Comparative modeling of cost-optimal energy system flexibility for Swedish and Austrian regions

Most of the renewable energy sources (RESs) growth is expected to come from volatile sources such as solar photovoltaics and wind energy. High penetration levels of volatile RES require additional and new flexibility measures to balance supply and demand at different timescales [1].

Energy systems have several potential sources of flexibility. The flexibility potential of district heating (DH) and district cooling (DC) systems is often suggested as key for facilitating high levels of RES in energy markets [1]. Combined heat-and-power (CHP) and heat pump (HP) technologies, power-to-heat, heat-to-power solutions, and thermal energy storage (TES) technologies can provide flexibility in DH and DC systems [2–5]. TES technologies can be subdivided into centralized or decentralized heat storage units [6], buildings' TES [7], and the heating/cooling network itself [8].

The DH/DC network is used as TES by circulating water in the network to buffer heating/cooling energy and smooth the supply. Advantages of network TES are minimal infrastructure investments (the network is already present) and that some storage capacity is available in DH/DC systems. Disadvantages of network TES include limited buffering capacity, and that increased return temperatures caused by the storage cycle decrease the efficiency of the generating units. Additionally, there is no standard method to estimate the capacity of the network TES or the size of the threshold needed for storage [9].

There are three types of centralized and decentralized TES: so-called sensible heat storage, in which a liquid or solid storage medium is used to store thermal energy; so-called latent heat storage using phase-change materials (PCMs); and thermochemical storage, which uses chemical reactions to store and release thermal energy [10]. The types of TES most applied in DH/DC systems are water-based sensible TES systems, which are cheaper compared to latent or thermochemical TES systems. TES systems can be centralized (e.g. a single-standing borehole TES) or decentralized (e.g. a hot water tank installed in a house). Decentralized TES units are used for short-term (intraday) storage, whereas centralized TES units can smooth out short-term and long-term demand variations. Centralized TES units have greater storage capacity than network TES, resulting in cost savings. However, the investment costs of centralized TES systems are substantial. In all, the use of centralized TES units can provide both economic and environmental benefits to DH/DC systems [6, 11–13].

Energy at the building level can be stored in individual water tanks, ceramic bricks, PCM heat batteries, and in the building's thermal inertia. Many studies have investigated building TES based on temporal over-or under-heating [7, 14, 15]. In comparison to a centralized TES, controlled by a DH/DC system operator, building TES alters the heat load by smoothing its variations and may involve end-users. The advantages of building TES are availability in DH/DC systems and lower investment than centralized TES units. The disadvantages are limited control and participation of the end-users.

Although the abovementioned studies provide valuable insights into the varying aspects of using different types of TES, there are no studies comparing the effects of these storage types on the operations of DH/DC systems (as in [16]). Moreover, most of the previous studies investigated the effects of different types of TES on the operation of DH/DC systems one at a time, without determining whether the TES types are complementary or competitive. The studies that investigated the effects of centralized and building TES reported that the cost and efficiency of the heat supply in DH systems [17, 18] were positively affected by centralized and building TES in the same system. These are interesting discoveries that should be further investigated and validated under different circumstances.

The aim of this study is to further understand how flexibility in the DH energy system can be optimized. The following questions are addressed:

• What are cost-efficient interactions between DH/electricity networks and the building sector?
• What are optimal investments?
• What are the potentials for expanding the flexibility from demonstration scale to city/region scale?

To that aim, a replicable method is developed for estimating the potential for cost-efficient flexibility in a local or regional energy system and is applied comparatively to two different case studies: Eskilstuna city (Sweden) and the region of Lower Austria (Austria). These demo sites have DH/DC companies that operate in varying heat markets, i.e. more mature in Sweden with essential interconnection with the electricity market, and less mature in Austria. Additionally, the case-studies differ in their size and in both their current technological supply and the flexibility options available.

Eskilstuna had 67 359 residents in the city (2015) and 100 092 in the municipality (2014). The total energy use of the Eskilstuna municipality was around 2322 GWh in 2015 [19]. The energy system (including housing and services, and DH) used around 66% of the total energy supplied. DH is the major supplier of heating, and it provides about 65% of the heating. The annual demand of the city for heating is around 700 GWh, of which more than 50% is used in multifamily buildings [20]. The demo site has a DH grid and a biomass CHP plant. Flexibility measures tested in the demo site include building side flexibility through utilizing building TES and flexible operation of an exhaust air HP, all operated by testing machine learning based demand forecast, operational co-optimization (flexibility and production side), and optimizing with electricity trading (day-ahead and intraday balancing).

