Microsoft Research has developed a machine-learning system that can forecast power-grid risks from space we…
Microsoft Research has developed a machine-learning system that can forecast power-grid risks from space weather 30–60 minutes in advance.
It combines solar-wind data, geological conductivity and location information to generate risk estimates for 66,935 substations across the continental US. In evaluation, the system detected about 76.5% of major events and 81.2% of severe events, though Microsoft says more validation is needed before operational deployment.