Research on fault localization of voltage transformer monitoring device based on wavelet transform

Introduction Voltage transformers (VTs) are critical to power system stability, with dielectric loss anomaly, capacitor breakdown, and compensating reactor short circuit accounting for 68% of total failures. Traditional Fourier transform-based methods lack sufficient time-frequency localization capability to detect transient fault signatures, necessitating a precise, standard-compliant fault time localization technique. Methods A fault time localization algorithm based on Daubechies 4 (db4) wavelet transform was proposed. The Mallat algorithm was used for 5-layer decomposition of 2000 Hz sampled normal and faulty voltage signals. The third detail layer (D3, 125–250 Hz) was selected as the optimal frequency band, with adaptive threshold detection applied to identify fault onset and termination. The algorithm was validated using 100 groups of field data from State Grid Ningxia Electric Power Co., Ltd., and compared with Sym4 wavelet and FFT-based STFT methods. Results The D3 layer exhibited the highest sensitivity to transient fault features across all three fault types, with a signal-to-noise ratio 28.6%–52% higher than other layers. The proposed method achieved an average time localization error of