Real-time, AI-based water intelligence for U.S. energy systems

Motivation:

Water availability limits U.S. energy production and infrastructure resilience, yet current prediction systems are too slow and too coarse for real-time decision-making, constraining hydropower operations and flood response.

Approach:

• Develop ORBIT, a unified AI foundation platform that transforms global weather forecasts into sub-kilometer environmental intelligence.
• Train on DOE exascale supercomputers and deploy anywhere to deliver real-time predictions.

Impacts:

• Enables real-time water forecasting for energy systems
• Strengthens grid reliability and infrastructure resilience
• Democratizes access to high-resolution prediction
• Turns days of supercomputer simulation into seconds on 
  a standard computer

AI Advancements for Water for Energy:

• Global prediction at 0.9 km resolution
• Orders-of-magnitude speedup: days à seconds
• Up to 52% higher accuracy at extended forecast horizons
• Achieve up to 99% accuracy for global prediction
• Being integrated into NVIDIA Earth-2

S&T Challenge: Aligns with “Predicting U.S. Water for Energy”