Getty Images
PJM proposes ride-through rules for large data center loads
PJM's proposed ride-through rules would require large data centers to stay connected during grid disturbances rather than switching to backup.
PJM is proposing new ride-through rules for large computational loads after a nearly 4,000 MW data center load in Northern Virginia was disconnected during a July transmission disturbance.
For data center operators, the rules could mean changes to how facilities respond to grid disturbances, including how uninterruptible power supply (UPS) systems, backup generation and other electrical controls are configured. PJM is a regional transmission organization that operates the high-voltage power grid and wholesale electricity markets across 13 states and Washington, D.C. It's asking large data center loads to demonstrate they can operate predictably when the grid is under stress, rather than simply switching to backup power to protect their equipment.
"The reliability problem is no longer simply whether the grid can supply the load," said Neil Osnato, founder of Persistence Analytics Group, an infrastructure advisory firm. "It is whether the load will behave as the grid assumes once it's connected."
PJM presented the proposed rules September 8 as part of its response to a June Federal Energy Regulatory Commission (FERC) order directing the grid operator to examine whether its tariff adequately addresses the reliability impacts of large-load interconnections.
FERC specifically raised questions about ride-through performance, ramp-rate limitations and load-restoration behavior.
Why it matters
The July 22 event showed why the issue matters. Nearly 4,000 MW of Northern Virginia data center load disconnected and transferred to backup generation. PJM and Dominion Energy Virginia then had to manage the resulting imbalance between generation and load, along with increases in system frequency and transmission voltage.
"A data center understandably designs around protecting its own equipment and maintaining computing availability," Osnato said. "But at multi-gigawatt scale, a protection action that makes sense from the facility's perspective can become a bulk-system event."
PJM isn't alone. FERC has asked it to review ramp and ride-through rules, while the North American Electric Reliability Corporation (NERC) is developing a mandatory reliability standard for large loads as part of its work developing and enforcing reliability standards for the bulk electric system across most of the U.S., Canada and part of Mexico.
"The behavior of large loads has broader reliability impacts for the grid, and so they can't simply act as though they can switch to backup power whenever it suits them," said Elizabeth K. Whitney, managing principal at Meguire Whitney, a federal energy-policy lobbying and strategic advisory firm. "They will need to design more intelligent systems and better communicate with grid operators."
What PJM wants data centers to do
Under PJM's draft tariff and manual language, a large computational load would include data centers and other computational facilities connected at 100 kilovolts or higher and aggregating to at least 50 MW.
New facilities entering the PJM study process on or after November 1 would have to demonstrate compliance as a condition of interconnection and operation. Material modifications to existing facilities could also trigger the rules.
The proposed voltage rules would require large computational loads to remain connected during specified voltage disturbances rather than experience an involuntary block loss of load.
During a voltage disturbance below 0.90 per unit but above 0.50 per unit, a facility could reduce its power use in proportion to the voltage reduction. Once voltage returns to the normal range, the facility would have to restore at least 90% of its load within two seconds.
The draft also limits current during voltage sags and addresses protective functions, filtered measurements and intentional time delays.
PJM would restrict schemes that automatically disconnect a computational load or transfer it to backup generation solely because a specified number of voltage sags or swells have occurred when the facility is otherwise expected to ride through the disturbance.
The rules would also require large computational loads to stay connected during certain frequency disturbances and keep their power use within 10% of their normal level. The goal is to prevent a grid disturbance from being amplified when large blocks of data center demand disappear at once.
PJM's proposal is similar to rules being developed elsewhere. The Electric Reliability Council of Texas (ERCOT) has approved ride-through rules for large computational loads that include a two-second recovery requirement under specified voltage conditions. PJM's presentation also compares its proposal with a draft MISO framework and a NERC working group proposal.
Osnato said the key question isn't whether two-second recovery is theoretically possible. It's whether a particular data center, as actually designed and configured, can demonstrate that it will perform that way.
That could require operators to examine the interaction among UPS systems, protective relays, transfer logic, backup generation, cooling systems and restart controls.
"PJM is prescribing the performance outcome, not a particular piece of equipment," Osnato said. That also raises a question about how operators will demonstrate compliance.
"A model, an engineering representation, a commissioning test and demonstrated operating behavior aren't necessarily equivalent forms of evidence," Osnato said.
Rules could extend beyond new data centers
PJM isn't limiting the proposed rules to future greenfield campuses. The draft creates a transition category for facilities that had already entered an interconnection study, signed an agreement, started construction or reached commercial operation before November 1.
Those facilities generally would have until Nov. 1, 2027, or one year after commercial operation, whichever is later, to comply.
PJM could require faster compliance from an existing facility if it determines that the load is contributing to transient voltage instability, cascading loss of load, voltage-collapse risk or another reliability violation.
Whitney doesn't expect the rules to have major effects on data center site selection or project timelines. But, she said, the rules could increase project costs by requiring additional equipment or software.
The rules could also push developers toward more sophisticated controls rather than fundamentally different facility designs.
"Smart data center developers probably saw the writing on the wall that just having backup generation or on-site batteries for their own use wouldn't long be sufficient," Whitney said. "These requirements basically tell data centers that just because they have their own juice, that doesn't mean they can just take their straw out of the grid's milkshake whenever they like."
Whitney said the changes are likely to be more about facility behavior and additional technology than a complete redesign of data centers. Going completely off grid would create a separate set of challenges, she said.
The ride-through proposal is only the first part of a broader set of rules PJM is developing for large computational loads. The grid operator plans to develop large-load ramp-rate rules in the first quarter of 2027 and a large computational load design envelope beginning in the second quarter.
The rules could affect how developers design and test their electrical systems, adding another layer to the interconnection process. More broadly, they could change what PJM expects from a data center after it connects to the grid.
"Energization is no longer the final validation gate," Osnato said.
PJM plans to continue reviewing the proposed tariff and manual language with stakeholders in September and October and target a filing with FERC in November.
Shane Snider is a senior news writer at TechTarget, covering AI infrastructure, hyperscale data centers, cloud platforms, and the power and energy systems driving modern compute expansion. You can reach Shane at [email protected] or on LinkedIn.