Machine Learning for Power System Modeling and Control

Machine learning has found many applications in power grids, e.g., generation or load forecasting. Using data for dynamic modeling, or system identification, is particularly useful for power grid applications. Current examples include using phasor measurement units (PMU) event data to identify system oscillation modes, oscillation sources, etc. In industry, dynamic models are required to be benchmarked with real event data. There are certainly many more applications, e.g., reduced-order power grid models using measurements, inverter-based resource modeling using measurement data, design of feedback control. This panel brings researchers working on machine learning for power systems together to discuss the current applications and the potential applications of data-drive dynamic modeling and control, including dynamic data generation, black-box model identification approaches, gray-box model identification approaches, and sparse wide-area control. In addition, university curriculum design to combine data science, machine learning and dynamic modeling for effective learning will be discussed.