Distribution network topology identification method based on SMOTE-TCN

With the construction of new power system, distributed energy and charging piles are connected in large quantities, which makes the topology of distribution network more complicated. At present, there is a lack of effective topology identification methods for distribution network stations. In this paper, a time convolutional neural network is proposed to identify the dependency relationship between the user and the station area, it provides the basis for judging the three-phase unbalance of the platform area. In view of the problems of low quality and imbalance data, SMOTE data processing algorithm is used to deal with the imbalance data. The experimental results show that: The combination of SMOTE data processing and temporal convolutional neural network algorithm can effectively identify the user's station area. The average identification accuracy rate of all stations reached 89.31%, and the recall rate reached 89.36%.