Fault Level Prediction Method for Urban Distribution Network Considering Class-Imbalance Problems

Urban Distribution network (UDN) fault diagnosis and risk warning are of great significance to the safe and stable operation of power system. At present, the class-imbalance problems greatly affect the accuracy of fault analysis and prediction in UDN. In this paper, a method of fault level prediction (FLP) based on improved natural breaks is proposed. The method of improved natural breaks is first introduced to achieve dynamic division of FLP and solve the class-imbalance problem of fault categories. Then, the precision and recall are selected as evaluation indicators for FLP. According to a certain area of UDN, the normal operation data and the fault simulation data are obtained and analyzed. Finally, the prediction results of different division methods under different data sets are compared. The results show that FLP method based on improved natural breaks can effectively improve the prediction precision and recall rates.