Microsoft AI solves grid optimization math 1,000x faster, close to the same answer
~1,000x faster than a full AC solver, with a 2.23% median cost gap from the exact answer
Microsoft Research released GridSFM, a small foundation model that solves the AC optimal power flow problem about 1,000 times faster than the standard solver grid operators use today, with a median cost gap of 2.23 percent versus the exact answer. The work targets a class of decisions tied to up to $20 billion a year in U.S. congestion costs and 3.4 terawatt-hours a year in curtailed renewable energy. The model, paper, and code were open-sourced.
Microsoft Research — GridSFM: A new, small foundation model for the electric grid

Social channels
The YouTube channel is live. More show channels will be added as they come online.