From Nadir to Nominal: The Four Stages of Frequency Response After a Generation Loss in US Interconnections

November 05, 2026 | Blog

Part A — Frequency Response at a Glance

A large generating unit trips. In the first fraction of a second nothing else on the system has changed: the same load is still connected, the same generators are still running, and the lost megawatts must come from somewhere. They come from the rotating mass of every synchronous machine in the interconnection, which slows down. Frequency falls.

Over the next ten minutes, a chain of responses takes over, each faster to arrive than the next and each with a different job. Inertia limits how fast frequency falls. Primary frequency response from governors, inverter-based resources and load arrests the fall and holds frequency at a new, lower level. Secondary control — automatic generation control (AGC) — returns frequency to 60 Hz and restores scheduled interchange. Tertiary actions — reserve deployment, re-dispatch and unit commitment — replace the reserves that were used up, so the system is ready for the next contingency.

That chain answers a question that confuses many people outside system operations: can frequency recover to 60 Hz while the lost generator is still offline? Yes. Frequency recovery does not mean the failed unit came back. It means other resources have replaced its output — first automatically and partially, then deliberately and completely. The generation shortfall at the failed plant persists; the system-wide imbalance does not.

This guide walks through that chronology in US terms: the four interconnections, the NERC BAL standards that define each stage, the measurement points that NERC uses to grade frequency response (Point A, Point C and Value B), the area control error that drives AGC, and the role of solar, wind and battery storage at each stage. Keentel Engineering's own single-area simulation of a 2,750 MW resource loss makes each stage concrete.

Keentel's companion article on grid inertia and RoCoF covers the swing equation, inertia constants and synthetic inertia in depth. This article focuses on the complete response chain — from the first second to the tenth minute — and on how US standards measure and enforce it.

The four stages in one table

The source of much confusion is that the stages overlap. Inertial response begins at the instant of the trip and continues — with declining contribution — until the nadir. Primary response starts within a fraction of a second, overlaps inertia completely and remains in place until AGC relieves it. Secondary control begins within seconds of the event in the AGC cycle, but its effect builds over minutes. A well-drawn frequency plot shows one continuous trajectory, not four separate steps.

What Keentel's analysis shows

Keentel modeled a 60 Hz, ERCOT-scale system with about 50 GW of load and a 2,750 MW resource loss — the size of the ERCOT resource contingency used in BAL-003. Governors have 5% droop and a ±0.036 Hz deadband, consistent with FERC Order No. 842, with about 7,050 MW/Hz of combined governor response; load damping adds about 1,250 MW/Hz. AGC acts on area control error with a slow integral characteristic. Three cases were run.

Figure 1. Keentel's single-area simulation of a 2,750 MW resource loss. Top: the full 10-minute recovery, with the BAL-003 Value B averaging window (20–52 s after the event) and AGC restoration. Bottom: the first 30 seconds, showing Point A (pre-event frequency), Point C (the nadir) and settling. The knee in the recovery near 59.96 Hz occurs where frequency re-enters the governor deadband and AGC alone completes the restoration. Case B is shown without load shedding to illustrate the margin lost; in reality, under-frequency load shedding would operate.

Five results stand out:

• Inertia changes the path, not the settling point. Cases A and B settle at the same Value B, because settling frequency depends on droop and load damping, not inertia. But case B falls 2.5 times faster and reaches a nadir below the first load-shedding stage.
• Fast frequency response rescues the nadir. 1,000 MW of battery response, at full output within 0.5 s of a 59.7 Hz trigger, lifts the low-inertia nadir by 0.2 Hz — back above the load-shedding threshold — and raises Value B as well.
• Primary response leaves an offset; it does not restore 60 Hz. With 2,750 MW lost and about 8,300 MW/Hz of combined response, frequency settles about 0.36 Hz low (including the deadband effect). Only AGC removes that offset.
• Recovery happens while the shortfall persists. In every case the failed unit stays offline. Frequency returns to 60 Hz because other units' output has been raised by 2,750 MW — first by governors, then by AGC and reserve deployment.
• The clock is set by standards. In this model AGC restores frequency in under nine minutes, inside the 15-minute Contingency Event Recovery Period in NERC BAL-002-3. A balancing authority that lacks reserves cannot meet that clock, regardless of how good its governors are.

