Partial Discharge Signal Detection in Generators Using Wavelet Transforms

This paper presents an algorithm developed to detect and locate Partial Discharge (PD) signals in generators using wavelet transform techniques. The signals obtained from the sensors in the generator are de-noised by Discrete Wavelet Transform (DWT). The original signal is decomposed into high and low frequencies to assist in the elimination of noise. The filtered signal is passed through a localisation process utilising the continuous wavelet transform (CWT), which will be shown in the time-frequency domain. The signals obtained at locations in the time domain can be analysed to determine their frequency content allowing a comparison of their characteristics to the standards of partial discharges in generators. The experimental results indicate that the proposed technique could efficiently decrease noise through application of the Daubechies 4 (Daub4) as a mother wavelet and lowering the signal to level 12. The results also show that this condition could lead to accurate performance in locating the time of the signal and determining the frequency of the PD signal via the CWT.