Narrow-band interference suppressing of transformer on-line insulation monitoring

In partial discharge (PD) on-line measurement of transformer insulation monitoring there is always a lot of interference, of which periodicity narrow-band interference demonstrates strongly in both time domain and frequency domain, so it must be filtered in advance. The usual adaptive filter is one of the best algorithms in suppressing narrow-band interference. But it is difficult in setting the parameters of the adaptive filter due to wide frequency range of narrow bandwidth noise, so some improved algorithms can´t gain better result. Based on the research of the threshold value of frequency domain and Wavelet filtering method, an improved algorithm is presented in which the narrow-band interference can be suppressed effectively in a wider frequency band by use of least square support vector machine (LS-SVM) regression. Finally this improved algorithm is used in interference suppressing of factual PD signals, the result shows that the method is effective.