Engineering review of common statements about frequency recovery

Five lessons

• Think of four jobs, not four steps. Inertia buys time, primary response arrests and holds, secondary response restores and tertiary action resets. Each job needs its own resources.
• Grade the nadir and the settling frequency separately. The nadir is set by inertia and response speed; Value B is set by the amount of droop response and load damping. Different fixes apply to each.
• AGC is an accounting system as well as a control system. ACE tells each balancing authority whether it is causing or helping the problem. The BA that lost the unit carries the obligation to replace it.
• Headroom is the price of response. Governors, solar, wind and batteries can only respond with the capacity they hold back. Reserve requirements are where frequency stability becomes an economic decision.
• Inverter-based resources can be the fastest responders on the system — if they are specified, set and verified to respond.
  

Part B — Technical Guide: From Nadir to Nominal

1. The US frequency control landscape

North America has four synchronous interconnections. Within each, every generator turns in lockstep, and a single frequency describes the whole system. Between them, power flows only through DC ties, which do not share frequency.

Each interconnection is divided into balancing authorities (BAs), which balance generation, load and scheduled interchange within their areas. Reliability coordinators oversee wide-area reliability. Frequency is a shared, interconnection-wide quantity, but the obligation to respond and to replace a lost resource is assigned to individual BAs by the NERC BAL standards.

The UFLS values above are typical first-stage settings. Confirm the current regional criteria under NERC PRC-006 and the applicable regional standard before relying on them.

2. What happens at the instant of the loss

2.1 Power balance becomes energy balance

Before the trip, mechanical power into all generators equals electrical power out, plus losses. When a unit trips, the electrical load does not change in that first instant, but the mechanical input falls by the unit's output, ΔP. The difference is drawn from the kinetic energy stored in the rotating masses of every synchronous machine. The center-of-inertia swing equation describes the result:

where E_k is the total stored kinetic energy of online synchronous machines (MW·s), f0 is 60 Hz, D is the load damping in MW/Hz and Δf is the frequency deviation. Stored energy is the sum of each machine's inertia constant H times its MVA rating:

2.2 Initial rate of change of frequency



At t = 0, the governors and load have not yet changed, so the initial RoCoF depends only on ΔP and E_k:

Worked example 1 — RoCoF in an Eastern-scale and an ERCOT-scale system



For an Eastern-scale illustration, take a 4,500 MW loss — the Eastern Interconnection resource contingency used in BAL-003 — and an illustrative stored energy of 2,500 GW·s:

Even at 1,500 GW·s, the initial RoCoF is only 0.09 Hz/s. The Eastern Interconnection is so large that inertia rarely limits it.

For an ERCOT-scale system, a 2,750 MW loss with 250 GW·s gives 0.33 Hz/s; with 100 GW·s — a low-inertia, high-renewable hour — it gives 0.83 Hz/s. If frequency kept falling at the initial rate, it would reach 59.3 Hz in only 2.1 s and 0.85 s respectively. Real trajectories curve as response arrives, but the numbers show why ERCOT watches inertia closely and why the speed of the next stage matters so much more there than in the East.

Keentel's companion article on grid inertia covers inertia constants, synthetic inertia and synchronous condensers in detail. The remainder of this guide follows what happens next.

3. Primary frequency response: arresting the fall

3.1 The sources of primary response

Primary frequency response is any automatic, locally controlled change in power that opposes the frequency deviation. In the US it comes from:

• Governors on synchronous units. Steam, combustion turbine, combined-cycle and hydro governors respond in proportion to frequency deviation outside a deadband. Response begins within a fraction of a second; full response takes several seconds (combustion turbines and hydro) to tens of seconds (some steam units). Units running at maximum output, or with governors blocked or operating in load-control modes, contribute little or nothing.
• Inverter-based resources (IBRs). Solar, wind and batteries with frequency droop enabled change output through their plant controllers or inverter controls. Batteries can respond in both directions within a fraction of a second. Solar and wind can reduce output immediately, but can increase output only if they are operating below their available power.
• Fast frequency response (FFR). Very fast, often triggered, injections or load reductions — typically from batteries or interruptible load — that reach full output within about 0.5 s or less.
• Load damping. Motor loads draw less power at lower frequency. A typical figure is 1–2% of load per 1% change in frequency.

