Working note, 30 October 2026. The numerical example uses illustrative inputs, not the official input values behind the announced cap.
1. Roadmap
This note explains, from the bottom up, how the Danish consumer price cap on district heating (forbrugerprisloftet) is computed and how it is used. The cap is not a statistic about district heating at all. It is the cost per unit of heat that a household would face if it left the district heating grid and heated its house with its own air-to-water heat pump. The regulator then compares each district heating company’s average price with this alternative cost. Section 2 sets out the legal frame and who computes what. Section 3 builds the benchmark cost per GJ term by term and ends with a single formula. Section 4 works through an illustrative numerical example and shows which inputs drive the cap. Section 5 describes how the benchmark is compared with the companies’ prices and closes with the parts of the construction that are less clean than the formula suggests.
2. Legal frame and division of labour
The cap rests on the Heat Supply Act as amended by L 78 (2024/25) and on Executive Order no. 1740 of 30 December 2024, whose Annex 1 lists the cost components [1, 3]. The executive order calls the benchmark “costs based on individual heat supply” rather than a price cap, and that wording is more accurate: nothing in the order prohibits a company from charging more than the benchmark. The order instead attaches publicity and planning obligations to exceeding it, as described in Section 5.
The computation has been carried out by two different authorities. For 2023–2025 the Danish Energy Agency (Energistyrelsen, ENS) computed the benchmark and added a precautionary margin of 15 per cent; the resulting caps were 486, 410 and 412 kr./GJ including VAT, each in its own year’s price level [5, 7]. From 2026 onwards Forsyningstilsynet computes the benchmark each year and announces it by 1 November for the following calendar year, without the 15 per cent margin. ENS delivers the technology data by 1 October. The cap for 2026 has been announced at 395 kr./GJ including VAT [8].
The definition of the reference house changed on 1 July 2025. Until then the standard house was a detached single-family house of 130 m² with an annual heat demand of 18.1 MWh. From the 2026 cap the house is 125 m² with a demand of 13.4 MWh, the average of the median consumption of district-heated detached houses in 2017–2023 according to BBR data [6, 7]. One secondary source reports the new demand as 14 MWh [2]; the note uses the ENS figure. As Section 4 shows, this seemingly technical change matters a great deal for the level of the cap.
3. Building the benchmark cost
The benchmark is built in three layers: the annual cost of capital, the annual operating cost, and the conversion of the annual total into a price per GJ. Each layer is introduced with its own notation before it is used.
3.1 The capital cost as an annuity
The household must buy and install the heat pump before it produces any heat, so the first task is to turn a one-off outlay into an annual cost. Let \(I_t\) denote the purchase-and-installation price of an air-to-water heat pump for the standard house in year \(t\), in kroner excluding VAT. From 2026 this is a simple average from the ENS technology catalogue for individual heating; for 2023–2025 ENS used realised installation prices from the heat pump subsidy scheme, trimmed to 50,000–250,000 kr. and with the cheapest and dearest 2.5 per cent removed [5, 8].
The outlay is spread over the lifetime of the unit with a standard annuity. Let \(T\) denote the technical lifetime in years, taken from the technology catalogue, and \(r_t\) the interest rate, defined as the long construction or bond rate plus one percentage point. The annuity factor is
$$a(r_t,T)=\frac{r_t}{1-(1+r_t)^{-T}}, \tag{1}$$
the constant annual payment that repays a loan of one krone over \(T\) years at rate \(r_t\). The annual capital cost is therefore \(I_t\,a(r_t,T)\). Because \(a\) is increasing in \(r_t\) and decreasing in \(T\), a higher interest rate or a shorter lifetime raises the cap. The factor is evaluated in closed form.
3.2 The operating cost and the role of SCOP
The operating cost has a fixed part and an electricity part. Let \(F_t\) denote the fixed annual operation and maintenance cost from the technology catalogue, in kroner per year. The electricity part requires knowing how much electricity the heat pump draws to deliver the house’s heat.
