To solve the problem that feeder-consumer connectivity is difficult to be efficiently identified and accurately checked due to the monotonous and large-scale measurement data, a method of identifying feeder-consumer connectivity in low-voltage distribution network based on multivariate data feature is proposed. Firstly, the correlation of outage equipment is analyzed, and the abnormal data preprocessing method based on fuzzy C-means algorithm, threshold division and Neville interpolation is proposed for smart meter sampling anomalies. Secondly, a clustering method of smart meters based on outage correlation and Hausdroff distance is proposed. Thirdly, based on Kirchhoff's current law, a quadratic programming model for the identification of feeder-consumer connectivity is established, which is effectively solved by a solver after transformation. Finally, the effectiveness and superiority of the proposed identification method of feeder-consumer connectivity are verified by a practical example.