Ultra-high frequency (UHF) technology has been widely used in partial discharge (PD) source localization based on the time difference of arrival (TDOA). However, the TDOA errors will then lead to the ill-conditioned problem of matrix ${A}$ . A larger condition number will then lead to a more ill severe condition of linear equations $\textit {AX}={b}$ , and resulting in a decline in the localization accuracy. In this paper, a novel method for measuring the PD source has been developed based on the UHF antenna and the truncated singular value decomposition (TSVD) regularization method. First, the PD signals are obtained through the cooperation of five UHF antennas. Second, a spherical model linear transformation method is proposed for non-linear measurement equations. Then, to address the ill-conditioned problem of the measurement equations caused by the UHF antenna installation position, the linear measurement equations are improved with the pretreatment method. Finally, the TSVD regularization algorithm is applied to solve the UHF measurement equations with the ill-conditioned problems. The regularization parameter of the TSVD is then optimized by generalized cross-validation (GCV). In addition, the algorithm is verified by theoretical verification and field tests. The experimental results reveal that the proposed localization method can accurately locate the PD source.