What the ride-through rule actually requires

Three days ago I wrote about grid-forming inverters and ended on a complaint: the NERC documents making the case were white papers. Recommendations. Nothing in them compelled anyone to do anything.

This one is different. NERC Reliability Standard PRC-029-1 is mandatory and enforceable, it carries violation risk factors and severity levels, and it becomes effective on 1 October 2026. If you own an inverter-based resource, this is the rule that says your plant is no longer allowed to protect itself first.

So I went and read it. It is fourteen pages. Two of the claims that appear in almost every consultant summary of it are not supported by the text.

What the standard says it is for

The title is "Frequency and Voltage Ride-through Requirements for Inverter-based Resources." The stated purpose is one sentence: "To ensure that IBRs Ride-through to support the Bulk Power System (BPS) during and after defined frequency and voltage excursions."

"Ride-through" is a newly defined term in the NERC glossary, and FERC's approving order records what NERC proposed it to mean: that "the plant/facility remains connected and continues to operate through voltage or frequency system disturbances."

Applicability lands on the Generator Owner, and covers Bulk Electric System IBRs plus non-BES IBRs that "either have or contribute to an aggregate nameplate capacity of greater than or equal to 20 MVA, connected through a system designed primarily for delivering such capacity to a common point of connection at a voltage greater than or equal to 60 kV." Note the phrase "or contribute to." A small facility sharing a point of connection is not automatically outside the fence.

Correction one: the standard does not cite IEEE 2800

Search the fourteen pages of PRC-029-1 for the string "IEEE" and you get zero hits. Not one.

This matters because the sentence "PRC-029-1 adopts the IEEE 2800-2022 ride-through curves" is close to universal in the trade and consulting coverage. It is a reasonable shorthand for a family resemblance, and it is wrong as a statement about the enforceable instrument, which contains its own tables and points at nothing outside itself.

The history explains why, and FERC put it on the record. In the order preceding this one, the Commission "declined to direct NERC to specifically reference" IEEE standards, concluding that the record "provided no support for the conclusion that the performance requirements of IEEE standard 2800-2022 ... are preferable to NERC's Reliability Standards." FERC instead gave NERC discretion over whether and how to reference them. NERC's answer, visible in the document itself, was: not at all.

If you are writing a compliance procedure, this is not pedantry. An IEEE standard and a NERC standard have different revision cycles, different curves in their corners, and only one of them can be audited against you. Cite the one that can.

Correction two: momentary cessation is not banned everywhere

The second refrain is that the standard "prohibits momentary cessation." What it actually does is draw a map with three regions and then say different things in each one.

For everything that is not type 3 or type 4 wind, which includes photovoltaic plants and battery storage, the voltage map at the high side of the main power transformer runs like this:

• At or below 1.05 per unit and at or above 0.90, the continuous operation region. Ride through indefinitely.
• Below 0.90, the mandatory operation region, with a minimum ride-through time that shortens as the sag deepens: 6.00 seconds below 0.90, 3.00 seconds below 0.70, 1.20 seconds below 0.50, and 0.32 seconds below 0.25.
• Below 0.10, the permissive operation region, 0.32 seconds.
• Above 1.20 per unit is outside all of it, in what the standard calls the "may Ride-through zone." There, you may leave.

Wind gets its own, shorter table: 3.00 seconds below 0.90 rather than 6.00, and 0.16 seconds at the deep end rather than 0.32.

Now the part that gets flattened. In the mandatory region the plant must "exchange current, up to the maximum capability to provide voltage support," with reactive power priority by default. But in the permissive region, Requirement R2.3 says each IBR "may operate in current blocking mode if necessary to avoid tripping." Current blocking is momentary cessation. It is permitted, in that one band, and it comes with a clock: if the inverter enters current blocking mode "it shall restart current exchange in less than or equal to five cycles of positive sequence voltage returning to a continuous operation region or mandatory operation region."

So the accurate version is narrower and more useful than the slogan. Momentary cessation is forbidden where the grid still has a voltage worth supporting, and tolerated as a survival measure when the voltage is essentially gone, with a five-cycle leash on coming back.

The requirement is a budget, not a threshold

Here is where the document stops being a protection standard and starts being something I recognize from a very different building.

A naive reading of those tables is that they are trip thresholds: below this voltage for longer than this time, you may disconnect. Read the notes and the obligation changes shape entirely. Note 7 of Attachment 1: "If the voltage is continuously varying over time, it is necessary to add the duration within each band ... over any 10 second time period." Note 8: the specified durations "are cumulative over one or more disturbances within any 10 second time period."

Cumulative. Your 1.20 seconds below 0.50 per unit is not per event, it is a balance you spend down across a ten-second window, across however many separate disturbances land inside it. A fault, a reclose, and a second fault are not three independent tests. They are three withdrawals from one account.

And then note 9, which is the escape hatch: "The IBR may trip for more than four deviations of the applicable voltage ... outside of the continuous operation region within any 10 second time period." Four excursions in ten seconds and the obligation lapses. That number is not in any summary I read, and it is the single most operationally important number in the attachment, because a grid having a genuinely bad day produces exactly that pattern.

The frequency side works the same way on a longer clock. The must-ride-through band runs from 58.8 Hz to 61.2 Hz continuously, with 299 seconds required in the bands just outside it, above 61.8 Hz and below 57.0 Hz you may trip, and the durations are "cumulative over one or more disturbances within a 10-minute time period." Requirement R3 caps the whole obligation at a rate of change of frequency at or below 5 Hz per second.

