With the high proportion of low-voltage distributed PV access, traditional transformer-customer relationship identification methods based on voltage and power characteristics are largely affected by photovoltaic input. Given the electrical characteristics of the low-voltage distribution network when there is large-scale distributed PV access, a data-driven identification method for transformer-customer relationship in the low-voltage distribution network is proposed. First, using voltage correlation, customers are preliminarily divided based on an improved fuzzy C-means (FCM) clustering algorithm. Then, based on the initial clustering, a minimum mutual information algorithm is used to eliminate the influence of photovoltaic output, and an identification model of the connection relationship between distribution transformers and customers is established. Regression analysis is used to solve the problem, so as to realize accurate identification. Finally, the effectiveness of the identification method of transformer-customer relationship based on improved FCM algorithm and minimum mutual information is verified by analyzing the actual data of a certain place. Experiments show that this method has higher accuracy than the traditional discrimination method using voltage and power characteristics.