To solve the problem of poor fault location accuracy caused by insufficient high-speed communication measurement points and complex fault characteristics in DC distribution networks, a new DC two-pole short-circuit fault location algorithm is proposed. First, a high-frequency impedance equivalent model containing a level shifter and a DC/DC converter is constructed to provide a stable impedance value for the fault process according to the high-frequency transient current loop. Secondly, the BCS theory is used to derive the high-frequency transient voltage equation of the node corresponding to the sparse measurement point. Finally, the node's high-frequency transient current sparse vector is determined to achieve fault location, combining the node's high-frequency transient voltage equation and Bayesian compressive sensing theory. The experimental results show that the proposed algorithm has a low requirement for the number of measurement points, does not need to measure data strictly synchronously, and is not affected by the control strategy and transition resistance of the converter, and has high fault location accuracy. This work is supported by the Key Research Project of Hebei Provincial Science and Technology Department (No. SCG2017442) and the Science and Technology Project of State Grid Hebei Electric Power Co., Ltd. (No. SGHEHS00FCJS1900440).