Power System Fault Classification with Imbalanced Learning for Distribution Systems

Faults are inherent for modern day grid operation. Operators need to identify from the voltage and frequency measurements available at the SCADA for appropriate identification and categorization of faults. Rapid use of high resolution micro PMU measurements paves the way for real time categorization of faults with an accepted level of accuracy. This paper proposes a critical comparison among different machine learning techniques to classify faults as like LLL-G, LL-G, L-G. With L-G, LLL-G faults being most and the least occurred ones, it produces an imbalanced dataset developing a biased classifier. This work uses SMOTE based artificial data generation for an impartial classification with the machine learning algorithms. Power system fault data were collected from the Drexel University's Reconfigurable Distribution Automation and Control (RDAC) software/hardware laboratory.