Traditional hardware zone user identification of electric power company includes host and handheld terminals. The host is installed on the three-phase low-voltage side of the distribution transformer. The staff need to hold the terminal to carry out the identification work on the user side, which consumes manpower and resources for the deployment of dedicated hardware. A method of undisturbed station identification based on the zero-crossing time and SNR collected by various sensors is designed in this paper in order to replace the traditional station area identification. In this paper, machine learning is learned to obtain the probability distribution of the meter station area. Several kinds of station identification algorithms such as clustering and deep learning were tested. The optimal path method is used to identify station area through comparative analysis, which is a station identification algorithm with high recognition accuracy and low cost.