Grid edge

Seven ways utilities are exploring AI for the grid

The grid is entering its automation era.

Listen to the episode on:
Apple Podcast LogoSpotify Logo
A graphic of transmission lines juxtaposed over a floating computer chip, on a turquoise background.

AI-generated image credit: Gold Flamingo

A graphic of transmission lines juxtaposed over a floating computer chip, on a turquoise background.

AI-generated image credit: Gold Flamingo

As artificial intelligence gets embedded in a growing number of consumer products, the technology may feel new. But the rapid spread of AI in 2023 is the result of more than 70 years of research, systems learning, and advancements in computing power. 

This is especially true in the power sector, where AI applications are building upon decades of work on digitization across the grid. 

Analysis from Indigo Advisory group indicates that the market for AI in the energy sector could be worth $13 billion. There are now 100 vendors with AI solutions embedded in their products, serving more than 50 possible use cases, from grid maintenance to load forecasting. 

According to Indigo Advisory Group managing director David Groarke, who is partnering with Latitude Intelligence on upcoming utility AI research, we’re entering a new era of grid tech.

“In the past 20 years, there was deployment of a lot of sensors, smart meters, and communication networks on the grid,” Groarke said. “We are now in this period where it is about self optimization and automating.” 

The chart below illustrates the different categories of AI that are in use — or could soon be in use — across the power sector.

Image credit: Indigo Advisory Group

Progress in AI is coming at a critical moment, as extreme weather intensifies, more variable renewables hit the grid, the utility workforce ages out, security threats increase, and decarbonization pressures mount. Below are seven critical areas where AI is already having an impact. 


Monitoring is a foundational layer of any grid technology. Only with accurate, timely, and granular data about the physical grid can utilities apply the additional layers of analytics needed to, for instance, detect cable faults and analyze power quality.

While AI is most often associated with software solutions, Latitude Intelligence analyst Fei Wang said monitoring really centers on hardware. 

Industry incumbents like Siemens and Schneider have monitors for a wide variety of grid parameters, Wang said, while newer startups tend to focus on a more compact, single device, all-in-one solution. But those smaller companies have a challenge ahead of them. Single device solutions are still very expensive, and the major players already have significant market share and existing relationships with utilities.


As a result of the 2021 infrastructure law, the federal government’s Grid Resilience and Innovation Partnerships Program is disbursing $10.5 billion to “ensure the reliability of the power sector’s infrastructure.” And for utilities managing an aging grid, much of reliability comes down to fault and outage detection. Utilities and vendors alike are increasingly turning to AI to bolster these capabilities, using the technology to flag potential downtime before it occurs, and ultimately minimize both repair costs and the impact on customers.

Sentient Energy's sensors and fault detection algorithms, for example, communicate fault information for both above and underground power lines back to utility control centers in real-time, providing accurate location data and thereby reducing repair times and outage durations.

Enhanced reliability is also one of the central offerings of a new generation of AI-equipped smart meters, which boast real-time information about what’s happening on the grid. Utilidata, for example, makes an NVIDIA-powered smart chip that attaches to meters; the company’s president and chief operating officer Jess Melanson told Latitude Media that the tech can bolster overall grid reliability by enabling utilities to locate and respond to outages more quickly.

Listen to the episode on:
Apple Podcast LogoSpotify Logo


When it comes to maintaining the grid’s smooth operation day-to-day, there are two main ways that AI is poised to play a key role: asset management and work management.

Asset management tools focus on the physical environment of the grid. For example, Pano AI’s wildfire detection technology applies deep learning AI and computer vision to images from high definition cameras, field sensors, and other data sources to identify and verify fires. Meanwhile, tools like IBM’s AI-powered suite of vegetation management tools use satellite data, LiDAR sensors, and weather insights to predict power interruptions resulting from plants growing near infrastructure.

AI tools can also help prioritize work orders and optimize the internal workflow of other administrative tasks flagged by frontline workers. Datch AI, for example, uses natural language processing to capture information from field workers and simplifies the process of creating, updating, and closing out work orders on-site from a phone. The company’s “voice-visual interface” amounts to a voice layer integrated with a utility’s database that can run on workers’ cell phones: essentially a pocket voice assistant for the industrial power sector.


