Traditionally and across most independent system operators, resources must be “deliverable” in order to count towards resource adequacy and participate in capacity markets. On the surface, this structure makes sense. If a resource cannot provide power during peak demand, why would it count towards resource adequacy?
We sought to take a closer look at the analytical aspects of this construct and understand if the approach mirrors operational risks. Are resources likely to be curtailed at times when power supply is short and the grid faces resource adequacy shortfalls? Generator deliverability assessments live in the interconnection world with their own set of analysis approaches, relying on snap shots to approximate the amount of power a generator can export to the grid under stressed conditions. Resource adequacy analysis, by contrast, lives elsewhere in the planning world and relies on an hour-by-hour analysis spanning a full calendar year to understand when resources may be insufficient for serving demand.
GridLab teamed up with Telos Energy (now part of PowerGEM) to examine the link between deliverability and resource adequacy using PJM, the country’s largest grid, as a case study. In an era of load growth, uncertain forecasts, and a need for speed, reexamining approaches more rooted in tradition than reality is critical.
How does deliverability shape resource adequacy in PJM?
Recent analysis from Lawrence Berkeley National Lab (LBNL) revealed that, with a few exceptions, most generators interconnect via network resource interconnection service (NRIS), which requires the project to be studied under extreme conditions including simultaneous peak load and outage conditions. If any system upgrades are required to allow the new resource to deliver 100% of its energy during these conditions, that upgrade must be constructed prior to bringing the resource online. A resource’s NRIS status is what allows it to participate in capacity markets or qualify for resource adequacy programs and collect capacity revenues, so almost all new interconnection requests across the country are NRIS requests.
Using PJM as a case study we first looked at historical market prices to see how closely resource adequacy events track with transmission congestion. With a few exceptions, they don’t. This makes intuitive sense: resource adequacy risk shows up when overall supply is low, which isn’t typically when the transmission system is clogged.
We then examined how PJM actually factors deliverability into its resource adequacy modeling. Two things stood out. First, PJM (like many ISOs) excludes ERIS resources from its model entirely. In PJM’s case, this likely isn’t a major driver of the results, since LBNL’s analysis shows ERIS resources make up only about 4% of interconnected capacity. The bigger factor turned out to be that PJM assumes that no resource can ever deliver more than its deliverability limit. But if our market analysis holds up, a meaningful amount of energy is likely being delivered to the system in practice that never shows up in PJM’s resource adequacy accounting.
What happens when PJM’s deliverability limits are removed from resource adequacy modeling?
To test this, we replicated PJM’s resource adequacy model and ran two versions: one that mirrored PJM’s existing deliverability limits, and one that removed them entirely. The results were insightful. Capacity contributions based on ELCC scores didn’t shift much for wind and solar, but they changed dramatically for storage — because without deliverability limits, more energy is available during off-peak hours to charge batteries. Even more notable: the total resources needed to meet a 1-day-in-10-years reliability standard dropped by more than 2 GW. That is a significant number in an era when the industry is struggling to build new resources fast enough.
We learned that deliverability assumptions really matter in resource adequacy analysis. To be clear, this work was focused on highlighting the technical implications of current deliverability modeling assumptions, and not thornier issues like transmission cost allocation. But the current analytical approach of excluding ERIS resources altogether and applying deliverability limits to wind and solar across every hour of the year in resource adequacy models doesn’t work either. Alternative approaches include incorporating historical curtailment data in place of static deliverability limits or shifting from zonal to nodal transmission constraints within resource adequacy models.
Expert takeaways from ESIG’s Deliverability and resource adequacy webinar
The analysis was presented on a September ESIG webinar by Derek Stenclik of Telos Energy, who was joined by Bruce Rew, formerly of Southwest Power Pool, Tyler Norris, head of Market Innovation of Advanced Energy at Google, and Zander Bischof of independent power producer MN8 Energy. Following a great presentation by Derek, James Okulla of ESIG masterfully navigated an intricate conversation digging further into the relationship between transmission deliverability studies, resource adequacy and real-time operations.
The panel discussion started with an early reminder that models are always an abstraction and in real-time, SCED does not distinguish between ERIS and NRIS resources. Deliverability study criteria are disparate across jurisdictions and undefined by FERC or NERC, which means that the links between deliverability and RA are very jurisdiction specific. It’s important to remember that the distinction between firm and non-firm resources is fundamentally about incentivizing building transmission and a reliable system depends on being able to build and pay for new transmission. As one panelist put it, “recognizing the value that all resources bring” is the most important thing to change about deliverability.
It was a great discussion but don’t take our word for it — watch the webinar for yourself! We’re continuing this work with nodal analysis to better understand how curtailment risk correlates with resource adequacy events. Hopefully this is one of a series of subsequent conversations we’ll have on deliverability and resource adequacy.