Abstract: The frequent occurrence of leakage faults due to incorrect wiring of neutral and ground wires in low-voltage distribution networks is a major reason for the difficulty in deploying leakage protection in the distribution area. The tracing of leakage faults relies on regional power outages following the tripping of protective devices and the experience level of maintenance personnel, resulting in an ambiguous scope and low efficiency in fault troubleshooting. Based on the synthetic characteristics of current phasors in the distribution area, the shortcomings of the linear regression identification method for wiring faults are analyzed. Utilizing the multi-source electrical data provided by the smart electric meter in the distribution area and combining with physical constraints, a multivariate regression model for wiring faults in the complex domain is constructed. The primal-dual interior point method is used to iteratively calculate the optimal complex correlation coefficient of each user's load current with respect to the residual current in the distribution area, which accurately locates and identifies abnormal users with wiring faults. Further comparison experiments with the amplitude multivariate regression identification method show that the proposed method significantly outperforms the amplitude multivariate regression in terms of identification reliability under various fault scenarios, and can still effectively distinguish abnormal users in complex fault scenarios involving wiring faults of multiple users.