Skip to main content
v2026.11,986 entries · CC-BY 4.0

Editorial · CASRAI · Research data infrastructure

PJM Proposes Mandatory Grid Ride-Through Rules to FERC for Large Data Centers

PJM filed proposed tariff rules with FERC requiring 50 MW+ data centers to ride through grid voltage and frequency disturbances to prevent mass outages.

Published 16 Nov 2026· 7 minute read

Ask CASRAI · free to try

Ask about this story

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

CASRAI is the reference for research administration — bookmark it for the next question.

On November 16, 2026, regional transmission organization PJM Interconnection filed proposed tariff revisions with the Federal Energy Regulatory Commission (FERC) to establish mandatory voltage and frequency ride-through standards for Large Computational Loads. Prompted by a July 2026 disturbance in Northern Virginia where nearly 3.9 gigawatts of data center demand abruptly disconnected, the rules mandate that facilities exceeding 50 megawatts remain connected during routine transmission faults, directly impacting commercial hyperscalers and university high-performance computing centers.

The July 22 Virginia Disturbance and the Catalyst for Reform

The regulatory filing responds directly to a major power system disturbance on July 22, 2026, within Dominion Energy’s transmission zone in Northern Virginia—the world’s densest concentration of data centers. At 7:56 a.m., a mechanical failure removed a 230-kilovolt transmission line from service. While utility protection equipment cleared the electrical fault within normal clearance cycles, the localized voltage dip caused an automated, simultaneous disconnection across neighboring computational facilities.

Rather than riding through the momentary voltage fluctuation, automated protection systems—including static transfer switches and uninterruptible power supply (UPS) modules—instantly detached server halls from the grid and transferred them to on-site diesel generators and battery banks. Dominion Energy and PJM operational data revealed an immediate initial drop of 2,970 megawatts, followed moments later by an additional 1,099 megawatts, shedding nearly 3.9 gigawatts of demand without warning.

The sudden loss of demand caused PJM’s system load to plunge from 99,984 megawatts to 96,205 megawatts. Because generation momentarily exceeded consumption across the Eastern Interconnection, system frequency jumped to 60.092 hertz, while local voltages spiked to 1.138 per unit (p.u.). PJM system dispatchers stabilized grid parameters within nine minutes using reactive power reserves and generation curtailment. However, engineers warned that under tighter system conditions, such multi-gigawatt disconnects could trigger cascading generator trips across the Eastern Interconnection.

Proposed Tariff Revisions and the Large Computational Load Definition

To prevent future load-shedding shocks, PJM’s November 16 filing amends the PJM Open Access Transmission Tariff and Manual 14B (Attachment D-2) to establish a new operational category: Large Computational Load (LCL). Under the revised tariff language, an LCL is defined as any customer facility—including data centers, cryptocurrency mining installations, and artificial intelligence compute campuses—with an aggregate demand of 50 megawatts or greater that serves at least 25 megawatts of its maximum load through power electronic conversion equipment.

The focus on power electronics addresses the unique physical characteristics of computational infrastructure. Unlike conventional industrial facilities that rely on heavy induction motors with mechanical inertia, modern GPU clusters and server racks utilize solid-state rectifiers and switch-mode power supplies. When transmission voltage sags, power supply controls can trip within milliseconds to protect sensitive silicon, creating massive bulk-power disturbances when replicated across dozens of adjacent facilities. PJM currently serves approximately 10 gigawatts of large computational loads and forecasts that regional computational demand will expand to 30 gigawatts by 2030.

Mandatory Voltage and Frequency Ride-Through Thresholds

Under the proposed framework, all current and prospective Large Computational Loads must demonstrate the capability to remain electrically connected and maintain continuous operation across defined voltage and frequency boundaries rather than transferring to backup generation during routine transmission faults.

Disturbance Parameter Operating Band / Threshold Minimum Required Ride-Through Duration Facility Operational Mandate
Overvoltage (High Surge) V > 1.20 p.u. Discretionary May ride through or disconnect to protect facility equipment
Overvoltage (Moderate) 1.10 p.u. < V ≤ 1.20 p.u. 1.0 second Must remain connected; continuous operation required
Nominal Voltage Range 0.90 p.u. ≤ V ≤ 1.10 p.u. Continuous Standard continuous normal operations
Undervoltage (Mild Sag) 0.80 p.u. ≤ V < 0.90 p.u. 2.0 seconds Must remain connected; active power scale-down permitted
Undervoltage (Moderate Sag) 0.50 p.u. ≤ V < 0.80 p.u. 0.5 seconds Must remain connected; active power scale-down permitted
Undervoltage (Severe Fault) V < 0.50 p.u. 0.25 seconds Must ride through cleared transmission faults before tripping
Overfrequency f > 61.8 Hz Discretionary Permitted to disconnect or scale back load
High Frequency Band 61.2 Hz < f ≤ 61.8 Hz 299 seconds Mandatory ride-through without automated load shedding
Nominal Frequency Band 58.8 Hz ≤ f ≤ 61.2 Hz Continuous Standard continuous operating frequency
Low Frequency Band 57.0 Hz ≤ f < 58.8 Hz 299 seconds Mandatory ride-through to support bulk grid stability
Underfrequency f < 57.0 Hz Discretionary Permitted to trip to prevent electrical equipment damage

