TSO-DSO Coordination — The Central Design Problem

The electricity grid is physically unified. Electrons don’t know if they’re in a distribution network or a transmission network. But the organizations managing those networks are separate — with different ownership, different markets, different regulatory authorities, and different time horizons. The challenge of aligning their actions around a shared resource pool is the central design problem in modern flexibility market architecture.

Why the problem exists

The traditional grid was designed for one-way power flow: generation at transmission level, consumption at distribution level. In this world, TSO and DSO had largely separate operational domains. The TSO balanced generation and load system-wide; the DSO delivered power to end customers. Coordination was minimal.

The energy transition breaks this separation:

• Distributed energy resources (solar, batteries, EVs, heat pumps) connect at distribution level but contribute to system-wide balancing
• Distribution constraints shape how much demand can grow, which affects system-level adequacy
• A flexibility resource — say, a large battery in a distribution substation — can simultaneously provide TSO balancing services (FCR, mFRR) and DSO congestion relief

When the same resource can be claimed by both TSOs and DSOs, someone has to decide who has priority. That decision, and the process for making it repeatedly, at scale, in real time, is the TSO-DSO coordination problem.

The Swedish version: the subscription mechanism

Sweden currently has a proxy for TSO-DSO coordination built into its infrastructure pricing: the subscription mechanism (abonnemang mot överliggande nät). Regional DSOs buy an annual subscribed power capacity from Svenska kraftnät; when they forecast a risk of exceeding it, they activate local flexibility markets or Villkorade Avtal.

This mechanism inadvertently creates a TSO-DSO coordination framework:

• If Svk denies a subscription raise, the DSO must use its local market → demand for flexibility is high → markets develop and function
• If Svk grants a temporary subscription (retractable at any time, pricing ~240–280 SEK/MWh), the DSO can avoid activating the local market → demand evaporates → thin markets, rare activation

The CoordiNet demonstration made this structural effect empirically vivid: Uppland (Svk frequently denied raises) developed a stable, high-volume market averaging 248 SEK/MWh over three winters. Skåne (Svk routinely granted temporary subscriptions) produced a sparse, high-price market averaging 2,285 SEK/MWh over the same period with 50× lower volume. The market design was similar; the TSO policy was different. (Source - CoordiNet D4.7.2 Swedish Demonstration (2022))

The implication: Sweden’s existing TSO-DSO coordination is not a formal framework — it is a price signal embedded in how Svk manages the subscription market. This is efficient under some conditions, but it is opaque, non-transparent, and makes local flex market liquidity contingent on TSO administrative decisions rather than grid economics.

Post-CoordiNet status: The DSO Entity’s 2026 knowledge-sharing report explicitly confirms that CoordiNet coordination between Svk and DSOs was discontinued after the CoordiNet project ended. As of early 2026, no formal operational TSO-DSO coordination framework for shared flexibility resources exists in Sweden. The SO GL Art. 182 agreement between Svk and DSOs — which would establish a formal cooperation framework for reserve delivery from distribution-connected resources — remained pending as of early 2026 (a June 2026 draft now exists; remiss autumn 2026, in force spring 2027 — see below). Sweden’s current state is therefore the “Separate SO” model with an informal subscription-price signal as the only coordination mechanism. (Source - DSO Entity Distributed Flexibility Practices (2026), Source - Svk Energiföretagen SO GL Art 182 Avtal (2026))

EU taxonomy: four market coordination models

The European Commission’s 2025 LFM study (VITO), surveying 37 LFM initiatives across Europe, provides the clearest EU-level taxonomy of how local flexibility markets relate to TSO balancing markets. Four models are identified, ranging from full separation to full integration:

Sweden’s position: E.ON SWITCH operates as a sequential market — E.ON’s market gate closures are timed to allow bids to be forwarded or resources rebought from Svk’s balancing market before the next balancing gate. The CoordiNet battery demonstration (0.48 MW/1 MWh, Skåne) is a concrete example: if cleared on SWITCH day-ahead at 10:00, the operator could send an FCR-D rebuy request in Svk’s D-1 auction at 18:00. This is structural sequencing — not accidental.

GOPACS as the benchmark: the Netherlands’ common market GOPACS (Grid Operator Platform Alleviating Congestion on the Subnet) is consistently cited in the EC study as the highest-scoring coordination model. It operates a shared order book where Dutch DSOs (Stedin, Liander, Enexis) and the TSO (TenneT) clear bids from the same pool of resources in a single session. Resources are allocated to the SO with the highest willingness to pay, maximizing overall value. GOPACS scores highest on value stacking and market efficiency among all 37 LFM initiatives.

Technical specification (from Energiforsk 2025:1088):

The 150% overbooking ensures that the market has sufficient volume even when some bids are not delivered, while the shared order book means resources are always allocated to the SO with the greatest need.

Finland’s FinFlex as a Nordic data point: on FinFlex, Fingrid (TSO, 400 kV) and Helen Sähköverkko (DSO, 110 kV in Helsinki) buy from the same ShortFlex bids but in sequence: the DSO’s auction closes at D-1 10:00, before day-ahead gate closure, and the TSO then trades continuously from D-1 15:00 until one hour before delivery. The operator reports ~69 MWh reserved under availability contracts at ~192 EUR/MWh on average and 8 MWh activated at ~99 EUR/MWh, a cumulative figure with no period or per-buyer split (Source - Fingrid FinFlex TSO-DSO Congestion Market (web)).

Denmark’s intermediate step — Cerius-Radius / TREFOR El-net pilots: two Danish DSO groups — Cerius A/S + Radius Elnet A/S (jointly) and, separately, TREFOR El-net — notified a temporary pilot methodology to Forsyningstilsynet (the Danish Utility Regulator) in 2025 for procuring flexibility as an explicit alternative to reinforcing main transformer stations, each testing a distinct model:

Legal basis: Elforsyningsloven §73a stk. 2 (added by Lov nr. 1677/2024, transposing the EU Electricity Market Directive) and §45b, tied to EU Electricity Market Regulation Art. 18 (tariff methodologies must incentivize flexibility use) — the same EU provision underlying Sweden’s effektavgift design requirements. Both pilots are explicitly framed as temporary and learning-focused, aimed at testing (a) whether customers can deliver the needed flexibility, (b) how reliably, and (c) whether it beats reinforcement cost (compared via a 40-year depreciation, 5% WACC annuity) — ahead of a permanent, sector-wide Danish DSO flexibility product. Precedent: Energinet’s 2019 voltage-regulation pilot tender. As of the notification, Forsyningstilsynet approval was pending; an addendum of 27 February 2026 sets the pilot duration.

The TREFOR El-net model — a fixed-price standing capacity ceiling — is structurally closer to a Swedish villkorat avtal than to a dispatch product, while Cerius-Radius’s availability-plus-activation structure resembles a Swedish LFM-e/energy-activation product. Denmark’s approach differs from the GOPACS common-market model above in scale and maturity: this is a two-station, single-DSO-group pilot rather than a national shared platform, closer in spirit to Sweden’s own early single-DSO pilots (e.g. Kinnekulle Energi‘s Götene Flex) than to GOPACS’s mature common order book. (Source - Cerius Radius Trefor Markedsbaseret Fleksibilitet Pilot (2025-2026))

Maturity trajectory: the EC study characterizes the natural evolution as Separate → Sequential → Common. No EU market has yet demonstrated full P2P at scale outside demonstration conditions. Sweden’s 168 DSOs make the transition to a common market structurally harder than the Netherlands (3 large DSOs + 1 TSO), but the sequential model is a viable intermediate stage that allows progressive integration as national FIS infrastructure matures.

A finer-grained taxonomy — where Sweden’s own model actually sits: the CoordiNet project’s own coordination-scheme classification (Source - Ruwaida et al TSO-DSO-Customer Coordination Swedish Demonstration (2022)) predates and refines the four-model EU taxonomy above into seven schemes along four dimensions (flexibility need — local/central/combined; who buys; number of markets; access level): Local Market (DSO-only buyer), Central Market (TSO-only buyer), Multi-level Market (separate central + local markets, TSO retains access to distribution-connected resources), Fragmented Market (same but the TSO lacks that access, requiring bilateral exchange agreements instead), Common Market (single shared market), Integrated Market (Common Market plus non-SO buyers, e.g. BRPs), and Distributed Market (P2P). Sweden’s own CoordiNet demonstrator explicitly ran the Multi-level Market Model — which maps onto the EU taxonomy’s Sequential tier, but the extra dimension (TSO access to distribution resources) makes explicit something the coarser four-model taxonomy leaves implicit: Sweden’s design was always built so the TSO could reach distribution-connected FSPs via forwarded bids, not just observe them.

(Source - EC LFM Specification and Design Criteria (VITO, 2025), Source - CoordiNet D4.7.2 Swedish Demonstration (2022))

What the current legal framework provides

SO GL Art. 182 establishes the existing TSO-DSO cooperation obligation for reserve delivery from distribution systems. It requires TSOs and DSOs to:

• Cooperate to enable reserve provision from distribution-connected resources
• Allow DSOs the right to set limits or exclude delivery based on technical grid safety reasons
• Process prequalification for distribution-connected reserve providers within 3 months

This is a framework but not a detailed operational protocol. In Sweden, the Art. 182 TSO-DSO cooperation agreement was not formally concluded as of early 2026 and was explicitly deferred from the FNA 2026 data requirements. (Source - FNA Överenskommelse Svenskt genomförande 2026 (2025))

Now drafted (June 2026): Svk and Energiföretagen have produced a draft Art. 182 agreement — the concrete Swedish grid-prequalification + temporary-limits mechanism for ancillary services connected to region/lokalnät (Source - Svk Energiföretagen SO GL Art 182 Avtal (2026)). Design points: on a prequalification application Svk contacts the affected grid company if it has signed; the DSO runs a nätanalys and may limit or wholly deny market participation, and may set temporary limits or deny delivery in operation with a stated forecast/operations reason. Svk’s BSP agreement is updated accordingly. The agreement is voluntary for grid companies, non-retroactive (applies at new/renewed prequalification; all facilities covered within the 5-year reassessment cycle), and a BSP is not compensated when limited. Timeline: remiss 30 September to 26 October 2026 → in force spring 2027, phased. The remiss documents (draft main agreement, Annex 1 and Svk’s guidance) add the operating rules: the agreement covers FCR, aFRR and mFRR units or groups with at least 0.5 MW prequalified per grid area; Svk sends a fixed data set within 5 working days; the grid company decides within 4 working weeks (extendable by 4 with Svk’s approval) as approved, approved with conditions, or not approved, and a missed deadline counts as approval; a full approval lasts as long as Svk’s prequalification (5 years) while conditional and negative decisions last 2 years and are re-reviewed (also deemed approved if not re-reviewed); temporary limits last at most two days, must be non-discriminatory and reach the BSP and Svk by 14:00 on D-2; each party bears its own costs, there is no right of withdrawal once signed, and a revision comes at the earliest with the autumn 2027 BSP-agreement update. (Source - Svk Energiföretagen SO GL Art 182 Avtal (2026)) This is also the mechanism that closes the November-2025 DSO-visibility gap noted next.

A concrete gap raised in November 2025: Per Wikström (Ellevio) raised at the Elmarknadsrådet that DSOs currently have no visibility into which actors are delivering ancillary services in their network — they can see the resource is connected but not whether it is actively bidding into balancing markets or what the TSO’s activation plans are. Svk confirmed this gap exists and that the forthcoming Art. 182 agreements will specifically address it: one function of those agreements is to ensure DSOs receive this information so they can assess grid safety implications and — where necessary — stop deliveries that would cause network problems. (Source - Elmarknadsrådet Meetings 3 and 4 2025 (Sep-Nov))

Ei2025:01 addresses multi-level congestion responsibility within the DSO tier: the DSO in whose own network the physical constraint exists bears the obligation. When a local DSO’s customers are constrained by an overlying regional grid bottleneck, the regional DSO (or TSO) must resolve it. This clarifies intra-DSO coordination but does not address TSO-DSO coordination for shared resources across balancing and local markets. (Source - Ei Ställningstagande Ei2025-01 Villkorade avtal (2025))

The NC DR’s coordination architecture

The Network Code on Demand Response replaces the informal subscription-based coordination with a structured framework (Title VII, Arts. 45–52). It establishes five interlocking instruments:

1. Observability areas

DSOs define their observability area — their own grid plus relevant parts of adjacent systems needed for congestion forecasting. Must be established within 6 months of T&C approval and reviewed with each DNDP update. This formalizes who is responsible for monitoring which parts of the network and what data they need from neighboring operators.

The concept predates the NC DR as regulation: the EU’s earlier TDX-ASSIST project (2017–2020) built its Portuguese TSO-DSO demonstrator — REN (TSO) and EDPD (DSO) — explicitly around each operator defining an observability area for the other, years before NC DR made it a mandatory obligation. That demonstrator’s architecture is a concrete illustration of what an observability area looks like operationally: a real-time SCADA-to-SCADA link (ICCP) for system-operation data, plus a separate non-real-time channel (SFTP/web-services over a segregated VLAN) for forecasts, maintenance plans, and short-circuit levels — generation forecasts at ~80 interconnection points, 15-minute sampling, 72h horizon, refreshed daily. (Source - TDX-ASSIST Portuguese Demonstrator (CIRED 2019))

2. Congestion forecasting hierarchy

Formal requirement for congestion forecasting across multiple timeframes:

• Day-ahead (before market gate closures)
• Intraday (after day-ahead, before real-time)
• Real-time At ≤1 hour granularity. Both TSOs and DSOs must publish congestion forecasts to enable service providers to position their resources.

3. Grid prequalification — the conflict arbiter

This is the core TSO-DSO coordination mechanism. Before or parallel to product qualification, connecting and impacted system operators assess whether activating an SPU/SPG would compromise grid safety in their systems.

Three outcomes (Art. 49):

• Approved: resource can activate freely
• Conditionally approved: activation only under specified time and/or quantity conditions (a temporal or volume limit)
• Not approved: requires full justification, including explanation of why temporary limits cannot resolve the issue

The auto-approval default (silence = approved) prevents SOs from blocking market access through inaction. Annual NRA reporting on non-approved and conditional cases creates transparency about where operators are constraining markets.

Grid prequalification does not allocate a resource to one operator or the other — it defines under what conditions each market can activate it. A resource can be conditionally approved for both TSO balancing and DSO local services, with temporal conditions specifying when each takes priority. This is the formal successor to the informal subscription-mechanism’s implicit priority signal.

4. Temporary limits

Short-term operational procedure for cases where congestion is forecast too late for grid prequalification to resolve. The SO must communicate temporary limits at least 1 hour before balancing energy gate closure. SOs must minimize market impact — they cannot routinely use temporary limits as a substitute for proper grid prequalification.

5. Trade position consistency

When a local service is activated after day-ahead gate closure, the activating SO must either cancel an opposite-direction position or accept a net energy imbalance — and the TSO must calculate an imbalance adjustment for affected balance responsible parties (Art. 51). This formalizes the financial reconciliation that currently happens informally (if at all) when DSO activations interact with balance responsible party positions.

The multi-level DSO complication

Sweden’s distribution grid has multiple ownership levels: Svenska kraftnät (transmission, stamnätet) → regionnät (regional DSO, 10–130 kV) → lokalnät (local DSO, low/medium voltage). The NC DR’s “procuring system operator / connecting system operator / impacted system operator” roles must be mapped to this Swedish reality.

A resource connected to a lokalnät serving a regionnät serving Svk may be:

• Procuring SO: the lokalnät (if congestion is in the lokalnät grid)
• Impacted SO: the regionnät (if activation affects its connection point capacity with Svk) and Svk (if transmission is affected)
• Connecting SO: the lokalnät (where the CU is connected)

All three must participate in grid prequalification. In practice, this means three separate SOs — potentially with different IT systems, different observability models, and different operational priorities — must agree on conditions for each resource. This is a significant operational overhead that doesn’t exist in countries with fewer grid ownership layers.

Sweco (2025) explicitly calls out this problem: investigating and documenting how the NC DR’s “system operator” concept should be interpreted for Sweden, given the extra grid level and extreme DSO size range, is one of their five Ei-directed recommendations (AP1). (Source - Sweco Kartläggning av lokala flexibilitetsmarknader (Ei, 2025))

Quantifying the problem — the Uppsala 70 kV case study. An academic techno-economic model of the same CoordiNet Uppsala demonstration puts numbers on the multi-level complication: the regionnät DSO there (Vattenfall Eldistribution) sits behind two separate TSO-DSO interfaces into the same subtransmission grid (2×315 MVA / 205 MW subscription; 250 MVA / 87 MW subscription) — a genuinely meshed-to-meshed topology with more than one boundary to coordinate across, not the single clean interface most coordination-scheme literature assumes. Without local flexibility, the DSO’s full modelled cost (1,674 k€/year) is subscription-penalty exposure — direct evidence that Sweden’s virtual-congestion mechanism shifts cost from TSO to DSO rather than resolving the underlying capacity constraint. Procuring local flexibility cut that cost 52% (to 802 k€/year), and scaling available DER flexibility to 1.6× the demonstration’s level eliminated the penalty entirely. The same model found that a 10 GWh/year increase in activated local flexibility let the grid absorb 145 GWh/year of additional demand without congestion — i.e., local flexibility measurably defers subtransmission reinforcement, not just balances it. (Source - Lind et al TSO-DSO Coordination Meshed-to-Meshed (2023))

Value stacking: the opportunity that requires coordination

The most concrete demonstration of what good TSO-DSO coordination enables is value stacking: a single resource participating in multiple markets simultaneously, earning revenue from each.

CoordiNet demonstrated this with a 0.48 MW/1 MWh battery in Skåne. The battery participated simultaneously in:

• CoordiNet local congestion market (DSO flexibility procurement, day-ahead and intraday)
• FCR-D (TSO frequency containment reserve, Nordic system-wide)

How the conflict was managed: FCR-D has two auction windows (D-2 and D-1). If the battery was cleared day-ahead on CoordiNet, the operator sent a rebuy request in FCR-D’s D-1 auction at 18:00, freeing the resource for local activation. This required:

1. The battery operator knowing its clearing status on CoordiNet before the FCR-D D-1 deadline
2. Svk’s balancing market accepting the rebuy request on short notice
3. A time sequence that worked: CoordiNet day-ahead clears 10:00, FCR-D D-1 auction at 18:00 → 8 hours of buffer

This is a manually managed workaround, not a designed system. At scale, with thousands of resources and multiple markets, the NC DR’s formal TSO-DSO coordination framework replaces this improvised sequencing with systematic grid prequalification, conditional approvals, and temporary limits. (Source - CoordiNet D4.7.2 Swedish Demonstration (2022))

The Norwegian precedent — NorFlex → Armering

Norway has gone furthest in turning this manual workaround into a designed product — the most concrete operational example of DSO+TSO co-participation in the Nordics. The NorFlex demonstration project (~2019/20–2023; Agder Energi, Glitre Energi, NODES, Statnett — Enova-funded) ran the first DSO + TSO co-use of local flexibility: on 6 January 2022, 3 MW of flexibility from the NODES marketplace was activated in Statnett’s mFRR market — the same resource offered simultaneously to local grid companies and the TSO, with unbought local volume aggregated into ≥1 MW mFRR blocks under a regulatory exception (NODES relaying Statnett’s e-bestill activation signals, since the TSO cannot handle sub-MW bids directly). Over its life NorFlex traded ~1,394 MWh across 4,000+ assets and 30,000+ activations. (Source - NorFlex Project (NODES, 2019-2023))

Its commercial successor, Norway’s Euroflex market (from 2024/25), productizes this as Armering (“arming”): an overlay on a local-flex contract that lets a resource hold availability in both the local market and Statnett’s markets without double-activation risk. When the DSO foresees a need it notifies ~2 days ahead and “arms” the contract, so capacity is withheld from Statnett only on the days the DSO expects to use it — letting the provider earn availability in both markets roughly nine days in ten. This is the first Nordic designed mechanism for co-participation-without-double-activation, and a working reference for exactly what the NC DR’s grid-prequalification / conditional-approval architecture (above) aims to standardize across the EU. In EU-taxonomy terms it pushes NODES markets from the Sequential tier toward Common-market coordination. (Source - Euroflex Norwegian LFM (NODES, 2026))

The NODES MCP — sequential coordination in practice

NODES’s Market Communication Platform (MCP) is the operational implementation of sequential coordination that the NC DR architecture formalizes in regulation. When any SO submits a buy order, the MCP routes a trade approval request to affected counterpart SOs before activation. A traffic light signal is returned:

• Green: no grid impact → trade proceeds
• Yellow: impact possible → accepted but conversation initiated; post-acceptance withdrawal is possible under strict rules
• Red: critical negative impact → trade rejected; negotiation phase opens

Pre-configured node limits can be set per grid node — DSOs and TSOs buy freely up to the limit, triggering the traffic light flow automatically above it. This is the operational analogue of the NC DR’s temporary limits mechanism.

The Yellow state is a notable design choice: it avoids hard refusal while retaining the ability to abort after initial approval, reflecting the reality that grid impact assessment under time pressure is imperfect. (Source - NODES FPM Presentation (2026))

Norway’s Glitre Nett uses the traffic light model (trafikklysmodellen) operationally in Euroflex to colour-code congestion areas and coordinate with Statnett — reporting activation volumes in the hundreds of transactions during cold spells. (Source - Euroflex Norwegian LFM (NODES, 2026))

The “chicken race” between grid levels

A structural barrier Sweco identifies: when a capacity constraint spans multiple DSO levels (e.g., a regional DSO constraint blocking local DSO connection requests), neither operator voluntarily bears the cost of resolving it. Each waits for the other. The local DSO cannot solve an overlying constraint; the regional DSO may lack visibility into the downstream problem; the TSO doesn’t see distribution-level congestion.

Ei2025:01 provides the legal obligation rule (the operator with the physical constraint must act) but not the economic incentive (who pays?). Ei’s methodology for cross-level cost attribution is still pending — it was on Sweco’s recommendation list for immediate Ei attention. (Source - Sweco Kartläggning av lokala flexibilitetsmarknader (Ei, 2025))

The NC DR helps by formalizing the impacted SO role — a regional DSO with an adjacent lokalnät running a flexibility activation must participate in grid prequalification. But it does not resolve the cost attribution question. That is national regulation territory.

What still needs to happen in Sweden

Related pages

• Network Code on Demand Response — NC DR coordination architecture (Arts. 45–52)
• Congestion Management — subscription mechanism mechanics; multi-level DSO responsibility
• Balancing Markets — FCR/aFRR/mFRR; value stacking demonstration
• Flexibility Market — local services markets and how they interact with balancing
• Villkorade Avtal — the current backstop mechanism and how NC DR constrains it
• NC DR Implementation in Sweden — What Changes and for Whom — actor-level implementation map
• Distribution Network Development Plan — observability area and DNDP update linkage
• The Signal Stack — Price Signal Collisions at the Customer DER — the asset-level mirror of this page: the same coordination problem, viewed from the DER where all the institutions’ price signals recombine