Lower Austria has a land area of 19 186 km² and a population of 1.612 million people. The region is supplied by many small plants, of which the most typical has been selected as demo site. The plant, one of more than 700 plants in the region and of more than 2000 plants all over Austria, supplies 30 heat consumers by a 1.5 km DH grid, i.e. restaurants, hotels, schools, public buildings and multifamily buildings with a heat demand of 120–215 MWh yr⁻¹ each. The heat is generated by two biomass boilers on 600 kW each; furthermore, a small biomass CHP is planned to cover the base load and prevent inefficient part load operation of the biomass boilers, possibly acting on the electricity balancing markets. In addition to this there is a current flexibility potential of an 8 m³ storage tank. Flexibility measures tested in the demo site include an optimized operation of the future biomass CHP plant and other renewable heat sources by active and flexible management of grid temperature, buffer storage temperature and TES in buildings.

Our general methodology is illustrated in table 1.

First, we develop a method to define future scenarios that meet climate targets and estimate the cost-efficient flexibility potential in the demonstration sites [21]. We design scenarios that compare future energy systems, energy markets, and climate and energy policy in the demonstrator regions by 2050, following the steps shown in figure 1. The literature review focused on SSPs, for being the only framework that directly links to global warming temperature datasets, which was required in related studies in the same project [22–27]. To our work, the SSP scenarios provide the narrative storyline, with socioeconomic challenges for climate adaptation and mitigation summarize through key elements (GHG emissions, energy demand, energy conversion and supply from fossil fuels). Following existing scenario methodologies at global [28–32], regional [29, 31–33], national [34], and urban levels [35–37], the scenario narratives are refined with local visions, development plans, and input from stakeholders, resulting in two framework scenarios, '100% RES' and 'Maximized flexibility', presented in the appendix. These storylines are refined and matched with the narratives and quantitative in-data assumptions relevant for the study cases, that is IEA World Energy Outlook (IEA WEO) [38–41], JRC-EU-TIMES model [42], and the Nordic Clean Energy Scenarios (NCES) 2020 [43].

Zoom In Zoom Out Reset image size

Second, flexibility potentials are characterized for energy system assessment by upscaling the potentials obtained at the demonstration sites [44, 45], utilizing existing models and data appropriate for the flexibility measures in focus. For instance, in Eskilstuna the focus is on soft-linking models to increase the understanding of TES in buildings (e.g. ECCABS model), whereas in Lower Austria the focus in soft-linking electricity sector models (e.g. MILP and BALMOREL models). The hourly resolution of these models allows for the assessment of flexibility options relevant at the hourly scale, e.g. building TES, and DSM (day-ahead and intraday balancing).

Finally, cost-efficient flexibility options when operating DH systems are identified with a TIMES model generator [46, 47] which minimizes the cost of providing the total heating demand of the city, considering both the energy supply and energy demand of the system, and encompassing the period between the base year and 2050 (with a 5 year step). The model optimizes heat supply in the city taking the perspective of a so-called 'system planner', that is, one single decision maker, which can decide for all the modeled 'customers' on their heat supply option for them, while making sure that the total cost of heat supply over the whole city and time horizon is minimized. The TIMES model requires exogenous heat demand projections by sectors (residential and services), building type (apartment building or single building), and final end-use energy services (space heating and domestic hot water). The simulation takes a reference year as starting point, for which data is best available and is most representative of normal conditions, i.e. 2018 before the COVID pandemic. The details of these stages are presented below for each case study. Schematic and detailed descriptions of the TIMES models used are available [48]; the impacts of their minor structural differences in the results are further discussed in section 5. The resolution of this model (72 or 12 slices per year) allows for the assessment of seasonal or large-scale flexibility options.

3.1. Eskilstuna

The modeling of flexibility options in Eskilstuna combines the use of the building-stock ECCABS model, to provide a physical description of the building stock and an upscaled estimate of the TES potentials in buildings from demo sites, with optimization with TIMES model of such upscaled TES potentials in buildings, together with other existing and potential flexibility options.

First, we use the ECCABS model [49, 50], developed to investigate energy use reductions in Swedish residences [51], and used to map opportunities and costs for the transformation of residential and nonresidential buildings in Sweden [52, 53] and several European countries [54]. Only some basic features of the model are used in this work (figure 2), i.e. physical building data, climate data, energy system data, and other details are the input data to determine the energy performance of the building stock through representative building typologies. To represent the building stock of Eskilstuna in the model, 142 archetype buildings are defined in three steps [55]: segmentation, quantification, and validation. Calculations were required when the data was not available at the city level but was available at the regional and national levels. The building stock is segmented by use (single family or multifamily building), dwelling size (⩽90 m², 90–130 m², >130 m²), construction year (before 2006, 2007–2010, 2011–2019), main source for heating (DH, electrical, other) and type of ventilation system ³ . These categories are given by data availability; for instance, data for the buildings constructed up to 2006 is provided by BETSI [56], as reported and validated in [49, 51, 57–59], whereas for newer buildings energy performance and average heat transfer coefficient per house type/area are given in the building code BBR 29 [60] ⁴ . The combinations of possible parameters in each category results in 142 archetype buildings, all of which are modeled with ECCABS.

