Researchers from Duke University have said that integrating more flexibility into U.S. power grids could help provide the energy needed to power future load growth, particularly the electricity needed to support artificial intelligence and data centers.
The group in a Feb. 19 webinar discussed their findings, which are contained in their recent report titled “Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in US Power Systems.” The report comes from Duke’s Nicholas Institute for Energy, Environment & Sustainability.
The study was coauthored by Tim Profeta, senior fellow at the Nicholas Institute and associate professor of the practice at the Duke University Sanford School of Public Policy; Dalia Patiño-Echeverri, associate professor at the Nicholas School of the Environment; Tyler Norris, a Ph.D student at the Nicholas School of the Environment; and Adam Cowie-Haskell, a graduate student at the Nicholas School.
The study looked at 22 of the nation’s largest balancing authorities, which together serve about 95% of the country’s peak power load. The researchers said that gigawatts of new load could be added to the U.S. grid in each balancing authority before the total load would surpass what system planners are prepared to serve. The key point of the study is grid flexibility, so that the new load can be temporarily curtailed as needed to avoid problems with power delivery.
Profeta said, “I think we’re really looking to use this analysis to inform the relevant decision-makers regarding how we service loads. That’s the regulators at the state level, at the regional level, and also the sources of load, and also the utilities, the federal policy-makers overseeing the system. I think that’s all relevant, but I think what we really want to do is introduce a process, to use this headroom in planning processes.”
Profeta added, “And then possibly maybe provide motivation for regulators to open proceedings to explore recommendations for how they should potentially enable this approach across their entire planning process. This [load growth] is presenting unique challenges to regulators. We want to show how flexibility can mitigate these additional loads.”
Curtailing Loads
The study found that 76 GW of new load, which is equivalent to about 10% of the nation’s current aggregate peak demand, could be integrated with an average annual load curtailment rate of 0.25%. In other words, that would mean new loads can be curtailed for 0.25% of their maximum uptime. The group also said that 98 GW of new load could be integrated at an average annual load curtailment rate of 0.5%, and 126 GW at a rate of 1.0%.
The report said the five balancing authorities with the largest potential load integration at 0.5% annual curtailment include PJM (Figure 1), which serves 13 states and the District of Columbia, at 18 GW; and the Midcontinent Independent System Operator, or MISO, which serves much of the Midwest, at 15 GW. The study said ERCOT, which oversees the Texas grid, could curtail 10 GW. The Southwest Power Pool, or SPP, serving the Southwest U.S., also could curtail 10 GW, while Southern Co.—the group that oversees several utilities in the Southeast U.S.—could curtail 8 GW.
Norris said, “Our study demonstrates that existing U.S. power system capacity—intentionally designed to handle extreme peak demand swings—could accommodate significant load additions with modest flexibility measures. Overall, the findings suggest that load flexibility offers a promising near-term strategy for regulators and market participants to more quickly integrate new loads, reduce the cost of capacity expansion and enable greater focus on the highest-value investments in the electric power system.”
The study said the average duration of load curtailment would be about 1.7 hours at 0.25%, 2.1 hours at 0.5%, and 2.5 hours at 1.0%. The group said about half of the new load on the grid would be retained during about 90% of the time that curtailment is required.
“There’s gold in the hills, and we should be hunting for it,” said Profeta. “Load growth is really on a collision course with the grid … it’s really intensified with the rise of AI [artificial intelligence] and data centers. We need every tool in the toolkit to address this problem, and we need every tool to address the delays we might see.”
Georg Rute, CEO of Gridraven, a company focused on optimizing the power grid, noted it’s important to recognize the challenges that exist along the grid. “While the Duke study highlights that the existing U.S. power grid has significant untapped capacity, it doesn’t address existing transmission constraints that can limit power delivery. Congestion is a problem, so a majority of the potential could easily be wiped out if that’s taken into account,” said Rute.
Rute told POWER, “However, there are readily available solutions to reduce congestion and unlock that potential. For example, even a slight wind blowing around 4 mph can increase transmission line capacity by 30% through cooling. In Germany, grid operators have implemented sensor-less dynamic line ratings (DLR) across their networks, boosting capacity and saving billions in congestion costs. Combining flexible connections with DLR is a critical step to ensure we can power AI data centers without overwhelming the grid.”
Importance of Flexibility
Profeta said the Duke group “particularly wanted to look at load flexibility to see whether it could be a near-term solution to handle load growth in this country amid some of the [grid] constraints. We wanted to look at how we can stabilize the grid while handling this load growth.”
Norris noted that load factors during the winter were on average lower than during summer months (Figure 2). “The highest demand occurs during a very short period of hours … during extreme weather events, or the hottest parts of summer or the coldest parts of winter,” said Norris. “Our systems are planned around these peaks. It’s assumed that when new load is added, it will happen during those extreme events.”
The group said those periods are when flexibility (Figure 3) would be important, as end-use customers could temporarily reduce their electricity consumption from the grid through the use of on-site power and energy storage, temporal flexibility, spatial flexibility, and reduced operations.
Norris said some utilities and other groups already are making adjustments to account for power demand from data centers. American Electric Power (AEP), which last fall said it expects to bring 4.7 GW of new data center capacity online this year, wants new data center customers to pay for a minimum of 85% of the energy they say they need each month, even if they use less, to cover the cost of infrastructure required to bring electricity to those facilities.
Norris noted that AEP has said it expects 30 GW of new load from data centers in the next few years. Bill Fehrman, CEO of Ohio-based AEP, last week said “It continues to be full speed ahead” for his group, and said the company is considering adding $10 billion to its record $54-billion capital expenditure plan through the end of the decade.
Demand Response
The study noted that the number of hours during which curtailment of new loads would be necessary each year, on average, is comparable to the curtailment from existing U.S. demand response programs. The group said their research suggests the U.S. grid is sufficient to accommodate significant constant new loads, as long as those loads can be safely scaled back during some hours of the year. They also said there is the potential to use flexible load to complement investments in new power capacity, enabling the grid to grow while also mitigating the need for large expenditures on new capacity.
The researchers, citing the North American Electric Reliability Corp.’s 2024 Long-Term Reliability Assessment, said that aggregated U.S. winter peak load is forecast to grow by 21.5% over the next 10 years, from the 2024 total of 694 GW to 843 GW by 2034. The group said demand from large commercial customers, including data centers, would be responsible for much of that load growth.
Norris during Wednesday’s webinar was asked how load flexibility could be translated into system planning processes, such as interconnection studies and resource adequacy studies that actually allow for using the headroom on the transmission and distribution system.
“This is a really important point,” said Norris. “I want to emphasize that because we’ve been asked … how is this different from existing demand response programs. The existing demand response programs, and they’ve obviously made very meaningful contributions in some markets, is that most of the loads that participate, especially the large loads—which are about 70% of participants in demand response programs—most of those loads are treated as firm loads for the purposes of resource adequacy planning and interconnection and transmission planning. And they opt to participate in demand response for economic purposes given the associated economic incentives.”
Norris continued, “What we’re talking about here can include that, but it’s a little different that we’re talking about pulling up the considerations earlier in the planning process from the beginning and recognizing the profile of the load and its capabilities before an interconnection study. And [then] a transmission study and a capacity expansion plan is developed to accommodate that new load. And that is, we think, a bit different than the existing programs we’ve seen, and it does arguably require a little bit more of a sophisticated approach to system planning.”
—Darrell Proctor is a senior editor for POWER.