Insulation deterioration of electric equipment is often accompanied by partial discharge （PD）.How to improve the detection accuracy of the partial discharge of power transformer is of great significance to the study of its insulation state.The ensemble empirical mode decomposition（EEMD）has been applied very well during the de-noising period of partial discharge.However

the key is how to accurately select intrinsic mode functions（IMFs）during the decomposition of EEMD and restore the real signal of PD.In view of this problem

a kind of PD de-noising method of power transformer of EEMD based on correlation coefficients is proposed in this paper.The method is based on PD signals which are simultaneously collected by two circuits.Their PD signal and the background signal are separated respectively in accordance with the abnormal detection calculation.The background noise component is filtered out from the IMF component of PD through the comparison of the threshold of the related coefficient and

at the same time

the unwanted IMF noise component introduced during the process of EEMD calculation is filtered out.It is shown by the test the method can achieve better de-noising effect

thus providing accurate discharge signal for the subsequent analysis of PD and mode identification of the PD of the power transformer.