Learning-augmented control and decision-making:
Theory and applications in smart grids

America’s Department of Energy reckons that ai and other improvements to the country’s existing grid could liberate as much as 100 gigawatts (gw) in transmission and distribution capacity over the next three to five years without the need to build new lines.

That is about 13% of current peak demand of around 740gw.

Some of these “grid-enhancing technologies” are now being rolled out, thanks to doughty startups developing them, their financial backers and utilities, which are becoming less resistant to innovation. gets, as they are known for short in the industry, fall into two main categories: hardware upgrades to transmission grids and software upgrades to those grids’ brains.

There is an even less intrusive approach to boosting transmission capacity than stripping out old cables or installing new batteries. Known as dynamic line rating (DLR), it uses sensors to monitor temperature, wind and other local conditions to determine how much power it is safe to channel down a line at any one time.

Cognite, a Norwegian software firm, has developed a program that makes decades’ worth of grid data, which in the past would have been hard to track down and manipulate, available readily to operators. Software made by Envelio, a startup controlled by e.on, a big German utility, creates detailed digital twins of local grids that can be used to pinpoint the best sites for distributed-energy resources (which can include anything from wind turbines and solar panels to heat pumps and batteries of parked electric vehicles). Envelio’s technology is used to manage such resources on roughly half the German power-distribution network.

The idea underlying all these software efforts is to make existing grids more flexible. Users could shift their power needs to quieter periods, when electricity is both more abundant and cheaper. That would lower peak demand, sparing utilities from installing a lot more new capacity.