Zoom In Zoom Out Reset image size

In the quantification, the frequency of each archetype in the stock is quantified using a weighting factor. To do this, the segmentation categories above presented are calculated as shares and combined with the total number of houses in Eskilstuna from SCB [61], as follows:

$$\begin{equation}{W_{i,j,k,l}} = N{D_{{\text{Tot}}}} \times {P_{{\text{CY}}}}_i \times {P_{{\text{HS}}}}_j \times {P_{{\text{VS}}}}_k \times {P_{{\text{DS}}}}_l \times {P_{{\text{VS}}}} \times {P_{{\text{DS}}}}\end{equation} \tag{ 1 }$$

where ${W_{i,\,j,\,k,\,l}}$ is the weighting coefficient for each archetype, ${\text{N}}{{\text{D}}_{{\text{Tot}}}}$ is the total number of dwellings, ${{\text{P}}_{{\text{CY}}}}_i$ is the percentage of buildings per construction period $i$, ${{\text{P}}_{{\text{HS}}}}_j$ is the percentage of buildings per heating system $j$, ${{\text{P}}_{{\text{VS}}}}_k$ is the percentage of buildings per ventilation system $k$, and ${{\text{P}}_{{\text{DS}}}}_l\,$is the percentage of buildings per dwelling size $l$. The shares for the different segmentation categories have been calculated from statistics from SCB and the Swedish Energy Agency [62–71].

In the validation, the model results for the number of dwellings, the heated floor areas, and the final energy use are compared to statistical data [72]. The number of dwellings in ECCABS are fully aligned with statistics [61], whereas the final energy use in ECCABS is 10% higher than statistics [72] (table 2). As for the heated floor areas, these are not in the statistics for Eskilstuna, so they were estimated by dividing the final energy consumption of Eskilstuna city [73] by the average energy used for heating (in kWh m⁻²) in the same temperature zone [74]. The heated floor areas in ECCABS are 7% lower than this estimate (table 2).

Table 2. Comparison of model results for the heated floor areas and final energy use in Eskilstuna with statistics.

	Heated floor area (1000 m2)			Final energy use (GWh year−1)
	ECCABS	Statistics	Match	ECCABS	Statistics	Match
Eskilstuna all dwellings	5 094	5 473	93%	772	701	110%
Single-family houses	2 626	2 667	98%	402	331	122%
Multifamily buildings	2 469	2 845	87%	370	370	100%

In the second step, the TES potentials in buildings obtained from the demo sites have been upscaled and characterized as required by the TIMES model, i.e. maximum flow $({\text{Flo}}{{\text{w}}_{{\text{Max}}}})$ and maximum capacity (${\text{Capacit}}{{\text{y}}_{{\text{Max}}}}$), as follows:

$$\begin{equation}{\text{Flo}}{{\text{w}}_{{\text{Max}}}} = \sum\limits_1^i {\left( {{C_{\text{F}}} \times {\text{Energy signatur}}{{\text{e}}_i}} \right)} \end{equation} \tag{ 2 }$$

$$\begin{equation}{\text{Capacit}}{{\text{y}}_{{\text{Max}}}} = \sum\limits_1^i {\left( {{C_{\text{c}}} \times {C_{\text{F}}} \times {\text{Energy signatur}}{{\text{e}}_i}} \right)} \end{equation} \tag{ 3 }$$

where ${C_{\text{F}}}$ and ${C_{\text{C}}}$ are constants, and ${\text{Energy signatur}}{{\text{e}}_i}$ is the energy signature of the buildings used (kWh °C⁻¹). The constants (${C_{\text{F}}}$ = 10 °C and ${C_{\text{C}}}$ = 5 h) are based on previous research [49, 75] and have been verified in both the demonstration work and the EU-project TEMPO (www.tempo-dhc.eu/). In equation (3), the energy signature for each building is derived from their energy demand as obtained with the ECCABS model. The energy signature can be evaluated in different ways depending on the available data. As already done in the literature [76], the energy signature for Eskilstuna was calculated using output data from the ECCABS model for each used building separately, as a linear regression where the hourly heat demand is expressed as a function of the outdoor temperature by converting the (°C) to (kW). We would like to stress that in this step lies the novelty of our methodological development, i.e. puts together different models (ECCABS and TIMES), methods [49, 75, 76] and real data from the demo sites to allow for a refined characterization of TES in buildings, which allows for locally tailored results.

In the third step, the cost-efficient flexibility potential in DH systems at city level is investigated using a TIMES model. In the base year, the centralized and decentralized heating technologies reflect the existing generation units and fuel mix of the modeled city. Each year is divided into 72 time slices, representing combinations of 12 months, workday and weekend, day and night, and peak time (12 × 2 × 3 = 72).

In the model, the DH system of Eskilstuna in the base year 2018 consists of a biomass-fired CHP plant, a biomass-fired heat-only boiler (HOB), four bio-oil-fired HOBs, and four oil-fired boilers that are typically used as reserve capacity. The total heat generation capacity of Eskilstuna's DH system is 465 MW. There is a centralized TES connected to the network with a capacity of 900 MWh. The supply and return temperatures of the DH network are 75 °C–100 °C and 40 °C–50 °C. The length of the DH network is 33 km, the annual losses of the DH network are 10%.

