Real-time AI-based Line-Line Fault Detection and Localization in Power Electronics Dominated Grids

The power grid is rapidly changing into the Modern Power System due to the increased penetration of power electronics. As more distributed energy resources are added, detecting anomalies and faults in the grid is increasingly important. This works proposes an artificial intelligence system built upon deep learning approaches to detect line-line faults in the modern power system. This two-layer approach uses both long short-term memory along with graph neural networks to classify and localize all line-line faults in a fourteen-bus grid. This approach creates a framework which can learn the features from three-phase data and utilize them for the detection of anomalies across the modern power system.