Grid-Aware Waveform Analytics for Event Classification in Distribution Grids

This study proposes a new method for event sit-uational awareness in distribution grids using Synchro Wave-form Measurement Units (SWMUs). An efficient feature-extracting technique named the Short-Time Matrix Pencil method (STMPM) is used to capture the oscillation modes and distortion of voltage and current measurements under common events such as transient switching events and challenging high impedance faults. The extracted features from the waveform data are then used as the input to Graph Neural Network (GNN) as the event classifier. GNN captures the spatial relationship between SWMUs and physical features of the network to enhance event classification accuracy. The proposed grid-aware waveform analytics is tested for classifying different events, and the superior performance of the proposed method with respect to other approaches is verified using the classification merits, such as accuracy, Fl-score, precision, and recall.