The modeled heat generation technologies can be divided into existing investments and potential new investments. The technical parameters of the existing individual heat-generating units were assumed to be identical to those in the TIMESCity model developed and implemented in the ERA-NET SureCity project [77]. The techno-economic parameters for the centralized and individual heat generation systems available in the model as investment options were extracted from the Danish Technology Catalog [78]. The equilibrium between heat supply and demand is maintained on an annual and a subannual basis, or in each time slice. The approach used to separate the space heating demand profile from the hot water demand profile using a single profile from the DH substations has been described in the literature [79]. The hourly electricity price profiles and the CO₂ emission factors for electricity consumption and generation in the price region of Sweden were obtained from the NordPool wholesale electricity market [80] and BALMOREL modeling performed by Ea Energy Analysis A/S [81]. IEA WEO scenarios are used to predict fuel and carbon prices [28, 29, 36, 82], whereas electricity prices extracted from the BALMOREL model for the CNN scenario [43].

Existing TES technologies in the DH system of Eskilstuna are a hot water tank and TES in the DH network. All the TES technologies considered in this study and their parameters are shown in table 3.

Table 3. Assumed techno-economic parameters of the investigated TES options, both existing and potential investments.

	Existing TES		Investment TES options
	TES in DH network	Existing water tank	TES in multifamily houses	TES in single-family houses	Large water tank	Cavern TES	Pit TES	Borehole TES
Cycle efficiency (inflow/outflow) (%)	100	98	100	100	98	98	70	98
Daily losses (%/day)	24	0.19	14	22	0.19	0.07	0.14	0.12
Maximum inflow/outflow (MW)	13 a	60	20	32	—		—
Maximum capacity (MWh)	30 a	900	102	160	—	—	—	—
Lifetime (yr)	—	40	—	—	40	40	20	40
Investment cost (kEUR TJ−1)	—	—	15.6 b	23.5 b	823	500	161	53
Fixed O&M cost (kEUR TJ−1)	—	—	—	—	2.4		0.8	—

^a Considering temperature difference in the supply water pipe of 5 K. ^b Values in kEUR°MW⁻¹, calculated using the estimated number of substations, the data, and approach by [16].

Four scenarios are investigated that share the same CNN central storyline from the earlier mentioned NCES project [21], which affects the modeling via the assumed electricity price profiles, CO₂ emission factors associated with power consumption, fuel prices, and price of CO₂. The sole difference between the scenarios is the availability of TES (both centralized and decentralized) in the studied DH system:

• (1)  
  'No-TES'—none of the TES options are included in the model (no TES in the city),
• (2)  
  'Exist-TES'—existing TES options are included in the model,
• (3)  
  'New-TES'—existing TES options are excluded from the model (assumed to be unavailable), but the model can invest in new storage capacities,
• (4)  
  'All-TES'—both types of TES, i.e. existing and new investment options, are available in the model.

As there are two TES options already available in the city of Eskilstuna (a large water tank connected to the DH system, and the DH network itself), the 'No_TES' and 'New_TES' scenarios are not realistic. However, to investigate the benefits of having a TES in the DH system, the operation of the system without a TES should be compared to the operation of the system with a TES available.

3.2. Lower Austria

The modeling of flexibility options in Lower Austria integrates an assessment of the effects of flexibility measures in the electricity market, a forecast of future electricity prices in Austria, and a forecast heating sector development. First, the effect of DSM as a flexibility measure for the demo site is assessed using an MILP optimization model. Second, future electricity prices in Austria are estimated using the BALMOREL power system model [83]. Then, as for Eskilstuna, the TIMES model is used to assess the effect of the flexibility measures in the heat sector in Lower Austria, scaling up the measures of the demo site [84].

The TIMES model represents Lower Austria from the selected base year (2017). Each year is divided into 12 time slices that represent the average day, night, and peak demand for every one of the four seasons (figure 3). Fuel prices are determined by the Austrian JRC-EU-TIMES model [42], which considers national targets by 2030 based on the Austrian National Climate and Energy Plan [85].

Zoom In Zoom Out Reset image size

Flexibility options considered include sector coupling via HPs and CHP plants, TESs for DH (pit TES and large-scale water tanks for DH) and DSM. The DSM measures are implemented in the modeling by modifying the proportion of heating load in the time slices. Figure 4 shows the capacity load ratio by time slices for the residential area in two cases, i.e. with and without DSM implementation. The capacity load ratio is defined as the ratio between the maximum capacity required to cover the heat demand in each time slice, and is highest when non-DSM is applied.

