The power grid in the Eastern US, known as the "Eastern Interconnection" (EI), is a technological marvel: an impossibly large, sprawling, and complex machine that’s been operating continuously for over a hundred years, now serving around 240 million people. When considered together with the Canadian EI, it forms what the National Renewable Energy Laboratory calls "the largest coordinated power system in the world."
Here it is, with all its transmission lines:
The colored outlines (MISO, PJM, SERC, etc.) denote areas managed by different Regional Transmission Organizations (RTOs). The RTOs are somewhat autonomous, but their grids are synced up on a common frequency and form part of a continuous whole.
The EI was built around coal, nuclear, natural gas, and hydro power, which can be deployed whenever grid operators need them. (They are "dispatchable.") The pressing question facing today’s policymakers is whether the EI can, relatively quickly, accommodate much more renewable energy, which is variable, i.e., only available when the wind blows or the sun shines.
Prior to the question of how that might be accomplished socially or politically is the simple question of whether it’s technically possible.
That is the question examined by NREL in its newly released Eastern Renewable Generation Integration Study (ERGIS, if you’re nasty).
The study is a remarkable technical achievement, marrying enormous datasets with enormous computing power to produce incredibly rich scenarios (one reason it stretches to 220 pages, with six appendices).
But the basic conclusion of the study can be summed up in four words: It can be done.
The EI can accommodate lots of renewables, quickly. Specifically, it can accommodate 30 percent "variable generation" (VG) — on- and offshore wind, utility and distributed solar — by 2026, using tools available today. (It will require space, money, and transmission lines, but no new advances in energy storage or demand management. More on that later.)
It’s worth pausing to emphasize what this means. There are plenty of arguments to be had about the costs of quickly ramping up renewables, or the right policies to get there. But anyone who says that the densely populated eastern US can’t do it without threatening service and reliability is, according to the best available research, simply wrong.
Building a virtual EI
The EI is incredibly complex: 5,600 generators, 50,000 nodes, and 60,000 transmission lines, all being kept in constant balance, with "dispatch" of resources happening in five-minute intervals, focused on minimizing costs within constraints (available transmission, reserves, etc.).
Simulating the behavior of that machine under different resource scenarios involves running millions of equations, for every one of those nodes, at every interval, ranging over gigantic datasets. It wasn’t even possible until recently, which is why analysis has relied so heavily on assumptions and educated guesses.
The NREL team put a giant, liquid-cooled supercomputer on it, which broke the task down into enough manageable parallel calculations that it could be done relatively quickly. (If you’re interested in hearing more, you’ll definitely want to check out Time Domain Partitioning of Electricity Production Cost Simulations.)
NREL ran four scenarios, based on varying penetrations of VG.
- LowVG, a baseline scenario, which held VG to current levels — no new build after 2012.
- RTx10, or Regional Transmission x 10, which involved a buildout of internal EI transmission and VG penetration of roughly 10 percent.
- RTx30, or Regional Transmission x 30, which involved internal EI transmission and total VG penetration of 30 percent.
- ITx30, or Inter-regional Transmission x 30, which involved a buildout of both internal EI transmission lines and EI transmission connections to surrounding regions, along with 30 percent VG.
Here’s how they break down:
Among other things, these scenarios produced some gorgeous visualizations. Here’s one showing ITx30 at work from May 11 to May 13, 2026, a time of "high variable generation." Things to note: The dots are generators; the color of the dot indicates the source; the size of the dot is the amount being generated; the line that crosses over the map is first light (sunrise); the conical arrows show transmission flows; on the right are RTOs, how much they’re generating, and from what.
Here’s one that shows the same days, with the same high VG, only RTx30 instead of ITx30, and instead of a map, a "kaleidoscope" visualization showing the flows of energy among RTOs:
Man, I could watch that shit all day.
Four lessons learned about integrating lots of renewables
Here are four lessons NREL draws from the study — these are things we already knew, but now know with an incredible degree of specificity and precision, in a form much more useful to grid planners.
1) Integrating lots of VG means that conventional power plants will be run differently. To wit, coal, natural gas, and hydro plants will be asked to ramp up and down much more quickly and more often, and to run for shorter periods of time (their "capacity factor" will decline). This will cause additional wear and tear, which NREL factored in.
(Interestingly, NREL modeled nuclear plants as entirely inflexible, unable to ramp, though there’s some evidence that’s not true.)
2) Sunrise and sunset will pose particular challenges, as they involve steep ramps in demand and in VG supply. The daily course of "net demand" — demand for power minus VG supply — starts to develop what energy nerds call a "duck curve." Watch what happens to the shape of everything beneath the green as more VG comes online:
Where conventional plants once had to ramp up smoothly to a level plateau, now they have to ramp up quickly to a morning peak, drop down, then ramp up quickly to an evening peak, then drop down quickly. Pulling that off will require changing the way grid operators work (see No. 4).
3) With more VG in the system, the flows of energy among RTOS will change more often and more quickly. NREL did build in some engineering and economic constraints on transmission flows, but even so its model is still deceptively smooth on this account. It models a single algorithm controlling energy flows, and a market structure governing dispatch. In reality, transmission is controlled by multiple regional entities in a mix of regulated (vertically integrated monopoly) and deregulated (market) areas. Achieving the kind of smooth inter-regional coordination envisaged in these scenarios will involve lots of changes in how the grid operates (again, see No. 4).
4) Technical potential is only technical potential. To realize it, generators, RTOs, regulators, and policymakers will have to change how they do things. The grid will need new operational rules. RTOs will have to cooperate better. Policymakers and regulators will need to change the incentives governing utilities and transmission operators.
These kinds of things — the political, social, and institutional changes — are more difficult than the technological challenge. And unfortunately, there is no model, no simulation, and no supercomputer that can show us how to induce today’s socioeconomic institutions and their leaders to behave in a more clueful fashion.
- It’s probably worth at least mentioning that increasing VG penetration to 30 percent would also reduce greenhouse gas emissions from the EI by 30 percent. That’s a lot.
- NREL did not model a substantial role for emerging technologies and practices like grid-connected battery storage and demand response. If thost techs flourish — and signs are good —it's likely the grid could accommodate much more than 30 percent in 10 years.
- Let’s pause a moment and give thanks for America’s extraordinary national laboratories, which continue pumping out high-quality research at a pace I can barely keep up with. (Special shout out to Aaron Bloom, NREL’s project leader on this study.)
- In particular, this study is a welcome step forward in transparency — NREL has released all of its data and tools alongside the study, on the open-source platform GitHub, so anyone can use or adapt them. (Scroll to the bottom here.) That’s a worthy example for other researchers to follow.