Balancing & Real-Time Markets

Two products, one objective

TSOs procure balancing capacity ahead of time — a paid commitment to be available — and activate balancing energy in real time from a merit order of bids. A jurisdiction's split between the two reveals its philosophy: heavy capacity contracting buys certainty; reliance on free energy bids trusts the market to show up.

Balancing is also, by design, the market of last resort — and its volume is a verdict on everything upstream. Every megawatt-hour of forecast error that liquid intraday trading absorbs voluntarily is one the TSO never activates; both markets feed on the same renewable forecast error, which grows with wind output and spikes when storms push turbines through cut-out. A healthy sequence shows balancing energy shrinking relative to renewable capacity even as the system gets harder to run — self-balancing doing quietly what activation would do expensively. When balancing volumes grow instead, the diagnosis usually sits in the intraday market, not here.

The legal spine: EBGL

Commission Regulation (EU) 2017/2195 (the Electricity Balancing Guideline) is the constitution of European balancing. It mandates standard products, marginal pricing for balancing energy, TSO-BSP model separation and the European platforms — turning what were once purely national TSO procedures into a harmonised market layer.

The role triangle

The TSO defines needs and activates; BSPs (balancing service providers) sell capacity and energy from prequalified assets; BRPs (balance responsible parties) carry the financial consequence of every MWh their portfolio deviates from schedule. Imbalance settlement is the hinge that transmits real-time system cost back into trading behaviour.

European platforms

PICASSO (aFRR energy) and MARI (mFRR energy) merge national merit orders into common European activation, subject to available cross-zonal capacity. Imbalance netting via IGCC cancels opposing system imbalances before any activation at all — the cheapest balancing energy is the activation avoided.

Dimensioning

How much reserve to hold is a design choice disguised as an engineering constant. FCR is dimensioned jointly for the synchronous area (the reference incident), while FRR dimensioning is increasingly probabilistic — sized against forecast-error and outage distributions that grow with renewable penetration.

Cost allocation

Capacity costs are typically socialised through grid tariffs; activated energy costs flow to the BRPs that caused the imbalance. Where the line is drawn — and whether passive imbalances that help the system are rewarded — shapes whether balancing is a cost centre or a genuine market.

FCR — frequency containment

The fastest layer: automatic, proportional response to frequency deviation, fully deployed within seconds, symmetric in most of continental Europe. Procured in joint auctions by the FCR Cooperation TSOs with ever-shorter contracting blocks — a deliberate design evolution to open the product to batteries, which now dominate several FCR markets.

aFRR — automatic restoration

Restores frequency and cross-border exchanges to setpoint via continuous automatic signals. Full activation times have been harmonised downward (5 minutes as the European standard), and PICASSO clears cross-border aFRR energy at marginal prices every activation cycle — the closest thing Europe has to a real-time energy price.

mFRR — manual restoration

The workhorse for larger, longer imbalances: scheduled or direct activation with a quarter-hour delivery rhythm. MARI's common merit order makes mFRR the product where cross-border competition bites hardest — a cheap Nordic bid can displace an expensive local activation when transmission capacity allows.

RR — replacement reserves

Slower reserves that relieve activated FRR and restore the system's ability to absorb the next incident. Not all TSOs use the product; its role shrinks as intraday markets trade closer to delivery and portfolios self-correct — and the European RR platform stopped clearing at the end of 2025.

Prequalification

The gate to the market: technical tests of ramp rate, availability and controllability per product. Prequalification rules quietly decide the supply side — the minimum bid size, pooling rules and metering requirements determine whether an aggregated portfolio of heat pumps can compete with a gas turbine.

Standardisation vs. specificity

Standard products enable cross-border exchange but flatten local nuance; specific products preserve it at the cost of fragmenting liquidity. EBGL's compromise — standard products for exchange, specific products by exception — is being stress-tested as storage and demand response ask for terms the standard products never anticipated.

