Classification of Faults in Power System with Probabilistic Neural Networks: An Imbalanced Learning Approach

Modern day grids with its inherent operating characteristics are susceptible to faults. Grid operators must detect as well as classify the current system conditions like normal or faulty from the current raw sets of measurement data available at SCADA (supervisory control and data acquisition system). With the rapid deployment of micro PMUs, faults are detected from the measurements in real time, but their classification in real time still possess a challenging task. This paper focus on a diligent comparison between several deep learning and machine learning methodologies for classifying faults (L-G, LL-G, LLL-G) in real time. In real life scenarios L-G fault being most frequent and LLL-G being rare, an imbalanced dataset is generally developed for supervised learning approach leading to a biased classifier. To mitigate this issue this paper proposes SMOTE based oversampling over the imbalanced dataset. The dataset used in this work is derived from the Drexel University's Reconfigurable Distribution Automation and Control (RDAC) software/hardware laboratory.