PJM, the nation’s largest grid operator, is planning to use artificial intelligence to overhaul how it runs its electricity system across 13 states, with major implications for renewables, fossil fuels and an interconnection backlog stalling projects.
In an announcement Thursday, Google said it is teaming up with PJM and Tapestry, an arm of Alphabet focused on grid innovation, to use new AI technologies to change the way the grid operator plans projects and brings them online. The idea is essentially to do for the electricity system what Google maps and searches did for the internet to speed up and simplify the interconnection process.
This will be “the first time that artificial intelligence is being used to manage the entire interconnection queue and process,” said Page Crahan, general manager of Tapestry, on a call with reporters about the project.
Today, developers are “largely using the same tools that we designed for last century’s grid. We know that we can do better than this,” added Crahan, who said her team has been working for seven years on developing the AI tools.
Tapestry, which plans to build on existing AI technologies such as Google DeepMind, said the plan would speed up reviews for projects applying to link to the grid. That could be significant for PJM, which has thousands of projects waiting to connect, particularly renewables, across its territory. PJM said it would eventually use the new technology for all interconnection requests.
The plan comes as the grid operator is trying to figure out how to address a looming electricity shortage because of retirements of coal and gas plants and a surge in demand from AI and data centers.
PJM expects it could lose about 40 gigawatts, or 21 percent of its capacity, by 2030, said Aftab Khan, PJM’s executive vice president.
New power resources are not coming online “at an adequate enough pace … to replace the risk of retiring resources,” Khan said.
Nationally, the grid backlog is more than double the size of the existing installed power fleet, according to Lawrence Berkeley National Laboratory.
It’s unclear how the plan might ultimately affect the electricity mix, but speeding up interconnection requests could bring more solar, wind and batteries on the grid, since those resources dominate the current backlog. In a press release, Google said the AI tools could “support the rapid and reliable integration of diverse energy sources onto the grid.”
Crahan said the AI technologies would consolidate what currently is a “laborious” PJM process. Under the existing system, officials have to review many documents and computer models to approve and connect one project. Crahan compared it to retail shopping before the internet, when everyone was faxing orders.
As an example, she said planners currently use separate software for everything from assessing how equipment connects to the grid to economic analyses of projects.
“Each of these software programs will generate a file, which creates its own unique model of the grid. Every time a change is made to that one model, it needs to be applied to all the other models,” Crahan said. Because everything is siloed, it is “extremely difficult” to keep track of details and move projects along quickly, she said.
With the new tools, grid planners would be able to assess projects in one place. An online demonstration resembled a mix of Google Maps and Microsoft Windows, where a user could drag and drop large amounts of information around one screen. The AI technologies also would eliminate many time-consuming steps to review and connect projects, such as verifying that land rights are associated with a given applicant.
PJM, Google and Tapestry said they do not know yet how much the AI shift could reduce the multiyear connection time to the grid, but results from other tests suggest the plan could be transformational. A grid planning tool that Crahan tested in other countries sped up the process of doing grid simulations by 86 percent, turning a process that formerly took weeks into minutes.
PJM and the companies said they planned to start using some of the tools this year, and others would be phased in over several years. Tapestry said it has other AI tools that can make the grid more resilient by using cameras that can more quickly determine where upgrades are needed.
The plan follows executive orders from President Donald Trump on Wednesday to try to bolster coal plants through various means, including repealing regulations, opening federal lands to mining, assessing coal use for AI and directing the Department of Energy to develop a process to keep plants online.
When asked how the new AI plan might interplay with that process, Khan said PJM is “fuel agnostic” and will “welcome any resource that can help our ultimate goal of reliability.” Coal plants currently make up about 20 percent of PJM’s capacity.
But Amanda Peterson Corio, Google’s head of data center energy, said the company does not support new coal plants, considering its target to reach net-zero emissions by 2030.
“We remain committed to our goals to decarbonize our electricity footprint,” she said.
The FERC factor
The AI solutions would have to sort through a mix of factors including developer finances, projects’ proximity to big PJM transmission lines, and state and local permitting, based on experts’ analysis of PJM’s queues.
Many of PJM’s backlog challenges date back to roughly 2010, when the hydraulic fracturing revolution led to a surge of increasingly cheap gas-fired generation that helped push older coal plants into retirement. About the same time, declining costs for renewable energy, particularly small solar projects, produced a flood of applications from developers seeking permission to hook into the PJM high-voltage system. In the 2010s, the number of proposed projects in PJM’s interconnection queue soared by fourfold, according to University of Wisconsin researchers.
The tide of applications overwhelmed PJM planners who were tasked with ensuring that each new project would not overload the surrounding transmission network. Developers faced costs of adding new lines or upgrading existing ones to prevent risks over grid load. Installations of new capacity fell further and further behind coal plant retirements, with increased warnings of potential power outages.
PJM said in a February 2023 report that of the 290,000 megawatts of renewable power projects awaiting approval to connect to its system, the historical success rate was only 5 percent. In 2022, PJM said it would not take new applications to join the interconnection queue until the next year, while it worked through the backlog. It trekked to the Federal Energy Regulatory Commission for a series of changes to its approval process, including closer scrutiny of which projects seemed most viable.
The most recent, PJM’s Reliability Resource Initiative, approved by FERC in February, invited up to 50 “shovel ready” new generation projects to apply for connection agreements in a process that favors large new gas-fired turbines over renewables.
Last month, PJM reported it had received 94 RRI proposals with a total capacity of nearly 27,000 MW, which will join 550 projects under review. But an update in February by Aurora Energy Research consultants determined that of 38,000 MW of projects that had been approved, only 6,000 MW were under construction. Another 20,000 MW were in engineering reviews, and 12,000 MW had been abandoned.
