Automatic frequency restoration reserve aFRR

What it is (precise)

SOGL Article 3(2)(7) defines the parent class: frequency restoration reserves are the active power reserves available to restore system frequency to the nominal frequency and, in a synchronous area with more than one LFC area, to restore power balance to the scheduled value. Article 3(2)(99) then defines automatic FRR as FRR that can be activated by an automatic control device.

Points (100) and (101) split the timing into two terms that are routinely conflated: the automatic FRR activation delay runs from the setting of a new setpoint value by the frequency restoration controller to the start of physical delivery, while the automatic FRR full activation time runs from that same setpoint to the corresponding activation or deactivation.

The Electricity Balancing Guideline (Regulation (EU) 2017/2195) adds a third clock. Its Article 2(30) defines full activation time for standard products as the period between the activation request by the connecting TSO — or the contracting TSO in a TSO-BSP model — and the corresponding full delivery of the product.

That is the definition the balancing platforms work to, and it is the one behind the 5-minute PICASSO figure. EBGL does not define aFRR itself: its Article 2 imports the definitions from Article 3 of SOGL, which is why any aFRR definition must be cited to SOGL and any standard-product timing to EBGL and the platform.

Why it matters in a real grid-scale project

The 30-second confusion has real cost. If a functional specification says aFRR 30 s per SOGL, the plant gets engineered and tested to start delivering within 30 seconds — which is what Article 158(1)(d) actually requires — while the market obligation it will be settled against is full delivery of the contracted volume within 5 minutes on PICASSO.

The two are not alternatives; the delay ceiling and the full activation time both apply, to different events. When you see a 30-second aFRR number, ask one question before anything else: delay or full activation? Then get the clause number that backs it.

Sizing follows from the second number, not the first. A 5-minute full activation time is a ramp obligation; the energy obligation comes from how long the TSO leaves you activated afterwards, which is a market and control-area question rather than a converter question.

That is what keeps aFRR compatible with revenue stacking — the ramp sits trivially inside a battery's capability, so the real constraint is state of charge headroom in both directions plus the product structure your TSO uses. PICASSO and MARI both support asymmetrical collection of bids, upward and downward separately, so a one-directional offer is a normal shape rather than an exception.

Typical values and standards

The operative figure: balancing service providers must be capable of delivering the full contracted aFRR volume within 5 minutes of receiving an activation signal from their connecting TSO, per ENTSO-E's description of the PICASSO standard product. That is not in SOGL.

SOGL sets a ceiling instead — Article 157(2)(c) requires that the automatic and manual FRR full activation times of an LFC block not exceed the time to restore frequency, and Annex III Table 1 fixes that at 15 minutes for Continental Europe, Great Britain, Ireland and Northern Ireland, and the Nordic area alike. As of mid-2026 no EU regulation fixes a single aFRR full activation time; it is delegated to the TSOs of each LFC block.

For contrast, the mFRR standard product on MARI is 12.5 minutes, and ENTSO-E states both in one sentence, which makes it the cleanest citation for the pair. Before harmonisation the spread was wide: an ENTSO-E study found European aFRR full activation times ranging from 2 to 15 minutes — 2 minutes in the Nordic LFC block, 3 minutes in Switzerland and Italy, 15 minutes in many other blocks.

Treat any pre-PICASSO activation time found in an older study as country-specific. The provider obligation itself sits in SOGL Article 158(1)(f): the providing unit or group must be capable of activating its complete automatic reserve capacity within the automatic FRR full activation time.

How it shows up in specs, studies and contracts

On the legal side, the European aFRR platform is mandated by EBGL Article 21, which required all TSOs to develop a proposal for an implementation framework; PICASSO is the implementation project endorsed by all TSOs through the ENTSO-E Market Committee. In practice you meet it through your connecting TSO's prequalification, not through the platform directly.

What you sell is split by EBGL Article 2 into balancing capacity — reserve you have agreed to hold and bid — and balancing energy, the energy actually used. Article 2(27) defines the balancing energy gate closure time, after which a bid for a standard product on a common merit order list can no longer be submitted or updated.

For equipment and controls, ask for measured end-to-end latency from setpoint receipt at the plant controller to power change at the point of interconnection, not a sum of component datasheet figures.

Then check which clock the vendor used: SOGL's aFRR terms start at the frequency restoration controller's setpoint, while EBGL's standard-product clock starts at the TSO's activation request, and the communication and validation steps between those two events belong to somebody. EBGL Article 2(29) even names the gap — the preparation period, between the request and the start of the ramping period. Write down which definition each guarantee is measured against.

Common pitfalls

Two 30-second numbers live in the same regulation and mean different things. SOGL Annex V gives 30 s as the FCR full activation time for Continental Europe, measured from the reference incident. SOGL Article 158(1)(d) gives 30 s as the maximum automatic FRR activation delay, measured from a controller setpoint and ending when physical delivery starts. Neither one is the aFRR full activation time. If a document cites 30 s for aFRR without a clause number, assume it has copied the wrong row and go back to the source before designing anything around it.

The second trap is treating 5 minutes as law. It is the PICASSO standard product parameter as described by ENTSO-E in its balancing-platform FAQ; the ACER decisions that approved the PICASSO and MARI implementation frameworks are the formal instruments and were not consulted for this entry. That matters when a contract says per applicable regulation — the regulation only caps the LFC block at the 15-minute time to restore frequency, so a regulatory pointer buys you 15 minutes, not 5. Write the actual product parameter, its source document and its date into the contract instead.