One platform per process

EBGL did not ask TSOs to harmonise their balancing markets. It asked them to build four specific pieces of shared infrastructure and connect to them: a platform for imbalance netting under Article 22 (IGCC), one for aFRR balancing energy under Article 21 (PICASSO), one for mFRR balancing energy under Article 20 (MARI) and one for replacement reserves under Article 19 (TERRE, running the LIBRA system). Each works the same way in outline. National TSOs post their needs and their balancing-energy bids into a common merit order; an activation optimisation function nets opposing needs, then matches the remainder across borders using whatever cross-zonal capacity the intraday market has left behind; TSO–TSO settlement moves the money afterwards.

Two boundaries are worth fixing early. The platforms exchange balancing energy, not balancing capacity — reserve procurement stays national or regional, so Europe integrated activation first and left the question of who pays to keep reserves available for later. And the platforms are second in line for transmission: they use the cross-zonal capacity remaining after intraday trading closes, which makes every decision about when intraday closes a decision about how much cross-border balancing is possible. That subordination is the thread running through everything below.

PICASSO — a price every four seconds

The aFRR platform went live on 1 June 2022 with the German and Austrian TSOs. Its activation optimisation function runs every four seconds: each cycle it nets demand, selects bids from the common merit order subject to available capacity, and produces a cross-border marginal price in each direction. Those four-second prices and volumes are then aggregated into the fifteen-minute market time unit for settlement. The result is the fastest-moving price in the European sequence and the closest thing the continent has to a genuine real-time energy price. Twenty TSOs were connected by early 2026, and Slovenia's ELES joined in July 2026.

Marginal pricing across a coupled area exports price formation along with the energy, and Italy was the first member to decide it had imported too much. Extreme swings in Italian imbalance prices were traced by market participants to single very high or very low bids setting the marginal price for every connected system at once, and in early 2024 the Italian regulator ARERA ordered Terna to suspend operational participation in the platform by 15 March. ACER's July 2024 decisions on the balancing implementation frameworks were the repair: the activation optimisation was adjusted, the platform's ±99,999 €/MWh technical price limits were replaced by transitional limits of ±15,000 €/MWh through July 2026, and TSOs were allowed to submit elastic demand — pricing a segment of their need so that a TSO can decline to activate above a threshold it considers unreasonable. Terna reconnected at the end of 2025. The episode is the clearest available evidence that coupling balancing energy is not merely a volume question: a platform with an inelastic demand side has no way to say no, and the marginal bid travels as far as the interconnection does.

MARI — quarter-hours and a common merit order

The mFRR platform went live on 5 October 2022 with the four German TSOs and ČEPS. Bids close twenty-five minutes before the start of the quarter-hour they serve; the standard product has a full activation time of twelve and a half minutes and can be activated on schedule, at the start of the market time unit, or directly within it when the need appears mid-period. Each cycle the optimisation nets TSOs' opposing needs first, then clears what remains against the common merit order. mFRR is where cross-border competition bites hardest, because a quarter of an hour is long enough for a cheap distant bid to displace an expensive local one whenever transmission allows — and because, unlike aFRR, the product does not require a continuous automatic control link to the activating TSO.

Membership, accession, connection

These are three different states, and only the third moves energy. MARI counted twenty-nine member TSOs at the start of 2026 and fifteen connected ones: the four German TSOs, ČEPS, APG, AST, Litgrid, Elering, REN, SEPS, Red Eléctrica, Elia, TenneT NL and Bulgaria's ESO, which went live in February 2026. The legal deadline for both platforms was July 2022.

The gaps are instructive rather than embarrassing. France sits on PICASSO but not on MARI: RTE's national derogation expired in July 2024 and it has been sharing cross-zonal capacity with the mFRR platform ever since without submitting bids to it — participating in the transmission sense while abstaining in the market sense. Poland is the mirror image, connected to PICASSO and expecting MARI at the end of 2026, having deliberately sequenced the two to manage implementation risk. Greece now expects 2028. The Nordic TSOs built their own mFRR platform first, live in March 2025, and are behind on both European ones. Switzerland is the sharpest case of all: Swissgrid's technical readiness is acknowledged by both projects, but Swiss participation turns on Articles 1(6) and 1(7) of the EB Regulation and is before the Court of Justice — a market boundary drawn by treaty law rather than by physics. Coverage claims about the European platforms should therefore always be read as operational connection, never as project membership.