Zoom In Zoom Out Reset image size

For Eskilstuna, the results show that the heating sector of the city will be dominated by electricity-consuming technologies, i.e. HPs, as early as 2040. Figure 5 shows the total heat generation mix of the city between the base year of 2018 and the year 2050 in the 'No-TES' scenario (a scenario without a TES unit). In 2050, slightly less than one-third of the heating demand of Eskilstuna will be met by individual heating systems. These are air-based HPs and biomass-fired HOBs. Individual HPs are used for the same reasons as centralized HPs, as above described. The primary reason for the investments in biomass-fired boilers is the limit on the transmission capacity of the DH network. For all scenarios, the model does not invest in expanding the DH network, as it is less expensive to supply one-third of the city's heating demand using individual heating.

Zoom In Zoom Out Reset image size

The Eskilstuna heating sector transforms from a net electricity generator to a significant electricity consumer (figure 6). In 2019, the cost-optimal operation of Eskilstuna's DH system results in almost 1000 TJ of generated electricity by the CHP plant, with the total electricity consumption for heating of 400 TJ. Beginning in 2040, the electricity consumption for heating exceeds 1000 TJ despite the cessation of electricity generation, and can use the excess, inexpensive electricity produced by the power sector. Electricity consumption significantly increases in the 'New-TES' and 'ALL-TES' scenarios, as HPs produce more heat to be stored between seasons. Such an increase is a significant challenge for the local power distribution grid.

Zoom In Zoom Out Reset image size

The tight connection between the heating and electricity sectors is shown in figure 7. The modeling results show that the variations in the applied electricity price profile are sufficient to affect the operation of the heat generation units in the investigated heating sector of Eskilstuna: the air-based HPs are dispatched to generate heat during the periods of lower electricity prices, while the biomass-fueled CHP plant is dispatched to generate heat during the periods of higher electricity prices. During the periods of increased heat demand, such as October in the graph (and other winter months), the impact of fluctuating electricity prices is not evident, because both the HP and CHP technologies generate heat at their rated capacity.

Zoom In Zoom Out Reset image size

In terms of cost-efficient flexibility (thermal storage) potential, modeling shows that the changes to the heat supply model of the city in the presence of a TES lead to reduced heating costs. Table 4 shows the total system cost of heat supply in the city from 2018 to 2050, as predicted by the model. Compared to the 'No-TES' scenario, the total cost of heat supply is reduced by 0.5% in the 'Exist-TES' and 'New-TES' scenarios and by 1% in the 'ALL-TES' scenarios. This shows that, even though TES requires an initial investment cost, the cost is paid off by reduced operational costs for generating heat.

The effect of the seasonal TES is shown in the duration curves of heat generation in descending order (figure 8). Compared to the 'No-TES' and 'Exist-TES' scenarios, the 'New-TES' and 'ALL-TES' scenarios considerably reduce the amount of heat produced during periods of peak heat demand. In contrast, between time slices 30 and 50 in figure 8, the heat generated in the last two scenarios is greater than that of the first two. This shows that heat generated during periods of low demand is stored in the seasonal TES and released during periods of peak demand. In the 'New-TES' scenario, the model invests in the building of TES, specifically in the TES in multifamily houses. In the absence of the existing TES options (hot water tank and network TES), developing TES investments becomes cost effective. At the same time, the model does not invest in TES in single-family houses, showing that their capabilities for storing heat are not important. In the 'New-TES' scenario, the model also invests in the building of TES, specifically TES in multifamily houses. Building TES investments become cost effective in the absence of existing TES options (hot water tank and network TES). At the same time, the model does not invest in the TES in single-family houses, leaving their capabilities for storing heat unused.

Zoom In Zoom Out Reset image size

For Lower Austria, the results show that sector coupling (biomass CHP and air-source HPs) will play a crucial role in decarbonizing the heating system. For the 100% RES and MaxFlex scenarios, the use of fossil fuel in the residential area ends in 2040, and building refurbishment significantly reduces energy consumption. In both scenarios, to cover future heat demand, the importance of HPs (air and geothermal) and DH increases. However, the use of HPs is higher in the MaxFlex scenario.

Figure 9 shows the projected residential fuel consumption and heat demand in urban zones. In both scenarios, heat demand is reduced by around 6.9 PJ (1917 GWh) in 2050. This year, building refurbishment significantly reduces the heat demand by around 1.8 PJ (500 GWh). In 2050, in the 100% RES Scenario, total fuel consumption is 7.2 PJ (2000 GWh), with DH and air-source HPs accounting for 69% and 30% of the fuel consumption, respectively. The MaxFlex scenario has a total fuel consumption of 7.3 PJ (2028 GWh), with DH and air-source accounting for 52% and 33% of the heat demand, respectively.

