The Signal Stack — Price Signal Collisions at the Customer DER

A single Swedish household battery in 2028 will receive at least six legally valid, independently issued price or control signals — and no part of the regulatory architecture says which one wins when they disagree. TSO-DSO Coordination — The Central Design Problem analyses the coordination problem between institutions sharing a flexibility resource pool. This page analyses its mirror image: the coordination problem between signals at the one place where they all physically recombine — the customer’s DER. The thesis, argued from first principles: signal collisions are not design accidents to be patched away; they are the direct consequence of Europe’s deliberate decomposition of electricity pricing, surfacing now because demand stopped being passive. Today the collisions are arbitrated by nobody — or more precisely, by private optimization code that no regulator reviews.

The stack — what one asset faces

Signals converging on a single connection point (battery, EV charger, heat pump) in southern Sweden, status mid-2026:

Plus the one non-price master signal: the customer’s own comfort and usage needs (the ±1 °C indoor boundary from Simris is the canonical example of comfort overriding dispatch).

Signals 1–6 can all individually move the same kilowatt in the same quarter-hour. Nothing above the asset coordinates them.

First principles — why collisions are structural

At any connection point, three distinct scarce goods exist simultaneously: energy (system-wide, time-varying), network capacity (local, peak-driven), and balance (system-wide, instantaneous). A theoretically complete market prices all three in one signal at one node — real-time locational marginal pricing with reserve co-optimization, the US RTO model, internalizes energy and congestion (and balance, via co-optimized reserves) into a single number the asset can follow.

Europe deliberately chose decomposition instead: a zonal energy price (no locational granularity below the bidding zone), regulated network tariffs as the only locational instrument below the zone, and separate TSO balancing markets. This decomposition was safe for fifty years because demand was passive — the decomposed signals never met a responder capable of seeing more than one of them. A household paid the tariff, consumed at will, and the signals coexisted without interacting.

The battery with a HEMS is the first asset that can see and respond to all of them. It is the physical point where the decomposed prices recombine — and where their mutual incoherence becomes behaviour. Each signal is locally rational: each issuer correctly prices the scarcity it is responsible for. The incoherence is emergent, and it emerges precisely at the DER. Signal collision is therefore not a bug introduced by bad tariff design; it is the original market design meeting a new kind of responder. Patching individual collisions (a dual tariff here, a carve-out there) treats symptoms of a structural property.

A second first-principles observation sharpens the worst case. The standard Swedish effektavgift prices an individual, non-coincident proxy (the customer’s own monthly peak) for a collective, coincident scarcity (the feeder peak). A customer’s private peak at 03:00 costs the grid nothing; the same kW at 17:30 on a cold January weekday is exactly what drives reinforcement. Ei‘s aggregate-load principle in Ei2025:06 — time-differentiation must reflect the sammanlagda belastningen on the DSO’s grid, not individual customer peaks — is the regulator recognizing this proxy error. (Source - Ei Ställningstagande Tariffer Ei2025-06) An automation layer optimizing against the proxy rather than the scarcity will dutifully flatten individual peaks while doing nothing for (or actively worsening) the coincident peak.

A taxonomy of collisions — documented Swedish cases

1. Energy vs capacity — the double edge. Low midday spot (solar) says charge now; the effektavgift penalizes the resulting peak regardless of what spot is doing. The signals point in opposite directions, documented by FlexAbility and acknowledged by Ei: the EIFS 2026:8 konsekvensutredning states plainly that users cannot reconcile the effektavgift with the spot price because the two signals are fristående och inte nödvändigtvis samstämmiga. Göteborg Energi Nät‘s dual-tariff test is the first design response. (Source - FlexAbility Delrapport 5 (2025), Source - EIFS 2026-8 Nätföretags Information till Elanvändare (2026))

2. Balance vs capacity — the recharge spike. A battery delivering FCR-D upward must recharge afterwards; if the recharge lands in the effektavgift measurement window, the balancing commitment directly creates a capacity charge. The same physics underlies the random-startup-delay carve-out: an anti-synchronization delay mandated for EV chargers is incompatible with Svk’s minutes-scale FCR/FFR activation requirement — one mitigation for one signal breaks the response to another. (Demand Response › Random startup delay)

3. Market activation vs balance responsibility. When an independent aggregator activates a resource on an LFM, the deviation lands in another party’s BRP portfolio. The entire NordREG Model 3/4 compensation architecture (Independent Aggregation in Sweden — The Implementation Gap) exists to settle this single pairwise collision — and it required a government assignment, two Svk reports, and a central data infrastructure (DHV) still years away. (Source - Svk Kompensationsmodell Delrapport 1 (2024))

