A de-noising method is presented for partial discharge（PD） online monitoring of transformers on the basis of an improved wavelet de-noising approach and a pulse-sequence analysis method. Dual couplers are used for detecting the PD pulse current flowing through the high voltage bushing of each phase and the grounding wire connected with the measurement terminal of the high voltage bushing. The white noise can be suppressed by the improved wavelet approach while the extracted pulses have small distortion. The extracted pulse sequences of each coupler from white noise contain the PD pulses and pulse-shaped noises. To separate the PD pulses from noisy pulses， the pulse-sequence analysis on the relationship between two pulse-sequences from the dual couplers is performed. The experimental signal processing results show that the measured partial discharges inside the transformers can be separated from white noise and partial discharges occurring on the conductors connected with the high voltage bushings.