Zoom In Zoom Out Reset image size

Figure 9 shows the projection of fuel consumption and heat demand for the service sector in rural zones in both scenarios. Heat demand is reduced by 4.6 PJ (1278 GWh) in 2050. In both scenarios, the intermediate years in rural zones are like in urban zones. In the 100% RES scenario, in 2050, total fuel consumption is 4.7 PJ (1305 GWh), and the prevalence of DH increases to 85% of the total with 4.0 PJ (1111 GWh). The remaining heat demand is covered by wood biomass. In 2050, for the MaxFlex scenario, total fuel consumption is similar to the 100% RES scenario, accounting for 4.7 PJ (1305 GWh). In the 100% RES scenario, total fuel consumption in 2050 is 4.7 PJ (1305 GWh) while DH prevalence increases to 85% with 4.0 PJ (1111 GWh). The remaining heat demand is covered by wood biomass. In 2050, MaxFlex fuel consumption is similar to the 100% RES scenario, accounting for 4.7 PJ (1305 GWh), and DH increases up to 80% of the total heat demand. In both scenarios, the use of biomass (wood and pellets) and HPs is critical in the long run for the decarbonization of the system. In the 100% RES scenario, fuel consumption increases to 11.8 PJ (3278 GWh) in 2050. This year, wood biomass and ambient heat account for 71% and 25% of the heat demand, respectively, with solar thermal accounting for the remainder. In the MaxFlex scenario, fuel consumption also increases, but less than in the 100% RES scenario, accounting for 9.8 PJ (2722 GWh) in 2050. This year, wood biomass is also the primary fuel, accounting for 74% of fuel demand, with air-source HPs accounting for the remainder. Biogas is also relevant in this scenario during the intermediate years up to 2035, as biogas CHP plants appear to be competitive.

When comparing figures 9 and 5, significant differences are observed between the case-studies in the transformation of the heating sector and the use of DH. The use of biomass in DH in Eskilstuna is less attractive in the long term because of the future high prices considered. On the other hand, Lower Austria can locally produce biomass at a competitive price, and more importantly, the smaller size of the DH plants makes it possible to cover its fuel demand from sources relatively close to the heating plants, all this facilitates the future use of biomass. In both locations, future electrification (HPs) is key for the decarbonization of the heating sector, being more relevant in Eskilstuna because the loss of relevance of DH driven by biomass.

Figure 10 shows the projection of fuel consumption of the DH sector in urban zones. In the 100% RES scenario, fuel consumption increases to 11.8 PJ (3278 GWh) in 2050. This year, biomass wood and air-source account for 71% and 25% of total heat demand, respectively, with solar thermal accounting for the remainder. In the MaxFlex scenario, fuel consumption also increases, but less than in the 100% RES scenario, accounting for 9.8 PJ (2722 GWh) in 2050. This year, wood biomass is also the major fuel (74% of total fuel demand), with air-source HPs accounting for the remainder. In this scenario, BGS is also relevant during the intermediate years up to 2035, as biogas CHP plants appear to be competitive.

Zoom In Zoom Out Reset image size

Figure 10 also shows the projected heat generation in the DH sector in rural zones for both scenarios. In the 100% Scenario, heat generation increases to 15.2 PJ (4222 GWh) in 2050. In 2050, wood-fueled CHP and heat-only plants, as well as ambient-heat technologies, will provide 51%, 25%, and 17% of the heat demand, respectively, while the remaining part is covered by new solar DH. As in urban areas, there is a significant increase in the use of CHP plants, from 15% of the heat generation in 2017 to 51% by 2050, at the expense of heat-only plants. CHP plants powered by pellets are used during the intermediate years. In the MaxFlex scenario, heat generation increases to 10.3 PJ (2861 GWh) by 2050. Wood-fueled CHP plants provide 70% of the heat (compared to 15% in 2017), with the remaining 30% coming from ambient heat technologies. CHP plants are increasingly being used at the expense of heat-only plants. CHP plants fueled by biogas are used during the intermediate years.

In terms of DH electricity generation, CHP plants have high long-term growth in both scenarios, with CHP biofuel biomass (pellets and wood) plants dominating (figure 11). CHP plants become more relevant in the mid-to-long term due to lower investment costs of these technologies, predicted future increase in electricity prices, and increased electricity demand e.g. because of an increase in the use of HPs in single-family buildings. In 2050, the 100% RES and MaxFlex scenarios boost electricity generation to 2.3 PJ (639 GWh) and 1.9 PJ (528 GWh), respectively. Wood-fueled CHP plants are replacing natural gas-fueled ones. Up until the intermediate years, pellet CHP plants in the 100% RES scenario and biogas CHP plants in the MaxFlex scenario take part in decarbonizing the DH system. However, they are ultimately replaced by wood biomass CHP plants. This is also the case for the rural zones.

Zoom In Zoom Out Reset image size

Figure 12 compares the relationship between heat production and installed capacity in 2017 and 2050 for both scenarios. The datasets are normalized using the base year as a reference. In both scenarios, compared to 2017, there is an increased capacity of DH to cover winter peak times in 2050 with a reduced installed capacity. This is due to the estimated efficiency improvement of future technologies as well as the contribution of DSM to reducing the winter peaks. This is important in the MaxFlex scenario because DSM is widely known as a flexibility measure.

