Abstract: For a long time, When single-phase ground fault in the distribution network dominated by 10 kV("distribution network "refers to 10 kV distribution network).Since 10 kV distribution network is a small current grounding system in which the neutral point is not grounded or grounded via the arc suppression coil.Judging the ground fault line by the zero sequence current, voltage or harmonics observed at the side of the substation is the mainstream ground fault selection principle at present.There are two main problems in this method: one is the existence of high resistance ground fault, which makes it impossible to achieve 100%correct rate of line selection; the other is that it is impossible to specify which section of the line is grounded.As a result, the ground fault lines are searched in line and wasted a lot of manpower and material resources.The judgment of ground fault line based on the observed electrical capacity of substation is based on the view of main network.But the ground fault is on the distribution network.Therefore, we should use the vision and resources of the distribution network to find the ground fault of the distribution network.That is to rely on distribution network big data for the search of the distribution network grounding fault, and even precise positioning.This article will explore this.