3.2 Droop and deadband

A droop characteristic sets how much a resource changes output for a given frequency deviation. FERC Order No. 842 requires newly interconnecting generators, including IBRs, to have primary frequency response capability with a maximum droop of 5% and a deadband no wider than ±0.036 Hz. Outside the deadband the response is:

where R is the droop (0.05) and f_db is the deadband. For each 0.1 Hz beyond the deadband, a 5% droop gives 3.33% of rating. Order No. 842 does not require a resource to hold headroom; it requires the capability and requires that the resource respond with whatever headroom it has.



3.3 Load damping

Load damping D is usually expressed as a percentage of load per percentage of frequency. For a 50 GW load with 1.5% per 1%:

Load damping is shrinking as more motors are driven through variable-frequency drives and more load is power electronics, which draw constant power regardless of frequency. Studies should not assume historical damping values for future systems.

3.4 The nadir: where inertia meets response

The nadir — Point C in NERC terminology — is the lowest frequency reached. It occurs at the instant when the combined primary response plus load damping first equals the lost power, so that df/dt = 0. Inertia determines how far frequency falls while response is building; the speed and amount of response determine when the fall stops. The nadir is therefore sensitive to:

• the size of the loss;
• system inertia;
• the delay and ramp rate of governors and IBR controls;
• governor deadbands and the share of units actually responding;
• FFR trigger levels and speed.

3.5 The settling frequency: Value B

Once the oscillations die out, frequency settles at a level where primary response plus load damping exactly equal the lost power. Ignoring deadbands:

where β is the frequency response characteristic of the interconnection, in MW/Hz. NERC expresses β in MW per 0.1 Hz, so a β of 8,300 MW/Hz is 830 MW/0.1 Hz. Inertia does not appear in this equation: it shapes the path, not the destination.

Worked example 2 — settling frequency with deadband

In Keentel's model, governors contribute 7,050 MW/Hz beyond the 0.036 Hz deadband and load damping contributes 1,250 MW/Hz with no deadband. Settling occurs where:

The simulation's Value B (59.65 Hz, averaged from 20 to 52 s) is slightly higher because AGC has already begun to act within that window. Without AGC, the model settles at 59.638 Hz — matching the hand calculation.

4. How NERC measures primary frequency response: BAL-003-2

4.1 Point A, Point C and Value B

NERC Reliability Standard BAL-003-2 evaluates frequency response using actual events. Each qualifying event is characterized by three points on the frequency trace:

A BA's measured frequency response for an event is the change in its net interchange (the extra power it exported to help) divided by the frequency change from Point A to Value B, in MW/0.1 Hz. NERC uses the median of a year of qualifying events to evaluate each BA. Confirm the current event-selection and averaging details in the BAL-003-2 procedure documents before relying on them.

4.2 The Interconnection Frequency Response Obligation

NERC calculates, for each interconnection, an Interconnection Frequency Response Obligation (IFRO): the minimum frequency response needed so that the design-basis resource loss will not cause the frequency to fall to the first UFLS stage, after allowing for statistical variation and the difference between the nadir and Value B. Conceptually:

where ARLPC is the adjusted resource loss protection criterion (the design-basis loss less any credit for load resources that trip on under-frequency) and MDF is the maximum allowable delta frequency between Value B and the starting point, in Hz, after deducting margins for the nadir-to-settling ratio and starting-frequency variation.

These criteria and thresholds come from the BAL-003 framework and regional programs; NERC recalculates the IFROs annually in its frequency response analysis. Confirm current values before using them.