TERRE — and why the RR platform is finished

TERRE delivered the LIBRA platform in January 2020. Eight TSOs signed up, six ever connected, and in its final year four were still clearing: RTE, Swissgrid, Red Eléctrica and REN. Great Britain, an original member whose industry had already built the settlement changes, was granted a derogation by Ofgem and left the project behind at the end of the Brexit transition. The platform then stopped for good: RTE and Swissgrid disconnected on 17 December 2025, Red Eléctrica and REN cleared until 30 December 2025, and LIBRA was decommissioned with the project formally closing in March 2026. It is the first European market platform to be switched off.

The decisive reason is a timing conflict the product could not survive. RR bids closed fifty-five minutes before the delivery hour, a lead time the standard product's thirty-minute full activation time made necessary, and the platform ran twenty-four hourly gates a day. The Electricity Market Design reform — Regulation (EU) 2024/1747, amending Article 8 of Regulation (EU) 2019/943 — moved the intraday cross-zonal gate closure time from sixty to thirty minutes before real time with effect from 1 January 2026. Under the old rule, an RR clearing at H−55 sat just after cross-zonal intraday trading had closed and could be given firm capacity. Under the new one it sits half an hour inside the open intraday window, and any cross-zonal capacity it took would be capacity withdrawn from the very market the reform had just opened up. The exchange that RR existed to organise had nowhere left to sit in the sequence.

Behind the timing sat two slower forces. Replacement reserves are optional under EBGL — a TSO must join the platform only if it actually uses the product — so membership could shrink, and it did, which is the opposite of every other European platform. And the product's economic space was being squeezed from both sides: intraday markets trading closer to delivery now do the pre-positioning that RR used to do, while MARI's common merit order covers the manual restoration need with less than half the lead time. TERRE was not a failed platform. It was a working platform for a product whose place in the sequence disappeared underneath it — and the lesson generalises. Every reform that shortens the distance between the last trade and real time takes territory from the slowest balancing product first, which is why an apparently technical parameter like the intraday gate closure time deserves to be read as a balancing-market decision.

Marginal or pay-as-bid?

EBGL settled the argument for balancing energy — marginal pricing, so bids reveal true costs and scarcity is priced honestly. Capacity procurement remains mixed, and the choice matters: pay-as-bid capacity auctions invite guess-the-clearing-price bidding that favours incumbents with better information.

The one durable real-world deployment of pay-as-bid is instructive: Great Britain adopted it in 2001 for its balancing mechanism only, after the regulator judged it unsuitable for the main wholesale auction — tolerable for thin residual volumes, corrosive for the reference price. Two decades of analysis of that experiment found the price effects of the format itself weak to indistinguishable; what moved GB prices was deconcentration, not the payment rule. That asymmetry — pay-as-bid tolerated at the residual margin, marginal pricing defended at the core — is now effectively the European consensus, written into EBGL for balancing energy.

Single or dual imbalance pricing?

Single pricing pays a BRP whose imbalance helps the system, recruiting every portfolio as a passive balancer. Dual pricing punishes all deviation, prioritising discipline over efficiency. European harmonisation has pushed firmly toward single pricing — with debate now centred on when deliberate imbalance becomes gaming.

Scarcity in the imbalance price

If the imbalance price never exceeds the cost of the last activated bid, no one has a reason to be available beyond contracted capacity. Scarcity components — value-of-lost-load-linked adders when reserves run thin — make the imbalance price the true investment signal for flexibility, at the cost of political exposure in tight winters.

