What happened
Researchers built an AI system that controls batteries in electrical distribution networks, using graph neural networks to understand how the physical layout of the grid affects decisions. In tests on small practice grids, the AI reduced voltage problems more reliably than older approaches and sometimes beat traditional optimization methods on cost.
Why it matters
This is a paper showing AI can handle a specific, real infrastructure problem — grid stability with distributed energy storage — better than existing methods; if deployed at scale, it suggests AI might solve coordination problems in power systems without the computational delays that make current optimization impractical.