Abstract: The power supply grid directly serves a large number of medium and low voltage users,and the location and capacity of the power supply point is a key link in the optimization of the distribution network structure. This paper analyzes the spatial distribution and characteristics of medium and low voltage loads based on big data mining technology,and proposes a cluster-based power supply grid power point selection method. Firstly,based on the open source massive information,the spatial distribution coordinates of the distribution network load are accurately analyzed,and two indicators-the local density ρi and the sample spacing δi,are defined to describe the spatial distribution characteristics of the load;the clustering data mining algorithm is used to carry out cluster optimization until the load of the cluster set meets the constraints of reasonable power supply radius and appropriate load capacity,so as to determine the location and capacity of the power point in the area;the average power supply radius of the grid power point obtained by the clustering optimization method is the smallest,which realizes the power supply in the cluster. The power user line loss is the smallest,which reflects the economy of the power point location method proposed in this paper. Finally,this paper demonstrates the practicability of the proposed method in combination with the power grid planning of the actual area.