The seasonal coefficient of performance connects heat delivered to electricity consumed. Let \(Q\) denote the annual heat demand of the standard house in MWh of heat, and \(\sigma\) the seasonal coefficient of performance (SCOP), the ratio of heat delivered to electricity consumed over a heating season. Annual electricity consumption is then \(Q/\sigma\) MWh. ENS uses the catalogue SCOP for a house with a radiator system, which is lower than for floor heating because radiators require a higher supply temperature [5]. The physical reason is that a heat pump’s efficiency falls with the temperature gap it must lift heat across; the ideal (Carnot) bound \(T_h/(T_h-T_c)\), with temperatures in kelvin, shows that a higher supply temperature \(T_h\) or a lower outdoor temperature \(T_c\) reduces the attainable efficiency.
The price of each kWh drawn has three components. Let \(p_t\) denote the electricity price from Forsyningstilsynet’s electricity price statistics, \(\tau^{n}_t\) the network and system tariffs of the DSOs and Energinet, and \(\tau^{a}_t\) the electricity tax on electricity for heat pumps. For 2023–2025 ENS weighted quarterly prices and tariffs towards winter, when a heat pump actually uses its electricity, for example with weights 0.8 and 0.2 on the first and second quarter [4]. Writing \(e_t=p_t+\tau^{n}_t+\tau^{a}_t\) for the all-in electricity price per MWh, the annual electricity cost is \(e_t\,Q/\sigma\).
3.3 From annual cost to a price per GJ
Adding the three annual cost terms gives the annual cost excluding VAT,
$$C_t = I_t\,a(r_t,T) + F_t + \frac{e_t\,Q}{\sigma}. \tag{2}$$
Dividing by the heat delivered converts this into a cost per unit of heat. Since 1 MWh equals 3.6 GJ, the heat delivered is \(3.6\,Q\) GJ. Let \(m\) denote the VAT rate, 25 per cent. Multiplying by \((1+m)\) and dividing by \(3.6\,Q\) gives the benchmark in kroner per GJ. Writing the result so that the two economically different parts are visible shows one part divided by \(Q\) and one part that does not depend on \(Q\) at all, because the \(Q\) in the electricity term cancels against the \(Q\) in the denominator.
Result. The benchmark cost per GJ of heat in year \(t\), including VAT, is
$$c_t=\frac{1+m}{3.6}\left[\underbrace{\frac{I_t\,a(r_t,T)+F_t}{Q}}_{\text{fixed cost per MWh}}+\underbrace{\frac{p_t+\tau^{n}_t+\tau^{a}_t}{\sigma}}_{\text{electricity cost per MWh}}\right], \tag{3}$$
expressed in the price level of year \(t\) via the net price index and rounded to whole kroner. For 2023–2025, ENS multiplied the result by 1.15. The fixed part falls in \(Q\); the electricity part is independent of \(Q\) and falls in \(\sigma\).
Annex 1 of the executive order lists these components in prose and does not write the formula out [1]. Equation (3) is therefore a reconstruction from the order, the ENS method notes and the Forsyningstilsynet announcement memo. It is consistent with all three, but details such as whether VAT is applied to an electricity price that already includes VAT must be checked against the regulator’s own calculation sheet.
4. An illustrative example
The example uses round, plausible inputs to show the size of each term. The input values behind the 2026 cap are not given in the published memo, so these numbers are not the official ones; they are chosen only to land in the same range as the announced 395 kr./GJ.
| Input | Symbol | Illustrative value |
|---|---|---|
| Purchase and installation, excl. VAT | \(I\) | 88,000 kr. |
| Lifetime | \(T\) | 18 years |
| Interest rate | \(r\) | 5 % |
| Fixed O&M | \(F\) | 1,800 kr./year |
| SCOP (radiator system) | \(\sigma\) | 3.4 |
| Electricity price | \(p\) | 0.85 kr./kWh |
| Network and system tariffs | \(\tau^{n}\) | 0.40 kr./kWh |
| Electricity tax | \(\tau^{a}\) | 0.01 kr./kWh |
| Heat demand of standard house | \(Q\) | 13.4 MWh |
Table 1. Illustrative inputs (not the official 2026 values).
