Fault Detection Method for Grounding Fault by using Dual-Switched Grounding Detector in Isolated System

Abstract

Electric vehicle high-voltage circuits commonly suffer two prominent technical bottlenecks: faint signals generated by ground faults and poor performance in pinpointing faulty segments. Targeting these practical limitations, this study develops an original fault localization scheme that combines sequential switching grounding detection with wavelet packet transform and support vector machine (WPT-SVM) algorithms. For the detection architecture proposed in this study, two controllable grounding test terminals are installed on the direct current bus and the drive motor branch circuit respectively; these terminals are switched to link with a unified grounding resistor in turn. Transient voltage waveforms captured in this process are used to reflect distinct fault signatures corresponding to each grounding access route. Wavelet packet decomposition is then adopted 
to compute normalized frequency band energy and energy entropy as fault characteristic parameters. Feature vectors extracted per grounding switching cycle are concatenated and input into a support vector machine for accurate fault region discrimination. Simulation validations show the proposed scheme can effectively distinguish ground faults in different high-voltage branches while retaining high classification accuracy and reliability under large transition resistance faults.