Noise Separation and Defect Recognition in AC Partial Discharge Pattern Signals Using Continuous Wavelet Transform

초록

Separating noise and partial discharge signals is essential for partial discharge analysis. However, the presence of conductive and radiative noise, such as switching pulses, electromagnetic interference, and network communication, makes this difficult in high-voltage systems. This paper presents an AC partial discharge pattern image analysis method using continuous wavelet transform and the element wise maximum method. This method combines the advantages of both frequency spectrum and AC partial discharge pattern analysis. Defect classification was performed using a convolutional neural network model under conditions with both noise and partial discharge signals. Additionally, explainable artificial intelligence technology was applied to visualize the regions influencing the pattern classification process. Extending the time-series partial discharge signals into the frequency domain effectively separates partial discharges from noise and classifies them under low signal-to-noise ratio conditions. © 2013 IEEE.

키워드