Distribution network station-area relationship identification has important applications in many fields. If the relationship is inaccurate, it will affect a series of problems, such as the success rate of data acquisition, line loss analysis, blackout area prediction, three-phase unbalanced governance, and so on. These problems are closely related to the interests of power enterprises and users. Based on the data collected by smart watt-hour meter, this paper proposes a new algorithm for the identification of desktop. Compared with the traditional algorithm, this algorithm reduces the amount of data calculation by at least two orders of magnitude, and can identify the corresponding phase relationship in the proposed household relationship, and the recognition accuracy is higher. Running on the large data platform, the algorithm calculates the relationship of 48 stations. The results show that the accuracy rate of single calculation is over 99% and the accumulative recognition rate is 100%. The algorithm can be directly used in the power grid to solve the problem of difficult identification of the relationship between stations in the distribution network.