Zoom In Zoom Out Reset image size

In general, the model invests in additional DH systems in both urban and rural areas. Table 5 shows the total heat and electricity generated by various technologies in the DH sector. Pit TES and large-scale water tanks are considered seasonal thermal storage in DH systems in the model. However, the model does not invest in these technologies. This can be explained by both the low availability of variable sources (solar thermal) in the DH system, and the high use of other flexible technologies such as CHP and HP.

The analysis of the economic viability of the flexibility potential in DH systems is performed using optimization models in the long terms and how this can evolve in Eskilstuna and Lower Austria. As with any modeling, the results of this study depend on the assumed input data, in particular electricity prices, heating demand, and technology and infrastructure investment.

Differences in fuel and electricity price assumptions are one of the major reasons why Eskilstuna and Lower Austria's models differ in results. In addition, subsidiary structures, business, and household taxation affect the results. Eskilstuna's assumed electricity prices extracted from the BALMOREL model for the CNN scenario are lower in 2030 and 2050 than in reference year, so that the model extensively invests in HPs (both centralized and decentralized). Although an analysis of future energy prices is beyond the scope of this study, it is known that increased renewable shares often lowers prices. Additional model runs with higher energy costs than in reference year could have resulted in (re)investments in biomass-fired CHP plants (biomass prices are also assumed to be relatively low in future years). The predominance of investments in HPs is due to assumptions on investment costs, variable and fixed operation costs, fuel prices, taxes, and subsidies. In the case of Austria, electricity prices were extracted from the national power system model by using the BALMOREL modeling tool. The simulated storyline follows the national electricity targets of 100% renewable electricity by 2030 (generation and consumption balance). Contrary to the lower electricity prices expected in Sweden, electricity prices are expected to increase in Austria by 37% by 2030 and 62% by 2050 compared to the 2017 level. Austria's electricity demand is predicted to rise. This led to model-specific growth of the CHP plants in the mid-to-long term due to higher electricity prices. In both scenarios, putting up a lot of CHP plants increases the amount of electricity made by 2050 from green fuels, such as wood biomass.

In both cases, strong dependence on technologies that consume electric power, such as HPs and electric boilers, will lock the entire sector into one technology type, jeopardizing supply security and increasing the pressure on the electric power sector. The discrepancy between recently observed high electricity prices and the common belief of field experts that future DH systems should contain CHP plants and support the electric power sector, and the modeling results pointing toward high HPs shares should be addressed in future research. In parallel, the future use of biofuel will be a key for the heating sector, where the prices, local availability and topology and capacity of the DH will determine its usage.

As for the heating demand, it may increase or decrease in the future. In Eskilstuna, heating demand is expected to decrease due to the implementation of heat-saving measures, and to increase due to population growth. A sensitivity analysis with both decreasing and increasing future energy prices shows that greater or lower heating demand will affect the amount of generated heat and the size of the installed heat generation and storage capacities, but not the generating mix and the overall composition of Eskilstuna's heating sector.

In terms of investment costs in technology and infrastructure, Eskilstuna's model does not expand DH network capacity. Instead, it (re)invests in heat generation units. Whereas in this study, we have assumed that the cost of DH network expansion is applicable to heat-density areas of 120–300 TJ km⁻², the costs for areas with other heat densities—expansion of the DH network—could be lower. The same result could have been observed if the investment or running costs of the individual heating technologies had increased. Additional model runs are required to establish more robust findings on the future development and flexibility potential of DH systems.

At the same time, the modeling approach and model structure have significant impacts on the results. For example, the model applied to Eskilstuna describes a given modeled year using 72 time slices, whereas the number of time slices in the Lower Austria's model is 12. Also, the approach to modeling flexibility in buildings differs as follows: (i) in Eskilstuna, the potential of buildings to serve as a TES is expressed as a maximum total capacity that can be optimized by the optimization model, (ii) in Lower Austria, the flexible heat demand of the buildings was precalculated and provided to the optimization model as several different heat demand profiles (with and without activated flexibility). Furthermore, while the Eskilstuna model refers to a city-concentrated model (100 000 people), the Lower Austria Model refers to a large region (1.6 million people), with different density characteristics in urban and rural zones. Finally, the two applicable models study different scenarios due to partnerships with national funding agencies and the slightly different foci of the models.

We lastly would like to reflect on how climate data is accounted for in the different models (table 1). TIMES models implicitly consider future climate (until 2050) aligned with limiting global warming, as take exogenous projections of demand, energy prices and CO₂ prices from other models [38–43] that generally explore climate neutral scenarios. ECCABS model uses climate data (1 year) from Meteonorm, an average of 30 years for the region [48], to calculate the energy signature for each building in equation (3). Nevertheless, in the same equation, both constants (${C_{\text{F}}}$, ${C_{\text{C}}}$) have been approximated empirically with the climate conditions in the demo sites in year 2020 measured by the Swedish Meteorological and Hydrological Institute (SMHI). A comparison of SMHI and Meteonorm datasets ⁵ shows a similar yearly pattern but a winter in 2020 warmer than the average. There is therefore a potential to adjust the energy signature values to warmer temperatures which has not yet been explored.

