PCS & grid

Frequency response FFR / FR

Frequency response is the rapid, automatic injection or absorption of real power to counter deviations of grid frequency from its nominal 50 or 60 Hz. In a grid-scale BESS the Power Conversion System measures frequency locally and slides active power along a defined frequency-to-power curve, acting within a fraction of a second to a few seconds.

The family runs from fast frequency response (FFR) and synthetic inertia in the sub-second range, through primary droop-based containment over seconds, up to slower secondary regulation. You meet it three times: as a P(f) line on a PCS datasheet, as a RoCoF-withstand value in an interconnection study, and as a paid product in a market rulebook. Batteries dominate because they act ten to a hundred times faster than thermal-plant governors.

Reviewed July 2026 by Sergey Syrvachev

New to BESS? Start free with the 7-email fundamentals course — no cost, no account.

What it is (precise)

Grid frequency is the real-time indicator of balance between generation and load across a synchronous AC system: lose a large generator and frequency falls, lose a block of load and it rises. How fast it moves — the rate of change of frequency, or RoCoF — historically depended on the stored kinetic energy of spinning machines.

Frequency response is the controlled real-power action that opposes the deviation: the BESS discharges into a low-frequency event and charges to absorb a high-frequency one. The delivered quantity is active power in MW, measured at a contractually defined reference point, usually the Point of Interconnection — not at the battery terminals, a distinction that quietly costs you MW.

The service family spans a spectrum of speed. Fast frequency response and synthetic inertia act sub-second to slow RoCoF and lift the frequency nadir; primary frequency response follows a droop curve over seconds to arrest and hold the deviation; secondary regulation then restores frequency over minutes. How the response is produced depends on control mode.

A Grid-following PCS measures frequency through a phase-locked loop and then commands a power change, while a Grid-forming inverter responds inherently through its droop characteristic without waiting for a measurement — which is why GFM assets are increasingly favoured for inertia-like products, where the first hundred milliseconds decide how deep the nadir goes.

Why it matters in a real grid-scale project

As synchronous generation retires, system inertia falls and events become faster and deeper — Ireland and South Australia already operate against explicit RoCoF constraints, and ERCOT has run at record-low inertia. Operators therefore procure dedicated fast products: GB Dynamic Containment (DC-L and DC-H), ERCOT Fast Frequency Response inside Responsive Reserve, AEMO contingency FCAS including the 1-second Very Fast service, and continental FCR under ENTSO-E rules.

For many batteries these services are a primary revenue stream, usually stacked with energy arbitrage, and price saturation of these shallow, capacity-limited markets is a standard downside case in the project financial model that a student should expect to see stress-tested.

The engineering consequences are concrete. The duty is power-intensive but short-energy, so it shapes C-rate, PCS sizing, and the inverter's headroom for fast power steps rather than total MWh. It drives thermal and auxiliary loads, erodes round-trip efficiency through frequent shallow cycling, and sets state-of-charge rules so the asset can deliver both upward (discharge) and downward (charge) response on demand.

In daily operation, SoC-recovery strategy and the availability and penalty clauses in the grid-service contract dominate far more than the raw power rating: an asset parked at full SoC can still earn a symmetric product on paper yet be physically unable to deliver its downward half.

Frequency response as a droop slope: real power corrects a frequency deviation.Interactive · bess.engineer ↗
Frequency response as a droop slope: real power corrects a frequency deviation. Open the interactive →
Key facts
Nominal frequency
50 Hz (Europe, GB, Australia) or 60 Hz (North America, ERCOT)
Typical droop
2-5%; FERC Order 842 caps new US units at 5%
Deadband
±10-15 mHz for fast products; ±36 mHz max under FERC 842
Full-activation deviation
FCR ±200 mHz; GB Dynamic Containment ±0.5 Hz
Full-response time
ERCOT FFR ~0.25 s (15 cycles); GB DC ≤1 s; FCR ≤30 s
Sustain duration
commonly ≥15 min for fast / limited-energy products
RoCoF withstand
1-2 Hz/s in modern codes (higher in island grids) — often the binding trip limit
UFLS first stage
~59.3 Hz on 60 Hz systems; ~48.8 Hz in GB
Governing market rules
GB Grid Code / SO service specs, ERCOT nodal protocols, AEMO NER (FCAS), ENTSO-E (FCR)
Capability standards
IEEE 2800-2022 (transmission IBRs); NERC BAL-003 (BA obligation, MW per 0.1 Hz)
Duty profile
power-intensive, short-energy — sizes C-rate and PCS headroom, not MWh
Safety standards (same system, separate role)
UL 9540 (ESS certification), UL 9540A (fire-propagation test data), NFPA 855 (installation)

Typical values and standards

Droop is commonly quoted at 2 to 5 percent; FERC Order 842 caps new US units at 5 percent droop and ±36 mHz deadband. Fast products pair a small deadband with a defined activation band: continental FCR uses roughly a ±10 mHz deadband with full activation at ±200 mHz within 30 seconds; GB Dynamic Containment uses about ±15 mHz with full delivery at ±0.5 Hz within 1 second; ERCOT FFR requires full output within about 15 cycles (~0.25 s) once frequency crosses 59.85 Hz, sustained at least 15 minutes.

