Microsoft Research Debuts AI for Power Grid Space Weather Forecasting

01 What happened

Microsoft Research has developed a machine learning system designed to estimate space-weather risks for 66,935 electrical substations across the continental United States.

02 Key details

• The pipeline produces location-specific risk estimates 30–60 minutes ahead in experiments, requiring further utility validation before live grid use.
• Inference for all 66,935 substations is performed in approximately 333 milliseconds.
• During a 2020–2026 evaluation, the model detected 76.5% of major, 81.2% of severe, and 64.1% of extreme space-weather events.

03 Why it matters

This technology enables utility operators to protect critical infrastructure by implementing proactive, localized adjustments during geomagnetic disturbances.

04 Who it matters to

Energy specialists, infrastructure engineers and artificial intelligence researchers.