The capital cost comes first. With \(r=0.05\) and \(T=18\), equation (1) gives \(a=0.05/(1-1.05^{-18})=0.0855\), so the annual capital cost is \(88{,}000\times0.0855\approx7{,}530\) kr. Adding \(F\) gives a fixed cost of about 9,330 kr. per year.
The electricity cost comes next. The heat pump draws \(13.4/3.4=3.94\) MWh of electricity per year. At an all-in price of 1.26 kr./kWh, or 1,260 kr./MWh, this costs about 4,970 kr. per year. The annual total from equation (2) is therefore about 14,290 kr. excluding VAT.
Converting to GJ gives the cap. The house uses \(3.6\times13.4=48.2\) GJ, so the cost is \(14{,}290/48.2\approx296\) kr./GJ excluding VAT and \(296\times1.25\approx370\) kr./GJ including VAT. Of the 296 kr., 193 are fixed cost spread over the house’s demand and 103 are electricity. In this example two thirds of the cap are capital and maintenance, not energy.
4.1 Why the size of the standard house matters
The fixed part of equation (3) is a hyperbola in \(Q\), so a smaller standard house raises the cap. Holding the other inputs at their Table 1 values, the cap including VAT is about 308 kr./GJ with the old \(Q=18.1\) MWh and about 370 kr./GJ with the new \(Q=13.4\) MWh. The change of reference house alone raises the cap by roughly 20 per cent in this example, since the fixed cost term is multiplied by \(18.1/13.4=1.35\). The electricity part stays at about 129 kr./GJ whatever \(Q\) is.
| \(Q\) (MWh/year) | 10 | 13.4 | 18.1 | 22 |
|---|---|---|---|---|
| Cap, kr./GJ incl. VAT | 453 | 370 | 308 | 276 |
Table 2. The cap as a function of the standard house’s heat demand, other inputs as in Table 1.
The same mechanism helps make sense of the step from 2025 to 2026. The 2025 cap of 412 kr./GJ contains the 15 per cent margin; without it, the cap would have been \(412/1.15\approx358\) kr./GJ. The 2026 cap of 395 kr./GJ is about 10 per cent higher than that, even though the margin was dropped. The smaller reference house pushes in exactly this direction. Inflation and new input data also changed between the two years, so this comparison does not isolate the effect of \(Q\).
4.2 Sensitivity to the electricity price
The electricity price enters linearly with slope \((1+m)/(3.6\,\sigma)\). With \(\sigma=3.4\), each 0.10 kr./kWh on the all-in electricity price moves the cap by \(1.25\times100/(3.6\times3.4)\approx10\) kr./GJ. The cap is therefore far less sensitive to electricity prices than to the level of fixed costs or the choice of \(Q\). Still, an error of 0.30–0.40 kr./kWh in a forecast moves the cap by 30–40 kr./GJ, which is not negligible for a company close to the line.
Result. Two derivatives summarise what drives the cap. The effect of the standard house’s demand is
$$\frac{\partial c_t}{\partial Q}=-\frac{1+m}{3.6}\,\frac{I_t\,a(r_t,T)+F_t}{Q^{2}} \lt 0,$$
so a smaller reference house raises the cap through the fixed-cost term only. The effect of the electricity price is
$$\frac{\partial c_t}{\partial e_t}=\frac{1+m}{3.6\,\sigma},$$
about 10 kr./GJ per 0.10 kr./kWh at \(\sigma\approx3.4\).
