Prediction of Cable Failures based on eXtreme Gradient Boosting

With the access of new energy sources, distributed power sources and diversified loads, the power flow changes from unidirectional to bidirectional, which makes it more difficult to diagnose and predict cable failures with traditional methods. The development of artificial intelligence makes the cable failure method based on big data widely used. In this paper, considering the influence of environmental factors and manufacturers on cable operation, the eXtreme Gradient Boosting (XGboost) method is proposed on this basis to judge the state of cable operation and the cause of fault. Compared with several common artificial intelligence methods at present, the superiority of XGboost in predicting cable failure is demonstrated. Finally, the method is verified by actual cable operation and failure data, and the reliability of the method is further demonstrated.