Transient phase-based faulty feeder identification for high-impedance faults via autocorrelation entropy difference

Abstract

In resonant grounding distribution systems (RGDSs), high-impedance faults (HIFs) generate severely attenuated currents and complex arcing behaviors, making faulty feeder identification particularly challenging. Existing magnitude-based methods suffer from significant performance degradation, while conventional phase-extraction techniques are vulnerable to noise and intermittent arcing. To address these limitations, a robust feeder identification method utilizing a novel autocorrelation entropy difference (ACED) index is proposed. First, based on equivalent models for both fixed-resistance and arcing HIFs, the initial phase disparity of zero-sequence transient currents inherently persists, remaining structurally decoupled from the extreme attenuation caused by HIFs. To circumvent the unreliability of direct phase estimation under low signal-to-noise ratios (SNRs), the ACED index is formulated to characterize the structural correlation discrepancies of transient signals, achieving robust feature extraction without relying on explicit phase calculation. Simulation and field-test results demonstrate that the proposed method maintains satisfactory performance under extreme fault resistances, nonlinear arcing, and diverse operating conditions. Specifically, under a SNR of 10 dB, the proposed method achieves an identification accuracy of 84.4%, outperforming existing methods, and reliably identifies HIFs with fault resistances up to 3000 Ω, demonstrating enhanced noise robustness and practical potential.