There is a large noise signal in the partial discharge (PD) signal of the switch cabinet.The traditional FIR filter will cause the phase shift of the filtered signal.Thus,the correct rate of fault identification using power spectrum is considered.To solve the problem,a PD fault identification method using zero phase error filtering (ZPEF) method is proposed.Firstly,PD signals are collected by high speed current transformer.Then,ZPEF is used to filter the noise in the discharge signal.Secondly,the power spectrum of the discharge signal is analyzed by FFT transform.Two bands power spectra of 1.0~1.5 MHz and 2.5~3.0 MHz are used to construct the eigenvectors.Finally,least squares support vector machine (LS-SVM) is used to identify PD faults.Experimental results show that the recognition rate of PD fault can be improved by using ZPEF.A corollary is also drawn:ZPEF can improve the speed of fault identification by ensuring the same fault recognition rate.