Worked example 3 — what an IFRO implies

The Eastern Interconnection IFRO has been on the order of 1,000 MW/0.1 Hz. With 4,500 MW and no load-resource credit:

That is the largest settled deviation the obligation allows. Actual Eastern frequency response has typically been much larger than the obligation — 2,500 MW/0.1 Hz or more — so a 4,500 MW loss settles only about 0.18 Hz low.

ERCOT's IFRO has been on the order of 380 MW/0.1 Hz, smaller than one might expect for a 2,750 MW criterion. The reason is the credit for load resources that trip automatically on under-frequency: subtracting about 1,400 MW of credit leaves roughly 1,350 MW, and 1,350 / (10 × 0.354) ≈ 381 MW/0.1 Hz. The IFRO is a minimum for compliance, not a design target. Real planning studies use full dynamic models.

4.3 How the obligation is shared

Each BA receives a Frequency Response Obligation (FRO), a share of the IFRO based on its generation and load. Each BA also sets a frequency bias, B, used in its ACE calculation (section 6). BAL-003-2 ties the two together: the bias setting should reflect the BA's actual natural frequency response, so that AGC neither fights nor ignores primary response.

5. Inverter-based resources in the primary stage

5.1 What IBRs can and cannot do

The difference between inertia and fast frequency response matters. Inertia responds to the rate of change of frequency without measurement or delay. FFR responds to measured frequency (or RoCoF) after a detection and ramp delay, typically a few hundred milliseconds. Both improve the nadir; FFR can deliver much more power than synthetic inertia, but only after its delay. Grid-forming inverters behave more like synchronous machines and provide genuinely immediate response, which matters most at the very start of the event.

Worked example 4 — a solar plant with headroom versus a battery

A 200 MW solar plant operates at 180 MW — 10% headroom — with 5% droop and the Order No. 842 deadband. A 100 MW battery provides FFR at full output within 0.5 s of a 59.7 Hz trigger.

The battery on FFR delivers five times more power, and much faster, than the curtailed solar plant — and the solar plant pays for its headroom with 20 MW of lost energy production every hour it is curtailed. That is why ISOs procure fast response from batteries and load resources, and why solar and wind usually provide upward primary response only when they are already curtailed for other reasons.

6. Secondary control: area control error and AGC

6.1 Why primary response is not enough

Primary response stops the fall, but it leaves the interconnection off-nominal and leaves every responding unit operating away from its economic dispatch point. It also leaves tie-line flows distorted: when a unit in one BA trips, governors throughout the interconnection respond, so neighboring BAs export power to the BA that lost the unit. Secondary control resolves all three problems.

6.2 The ACE equation

Each BA's AGC computes area control error every few seconds. Under NERC BAL-001-2:

where NI_A and NI_S are actual and scheduled net interchange (MW, exports positive), B is the BA's frequency bias setting in MW/0.1 Hz (a negative number), F_A and F_S are actual and scheduled frequency (Hz), and I_ME is a meter-error correction. BAs in the Western Interconnection add an automatic time error correction term, I_ATEC. A negative ACE means the BA is under-generating; AGC raises its generation.

The frequency-bias term is the clever part. If B matches the BA's actual natural frequency response, then a BA that did not cause the disturbance sees its extra exports exactly cancelled by the bias term — its ACE stays near zero, and its AGC does not withdraw the help its governors are giving. The BA that lost the unit sees an ACE equal to the lost megawatts, and its AGC replaces them.

Worked example 5 — ACE in two balancing authorities

Consider an interconnection with a total frequency response of 2,500 MW/0.1 Hz. Balancing Authority A, with a bias of −300 MW/0.1 Hz matching its natural response, loses a 1,000 MW unit. The interconnection settles at:

Area A's own governors supply its share, 300/2,500 × 1,000 = 120 MW. The remaining 880 MW flows in from neighbors, so A's net interchange falls 880 MW below schedule:

The rest of the interconnection, with a combined bias of −2,200 MW/0.1 Hz, exports 880 MW above schedule:

Area A's ACE is exactly its lost generation; every other area's ACE is zero. Area A's AGC and contingency reserves replace the 1,000 MW; as they do, frequency rises, the neighbors' governors back down, and tie-line flows return to schedule. Frequency recovers to 60 Hz while the tripped unit remains offline — the replacement megawatts come from other units in Area A.