Utilities increasingly need tools to optimize both their portfolios of distributed energy resources and their transmission capacity. And a wide swath of vendors are stepping up with AI-powered tools on offer. 

According to Wang, many companies in the demand response space have evolved to focus on resource optimization. One example is startup Leap, which is focused on connecting behind-the-meter load to wholesale energy markets via its Distributed Energy eXchange. Leap’s cloud-based platform aggregates DERs, then uses optimization models to bid into deregulated markets. (These, of course, are the kinds of tools that will be integral to setting up virtual power plants.) 

On the transmission capacity side of optimization, LineVision’s Dynamic Line Rating technology uses machine learning to accurately model transmission line ampacity, in some cases up to 10 days ahead. Transmission lines capacity depends on weather, and without real-time data, operators tend to be more conservative with how much amperage lines can take at any given time, explained Wang.

That’s a problem for wind and solar, which often get curtailed due to lack of line capacity, she added. That’s where AI solutions like LineVision’s could come in, to reduce curtailments and increase overall capacity by using non-contact sensors and software to provide accurate transmission line analytics.


AI-backed forecasting is having a moment in the spotlight, with Texas-based startup Amperon raising $20 million in Series B funding in October 2023, and Snowflake’s January acquisition of time series forecasting startup Myst.

Amperon integrates machine learning into their software to get play-by-play demand and supply analytics. This information helps utilities realize the scope of future energy generation. The company claims that their new net demand forecasts are 25% more accurate than those of independent system operators and “a critical grid management tool for the energy transition.” 

Myst AI’s forecasting solutions target load serving entities like utilities, renewable energy generators, DER aggregators, and grid operators alike, and forecast everything from energy prices to electricity load and demand.

Energy markets

Predicting the amount of energy generated from coal and fossil gas has long been an established and relatively simple task. But with renewables, far more factors come into play — including weather patterns, geographic location, wind speed, and cloud cover — that make predictions much more challenging.

One way to make renewables more profitable is to increase their trade value in energy markets. And AI may be able to help with that.

AI price forecasting tools reduce the need for human touchpoints, allowing traders instead to rely on algorithms trained on immense amounts of data (though challenges remain when it comes to the quality of that data.). Enspired, for example, runs an AI powered algorithm that uses 7.3 billion data points, consisting of everything from weather forecasts to regional battery constraints. Then there’s legacy player General Electric: GE’s Alpha Trader uses predictive analytics to make recommendations about which hours are most profitable for trading, as well as hourly market insights.


AI solutions focused on customers hail from both the hardware and software sides of the tech world. But all tend to be aimed at collecting and analyzing more detailed data. 

On the hardware side sit companies focused on smart metering, like Sense Energy Monitoring, which places a device into electrical panels and uses machine learning to disaggregate household usage data by learning the power signature of individual devices. (Sense also has a partnership with Schneider that includes integrating disaggregation capabilities into Schneider’s Square D Energy Center.) Another vendor is Copper Labs, which uses neighborhood-level electricity, gas, and water meter data to provide behavioral load management tools for utilities and in-app consumption insights to customers.

Meanwhile, software vendors are developing tools for things like billing and customer segmentation. BlastPoint, for example, has a suite of tools aimed at the energy industry that include billing and collections optimization (identifying customers likely to pay in full or default) and customer segmentation (zeroing in on best candidates for weatherization and appliance upgrades, for instance).

Wang said these customer-centered examples are where artificial intelligence is most prepared to impact processes in the energy sector (and in fact is already doing so). That’s in part because the stakes are lower — there’s less risk in handing admin tasks off to AI.

“Anything that doesn’t have to do with real-time operations has a lot of potential already, you can deploy now,” Wang said. 

But utilities are much more wary about implementing AI solutions that focus on grid operations themselves, she added. Anything that requires “nanosecond reactions” — like figuring out where faults are, maintaining frequency, and turning the flow of electricity on or off — is more challenging.

“Machine learning has a lot of potential, but anything to do with the physical layer of the grid is so tricky,” said Wang. “When it comes to actual grid operations, that’s where people become cautious. And you should be more careful, because there are more severe ramifications when things don’t go as planned.”

No items found.
No items found.
No items found.
No items found.
energy markets
demand response
Department of Energy (DOE)
energy transition
grid infrastructure
smart grid
virtual power plants