A large-scale blackout in power systems would result in millions of dollars in revenue loss. It interrupts businesses, and even poses risks to environment and public safety. Protecting power systems from large-scale outages is no doubt a top priority. Early detection of random component failures and prediction of cascading failures are critical to the prevention of large-scale blackouts, but they can be achieved only when the power grid is equipped with adequate capability for system understanding, situational awareness, and emergency response. Modern power systems are becoming increasingly complex with the addition of a variety of active controllers, renewable energy resources and storages. With the complexity and uncertainty involved, traditional approaches based on intensive computation to solve a system of model equations are no longer suitable for real-time analysis and control. One graduate student will be support in year 1 of this award.
In this project, we leverage recent advances in data science to improve power systems reliability, security, and resilience. In particular we propose to use deep neural networks and data-driven uncertainty quantification, to advance the core algorithms pertaining to the analysis and control of power systems. The proposed work includes three major thrusts: (1) real-time power flow analysis, (2) real-time anomaly detection and causal analysis, and (3) real-time contingency analysis and optimal emergency control. The project is expected to make a significant breakthrough in reliable energy delivery. It will not only benefit the power system research and operation, but also advance data science research by promoting physical law-assisted machine learning.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
