Dynamic ground fault location algorithm for distribution networks based on frequency-energy feature analysis
Artificial intelligence-driven signal feature analysis methods offer new technical pathways for fault location in power distribution networks. Addressing the accuracy limitations of traditional localization methods under weak ground fault conditions, this study constructs a dynamic ground fault location model for distribution systems based on frequencyenergy feature analysis. By modeling time-series data from multi-node sampled signals, the Fast Fourier Transform (FFT) is employed to extract energy characteristics across different frequency bands and form a frequency-energy feature vector. Building upon this, a feature-distance-weighted dynamic localization algorithm is designed, incorporating a multi-node time-series coordination mechanism to update location results. Tests conducted on a 10 kV, 20 km distribution line simulation system compared the proposed method with impedance and traveling wave techniques. Results show an average positioning error of 0.19 km—lower than the impedance method's 0.42 km and traveling wave method's 0.27 km—while achieving high accuracy within 0.038 s computation time. This validates the effectiveness of frequency-energy feature analysis for distribution ground fault location.