If B is set too small in magnitude relative to natural response, neighbors' AGC sees a positive ACE and pulls back their help prematurely, delaying recovery. That is why BAL-003 links bias settings to measured frequency response.

6.3 AGC in practice

AGC typically runs on a 2–6 second cycle. It filters ACE, allocates the required change among regulating units according to their participation factors and ramp rates, and sends raise or lower signals. In ISO markets, AGC dispatches regulation service, and the real-time market redispatches energy every five minutes. Restoring a large deficit therefore takes minutes, not seconds: the speed is limited by unit ramp rates, the amount of regulation and contingency reserve deployed, and deliberate filtering to avoid chasing noise.

7. How NERC measures secondary control: BAL-001-2

BAL-001-2 sets two performance measures for normal operation:

• Control Performance Standard 1 (CPS1) evaluates, over a rolling 12 months, how ACE correlates with frequency error. A BA that helps frequency when it is off-nominal scores well; one whose ACE pushes frequency further away scores poorly. The requirement is a CPS1 of at least 100%.
• Balancing Authority ACE Limit (BAAL) sets frequency-dependent bounds on ACE. The further frequency is from 60 Hz, the tighter the limit on ACE in the direction that worsens it. A BA must not exceed its BAAL for more than 30 consecutive clock minutes.

BAL-001 governs everyday balancing. Large disturbances are handled by BAL-002.

8. Recovering from the contingency: BAL-002-3

The Disturbance Control Standard, BAL-002-3, applies to Reportable Balancing Contingency Events — sudden losses of generation or import above a threshold defined in the standard. For each such event, the responsible BA (or its reserve sharing group) must:

• recover ACE to zero, or to its pre-event value if that was negative, within the Contingency Event Recovery Period of 15 minutes; and
• restore its contingency reserve within the Contingency Reserve Restoration Period, 90 minutes after the end of the recovery period, so that it can withstand the next event.

BAL-002 is what makes the answer to "can frequency recover while the unit is still offline?" an obligation rather than a hope. The BA must carry enough contingency reserve to cover its most severe single contingency, and it must deploy it within 15 minutes.

Worked example 6 — reserve deployment for a 1,000 MW loss

Area A in worked example 5 must bring 1,000 MW of replacement generation online within 15 minutes. If the reserve comes from spinning units ramping in parallel, it needs an average combined ramp of at least:

In practice the BA plans for a faster ramp, because AGC filtering, communication delays and slower units eat into the 15 minutes. If part of the reserve comes from quick-start combustion turbines that need 10 minutes to reach full load, the spinning portion must cover the shortfall until they arrive. After recovery, the BA has 90 minutes to restore its contingency reserve — by committing additional units, re-dispatching, arranging emergency purchases or, in extreme cases, shedding interruptible load.

9. Tertiary control and returning to normal

Tertiary control is the slower, mostly economic layer that follows AGC:

• Economic dispatch. In ISO markets, the security-constrained economic dispatch runs every five minutes and moves the replacement megawatts from regulating units to the least-cost available units, freeing regulation capacity.
• Reserve restoration. Units that supplied spinning reserve are released, and new reserve is designated. Quick-start units may be committed for the rest of the operating hour.
• Unit commitment. If the tripped unit will be out for days, the day-ahead and reliability commitment processes replace it.
• Time error correction. Frequency below 60 Hz for several minutes makes synchronous clocks run slow. Time error is now managed under industry business practices rather than a NERC reliability standard, using small scheduled frequency offsets; the Western Interconnection uses automatic time error correction in ACE.

