A multi-institution effort to make electric grids resilient to cyberattacks, equipment failures, poor-quality data and extreme weather — combining power systems, AI, cybersecurity, hardware security, communications, sensing, economics, visualization and workforce training.
A smart grid uses sensors, software, communications technologies and AI to make real-time operating decisions. A secure grid can keep operating through cyberattacks, poor-quality data, equipment failures and disrupted communications. GRID-AIR works on the second problem: making the AI that increasingly runs the grid something operators can trust under adversarial and degraded conditions.
The work spans the full stack — from hardware Trojans and side-channel leakage in the silicon that executes grid AI models, through secure sensing, forecasting and anomaly detection, to self-healing responses that recover after data or model poisoning, all evaluated on digital and physical microgrid twins in a cross-jurisdiction testbed.
The project is led by the University of Maine and brings together teams in four DOE EPSCoR jurisdictions, pairing the technical program with a workforce development effort that trains students across these disciplines.
Each theme is led jointly across the partner institutions and feeds results into the shared testbed in Theme 4.
Securing the models themselves and the hardware they run on, and establishing what safe behaviour means for grid AI.
Getting trustworthy data off the grid and turning it into forecasts, detections and decisions operators can act on.
Keeping the grid running when the AI itself is attacked — detecting compromise, recovering models, and failing safe.
Evaluating the secure AI technologies from the other themes under realistic grid conditions, in simulation and on hardware.
Training students and early-career researchers across power systems, AI, cybersecurity and hardware security — the combination the field currently lacks.
“No single field can solve this issue alone.” Twenty investigators across four DOE EPSCoR jurisdictions, spanning power systems, AI, cybersecurity, hardware security, communications, sensing, economics, visualization and workforce training.
GRID-AIR is a project of the Secure and Intelligent Edge Research Lab.
Back to SIEGE Lab