Load Comes Off Automatically at a Frequency Chosen Years Before the Moment It Trips

Alternating-current frequency is one shared number across an interconnected grid. Every generator, motor and protective relay inside the same interconnection sees the same value at nearly the same instant, and it moves whenever generation and load stop matching. For the Texas Interconnection, NERC's frequency-response analysis starts its arithmetic from a figure it labels the starting frequency, 59.970 Hz, and works down to a low-frequency limit of 59.300 Hz.

That limit is not an arbitrary floor. The report states that “the low-frequency limits to be used for the IFRO calculations (Table 2.2) should be the highest step in the Interconnection for regionally approved UFLS systems.” That is written as an instruction for choosing the number, not as a finding about any particular program, and the report does not say which program each figure comes from. The same four values are then carried into the calculation table beside the starting frequencies, where the row is called the minimum frequency limit. The gap between the two is printed there as 0.670 Hz for Texas, which is a distance on the table rather than the amount the obligation is sized on.

The same limit is 59.500 Hz in the Eastern and Western Interconnections and 58.500 Hz in Quebec. Those numbers are inputs to a calculation that decides how much frequency response each interconnection must be able to deliver.

The Chain

The sequence has five links, and only the last one is visible from a kitchen window.

First, an imbalance appears. A generating unit, a plant, or a transmission element carrying imported power is lost. Load does not change in that instant, so the difference comes out of the rotating mass already spinning on the system, and frequency falls.

Second, the decline is measured in a defined window. NERC's frequency response analysis does not use an intuitive definition of “how far it fell.” It fixes a pre-event baseline from the seconds before the event, then takes Point C, which it defines as “the lowest frequency experienced within the first 20 seconds following the start of a frequency event.” A stabilizing value is averaged from a later window. These are conventions, and they matter later.

Third, governors and inverter controls respond. Whatever arrests the fall has to arrive inside those first seconds. This is the quantity that gets an obligation attached to it.

Fourth, if the fall keeps going, it reaches the highest step of the load-shedding program. Underfrequency load shedding is not dispatched. It is armed in advance, and it operates on the frequency reading alone.

Fifth, a block of load is disconnected. The relay's job is to arrest the decline by removing demand faster than the system can lose more supply. NERC's continent-wide standard describes the purpose as being “to establish design and documentation requirements for automatic underfrequency load shedding (UFLS) programs to arrest declining frequency, assist recovery of frequency following underfrequency events and provide last resort system preservation measures.”

The Obligation Is One Division

Every year NERC publishes a Frequency Response Annual Analysis that recommends an Interconnection Frequency Response Obligation, or IFRO, for the operating year ahead. The obligation is stated in megawatts per 0.1 Hz: how much power the interconnection has to produce, automatically, for each tenth of a hertz the frequency drops.

The recommended values for Operating Year 2026 are built from three published inputs.

• Eastern Interconnection. Starting frequency 59.971 Hz; minimum frequency limit 59.500 Hz; maximum delta frequency 0.420 Hz; resource loss criterion 3,875 MW; obligation 923 MW per 0.1 Hz.
• Western Interconnection. Starting frequency 59.970 Hz; minimum frequency limit 59.500 Hz; maximum delta frequency 0.280 Hz; resource loss criterion 2,918 MW; obligation 1,042 MW per 0.1 Hz.
• Texas Interconnection. Starting frequency 59.970 Hz; minimum frequency limit 59.300 Hz; maximum delta frequency 0.405 Hz; resource loss criterion 2,805 MW, against which a load-resource credit of 947 MW is applied; obligation 459 MW per 0.1 Hz.
• Quebec Interconnection. Starting frequency 59.964 Hz; minimum frequency limit 58.500 Hz; maximum delta frequency 0.947 Hz; resource loss criterion 2,000 MW; obligation 211 MW per 0.1 Hz.

The four obligations reproduce exactly as one division. Subtract the load-resource credit from the resource loss criterion, then divide by ten times the maximum delta frequency. Eastern: 3,875 divided by 4.20 is 922.6. Western: 2,918 divided by 2.80 is 1,042.1. Texas: 2,805 minus 947 is 1,858, divided by 4.05 is 458.8. Quebec: 2,000 divided by 9.47 is 211.2. Each rounds to the published figure. The obligations are published as negative numbers, because frequency response is delivered while frequency is falling; the magnitudes are what compare.

