In order to solve the inconsistency of the traveling wave waveforms between primary and secondary sides in the power grid caused by the decay and noise interference in transmission process of the traveling wave sensor, an accurate detection method of fault voltage traveling wave combining the truncated singular value decomposition (TSVD) regularization and discrete deconvolution is proposed. This method adopts the transfer function model based on lumped parameters to construct a forward model of voltage traveling wave sensor, and the principle of deconvolution is used to establish the inversion model of traveling waves. Aiming at the ill conditions generated during the inversion of traveling wave waveforms, a waveform inversion method combined with TSVD regularization theory is proposed to realize the accurate detection of fault voltage traveling wave. The simulation shows that the detection method can realize the accurate inversion of the secondary traveling wave signals, and the waveform characteristics of the primary traveling wave signal obtained by the inversion can reflect the real fault waveform characteristics.