Three numbers define any product — deadband, activation band, and response time — and all are revised over time, so size against the current published terms.

Performance is governed by grid codes and market rules, not one global standard: the GB Grid Code and system-operator service specs, ERCOT nodal protocols, AEMO's NER and FCAS rules, and ENTSO-E network codes. Alongside them, IEEE 2800-2022 defines frequency-response and ride-through requirements for transmission-connected inverter-based resources, and NERC BAL-003 sets each balancing authority's obligation in MW per 0.1 Hz.

Under-frequency load shedding typically starts near 59.3 Hz on 60 Hz systems and 48.8 Hz in GB — fast response exists to keep events clear of those trips. These operational rules are separate from safety standards on the same system: UL 9540 certifies the ESS product, UL 9540A generates fire-propagation test data, and NFPA 855 governs installation.

How it shows up in specs, studies and contracts

On a PCS datasheet the function is the frequency-watt or P(f) line. Check the configurable deadband and droop ranges, the combined measurement-plus-response latency, and — the question most buyers forget to ask the vendor — whether the P(f) curve executes locally in the PCS or is commanded from the plant controller: routing the loop through the PPC adds hundreds of milliseconds and can disqualify a plant from any sub-second product.

Confirm the measurement point, because delivery at the Point of Interconnection is net of Transformer and auxiliary losses, so terminal-rated MW is not deliverable MW. Verify P-Q capability at the contracted power factor, since Four-quadrant operation runs the reactive duty simultaneously.

In interconnection studies the service appears both as mandatory capability (FERC Order 842, governor clauses in most grid codes) and as frequency Ride-through envelopes with RoCoF-withstand values, commonly 1 to 2 Hz per second and higher in island systems — that withstand figure is often the binding constraint, since an asset that trips delivers nothing.

Market prequalification runs injection tests against reference frequency traces with sub-second telemetry and baselining rules that decide whether you are paid. In contracts, scrutinise availability definitions, SoC-management responsibility, and penalty regimes; in the battery supply agreement, confirm the warranty counts shallow FR micro-cycles as fractional throughput rather than charging a full equivalent cycle per excursion.

Common pitfalls

Two conflations cause most of the confusion. First, frequency response is not frequency regulation: FFR and primary response are autonomous, local, droop-driven reactions to frequency error acting within seconds, while regulation (AGC or aFRR) is a centrally dispatched setpoint-following service over minutes — a plant can be excellent at one and unqualified for the other.

Second, response is not Ride-through: ride-through is the obligation to stay connected through a disturbance, a precondition rather than the service itself. A battery that trips on a 1 Hz-per-second RoCoF event delivers no response at all, however fast its controls, which is why the withstand number gates qualification before the P(f) curve ever matters.

The energy dimension is routinely under-scoped. A 15-minute sustain at full power needs only 0.25 hours of energy on paper, but symmetric products demand headroom in both directions, SoC recovery must finish inside the service window, and the usable SoC band shrinks the tradable range further.

EV traction batteries face a completely different duty and are a poor mental model for this shallow, high-frequency cycling. Do not design from yesterday's parameters: deadbands, droop slopes, activation bands and sustain times differ product by product and change regularly, so qualification is always against the current published service terms, not a textbook table you memorised last year.

Common misconception

A battery can provide frequency response as long as its PCS meets the power rating — the energy requirement is trivial, since a 15-minute product is only 0.25 hours.

In reality: The binding constraint is rarely peak power or energy; it is managed SoC headroom and response time. A symmetric product must hold mid-range state of charge to deliver both discharge and charge on demand, hit strict per-product response-time and sustain limits, and pass availability and telemetry tests at the Point of Interconnection net of losses. An asset that comfortably meets its MW rating but drifts to full or empty SoC, or answers 200 ms too slow, forfeits payment for the interval regardless of nameplate.

Visuals & further reading
Go deeper

Frequency response, in context.

The Grid-Scale BESS course covers frequency response — and the rest of the system — from the ground up, the way it actually gets deployed.

Browse the course