Gridware devices use AI to listen to power lines for potential problems.

On a plot of land on the eastern shore of San Francisco Bay, startup Gridware built a full-scale electrical grid, complete with 50-foot poles, 200-foot spans of wire, and 12.5 kilovolts of power flowing through it. Once it was finished, the company then set about destroying it. Over and over and over again.

“I’ve destroyed it thousands and thousands of times,” Gridware co-founder and CEO Tim Barat told us. “I throw trees on the lines, I cut live wires, I explode transformers.”

The goal of all of this mangling is to collect data on the kinds of things that might happen to power lines in the real world. Gridware has used it to inform the sensors the company now supplies to utilities. Its product, Gridscope, mounts onto electrical poles and listens for signs of threats or problems using machine learning.

“If you [test] enough, you start to recognize the patterns associated with how the equipment will behave,” Barat said. “And then what we discovered is—and we weren’t surprised by this—it matches what happens in the real world.”

Bolstering for potential damage has become increasingly important for utilities as the country’s aging grid infrastructure faces new threats from climate-related disasters like wildfires, as well as newfound stress from ongoing electrification efforts. Without AI to interpret patterns in a flood of data, monitoring the grid can also be tricky, according to Barat.

“They have an aging infrastructure, they have a diminishing labor force, they have pressures on rates,” he said. “They have increased frequency and severity of these weather events, and they have to totally change the way they operate, and yet they don’t have any tools to be able to do that.”

Continue reading at Tech Brew

For more utility related reading, check out the potential impact of the Supreme Court’s Chevron decision, TerraPower’s new reactor construction, and the demand for AI potentially driving more generation investment.