Partial discharge（PD）detection system is often affected by external interferenc e

and sometimes the partial discharge signals are submerged in noises.So the sig nals acquired must be preprocessed to obtain reliable partial discharge inform ation.The paper presents a method to fulfil automatic noise reduction.By selecti ng an optimal wavelet and processing high-frequency components using the soft th reshold method with an automatic thresholding rule

partial discharge signals are then reconstructed with modified detailed coefficients.Results illustrate tha t this method can extract characteristics of partial discharge signals effective ly and may be used in partial discharge on-line monitoring systems.