High Impedance Fault Location for 35 kV Feeders Based on SPWVD and Intelligent Image Recognition

Abstract

High-impedance grounding faults (HIFs) in 35 kV distribution networks generate extremely weak fault signatures, making traditional traveling-wave-based localization methods prone to large errors or failure. To address this challenge, this paper proposes a fault location method that integrates smoothed pseudo Wigner-Ville distribution (SPWVD) with an intelligent image recognition framework. The zero-sequence current signal is transformed into a time-frequency spectrogram via SPWVD, which suppresses cross-term interference while preserving high energy concentration. An improved projection analysis is then applied to accurately extract the traveling wavefront arrival time. The fault distance is calculated using the double-ended traveling wave principle. A 35 kV resonant grounded distribution network model is built in PSCAD/EMTDC, and simulations are conducted under various fault distances, transition resistances (up to 3000 Ω), fault inception angles, and noise levels. Results show that the maximum positioning error is 62.11 m under noise-free conditions and 109.26 m under 40 dB noise. This work applies intelligent image recognition to improve new power system protection and control, confirming its superior accuracy and robustness.