A Method for Identifying the Active Characteristics of Low Voltage Stations Based on Digital Twin

With the deepening of new power system construction, the increasing frequency of user-side participation in grid interactions has led to rapid changes in the morphological structure and dynamic characteristics of distribution power systems. This evolution has brought new challenges and problems to Low Voltage Distribution Stations (LVDS). Currently, user-side active equipment remains a "black box" for grid companies, which are unable to accurately grasp the status and characteristics of such equipment. This poses significant risks to the operation and maintenance of LVDS and hinders the auxiliary planning of the distribution network as well as the provision of precise customer services. To address these issues, this paper proposes a data-driven method for identifying the active characteristics of LVDS. The approach begins by utilizing digital twin technology to simulate the steady-state and transient behavior of LVDS under various topologies, new energy access modes, and typical daytime scenarios, thereby creating a large simulation dataset. Subsequently, the paper introduces an uncertainty learning technique for normal operating conditions and an abnormal condition identification method based on probability distribution modeling. By applying the Gaussian Mixture Model (GMM) for the active identification of LVDS, the joint probability distribution model of multi-dimensional measurement data from these stations is constructed using historical operational data. The Jensen-Shannon (JS) divergence of the joint probability distribution model is then calculated to compare similarities both horizontally (across different LVDS) and vertically (across historical data). This analysis helps determine the presence of distributed power supply and storage equipment within the LVDS area.