Abstract: Aiming at the problems of the traditional cable defect location methods based on frequency domain reflection,such as poor adaptive adjustment ability and multiple false defect signals,this paper proposes a method based on multi-stage singular value decomposition(MSSVD)and multiple signal classification (MUSIC)for cable defect location.Firstly,the broadband impedance spectrum (BIS)model of power cable is established based on the transmission line theory,and the principle of locating defects based on phase spectrum is introduced.Then,the phase spectrum in BIS is separated into component signals dominated by different defect information through MSSVD.By decomposing and projecting the feature space of each component signal through MUSIC,the power spectrum curves of each component signal are obtained.Finally,the defect localization spectrogram is obtained by normalizing and summing up the power spectrum curves of each component signal.To verify the universality and feasibility of the proposed method,the simulated cable models with different defect states and different measurement conditions are established in this paper.At the same time,the defect localization result on a 700m 10kV actual power cable is studied,which shows that the proposed method can accurately locate the defect in actual cables,and the relative positioning error is 0.31%.Compared with traditional cable defect localization methods,this proposed method has a stronger adaptive adjustment ability and can effectively reduce the reflected signals of false defect positions,improving the accuracy of the defect localization