The effective de-noising of partial discharge（PD） array ultrasonic signal is important for improving the partial discharge ultrasonic array detection. To avoid impacting array signal phase and amounts of calculation caused by the array signal denoising methods

we apply the matching pursuit algorithm to PD ultrasonic array signal de-noising. Moreover

we decompose and denoise each array element receiving signals on over-complete dictionary. There will be little phase distortion in the denoising process and only real signal will be extracted. The denoising results are evaluated quantitatively by the denoising evaluation index. The simulation shows that the direction-finding error of signal before and after denoising is more than 30° and less than 2°

respectively

and that the waveform evaluation index after denoising is available. The experiment show that the direction-finding error of signal before and after denoising is（6.525°

6.863°） and（2.9°

3.038°）

respectively.The diagrams of the arrival spectrum directions and contour maps indicate that signal after denoising has a single spectral peak

which is the proof of the energy concentration. The results of the simulation and experiment show that the matching pursuit algorithm can effectively improve the signal-to-noise ratio and remove noise in the partial discharge array ultrasonic signal.