Sparse measurements challenge fault location in distribution networks. This paper proposes a method for asymmetric ground fault location in distribution networks with limited measurements. A virtual injected current vector is formulated to estimate the fault line

which can be reconstructed from voltage sags measured at a few buses using compressive sensing (CS). The relationship between the virtual injected current ratio (VICR) and fault position is deduced from circuit analysis to pinpoint the fault. Furthermore

a two-stage recovery strategy is proposed for improving reconstruction accuracy of the current vector

where two different sensing matrixes are utilized to improve the incoherence. The proposed method is validated in IEEE 34 node test feeder. Simulation results show asymmetric ground fault type

resistance

fault position and access of distributed generators (DGs) do not significantly influence performance of our method. In addition

it works effectively under various scenarios of noisy measurement and line parameter error. Validations on 134 node test feeders prove the proposed method is also suitable for systems with more complex structure.