4. Contract vs market. A villkorat avtal or FCA curtailment can pre-empt the same capacity the customer has bid into an LFM. The Comillas co-design analysis enumerates the “red conditions”: ex-post curtailment notification blocks day-ahead market participation; LIFO curtailment priority undermines bid reliability. The recommended sequencing — market first, FCA as backstop — is a pairwise de-confliction rule. (Source - DSO Service Acquisition Interaction Comillas (2024))

5. Tariff vs tariff — the vertical cascade. Svk’s four-component transmission tariff (2027, effektavgift component +20%) shapes regionnät subscription costs, which shape lokalnät subscription optimization, which shapes the retail effektavgift design — three tariff layers, designed by three actor classes on three timetables, with no requirement of mutual temporal consistency.

6. The inversion — consensus is worse than conflict. From first principles, when signals disagree, the optimizer arbitrates and portfolio behaviour partially cancels — collisions are individually costly but systemically dampening. When all signals agree, every optimizer computes the same answer at the same time: the −13% drop at 07:00 and +16% rebound at 20:00 at Göteborg Energi’s tariff boundaries; the projected simultaneous mode-switch of solar+battery systems at the ~60 öre/kWh threshold; the Elforsk estimate that 100,000–700,000 simultaneously responding households break BRP forecastability. Collision dampens; consensus synchronizes. This inverts the intuitive design goal: perfectly harmonized signals maximize the synchronization externality that Security and Resilience of the Digitalized Flexible Grid identifies as the non-malicious twin of a coordinated DER cyberattack. (Source - Göteborg Energi Elektrifieringsrapporten nr 1 (2025), Source - FlexAbility Delrapport 5 (2025))

Who arbitrates? Four resolution architectures

The collisions can be resolved at four different layers. Sweden currently uses all four, unevenly and without naming the choice.

A. Merge the prices (re-integration). Make the grid signal co-temporal with the energy signal: coincidence-based or fully dynamic network pricing. The EIFS 2026:8 konsekvensutredning records DSOs already exploring per-kvart dynamic network prices built from a day-ahead grid-load forecast — economically, this is quasi-nodal pricing introduced through the tariff back door, reversing the original decomposition one DSO at a time. Strengths: one merged signal is exactly what automation needs; the effekttariff integration gap (AFRY) dissolves. Weaknesses: a regulated monopolist starts setting a quasi-market price from a private forecast, and household tolerance for price complexity is demonstrably low — the effektavgift complaint surge (837 tariff-design complaints in 2025) happened at far lower complexity than dynamic kvart pricing. (Source - EIFS 2026-8 Nätföretags Information till Elanvändare (2026), Source - AFRY Styr och Informationstjänster Konsumenter (2023))

B. Impose a hierarchy (the control stack). Rank the signals: grid security > contracted curtailment > market commitment > price response > comfort. The Network Code on Demand Response quietly builds the top of this stack — grid prequalification (approve / conditionally approve / refuse market participation per resource) and temporary limits encode grid-beats-market, decided ex ante rather than overridden in real time; mandatory staggered activation addresses the consensus problem at the dispatch layer. Villkorade avtal are the contracted-priority tier. What the hierarchy deliberately does not rank: tariff response vs market participation — both are voluntary economic choices, so the bottom of the stack is unranked by design. (Source - NC DR Amended Text (ACER Recommendation 01-2025 Annex 1))

C. Separate by design (de-confliction). Keep signals apart in time or product space: LFM-e seasonal availability windows, the Comillas market-first/FCA-backstop sequencing, Göteborg’s dual tariff splitting the capacity signal from the energy signal. Effective pairwise — but pairwise de-confliction of N signals needs N(N−1)/2 rules, and each new instrument (Art. 7a, dynamic tariffs) multiplies the rule count. It does not scale.

D. Delegate to the optimizer (the status quo). Let the HEMS/aggregator algorithm see all signals, weigh them in SEK, and decide. This is Sweden’s actual current answer for the unranked bottom of the stack. CheckWatt‘s value proposition is precisely multi-market weighing (FCR-D/N, mFRR, FFR, LFMs, spot); AFRY counted 44 consumer steering/information services from 172 actors — these are the de facto arbiters. The consequence deserves stating plainly: the effective merit order across Sweden’s public price signals is set by private, commercially confidential optimization code that no regulator reviews. An “algorithmic merit order” has replaced the missing institutional one. EIFS 2026:8 actively reinforces architecture D — from 2027 every DSO must tell customers that automated steering services exist. (Source - CheckWatt Website (2025-2026), Source - AFRY Styr och Informationstjänster Konsumenter (2023))

The conflict rent. Architecture D has a political economy: part of the aggregator/HEMS business case is the arbitrage across uncoordinated signals — value that would compress if architecture A merged them. The actors best equipped to navigate the stack profit from its complexity; the actors who suffer it (non-automated households, who face the raw collisions manually) are the ones the complaint statistics measure. This asymmetry is itself a distributional finding: signal complexity acts as a regressive filter, transferring the stack’s value to whoever can afford the optimizer.

The April 2027 effektavgift model is the arbitration decision

Track 2 requires Ei to propose a new effektavgift model by 12 April 2027 with criteria of transparency, non-discrimination, proportionality, and “correct incentives for consumption adaptation.” Read through this page’s lens, the model choice is really a choice of arbitration architecture:

• An individual-peak model (status quo cleaned up) keeps the grid signal independent of spot timing → institutionalizes architecture D: the optimizer arbitrates, the conflict rent persists, and automation optimizes against the proxy rather than the scarcity.
• A coincidence-based or dynamic model (Ei2025:06’s aggregate-load principle taken to its conclusion; the per-kvart exploration in the EIFS 2026:8 KU) moves toward architecture A: the grid signal merges temporally with the energy signal, becoming machine-readable by design.

Tommy Johansson’s automation-barrier statement — fragmented designs block automatiska styrningstjänster — and the EIFS 2026:8 invoice rule for weighted-average display of dynamic per-kvart charges (a rule that only matters if dynamic designs are coming) both point toward A. But A imports the synchronization inversion: the more coherent the merged signal, the stronger the consensus dynamics — so a move to A makes NC DR staggering and randomization (architecture B’s dispatch layer) more load-bearing, not less. The two architectures are complements, not alternatives: merge the prices, then rank and stagger the responses. (Source - Ei Effektavgifter Uppdrag (2026))

Implications

For DSOs: any effektavgift that ignores spot timing will be optimized against, not responded to — automation turns a behavioral signal into an arbitrage input. Design for coincidence with the system price or expect the peak measure to be gamed while feeder peaks persist.

For aggregators and HEMS providers: the stack is the business model. Track all eight signal classes; expect rent compression if the 2027 model merges grid and energy pricing; and expect the synchronization externality to attract regulatory attention to optimizer behaviour itself.

For Ei: the 2027 model decision is implicitly a decision about who arbitrates Sweden’s price signals. Separately, architecture D at scale raises a question no current instrument addresses: whether optimizer objective functions — once they steer hundreds of MW — need the same supervisory visibility that market bids have. Today they have none.

For Svk: optimizer consensus is a new disturbance class (rebounds, threshold cascades) that originates in price design, not faults. The system operator’s interest in how DSO tariffs are timed is therefore not academic — tariff design has become a frequency-stability input.

Data gaps

• No published Swedish quantification of a multi-signal conflict at a single asset (e.g., FCR revenue forgone to avoid an effektavgift peak) — aggregator settlement data would show this; candidate sources: CheckWatt, Flower public material
• Direction of Ei’s new effektavgift model (individual-peak vs coincidence/dynamic) — open until the 12 April 2027 proposal
• Which DSOs are exploring dynamic per-kvart network pricing — the EIFS 2026:8 KU confirms contacts but names none
• How commercial steering services actually rank conflicting signals — no public documentation of any Swedish HEMS/aggregator objective function

Related pages

• TSO-DSO Coordination — The Central Design Problem — the institutional layer of the same problem; this page is its asset-level mirror
• Demand Response — the double-edged price signal, rebound data, synchronization risks
• Security and Resilience of the Digitalized Flexible Grid — price-signal synchronization as the non-malicious cyberattack twin
• The Swedish BESS Business Case — Revenue Stacking and the FCR Saturation Problem — revenue stacking is the optimizer’s view of the stack (opportunity, not collision)
• Swedish DSO Tariff Reform — Three Parallel Tracks (2025–2027) — Track 2 as the pending arbitration decision
• Independent Aggregation in Sweden — The Implementation Gap — the Model 3/4 settlement of the market-vs-BRP collision
• Why Swedish Local Flex Markets Are Thin — Structural Causes — supply-side context; the conflict rent connects to aggregator economics
• Swedish Household Demand Response — Consumer Adoption and Barriers — the non-automated household as the party bearing raw signal complexity