9.1 What "recovered" really means

When frequency returns to 60 Hz and ACE returns to zero, the system has recovered its balance, not its security. Three things are still different from the moment before the trip:

• Reserves are depleted. The megawatts that replaced the lost unit came out of regulation and contingency reserve. Until those reserves are restored, a second contingency of the same size could not be covered. This is why BAL-002 has a separate 90-minute reserve restoration requirement.
• Flows have changed. The replacement generation is in different locations from the lost unit. Transmission flows, voltage profiles and post-contingency loadings may be worse than before, and operators may need to re-dispatch to respect system operating limits under the TOP and IRO standards.
• The failed unit is still out. Its owner must investigate the trip, and if the trip was caused by protection or control misbehavior during a disturbance, it may become a NERC event analysis or a misoperation review.

9.2 The same chain in reverse: over-frequency

A sudden loss of load — for example, a large data center or industrial site tripping offline during a transmission fault — produces the mirror image: frequency rises, inertia slows the rise, governors and IBRs reduce output, and AGC lowers generation to restore 60 Hz. Inverter-based resources are especially good at over-frequency response, because reducing output needs no headroom. As large loads grow, NERC and several ISOs are studying load-loss events and ride-through expectations for large loads, because the sudden loss of thousands of megawatts of load is now a credible contingency in some regions.

10. ERCOT: a closer look at a low-inertia interconnection

ERCOT is a single BA with limited DC ties, so it must handle every contingency internally, and its inertia varies widely with wind and solar output. ERCOT procures several ancillary services that map onto the stages in this guide:

ERCOT also monitors system inertia in real time against a critical inertia level, below which existing reserves might not prevent load shedding for its 2,750 MW contingency, and procures more fast response as inertia falls. Several ERCOT events in recent years — including large simultaneous losses of inverter-based resources during transmission faults — have reinforced how much the nadir depends on fast response and ride-through.

11. Under-frequency load shedding and ride-through: the safety net

11.1 UFLS

If primary response cannot arrest the fall above the first threshold, under-frequency load shedding (UFLS) disconnects blocks of load in stages. NERC PRC-006 requires planning coordinators to develop UFLS programs meeting regional criteria, and transmission and distribution owners to implement them. Typical first stages are about 59.5 Hz in the Eastern and Western Interconnections and 59.3 Hz in ERCOT, with total shed amounts of 25–30% of load spread over several stages.

UFLS is a last resort. A system that sheds load has saved itself from collapse, but customers have lost service, and the event becomes a reliability event under NERC event analysis.

Worked example 7 — nadir margin to the first UFLS stage

Using Keentel's simulation results and ERCOT's 59.3 Hz first stage:

Planners also add margin for measurement error, UFLS relay time delays, and uncertainty in governor response. A margin below about 0.1 Hz should prompt sensitivity studies with detailed models.

11.2 Ride-through

The resources that provide the response must stay connected through the event. NERC PRC-024 sets frequency and voltage protection setting boundaries for synchronous generators, and PRC-029-1 sets ride-through requirements for inverter-based resources. IEEE 2800-2022 requires transmission-connected IBRs to ride through specified frequency excursions and rates of change of frequency, and IEEE 1547-2018 sets ride-through categories for distributed resources. A resource that trips at 59.4 Hz, or on a RoCoF relay, during the event it is meant to help turns a design-basis loss into a cascading one.

12. Keentel's model and how to read it

Keentel's single-area model represents the interconnection as one equivalent machine:

• Inertia: 250 GW·s (baseline) or 100 GW·s (low).
• Governors: 7,050 MW/Hz of droop response beyond a ±0.036 Hz deadband, limited to 4,000 MW of headroom, with a 0.5 s valve lag and a 1.5 s turbine lag.
• Load damping: 1,250 MW/Hz.
• FFR: 1,000 MW triggered at 59.7 Hz, starting 0.25 s after the trigger and reaching full output 0.25 s later.
• AGC: integral control on ACE, beginning 30 s after the event.

The model is a teaching tool. It ignores network constraints, voltage dependence of load, individual governor models, unit headroom distribution and the reduction of governor response that some units exhibit after a few tens of seconds (sometimes called withdrawal). Real studies use dynamic models in positive-sequence simulation tools, validated against measured events.