The rules provide operational flexibility: during voltage sags between 0.50 p.u. and 0.90 p.u., facilities may scale down active power consumption proportionally to the voltage drop to reduce electrical stress. However, once system voltage recovers, the facility must restore at least 90 percent of its pre-disturbance active power demand within two seconds.

Regulatory Alignment with Federal Directives Under FERC and NERC

PJM’s tariff filing aligns with broader federal initiatives examining large-load grid integration. On July 16, 2026, FERC issued an order in Docket No. RD26-7-000 directing the North American Electric Reliability Corporation (NERC) to develop mandatory reliability standards for computational loads across the Bulk-Power System by December 31, 2026.

Federal regulators have compared the sudden growth of computational loads to the earlier integration of inverter-based resources like solar and wind generators, which necessitated FERC Order 901 to address uncoordinated inverter tripping during faults. PJM’s filing establishes baseline regional requirements in Manual 14B while allowing individual transmission owners—including Dominion Energy Virginia, American Electric Power, and Commonwealth Edison—to enforce more stringent local standards where necessary.

Industry Perspectives and Engineering Trade-Offs

The proposed mandates highlight an engineering tension between server uptime and grid reliability. Data center power systems have historically been configured to isolate servers from any power quality anomaly. Neil Osnato, founder of infrastructure advisory firm Persistence Analytics Group, observed that the core reliability challenge has shifted: the issue is no longer just whether the grid can supply energy, but whether large loads behave predictably once connected. Osnato noted that while disconnecting protects individual facility availability, simultaneous multi-gigawatt load shedding creates a systemic emergency for the entire interconnection.

Hyperscale operators and equipment vendors have participated in PJM Planning Committee sessions to assess implementation details. Compliance will require testing and firmware updates across modular UPS systems, static switches, and battery energy storage systems to prevent early tripping while safeguarding server power supplies from transient voltage surges.

Implications for Universities, Research Computing, and Research Offices

While discussions often focus on commercial cloud providers, PJM’s ride-through rules carry significant operational implications for research universities and high-performance computing (HPC) facilities. Multi-megawatt academic computing clusters—such as those funded under NSF CISE cyberinfrastructure grants—increasingly approach utility-scale thresholds when aggregated on campus substations.

First, university facilities planners and research computing directors must evaluate campus interconnection boundaries. If an institution aggregates supercomputing clusters, advanced data centers, and research laboratories to exceed 50 megawatts on a single utility interconnection (with 25 megawatts or more served through power electronics), the site could be designated an LCL. Campus engineering teams must review substation protection curves and uninterruptible power architectures to ensure compliance with PJM’s two-second ride-through mandate.

Second, riding through transmission disturbances requires enhanced computational resilience. Because grid sags may propagate as brief voltage fluctuations before clearing, academic computing administrators must ensure that mission-critical simulations and multi-day model training utilize automated checkpointing. Containerized workflows deployed on high-performance clusters—evaluated through frameworks like Apptainer vs Docker for HPC research—must save intermediate state to resilient shared storage. Similarly, automated data staging pipelines using secure cluster synchronization protocols (rsync vs scp on HPC clusters) should incorporate fault-tolerant retry logic to withstand transient substation voltage variations.

Third, university sustainability and energy offices must strengthen power telemetry. With transmission operators demanding active grid coordination, academic data centers will need detailed monitoring of power draw, power factor, and harmonics. Computational telemetry tools such as CodeCarbon, which track energy consumption and compute emissions across intensive GPU workloads, offer researchers and campus energy managers granular data to coordinate efficiency and power stability.

Implementation Timeline and Next Steps

PJM has proposed an effective date of January 16, 2027. Under this timeline, any new Large Computational Load submitting an interconnection request or executing an interconnection agreement after that date must demonstrate full ride-through compliance as a condition of commercial operation.

For existing facilities and projects currently under development, PJM proposes a phased transition: facilities must achieve compliance within one year of the effective date or within one year of commercial operation, whichever is later. During this transition, operators must coordinate with transmission owners to review protection settings and verify compliance documentation. Throughout 2027, PJM and its stakeholders will continue developing complementary performance standards, including maximum ramp-rate limits and active power recovery controls.

Sources

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →