Baseline Methods

The baseline is, in effect, a forecast with an adversary — which is why “good” forecasting for flex markets is defined by manipulation-resistance and recalculability, not just accuracy. See STLF for Flexibility Markets — What Counts as Good and How to Achieve It.

From activation to payment

Why baselines are hard for DERs

In traditional balancing markets, large generators commit to a schedule in the day-ahead market; the schedule serves as the baseline, and flexibility delivery is measured as deviation from it. For distributed energy resources (DERs), no such individual schedule exists. A heat pump, battery, or EV charger has no pre-committed output plan visible to the market. The flexibility buyer (DSO or TSO) must therefore estimate what the resource would have done — a fundamentally uncertain exercise. (Source - Lind et al Baseline Methods (2023))

A baseline’s reliability also depends on the historical interval data it is built from actually being measured, not silently estimated — see Load Forecasting › Data challenges in flexibility contexts for a documented case where DSOs fill gaps with a flat synthetic profile that is indistinguishable in format from a real reading.

Baseline design has direct consequences for:

• FSP revenue: a systematically high baseline inflates apparent flexibility delivery and over-rewards the FSP
• Market integrity: FSPs may be able to game certain methods by changing behaviour before the baseline measurement window (e.g. charging a battery just before an MBMA reading)
• Market participation: overly complex baseline methods are a barrier to entry for smaller FSPs

Methods

Ten methods are established in the literature and practice (the cited source’s own section header undercounts these as “nine” over what is actually a 10-row table — an error inherited from the source, corrected here):

MBMA and the battery integrity problem

MBMA (Meter-Before-Meter-After) reads the meter immediately before activation. If that reading is used directly as the baseline, a battery operator providing upward flexibility via increased injection can game the method: by switching from discharging to charging just before the pre-activation reading, the apparent baseline shifts toward higher consumption (less injection), making the subsequent injection appear larger than it really is — inflating measured delivery and payment. (Source - Lind et al Baseline Methods (2023))

The alternative Lind et al. propose for batteries is a zero baseline: any injection during activation counts as delivered, with no prior-period manipulation possible.

SWITCH offers a zero reference (noll-referens) alongside MBMA and own-plan uploads; its documentation does not say which resource types must use which method. (Source - SWITCH User Documentation (2026))

Capacity clearing vs. energy settlement

A common misconception: that capacity-based products (where clearing is in MW, not MWh) eliminate the need for a baseline. This is true for the clearing step — the DSO procures a capacity commitment and pays an availability fee in SEK/MW. But energy delivery validation still requires a baseline. When the DSO activates, the FSP must deliver the agreed energy volume; comparing metered output to the baseline determines whether the 75% delivery threshold was met and what the activation payment should be.

In SWITCH‘s TO (Tillgänglighetsordrar) and DO (Direktordrar) products: the market clearing is MW-based (capacity bids), but the FSP is still expected to deliver the energy amount awarded. A baseline is therefore required for post-activation settlement, even though clearing used capacity logic.

Method selection by DER type

Swedish market practice

SWITCH offers three methods (Source - SWITCH User Documentation (2026)):

1. Egen referensplan — FSP uploads own baseline (UI or API); deadline D-1 09:30 for DA, H-4 for ID
2. Noll-referens (zero baseline) — assumes 0 production/consumption; suited to resources that only produce/consume during activation
3. MBMA — automatic calculation from resource metering

The SWITCH documentation locks down baseline integrity: baseline values cannot be uploaded or changed after a flexibility trade has occurred on that market. Resource control was moved to the DSO after a 2022/23 risk assessment identified that FSPs could manipulate resource registration data; the control change was a preemptive measure, not a response to confirmed abuse. (Source - BeFlexible D5.1 Demo Planning and Deployment (2024))

NODES (via sthlmflex) used a 5-day rolling average as the standard method: same hour across the 5 preceding working days. FSPs could alternatively upload their own baseline or agree an alternative with the buying DSO. (Source - sthlmflex säsong 3 (2022-2023))

European LFM baseline practices

Sassone et al. (2025) document the baseline methodologies deployed across 7 European LFM case studies — the most comprehensive comparative dataset available, though only 4 of the 7 (GB, Netherlands, France, Slovenia) are labeled operational in the source’s own market table; the other 3 (Sweden, Portugal, Italy) are concluded or pilot-stage. (Source - Local Flexibility Markets in Europe Critical Review (2025))

The key limitation of historical baselines: weather-driven demand shifts between the observation period and the activation day can cause large errors. Correction factors (actual flow H hours before activation) partially mitigate this, but no single method suits all DER types. Swedish DSOs (sthlmflex) offered BSPs the user-nominated approach — own forecasting model, DSO-approved — placing the accuracy burden on the FSP while giving them maximum flexibility. The Netherlands GOPACS model avoids baselines entirely by defining each bid as a modification to a commercial schedule; this is conceptually similar to capacity-limitation products (see below).

Capacity-limit products as a baseline alternative

The European Commission’s 2025 LFM study (VITO) formally endorses capacity-limiting products (operating envelopes) as a structurally different approach that avoids baseline calculation entirely. Instead of measuring a flexibility volume (MWh change vs. counterfactual), the product defines a power cap at the connection point: the FSP commits to staying below (or above) a specified MW threshold. Delivery is verified against the cap, not against a counterfactual baseline.

This approach is particularly suited for:

• LV grid congestion where individual sub-metering is impractical
• Cases where baseline manipulation risk is high (batteries, weather-driven DG)
• Early-stage markets where settlement complexity is a participation barrier

The trade-off: capacity-limit products require different clearing algorithms and cannot be easily stacked with energy-based products in the same market session. They represent a different product architecture, not just a different measurement method. (Source - EC LFM Specification and Design Criteria (VITO, 2025))

MaxUsage™ — the Swedish worked example, and a baseline caveat. NODES MaxUsage™ at Effekthandel Väst is the clearest operational Swedish capacity-limit product: the FSP and DSO jointly set a consumption ceiling for fixed peak hours and the FSP is paid for staying under it — verified by direct metering, with no per-event baseline (Renova capped at 75 kW for 07:00–10:00; GoCo halved workplace EV-charging power 08:00–12:00). But it only appears baseline-free: the value is benchmarked against a historical-consumption reference (Renova’s ~300 kW historical draw in those hours), which is a counterfactual that drifts as efficiency and behaviour change. That drift is one of two reasons Kinnekulle Energi discontinued the product (GKT’s 450 kW reference shifted between pilot and active seasons, confounding validation; the other is that mandatory time-differentiated network tariffs from 2027 make the product redundant — see Kinnekulle Energi for the specific sourcing), even as Effekthandel Väst scaled it. The lesson for baseline design: a capacity-limit product removes the baseline from settlement but not from valuation unless the cap is anchored to a firm, non-drifting reference (e.g., contracted capacity) rather than historical consumption. (Source - Effekthandel Väst Produkter och MaxUsage (NODES, 2024))

FSP experience — baseline as a participation barrier

Palm et al. (2023) collected qualitative evidence from CoordiNet Uppland and Skåne FSPs and PFSPs showing that baseline calculation is a genuine day-to-day barrier, not just a theoretical design concern. (Source - Palm et al LFM Drivers and Barriers (2023))

Conceptual difficulty: Several FSPs found the concept of “what would we have consumed” genuinely confusing — “You could say that you had intended to consume something, but it may be untrue what you come up with. You might claim that you reduced the consumption a lot, but it might have happened anyway.” This uncertainty was particularly acute for consumption-side flexibility, where the counterfactual is less obvious than for a generator.

Baseline manipulation — the sports arena case: A story circulated among CoordiNet participants of a sports arena that turned on all its lighting, used that as its base load, and then earned money on the resulting falsified reduction in consumption (Palm et al. report it as a circulating story, not a verified incident, and give no timing or pilot). Read alongside the FSPs’ confusion about how baselines worked, it points to a concern about the integrity of a system that could be gamed.

This corroborates the SWITCH response (Source - BeFlexible D5.1 Demo Planning and Deployment (2024)): after a 2022/23 risk assessment, E.ON moved resource control to the DSO (not FSP) as a preemptive anti-manipulation measure — moving control of resource configuration (not baseline calculation: FSPs can still upload their own baseline before a trade) from the resource owner to the market operator.

D-1 timing and real-time information loss: FSPs also noted that committing to a baseline a full day ahead means losing all real-time information that becomes available between bid submission and activation. One FSP: “In CoordiNet, you lose all information that is added between the day before and the control occasion … you can have a very good forecast, but it is a dynamic system.” This is less a baseline method problem than a market timing problem, but it compounds baseline uncertainty for FSPs — their baseline was set without access to information that only became available on the activation day.

Submetering as enabling infrastructure for baseline accuracy

A structural solution to the multi-DER baseline problem is dedicated measuring devices (DMDs) — circuit-level meters that measure a specific DER (EV charger, heat pump, battery) independently of total building consumption. See Submetering for the full concept.

DMDs solve the attribution problem for mixed-DER portfolios:

• Without a DMD, the whole-building meter contains the activation signal plus all unrelated household load variations — making the baseline inherently noisy
• With a DMD, only the specific DER’s pre/post consumption is captured; the baseline is clean and device-specific
• Per-device attribution enables applying different methods to different DER types in the same portfolio: zero baseline for the battery’s injection output, MBMA for the heat pump’s consumption reduction

Art. 7b of Regulation 2024/1747 (EMD Reform) lets TSOs, DSOs and aggregators use DMD data, with customer consent, for flexibility observability and settlement; validation rules are set by Member States and implementation rules are pending. The Network Code on Demand Response T&C development is where Sweden may define how DMD data is used in baseline calculations. The Comillas/BeFlexible paper sees submeters as a complement to CAP-level baselines, not a replacement (Source - Submetering for Flexibility Services Comillas (2024)).

An EU-level review of IA–DER commercial practices confirms that baseline accuracy directly determines FSP revenue fairness: a systematically high baseline over-rewards the FSP; a low baseline under-rewards real flexibility delivery. Aggregators managing third-party DERs have a fiduciary interest in accurate baselines that cannot be provided by whole-building meters alone. (Source - Aggregators DR Relationships Comillas (2025))

Regulatory horizon

The Network Code on Demand Response will require standardised settlement processes for flexibility markets across the EU. ACER’s NC DR text (Art. 14) calls for baseline methods that must be “recalculable, transparent, precise, accurate, and unbiased” and requires a register of all MS-approved baseline methods. (A previously stated claim that “NC DR Art. 35 explicitly allows alternatives that do not require baselines” has been removed — Art. 35 actually covers pricing/settlement structure and requires settlement to use the relevant baseline; no article in the NC DR text was found to state a baseline exemption for capacity-limit products.)

The European Commission study (2025) sets a medium-term target of baseline methods coordinated at MS level and informed by an EU-level baseline library containing all approved national approaches — a structured convergence mechanism that avoids forced harmonisation while enabling knowledge sharing.

NC DR Art. 43–44 links settlement requirements to DNDP flexibility procurement, which will bring SWITCH’s and NODES’s current informal methods into scope of regulatory review. Sweden’s ~6-year operational experience with MBMA, noll-referens, and rolling-average methods will be directly relevant input to the national T&C baseline definition process. (Source - EC LFM Specification and Design Criteria (VITO, 2025))

ACER’s monitoring of baselining (DRESG, October 2026)

ACER’s draft monitoring report on baselining, discussed at the third Demand Response European Stakeholder Group meeting on 6 October 2026, proposes to say that in the relationship between a market participant and the system operator a baseline is needed mainly for two purposes: imbalance settlement (allocating responsibility between balance responsible parties at the same metering point) and settlement of a service whose activated volume is based on metered values. It argues that day-ahead and intraday market schedules are an appropriate baseline for balancing and other post-market actions, being committed, harder to manipulate than physical schedules and comparable across technologies, while for demand response physical schedules would be inappropriate and could deter participation. A declarative baseline is one where the market participant decides the calculation method or the data; ACER is considering calling it a “subjective forecast of the activation event”, in contrast to methods where method and data are agreed so that anyone can verify the result. ENTSO-E noted that the Swedish entry in the draft marks all methods as pilot although some are now business as usual, to be checked with Ei, and SmartEn noted that the same methods run through NODES in Sweden, Finland, Belgium and Norway. The slides are preliminary and unendorsed (Source - DRESG 3rd Meeting (2026-10-06)).