Fourier series method on eliminating narrow frequency noise in partial discharge detection

A new method to eliminate periodical narrow frequency noises in partial discharge detection of electric power system apparatus, Fourier series method, is presented. It uses the difference not only on frequency domain but also mainly on time duration between signals and noises. The anti-interference effect of this method was analyzed by experiments data on field. Several factors that can make influence to the affect of this method, such as the error of frequencies of the noise, the data length by which the Fourier coefficients are calculated, the error of A/D conversion, were discussed respectively. It was found that this method can almost completely banish narrow frequency noises, at the same time can retain the amplitude and polarity of partial discharge pulses, even if the main frequency of the partial discharge pulses are close to that of narrow frequency noises. The affect of this method was also contrasted with normal method, such as digital filter, adaptive filter and wavelet. The result shown that this new method has obvious predominance at remaining the amplitude and polarity of pulse signals.