Development of a Current Signal Based Model for High Impedance Fault Identification in Power Systems

Abstract

High-impedance faults (HIFs) generate low-magnitude, highly irregular arcing currents that closely resemble normal load behavior, causing conventional overcurrent-based protection to fail in their identification. This study proposes an adaptive current-signal correlation model designed to detect HIFs using time-domain waveform similarity analysis. The method utilizes a band-pass filtered current waveform, half-cycle window segmentation, and a correlation measurement against a reference pattern bank derived from varying ignition-angle scenarios. An adaptive threshold mechanism is introduced to improve robustness against noise, switching transients, and load fluctuations. The proposed model is validated through MATLAB/Simulink simulations and Real-Time Digital Simulator (RTDS) experiments, representing near-real operating conditions of distribution feeders. Results demonstrate a detection accuracy of 97.69%, false alarm rate below 1.5%, and a detection time within one cycle (20 ms). Compared to harmonic, wavelet, and ANN-based methods, the proposed algorithm shows superior speed, computational efficiency, and compatibility with existing current-based relay infrastructures. This approach enables practical field implementation without requiring additional sensors or complex feature extraction.