonic directional cosine values of each feeder are extracted to construct an 8-dimensional feature vector. A decision tree is then employed to achieve adaptive fusion of the dual criteria, establishing a physically transparent classification model. A 10 kV mine resonant grounding system simulation model is built in MATLAB/Simulink with four feeders and various transition resistances and fault angles for verification. Results show that the method accurately identifies fault lines across the full range of transition resistances (0.1 Ω–500 Ω) and fault inception angles (0°–90°), achieving 100% test accuracy. Compared with black-box models such as Random Forest and SVM, the decision tree model offers concise rules and strong interpretability, with its splitting rules directly corresponding to the physical mechanisms of transient polarity reversal and 5th harmonic direction inversion, facilitating on-site understanding and deployment.