13. What engineers should assess

Part C — Three Anonymous Case Studies

These composite case studies are based on the kinds of problems Keentel Engineering sees on US projects. Names, locations and details have been changed or combined, and figures are rounded.

Case study 1 — A battery fleet that was not responding the way the ISO expected

Situation. An independent power producer in Texas owned four battery storage plants, totaling 400 MW, committed to provide fast frequency response. During a large generation trip, frequency fell to about 59.6 Hz. The ISO's post-event review found that the fleet's measured response was less than half of its award, and that two plants had responded more than a second late.

Analysis. Keentel reviewed plant controller logs, disturbance monitoring records and the ISO's event data. Three issues emerged. First, two plants measured frequency at the plant controller with a heavily filtered signal, adding about 0.8 s of delay before the trigger was recognized. Second, one plant's energy management system had moved its state of charge below the level needed to sustain full output for the required duration, so the plant controller limited its response. Third, one site's frequency trigger had been set to 59.70 Hz but implemented as "below 59.70 Hz for 0.5 s", a common misreading of the service requirement.

Engineered solution. Keentel moved frequency detection to fast inverter-level or dedicated relay measurements, revised the trigger logic to remove the unnecessary time qualification, and added a state-of-charge reserve constraint tied to the awarded service. Each plant was then tested by injecting a simulated frequency signal into the plant controller and recording response at the point of interconnection.

Outcome. In the next qualifying event, the fleet delivered its full award within about 0.4 s of the trigger. The owner avoided a performance penalty and gained a documented test procedure for annual verification.

Case study 2 — A balancing authority with a bias setting that slowed recovery

Situation. A mid-sized balancing authority in the Eastern Interconnection noticed that after its own generator trips, ACE recovery took longer than expected, approaching the 15-minute limit in BAL-002-3 on two occasions. Its neighbors also reported that their AGC reacted strongly whenever the BA had a disturbance.

Analysis. Keentel analyzed several years of event data using the BAL-003 method. The BA's measured natural frequency response was about 40% larger than its frequency bias setting — the bias was based on an old generation mix with more governor-responsive steam units and had not been revisited after several unit additions. Because the bias was too small in magnitude, neighboring BAs saw positive ACE during the BA's events and withdrew their primary support early, while the BA's own AGC had to recover a larger deficit with reserves that ramped more slowly than assumed.

Engineered solution. Keentel recommended a revised bias setting consistent with measured response, a review of reserve units' actual ramp rates against the reserve plan, and an increase in fast-ramping reserve during peak hours. AGC filtering parameters were adjusted to respond faster to large ACE excursions while keeping normal regulation stable.

Outcome. After the changes, contingency recovery time for comparable events fell from about 13 minutes to about 7 minutes. The BA's CPS1 score improved, and its neighbors stopped seeing large AGC swings during its events.

Case study 3 — A planning study that found the nadir problem before it happened

Situation. A utility in the Western Interconnection was planning the retirement of two large coal units and the addition of about 1,500 MW of solar and 600 MW of batteries. The steady-state and transient stability studies were acceptable, but the transmission planner asked whether frequency performance for the design-basis loss would remain acceptable in low-load, high-solar hours.

Analysis. Keentel built a frequency-performance study using positive-sequence dynamic models validated against recent measured events. The study varied commitment patterns across a year of hourly operating conditions. In spring midday hours, online inertia in the study area fell substantially, and with governor deadbands and headroom represented realistically, the regional frequency response shrank. The modeled nadir for the design-basis loss fell to within 0.05 Hz of the first UFLS stage in about 4% of hours.

Engineered solution. Keentel recommended specifying fast frequency response capability and droop settings for the new batteries in their interconnection agreements, enabling grid-forming capability on part of the battery fleet, and holding a minimum amount of upward headroom on the batteries during low-inertia hours. Primary frequency response settings were verified for all new resources under Order No. 842.

Outcome. With the recommended battery response, the minimum modeled nadir margin improved to about 0.25 Hz in all hours. The utility incorporated the battery settings into its procurement specifications and added frequency-performance screening to its annual planning cycle.

