A study of power distribution system fault classification with machine learning techniques

Power system protection includes the process of identifying and correcting faults (failures) before fault currents cause damage to utility equipment or customer property. In distribution systems, where the number of measurements is increasing, there is an opportunity to improve fault classification techniques. This work presents a study in fault classification using machine learning techniques and quarter-cycle fault signatures. Separate voltage- and current-based feature vectors are defined using multi-resolution analysis and input to a two-stage classifier. The classifier was trained and tested on experimental fault data collected in Drexel University's Reconfigurable Distribution Automation and Control (RDAC) software/hardware laboratory. Results show: (1) non-linear, and even non-contiguous decision regions on a “fault plane”, using a phase voltage-based feature, and (2) an accurate classifier for determining the grounding status of multi-phase faults, using a neutral current-based feature.