5. How the benchmark is used
The benchmark is compared with each company’s average realised price, not with its tariff. Let \(j\) index district heating companies at the CVR level, \(R_{jt}\) company \(j\)’s realised revenue from consumption billing in year \(t\), and \(X_{jt}\) its realised heat sales in GJ. The company’s computed consumer price is
$$\pi_{jt}=\frac{R_{jt}}{X_{jt}}, \tag{4}$$
which includes fixed charges spread over the heat actually sold [1, 8].
Both sides of the comparison are averaged over three years. Let \(\bar\pi_{jt}\) and \(\bar c_t\) denote the three-year averages of the company price and of the benchmark, each adjusted to a common price level with the net price index. The executive order specifies three-year averages but does not write out the indexation formula, so the notation describes the principle rather than a verified algorithm. Company \(j\) is placed on the published list when
$$L_{jt}=\mathbf{1}\{\bar\pi_{jt} \gt \bar c_t\}=1.$$
A listed company must inform its customers, must publish within three months a plan for reducing its heat price over three years, and may be subject to a decision by Forsyningstilsynet, which also informs the municipality. Such decisions cannot be appealed to the Energy Board of Appeal [1].
5.1 Where the construction is less clean than the formula
The comparison mixes a reference house with an average customer. The benchmark is the cost per GJ for one specific house, whereas \(\pi_{jt}\) is an average over all of a company’s customers, including apartments and commercial customers. A company whose fixed charges are spread over customers with low consumption per installation will show a high \(\pi_{jt}\) for reasons unrelated to efficiency, just as the benchmark rises when \(Q\) falls. The two objects are not computed on the same basis.
The benchmark is announced before the year it covers. The cap for year \(t\) is published by 1 November of year \(t-1\), so the electricity price and tariffs are necessarily either a forecast or the latest available statistics. For 2025, ENS used its own electricity price projection; Forsyningstilsynet’s 2026 memo refers to its electricity price statistics [5, 8]. With ENS no longer producing the projection, whoever computes the cap must choose an electricity price method, and Section 4.2 gives the stakes of that choice.
The benchmark is an average cost, not a marginal one. The household in the benchmark pays the full annuity on a new heat pump. A household that already owns a working gas or oil boiler would compare district heating with the cost of replacing that boiler at the end of its life, which is a different calculation. The benchmark represents the long-run alternative, not the short-run outside option.
Several inputs are averages over a fixed collection window. The heat pump price, the interest rate and the electricity price are each averaged over a specified period before the deadline, and the SCOP and lifetime come from a catalogue that is updated irregularly. A new catalogue version can therefore move the cap from one year to the next without any change in the market.
References
- Klima-, Energi- og Forsyningsministeriet (2024). Bekendtgørelse om omkostninger baseret på individuel varmeforsyning og beregnede forbrugerpriser, BEK nr. 1740 af 30/12/2024, with Annexes 1–2. retsinformation.dk
- BDO (2025). Ændringsbekendtgørelse fastlægger nyt udgangspunkt i beregningsgrundlaget af forbrugerprisloftet for fjernvarme. 3 June 2025. bdo.dk
- Dansk Fjernvarme (2024). Gennemgang af lovforslag om ændring af den økonomiske regulering af fjernvarmesektoren (L 78). Notat, 12 November 2024. danskfjernvarme.dk
- Energistyrelsen (2024a). Metodenotat for beregning af prisloft for individuel varmeforsyning. 14 October 2024. Folketinget, KEF alm. del, bilag 33. ft.dk
- Energistyrelsen (2024b). Metode og forudsætninger for Energistyrelsens beregning af prisloftet for 2023–2025. J.nr. 2024-10297, 27 November 2024. ens.dk
- Energistyrelsen (2025a). Opdatering af repræsentativt varmeforbrug. 3 June 2025. ens.dk
- Energistyrelsen (2025b). Forbrugerprisloft på fjernvarme (web page and press release on the caps for 2023–2025). ens.dk
- Forsyningstilsynet (2025). Udmeldelsesnotat: Prisloft for fjernvarme 2026. 31 October 2025. forsyningstilsynet.dk