We have developed a reproducible method for estimating the cost-efficient flexibility potential of a local or regional energy system, and implemented it to two different case studies, representative types of DH networks in terms of size and integration with the electricity grid. Our main conclusion is that future energy prices, including biofuels and electricity prices, have a big impact on the shape of future heating systems, and lead to either a high degree of centralized HPs (low electricity prices) or a high proportion of CHP (high electricity prices and low biomass prices). The availability of large-scale energy storage also appears to play an important role. Even though DH businesses may have access to demand-side flexibility, under current assumptions the most cost-effective flexible solutions stay centralized.

For the case study of Eskilstuna, the modeling shows that the future heating sector will be dominated by both centralized (in the DH system) and individual HPs, regardless of the applied scenario or the availability of TES, due to the assumed decreasing future electricity prices. The city's heating sector becomes a net consumer of electricity and can assist the power sector by using excess electricity. However, relying only on technologies that consume electric power in the city's heating sector limits its potential to provide flexible services to the electric power sector—the retirement of the CHP plant eliminates the possibility of generating electricity during periods of high electricity prices. Moreover, recently observed high electricity prices and the national overall business consensus that the future DH system should contain CHP plants and be capable of supporting the electric power sector contradict the modeling results. Therefore, further research bridging the scientific, modeling, governmental, and corporate perspectives on the development of the heating sector in different countries is crucial.

The modeling results also show that the already available and invested TES solutions will be actively used in the future for storing heat between seasons and for reducing peak heat generation, i.e., for flattening the heat generation curve. This reduces the cost of the heat supply and increases the flexibility potential of the heating system. In addition, using HPs with TES units in DH systems will provide a reliable and flexible service that will be beneficial to the electric power industry.

In Lower Austria, in both the analyzed scenarios it is possible to decarbonize the heating sector by 2040 and limit the use of fossil fuels by 2035. In the residential sector, the use of fossil fuels ends in 2040 when demand is reduced through refurbishing of buildings and expanding HPs (air and geothermal) and DH. In the services sector, DH becomes more prevalent. The changeover relies on wood biomass and air-source HPs. Long-term use of wood biomass and air-source HPs is high in DH and the role of CHP plants increases. During the interim years, pellet CHP plants appear relevant in the 100% RES scenario, whereas biogas CHP plants appear relevant in the MaxFlex scenario. The massive use of CHP plants boosts electricity generation. In both scenarios, there is an increase in the use of DH to cover winter peaks, which is expected to lead to increased efficiency and better performance of future technologies, as well as wider use of DSM to reduce winter peaks.

The results depend on the assumed techno-economic parameters of the technologies, energy prices, and CO₂ prices, which have attempted to reflect the mid-to-long term plans and visions of the regions studied. In that sense, the heating sector and use of DH evolve differently in the regions. The use of biomass in DH in Eskilstuna in less attractive in the long term because of the future higher Swedish biomass prices, whereas in Lower Austria the smaller size of the plant makes it can be supplied by local low-cost biomass. Both locations show a higher electrification of the heating sector in the future, where HPs will play a key role, although inversely proportional to the price of electricity.

We gratefully acknowledge the contributions of Akram Sandvall and Johanna Nilsson at IVL to defining the scenarios, and of Ralf-Roman Schmidt and Sarah Wimmeder at AIT. We also thank Johan Kensby and Christian Johansson for their support in representing building TES via equation (3). The content and views expressed in this material are those of the authors and do not necessarily reflect the views or opinion of the ERA-Net SES initiative that has funded the project.

The data cannot be made publicly available upon publication because they are not available in a format that is sufficiently accessible or reusable by other researchers. The data that support the findings of this study are available upon reasonable request from the authors.

There are no conflicts of interest to declare

Érika Mata: study conception, methodology, formal analysis, writing/manuscript preparation: writing the initial draft, writing/manuscript preparation: visualization/data presentation, project administration, funding acquisition

Nicolas Pardo Garcia: study conception, methodology, computation, formal analysis, investigation: data/evidence collection, writing/manuscript preparation: critical review, commentary or revision, writing/manuscript preparation: visualization/data presentation, funding acquisition

Demet Suna: study conception, methodology, writing/manuscript preparation: critical review, commentary or revision, funding acquisition

Burcu Unluturk: computation, formal analysis, investigation: data/evidence collection

Anton Jacobson: computation, formal analysis, data curation, writing/manuscript preparation: visualization/data presentation

Olga Lysenko: writing/manuscript preparation: writing the initial draft, writing/manuscript preparation: visualization/data presentation

All authors are funded by the Flexi-Sync project the joint programming initiative ERA-Net Smart Energy Systems, with support from the European Union's Horizon 2020 research and innovation program.

Data available on request from the authors