Empirical mode decomposition (EMD) is a signal processing method widely used in the field of fault diagnosis.It has been successfully applied in the field of fault diagnosis,But the decomposition results are easily interfered by high frequency noise.Therefore, this paper puts forward the method of using the wavelet packet to remove the noise and the fault arc signal processing by EMD decomposition.First,the fault arc experimental platform is built according to the national standard GB/T31143-2014,and the fault arc signal is collected,using wavelet packet to remove the noise in the signal, Then the EMD method is used to decompose the normal arc and the fault arc to obtain a series of IMFs,the energy entropy of these components is obtained by calculation and the eigenvectors are obtained by normalization,finally,support vector machine (SVM) training, and then the fault type arc fault diagnosis.The experimental results show that the method can effectively and accurately determine the fault arc.