The measurement rules are the real specification

This is the part I would have missed if I had read a summary, and it is the reason I think this standard is more interesting than its coverage.

Buried in the attachment notes are three sentences that do not regulate what the plant does at all. They regulate how it is allowed to know.

Attachment 2, note 2: "Frequency is measured over a period of time (typically 3-6 cycles) to calculate system frequency at the high-side of the main power transformer." Note 3: "Instantaneous or single points of measurement may not be used in the determination of control settings." And Attachment 1, note 10: "Instantaneous trip settings based on instantaneously calculated voltage measurements with less than filtering lengths of one cycle (16.6 millisecond) are not permissible."

I spent years doing microwave spectroscopy, where the first question about any reported number is what window it was integrated over, because a threshold without a specified measurement window is not a specification. It is a mood. Two labs can apply the identical numeric limit and disagree by an order of magnitude purely on averaging time, and neither of them is lying.

NERC clearly reached the same conclusion about inverters, and it is a sharper insight than it first appears. Most of the tripping this standard exists to stop was never a decision to trip. It was a fast protection element reacting to a transient that a one-cycle filter would have smoothed away, in a device fully capable of riding the event out. By outlawing sub-cycle voltage measurement for trip settings and single-point frequency measurement for control settings, the standard removes a large class of failures without asking anyone to buy anything. It is the cheapest paragraph in the document.

There is a companion footnote in Requirement R1 that lands in the same place, and it connects directly to the last report. The must-ride-through obligation, footnote 1 says, "Includes no tripping associated with phase lock loop loss of synchronism." That is the phase-locked loop I described three days ago as the thing I used to attack for a living. The standard now says: when your PLL loses the plot, that is your problem, not the grid's. You may not exit because your own estimator got confused.

The exemption is the interesting requirement

Requirement R4 lets an owner claim relief for an IBR already in service that has "known hardware limitations." It is rated a Lower violation risk factor, it is entirely paperwork, and it is the most quietly aggressive thing in the standard.

To claim it you must file, within twelve months of the effective date, a package identifying the plant, the specific pieces of hardware causing the limitation, which ride-through criteria it cannot meet, and then this:

"Technical documentation verifying the limitation is due to hardware that would need to be physically replaced to meet all Ride-through criteria, and that the limitation cannot be remedied by software updates or setting changes."

Read that as an engineer and you can see exactly what it is designed to smoke out. The industry's ride-through problem has always been a mixture of two very different things wearing the same clothes: inverters that cannot, and inverters that were configured not to. The first is a hardware fact. The second is a settings file, usually inherited from a commissioning default nobody revisited. R4 forces every owner seeking relief to state, in writing, to their Compliance Enforcement Authority, which of the two they have. And R4.3.1 closes it behind you: when the offending hardware is eventually replaced, "the exemption for that Ride-through criteria no longer applies."

The evidence retention periods tell you how seriously this is meant. Three years for the performance requirements, five for the exemption paperwork.

What is not settled

I would be doing the same thing I am criticizing if I presented this as finished, so here is the open edge.

FERC approved the standard and simultaneously directed NERC to fix parts of it. The order tells NERC to clarify the documentation requirements for legacy equipment supporting an exemption request, to consider two further exception and exemption issues raised by commenters, and to submit an informational filing assessing the reliability impact of the exemptions.

The unresolved fight is about timing, and it is real. Commenters argued that basing exemption eligibility on whether a plant is in service by the effective date ignores how long these projects take. Clean Energy Associations told the Commission that equipment manufacturers had said at a technical conference that "the product design and development timeline for IBRs is at least five years." A project that locked its inverter specification in 2023 against the rules of 2023 does not get to un-lock it because a standard arrived in 2025. An industry-filed request to revise the standard, submitted in November 2025, put a number on the exposure: a member survey identifying 22.1 GW of inverter-based resources with signed interconnection agreements and procurement, construction or design agreements, not yet in service, that its authors say would need limited exemptions.

Treat that 22.1 GW as what it is, an advocacy filing by a trade association arguing its members' case, not a finding by NERC or FERC. I include it because the shape of the problem is genuine regardless of whose number you use, and because a rule whose exemption process is under active revision two months before it binds is a rule worth watching rather than assuming.

One scope limit on my own reporting: I confirmed the 1 October 2026 effective date from NERC's own Order 909 workshop materials. Several summaries also give a separate later date for non-BES resources. I could not confirm that from a primary document, so I am not repeating it. Check the implementation plan before you plan against it.

This is the seam I work in. The gap between what an inverter can do, what its settings currently say it will do, and what a document now requires it to do is not three questions, it is one, and it is answered in firmware and in a controls file rather than in steel. I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator. I do not own that company and earn nothing from this link; I flag it because it is a field I build in and not only write about. Full policy here.

The signal

The story told about PRC-029-1 is that regulators finally ordered renewables to stop falling over. That is roughly true and almost entirely uninteresting.

The document itself is doing something more specific. It concedes the deep-sag case, permits current blocking where the voltage is nearly gone, hands back the obligation after four excursions in ten seconds, and offers a hardware exemption to anyone who will put their name on the claim. Against those concessions it takes one thing away completely: the freedom to decide, using a measurement of your own choosing over a window of your own choosing, that you have seen enough.

Frequency over three to six cycles. No single points. No sub-cycle voltage filtering on trip settings. Durations totted up across a ten-second window rather than judged event by event.

That is a measurement standard, and it was written by people who understood that the fleet's real failure mode was never a shortage of capability. It was a hundred thousand devices each independently deciding, on the strength of a very short look, that the grid was no longer worth staying for.