Differential protection based on fault time self-synchronization principle requires to detect fault time accurately in distribution network. However, the error in the conventional phase current sudden-change-based fault time detection method is greatly affected by the fault occurrence time, distributed power generation fault characteristics and other factors. These can cause a large synchronization error in the differential protection. Therefore, an accurate fault time detection method is proposed, one which is suitable for the fault time self-synchronization principle in a distribution network. First, a data window is selected based on a conventional phase current sudden-change detection algorithm, and then an improved multi-resolution morphological gradient operator is used to process the data to obtain the gradient signal. This highlights the waveform mutation. Then the fault time is extracted through extreme point positions of the gradient signal. Finally, PSCAD simulation and field test data show that this method has high detection accuracy and good stability, and can effectively reduce the synchronization error caused by fault time detection. This work is supported by the Science and Technology Project of the Headquarters of State Grid Corporation of China (No.5100-202055018A-0-0-00).