Procurement timing

Contracting capacity a year ahead buys investor certainty but locks in yesterday's technology mix; daily procurement lets weather-dependent portfolios compete but exposes the TSO to tight days. The European trend is short: daily (or intra-day) capacity auctions with four-hour products are becoming the norm.

Who may participate?

Aggregation rules, minimum bid sizes, symmetric-product requirements and metering granularity together define the de facto entry barrier. The Electricity Directive requires market access for demand response and storage; implementation quality varies — and is one of the clearest markers separating leading from lagging balancing markets.

The live argument is whether the standard 1 MW minimum bid should fall to 100 kW to admit small assets directly. The DSO side of the debate argues it should not: aggregation into service-providing groups already carries kilowatt-scale resources to market — local flexibility platforms accept bids down to single kilowatts — and shrinking the standard product instead multiplies bids, prequalification load and settlement complexity for every TSO. Whether small assets reach balancing through smaller products or bigger aggregators is quietly one of the more consequential design forks in the demand-response network code.

TSO coordination depth

Platforms exchange energy; deeper designs share reserves themselves. Regional reserve-sharing (as pioneered in the Nordic cooperation) reduces total procurement but requires trusting a neighbour's reserves during your own scarcity — a governance question wearing a technical costume.

What a balance responsible party actually carries

Every megawatt-hour injected into or withdrawn from a European grid belongs to somebody's balance responsible perimeter. The BRP submits a schedule, the meter records what happened, and settlement charges or pays the difference over the imbalance settlement period. That obligation is not administrative housekeeping: it is what gives the balancing market a demand side. Without it nobody pays for the TSO's actions, and more importantly nobody has a financial reason to be accurate in the first place.

The BRP is not the BSP, and the distinction is worth holding firmly because the same company usually plays both parts. A balancing service provider sells capacity and energy to the TSO from prequalified assets and is paid for delivering them. A balance responsible party carries the financial consequence of its portfolio's deviation. When a BSP is activated, its own BRP perimeter moves — the plant is producing something other than its schedule — so settlement has to neutralise that movement, or the provider would be penalised for doing exactly what it was paid to do. Getting that correction right is one of the least discussed and most error-prone parts of the whole design.

Underneath sits a tension that explains most of the argument that follows. The imbalance price does two jobs at once: it recovers the cost of the TSO's balancing actions, and it sets the incentive facing every portfolio in the system. A price built to recover cost accurately is usually a poor incentive, and a price built to sharpen incentives usually leaves a residual that has to be socialised somewhere. No design escapes the trade-off; every national rulebook is a position on it.

Proactive and reactive balancing

Before any of the pricing detail, there is a prior choice about how the system operator behaves, and it does more to explain cross-border differences than any settlement formula. A reactive TSO waits for the imbalance to materialise and activates against what it measures. A proactive TSO anticipates it from its own forecast and activates ahead of time, typically leading with mFRR and holding automatic reserve as a last resort. The Netherlands is the reactive archetype, France the proactive one.

Each buys something and gives something up. Proactive balancing smooths operation and reduces reliance on the fastest and most expensive products, but the system operator has substituted its forecast for the market's, activation volumes rise, and the imbalance price carries less information because much of the correction happened before the imbalance appeared. Reactive balancing keeps the price informative and recruits every portfolio in the country as a passive balancer, responding to a live signal — but it only works if BRPs can actually respond, which requires liquid intraday trading right up to the gate, telemetry good enough to see the position, and a settlement period short enough to reward acting quickly.

So the choice is not free-standing. It interacts with single pricing, which is what makes passive balancing profitable; with intraday liquidity, which is what makes it possible; and with how much automatic reserve the TSO chooses to hold. A country with a thin intraday market and a proactive operator will find that neither mechanism is doing the work its design assumes: the market cannot self-correct and the TSO is paying to correct on its behalf. And the European platforms have quietly disturbed the balance — a proactive TSO that used to pre-position with manual reserve now finds automatic reserve clearing every four seconds against a continental merit order, which is one reason several systems have reported unfamiliar imbalance-price volatility after connecting to PICASSO.

