Ungrounded Fault Detection in Medium Voltage Distribution Network Based on Machine Learning

As medium-voltage distribution network substations often use the insulated neutral point, it will not cause obvious fault characteristics when single-phase bilateral non-grounded line breakage events happen. Therefore, it is still not possible to determine the fault by using the protection device in the transformers, and the disconnection can only be detected by the users or the inspection staffs. In this paper, we develop an intelligent system to detect disconnection based on electrical information system data and a variety of machine learning algorithms. After multiple comparisons, this method can significantly improve the predict performance, and can find out the main criteria for judgment. The evaluation vectors are {AVC, ACC, TPR, TNR} = {0.9727,0.9144,0.9154,0.9136}. This method can be applied to online diagnosis of disconnection fault in medium voltage distribution network to improve the reliability of power supply.