The design rule sitting behind the middle term is stated plainly in the same report: “A frequency event should not activate the first stage of regionally approved UFLS systems within the Interconnection.” That is what the maximum delta frequency encodes. It is a budget for how far frequency may fall on the worst credible resource loss, set so the fall stops above the relays.

The budget is built in stages, and the report prints each one. The base delta frequency is the plain distance between the two frequencies: 0.471 Hz in the East, 0.470 Hz in the West, 0.670 Hz in Texas, 1.464 Hz in Quebec. An adjustment of 0.010 Hz is then applied to the Eastern Interconnection alone, which gives calculated maximum allowable deltas of 0.461, 0.470, 0.670 and 1.464 Hz. The figures actually carried into the obligation are smaller again, on a row labelled as the maximum delta frequency per procedure: 0.420, 0.280, 0.405 and 0.947 Hz.

Subtracting those from the starting frequencies puts the sizing points at 59.551, 59.690, 59.565 and 59.017 Hz. In each of the four, that point sits above the low-frequency limit rather than on it. The distance left over runs from 0.051 Hz in the East to 0.517 Hz in Quebec, a spread of roughly ten times. Quebec's sizing point lands a little above 59.0 Hz while its limit waits down at 58.500 Hz.

One property of these numbers is easy to assume wrongly: the deltas are not re-derived each year. The report notes that “MDF is the Maximum Delta Frequency for the specific interconnection as determined in the 2017 Frequency Response Annual Analysis (FRAA).” The resource loss figures move; the frequency budget they are divided by has been carried forward. The full recommendation is set out in the North American Electric Reliability Corporation, 2025 Frequency Response Annual Analysis.

Set Points Are a Window, Not a Table

There is no national schedule of underfrequency set points. The continent-wide standard assigns the design work to Planning Coordinators and then constrains it; regional standards narrow the constraints further. What a region publishes is a window.

SERC's regional standard is a compact example. It requires a program to “shed load with a minimum of three frequency set points.” It bounds the top of the ladder: “The highest frequency set point for relays used to arrest frequency decline shall be no lower than 59.3 Hz and not higher than 59.6 Hz.” It bounds the bottom, at no lower than 58.2 Hz, and it requires spacing between steps of at least 0.2 Hz and no more than 0.5 Hz. On timing it is blunt: “Time delay setting shall be at least six cycles (0.1 seconds).”

Six cycles at 60 Hz is a tenth of a second. What the standard sets there is a floor rather than a design value: the delay has to be at least that long, and nothing in the requirement stops a program from setting a longer one. The delay actually used on a given circuit is a program parameter, not a number this standard publishes.

The window is also wider than the single figure the frequency-response calculation uses. SERC's region sits inside the Eastern Interconnection, where the low-frequency limit carried into the obligation is 59.5 Hz, while this regional standard would permit a highest set point as high as 59.6 Hz. Neither document identifies which program supplies the figure the calculation uses.

The window itself has moved. An earlier revision, PRC-006-SERC-02, set the top of the ladder between 59.3 Hz and 59.5 Hz and the bottom at no lower than 58.4 Hz; the later PRC-006-SERC-3 widens both ends. Which of the two governs is not settled by the documents themselves, because the later one states that it is “Effective for SERC Region applicable Registered Entities on the first day of the first calendar quarter after approved by FERC.” A set point outside the allowed range under one revision can be inside it under the other, which is one reason a set point is best read with a date attached. Both texts are posted: North American Electric Reliability Corporation, Regional Reliability Standard PRC-006-SERC-3, Automatic Underfrequency Load Shedding Requirements and the earlier North American Electric Reliability Corporation, Regional Reliability Standard PRC-006-SERC-02.

The Second Clock

Load shedding is one half of a race. The other half is what generators are permitted to do while frequency is low.

A separate standard sets the ride-through requirement. Its stated purpose is “to set protection such that generating resource(s) remain connected during defined frequency and voltage excursions in support of the Bulk Electric System (BES).” Its attachment gives, for each interconnection, a low-frequency band and a minimum time the resource has to stay connected inside it.