How the price is built

The Balancing Guideline harmonised the frame and left the details national. Single pricing — one price per settlement period per area, paid or charged whichever way the portfolio deviated — is the target model, and ACER's monitoring in July 2024 found twenty TSOs across seventeen member states using it, five using dual pricing and one moving to it; nineteen TSOs in sixteen member states add one or more components on top. The settlement period is now fifteen minutes across the Union, the last derogations having expired; Great Britain settles on thirty.

Everything else is a choice. Whether the price is the marginal one — set by the last and most expensive activation — or a volume-weighted average of everything activated. Whether it follows the activated merit order or the system's net position. And what gets added: a scarcity term, an incentive term, a factor restoring the operator's financial neutrality. The table below is what those choices look like in the systems that matter most.

Why the components matter more than the label

Almost every system in the table says “single pricing”, and almost none of them means the same thing by it. Two countries both on single marginal pricing can transmit opposite incentives depending on what is bolted on. An incentive term capped against the intraday index keeps the imbalance price tethered to an alternative the BRP could have traded, which is a deliberate decision to limit how far the signal may run. A scarcity term does the reverse: it makes the price diverge on purpose when reserves are thin, so that being short at the wrong moment costs more than any energy-market alternative. And a financial-neutrality factor, which exists because the TSO's costs and its settlement receipts never quite match, dilutes the marginal signal in proportion to how much residual there is to recover.

Read the table for those components rather than for the single-or-dual label and the map looks different. Germany's averaged price with a capped incentive term is a fundamentally gentler instrument than the Dutch marginal price, even though both are “single”. Britain's explicit scarcity repricing puts a value-of-lost-load number directly into a settlement price, which is a stronger statement about what shortage is worth than most continental systems make anywhere in their design.

The line between passive balancing and gaming

Single pricing pays a portfolio that deviates in the direction that helps. That is the intended behaviour: a BRP running long while the system is short is supplying balancing energy without ever bidding, and the design recruits it deliberately. The difficulty is that the same behaviour becomes something else once a portfolio is large enough to influence the price it is being paid, or once a participant deviates predictably enough that its position is a strategy rather than a response. Nothing in the settlement rules distinguishes the two, because the distinction is quantitative rather than qualitative, and no European regulator has yet drawn the line cleanly. It is the open question at the end of every imbalance-pricing reform, and it gets sharper as portfolios consolidate and as automated trading makes the response to the imbalance price faster than the price itself.

One market where Europe has two

The European sequence ends trading and then starts balancing: intraday closes at the gate, the TSO becomes the single buyer, and a separate market in reserves and activation takes over. A US market never hands off. The day-ahead market produces a financially binding schedule, and the real-time market re-clears the same commodity against actual conditions, dispatching generators and settling deviations at the price it produces. There is no gate closure in the European sense and no moment when the market stops and an operator starts.

The consequence runs through everything below. Because balancing is the real-time market rather than a market beside it, the balancing price is the energy price. Nobody constructs it from activation costs, adders and penalty terms, and the arguments Europe has about single versus dual pricing, about whether the imbalance price should carry a scarcity component, and about how much information the price should convey, simply do not arise in the same form. The American equivalents are arguments about how to price a market, not about how to build a settlement formula.

Five-minute dispatch, and then five-minute settlement

Security-constrained economic dispatch re-optimises the system every five minutes, issuing instructions and producing a price at each node. For years settlement lagged that: markets dispatched on five minutes and settled on the hourly average, so a generator following a costly instruction was paid an average that did not reflect it, and the incentive to follow dispatch was blunted. FERC Order 825 closed the gap in 2016 by requiring settlement intervals to match dispatch intervals. The principle — that a participant should be settled at the price of the interval it was actually asked to respond to — is the same one behind Europe's move to a fifteen-minute imbalance settlement period, arrived at a decade earlier and three times finer.