Part D — 20-Question Technical FAQ

Keentel Services, References and Contact

How Keentel Engineering Helps

Keentel Engineering supports utilities, balancing authorities, independent power producers and large-load developers with every stage of frequency performance:

• Frequency performance studies: aggregate and full dynamic studies of RoCoF, nadir, settling frequency and recovery for design-basis losses, across a year of operating conditions.
• Primary frequency response compliance: droop, deadband and headroom settings for synchronous units, solar, wind and batteries under FERC Order No. 842 and interconnection agreements, with commissioning tests.
• Fast frequency response and grid-forming design: trigger logic, measurement, response timing and state-of-charge management for battery fleets; grid-forming specification and EMT verification.
• BAL standards support: BAL-003 event analysis and bias-setting reviews, BAL-001 and BAL-002 performance analysis and reserve-ramp assessments.
• Ride-through and protection reviews: PRC-024, PRC-029 and IEEE 2800 settings reviews so responding resources stay online.
• Event analysis: reconstruction of frequency events using PMU, SCADA and plant data, with root cause and corrective actions.
• Model development and validation: governor, plant controller and IBR models benchmarked against field data for planning and interconnection studies.

Why Clients Choose Keentel

• System studies and plant engineering under one roof. We understand both what the interconnection needs and what a plant controller can deliver.
• US standards fluency. NERC BAL-001, BAL-002, BAL-003, PRC-006, PRC-024 and PRC-029; FERC Orders 842 and 901; IEEE 2800 and 1547.
• Evidence-based results. We validate models against recorded events and test settings in the field.
• From 4 kV to 765 kV. Distribution-connected batteries, utility-scale solar and wind, and transmission-level planning areas.
• Responsive senior engineers who stay with the project from study through commissioning.

Talk to Keentel Engineering

Planning a battery, solar or wind project with frequency response obligations, or reviewing frequency performance for your area? Keentel Engineering can help you study it, set it and prove it.

• Phone: (813) 389-7871
• Email: contact@keentelengineering.com
• Book a 15-minute call: calendly.com/keentel-engineering/15min
• Web: keentelengineering.com

References

• NERC Reliability Standard BAL-001-2, Real Power Balancing Control Performance.
• NERC Reliability Standard BAL-002-3, Disturbance Control Standard — Contingency Reserve for Recovery from a Balancing Contingency Event.
• NERC Reliability Standard BAL-003-2, Frequency Response and Frequency Bias Setting, and associated procedure documents.
• NERC, annual Frequency Response Annual Analysis.
• NERC Reliability Standards PRC-006 (Automatic Underfrequency Load Shedding), PRC-024 (Frequency and Voltage Protection Settings) and PRC-029-1 (Frequency and Voltage Ride-through Requirements for Inverter-Based Resources).
• NERC Reliability Guideline, Primary Frequency Control; NERC guidance on fast frequency response and inverter-based resource performance.
• FERC Order No. 842, Essential Reliability Services and the Evolving Bulk-Power System — Primary Frequency Response.
• FERC Order No. 901, Reliability Standards to Address Inverter-Based Resources.
• IEEE Std 2800-2022, Interconnection and Interoperability of Inverter-Based Resources Interconnecting with Associated Transmission Electric Power Systems.
• IEEE Std 1547-2018, Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces.
• ERCOT, Nodal Protocols and ancillary service methodology documents.
• Kundur, P., Power System Stability and Control, McGraw-Hill.
• Wood, A. J., Wollenberg, B. F., and Sheblé, G. B., Power Generation, Operation, and Control, Wiley.

Disclaimer

This article is provided by Keentel Engineering for general technical education. It is not engineering, operational or compliance advice for any specific system or project. The system data, inertia, governor and FFR parameters, simulation results, and case studies are illustrative or composite. Resource contingency criteria, frequency response obligations, UFLS thresholds and market products change over time; always confirm current values in the applicable NERC standards, regional criteria and ISO/RTO rules. Keentel Engineering accepts no liability for decisions made on the basis of this article without a project-specific engineering review.

© 2026 Keentel Engineering. All rights reserved. Written by Keentel Engineering.