In the Texas Interconnection the bands run: continuous operation above 59.4 Hz; 540 seconds at 59.4 Hz and below; 30 seconds at 58.4 Hz and below; 2 seconds at 58.0 Hz and below; instantaneous at 57.5 Hz and below. The Western Interconnection ladder is shorter at the top and longer at the bottom: continuous above 59.4 Hz, then 180 seconds in the band where Texas requires 540, then 30 seconds at 58.4 Hz, 7.5 seconds at 57.8 Hz, 0.75 seconds at 57.3 Hz, and instantaneous only at 57.0 Hz and below.

Reading a ladder against its own low-frequency limit is the point. In Texas that limit, 59.300 Hz, sits inside the band where a generating resource is still required to stay connected for 540 seconds, or nine minutes. That 540-second requirement does not run the whole way down: it is replaced by 30 seconds at 58.4 Hz and below, 2 seconds at 58.0 Hz and below, and nothing at all at 57.5 Hz and below. Load shedding is expected to arrest the decline in the band where the 540-second requirement still applies.

The standard is careful about what its curve does not say. Its attachment carries the note that “the area outside the ‘No Trip Zone’ is not a ‘Must Trip Zone.’” Falling past a boundary does not command a generator to disconnect; it removes the requirement that the generator stay. The full attachment appears in NERC Reliability Standard PRC-024-3, Frequency and Voltage Protection Settings for Generating Resources.

What the Design Has to Survive on Paper

Before any of this is armed, it has to pass a simulation with a stated severity. The continent-wide standard requires each Planning Coordinator to develop a program “that meets the following performance characteristics in simulations of underfrequency conditions resulting from an imbalance scenario, where an imbalance = [(load — actual generation output) / (load)], of up to 25 percent within the identified island(s).”

An island losing a quarter of its supply is the design case. One of the performance characteristics is a boundary curve the simulated frequency has to stay above: no lower than 58.0 Hz in the first two seconds, then a rising limit through the first minute, and no lower than 59.3 Hz after 60 seconds. The requirement holds “either for 60 seconds or until a steady-state condition between 59.3 Hz and 60.7 Hz is reached.” A second characteristic caps the overfrequency side, and a third limits volts per hertz at generator buses. The curves are floors and ceilings the design has to clear, not predictions of any particular event.

The review interval is written down as well. Each Planning Coordinator “shall conduct and document a UFLS design assessment at least once every five years that determines through dynamic simulation whether the UFLS program design meets the performance characteristics in Requirement R3 for each island identified in Requirement R2.” Separately, where a BES islanding event produces frequency excursions below the initializing set points of the program, an assessment of the equipment's performance and the program's effectiveness is due within one year of event actuation. The standard text is published as NERC Reliability Standard PRC-006-5, Automatic Underfrequency Load Shedding.

Five years and one year are the clocks on the paperwork. Six cycles is the shortest the relay's own clock is allowed to be in one region that publishes a floor. The gap between those timescales is where a setting quietly ages.

Where This Doesn't Apply

These numbers describe the bulk power system, not a service address. A relay arms and drops a distribution circuit, and which customers sit on which circuit is decided in distribution planning under a different set of rules.

Automatic underfrequency shedding is also not the rotating outage that gets announced during a shortage. Those are ordered by operators, they happen while frequency is still being held, and they run on a timescale of minutes to hours. The relay ladder described here sits below that, and reaching it means the ordered measures did not clear the imbalance in time.

The frequency-response figures are a recommendation for a single operating year and are revised annually; the resource loss criteria in particular move as the fleet changes. The regional set-point window quoted here is SERC's, whose region lies inside the Eastern Interconnection; it does not govern the Texas Interconnection, and other regions publish their own requirements. Quebec's interconnection lies largely outside the United States and its thresholds reflect a smaller, differently built system, which is part of why its frequency budget is more than twice any of the others.

Finally, Point C and the stabilizing average are measurement conventions rather than physical events. The analysis itself allows for a secondary dip that arrives after the measurement window closes and can be lower than the value the convention records. A frequency event that looks contained in the published statistics can still have gone somewhere the statistic was not built to see.