In distribution network, fault data self-synchronization method which takes the fault instant for reference can realize fault data self-synchronization without external time identifying facility, which makes a low investment. And it is also little affected by data communication channel or channel time delay. However, the precise detection of fault instant relates to the self-synchronization error directly. Therefore, a fault instant detection scheme for fault data self-synchronization method in distribution network is proposed. In this scheme, fault data is used to form Hankel matrix and Singular Value Decomposition (SVD) is used to decompose signals. In this way, detail signals reflecting singularity can be gained. And then singular point, i.e. fault instant, can be obtained by modulus maximum method. The simulation results verify that this scheme proposed can effectively reduce synchronous error caused by different delay of fault detection and has a high stability. This work is supported by Science and Technology Project of State Grid Corporation of China “Study on the Relay Protection Techniques in Distribution Network with High Permeability of Distributed Generation” and Science and Technology Project of State Grid Jiangsu Electric Power Co., Ltd “Research and Demonstration of Fault Location Technology with High Efficiency based on Multi-source Information”.