WAVELET-BASED PARTIAL DISCHARGE SIGNAL ANALYSIS FOR CONDITION MONITORING OF HIGH-VOLTAGE POWER CABLES

Abstract

Partial discharges (PDs) are widely recognized as early warning indicators of insulation deterioration in high-voltage power cables, making their timely detection essential for maintaining power system reliability. In this study, a wavelet-based signal analysis framework was presented to enhance the detection of weak PD activity under noisy measurement conditions. Simulated PD signals were generated as Gaussian-modulated sinusoidal pulses with a central frequency of 300 kHz and a pulse width of 50μs, sampled at 2MHz. To reflect realistic operating environments, the signals were further corrupted with 30% additive white Gaussian noise. Both Continuous Wavelet Transform (CWT) and Wavelet Packet Transform (WPT) were employed to extract detailed time–frequency features from the PD signals, while Discrete Wavelet Transform (DWT)–based denoising was used to suppress noise and improve signal clarity. The denoising process led to a noticeable improvement in signalto-noise ratio, allowing clearer visualization and identification of individual discharge pulses.
Quantitative analysis showed that the denoised PD signals exhibited an average normalized pulse amplitude of 0.48, a mean signal energy of 0.0.0007, and a total envelope energy of 5.4204. About 16 distinct PD events were successfully detected within the 2ms observation window. Frequencydomain energy analysis further revealed that PD activity was predominantly concentrated in the 250–400 kHz band, which aligns well with typical discharge behavior observed in high-voltage cable insulation. The findings confirm that wavelet-based techniques provide a robust and noise-resilient approach for PD feature extraction, offering strong potential for insulation condition assessment and early fault detection in high-voltage power systems