EAGER: Transition to Practice: AI4OPT Energy Technologies

Artificial Intelligence (AI) holds great promise for transforming critical industries across the Nation by driving efficiency and productivity. While computational power has increased, there remains a gap between fundamental science-driven advances in AI methods and their practical usage in real test cases. As part of NSF's commitment to advancing fundamental AI research, the national AI Institute for Optimization (AI4OPT) has pioneered new data-driven, AI-enabled methods that address unique challenges for power grids that must operate reliably and efficiently under highly variable conditions due to their diverse energy generation sources (e.g., natural gas prices, uncertain demand, and evolving capacity of battery and storage technologies). When historical power systems data is examined, the AI4OPT unit commitment optimization has shown significant advantages over state-of-the-art algorithms. This EArly-concept Grant for Exploratory Research (EAGER) project aims to translate these foundational advances into an AI-assisted platform deployed in practice and usable by power system operators, generators, load-serving entities, and energy traders to model, assess, and jointly optimize risk and costs in their planning and real-time operations.

In conjunction with Southern Company, the platform looks to integrate fundamental innovations in trustworthy optimization learning methods via primal and dual optimization proxies, temporal fusion transformers, scenario generation, and stochastic optimization for high dimensional time series forecasting, on software and hardware infrastructures in daily use by planners and operators. The project seeks to produce commercially-viable tools for Unit Commitment and near optimal market clearing algorithms, as well as real-time risk management simulators that evaluate system-level, asset-level and financial risk for meeting real-time constraints. The platform is expected to offer orders of magnitude improvements over state of the art in market clearing algorithms that deliver both feasible solutions in milliseconds as well as certificates of quality. The project intends to produce significant economic benefits in terms of energy savings, emission reductions, and increases in equipment reliability for the southeastern United States and looks to serve as a demonstration of effective technology transition from proof of concept research to practical deployment.

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.

Award Number: 2515977
Principal Investigator: Pascal Van Hentenryck
Funds Obligated: $299,839
State: GA