Co-optimisation, and the price of holding reserve

This is the deepest structural difference and the one European designers most often point to. A US real-time market clears energy and operating reserves in the same optimisation. A unit held back to provide reserve is not producing energy, and the optimisation prices that forgone output automatically: the reserve price emerges as the opportunity cost of the energy the unit did not sell. Nobody has to procure the reserve separately or guess what it should cost.

Europe procures balancing capacity in its own auctions, frequently a day or more ahead, and clears balancing energy later on a different platform. The two are sequential rather than simultaneous, so the opportunity cost is borne by the provider and expressed in a capacity bid rather than computed by the market. That is defensible when reserve requirements are stable and known in advance, and it becomes expensive when the system is tight, because the capacity auction has already committed the unit at a price set before anyone knew what energy would be worth. ERCOT spent years without real-time co-optimisation and has been implementing it precisely to close that gap.

Scarcity by demand curve rather than by adder

When reserves run short, something has to lift the price, and the two systems do it in opposite directions. Europe mostly adds: a scarcity component is appended to an imbalance price that was built from activation costs. The American approach defines a demand curve for reserves in advance — an operating reserve demand curve — which states what the system is willing to pay for the next megawatt of reserve at each level of remaining margin. As margin erodes the curve lifts the reserve price, and through co-optimisation it lifts the energy price with it.

The difference matters more than it sounds. A demand curve is a published, ex-ante statement of the value of reliability, so a generator or a battery can compute in advance what a tight hour is worth and invest against it. An adder is a rule applied after the fact. ERCOT's curve is the most consequential example, because it is an energy-only market with no capacity mechanism: the scarcity price is the entire investment signal, which is why the shape of that curve is argued over more fiercely than almost anything else in American market design.

Nodal prices, and where congestion goes

US real-time prices are locational, computed at thousands of nodes, so congestion is inside the price rather than handled after it. A constrained network produces different prices on either side of the constraint, and a generator behind it sees directly what its location is worth. Europe's zonal design produces one price across a large area and resolves the resulting physical infeasibility afterwards, through redispatch and countertrading paid for outside the market.

This is the oldest and least settled comparison, and it is worth being precise about what it does and does not prove. Nodal pricing produces better short-run signals and avoids paying twice for the same megawatt. It also produces thousands of prices, makes hedging harder, requires a financial transmission-rights market to make long-term contracting possible, and redistributes value in ways that are politically difficult to do to an existing market. The relevant question for Europe is not which is better in principle but whether the redispatch bill has grown large enough to justify the transition cost, which is a question about magnitude rather than theory.

The Western Energy Imbalance Market

One American development maps almost directly onto European experience. The Western Energy Imbalance Market extends real-time dispatch across a footprint of utilities that are not part of a single organised market, clearing imbalance energy every fifteen and five minutes across balancing authority areas that keep their own dispatch otherwise. It is, structurally, a platform for sharing balancing energy between neighbouring systems — the same thing PICASSO and MARI do, arrived at from the opposite direction.

The sequence is instructive. The imbalance market came first and proved the value of sharing in real time; the day-ahead extension followed years later, once participants trusted the mechanism. Europe integrated day-ahead first and is still completing balancing integration more than a decade on. Both ended up in a similar place, and neither took the route the other did.

What the comparison is actually for

It is tempting to read all of this as an argument that one design is better. It is more useful as a demonstration that several European debates are artefacts of a structural choice rather than genuine open questions. Dual imbalance pricing exists because the imbalance price is constructed rather than cleared. The difficulty of valuing reserve exists because capacity and energy are procured sequentially. Redispatch cost exists because the price is zonal. Each of those is downstream of the decision to separate energy from balancing and to price over an area rather than a point.

None of that makes the American design transplantable. It rests on a single regulator per market, on nodal networks built alongside the market, and on a willingness to let prices reach values European regulators would not tolerate for long. But when a European proposal claims to solve a balancing problem, it is worth asking whether the problem is intrinsic or inherited — and the real-time markets are the cleanest way to tell the difference.