Time Series of Heat Demand and Heat Pump Efficiency for Energy System Modeling

Abstract

With electric heat pumps substituting for fossil-fueled alternatives, the temporal variability of their power consumption becomes increasingly important to the electricity system. To easily include this variability in energy system analyses, this paper introduces the “When2Heat” dataset comprising synthetic national time series of both the heat demand and the coefficient of performance (COP) of heat pumps. It covers 16 European countries, includes the years 2008 to 2018, and features an hourly resolution. Demand profiles for space and water heating are computed by combining gas standard load profiles with spatial temperature and wind speed reanalysis data as well as population geodata. COP time series for different heat sources – air, ground, and groundwater – and different heat sinks – floor heating, radiators, and water heating – are calculated based on COP and heating curves using reanalysis temperature data. The dataset, as well as the scripts and input parameters, are publicly available under an open source license on the Open Power System Data platform. Machine-accessible metadata file describing the reported data: https://doi.org/10.6084/m9.figshare.9878849

Full text

1
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
Time series of heat demand and


 & Aaron Praktiknjo












In view of the global energy transition, open energy data are more important than ever1. is includes data on
electric building heat pumps, which constitute a cornerstone of sustainable energy scenarios2. eir power con-
sumption is naturally highly variable. On the one hand, there are uctuations in the heat demand to be fullled
by the heat pumps. On the other hand, the COP of the heat pumps, which is dened as the varying ratio of their
heat generation and electricity consumption, changes over time. is variability will be essential for the future
electricity system balance and needs to be considered in related system and market analyses3,4.
Against this background, this paper introduces the When2Heat5 dataset comprising the rst ready-to-use
national time series of both the heat demand and the COP of building heat pumps. e strengths of the dataset
include:
(1) Validity: Historic time series are presented, thus excluding uncertain assumptions on future developments.
e heat demand is computed using standard load proles, which are permanently used by German gas
suppliers, and internationally validated with measurements from the UK as well as building data from
the EU. e COP calculation is parametrized on manufacturer data and additionally validated with eld
measurements.
(2) Accuracy: e dataset considers particularities of dierent heat demands (space and water heating), dif-
ferent heat sources (air, ground, and groundwater), and dierent heat sinks (oor heating, radiators, and
water heating).
(3) Comprehensiveness: e time series cover a large geographical area of 16 cold-temperate-climate EU
countries (Fig.1), which is relevant for modelling the balance of the more and more integrated European
electricity system. Furthermore, eleven years (2008–2018) are included to enable weather year sensitivity
analyses.
(4) Applicability: e data are spatially highly aggregated (on the national level), while being temporally highly
resolved (hourly). Hence, the format matches those of many European electricity market models.
1Institute for Future Energy Consumer Needs and Behavior (FCN), RWTH Aachen University, Aachen, Germany.
2Hertie School of Governance, Berlin, Germany. 3Neon Neue Energieökonomik GmbH (Neon), Berlin, Germany.
4Mercator Research Institute on Global Commons and Climate Change (MCC), Berlin, Germany. Correspondence and
requests for materials should be addressed to O.R. (email: oliver.ruhnau@rwth-aachen.de)
Received: 23 April 2019
Accepted: 20 August 2019
Published: xx xx xxxx


Content courtesy of Springer Nature, terms of use apply. Rights reserved

2
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
e When2Heat dataset is a contribution to the Open Power System Data project and follows the frictionless
data principles6. Focusing on the representation of heat pumps, the aim is to improve eciency, transparency, and
reproducibility of electricity market models, which might be part of more general integrated energy system anal-
yses. Furthermore, it may serve as a valuable benchmark for alternative heat demand and heat pump modelling
approaches on the national level. Existing limitations of the dataset are critically discussed in the Usage Notes section.
e heat demand time series are based on the German gas standard load prole approach dened by BGW7
and BDEW8. e methodology is combined with three super-national datasets to estimate national time series
for 16 European countries. First, temperature and wind speed data from the global ERA-Interim reanalysis
serve to generate separate demand time series for space and water heating at all available locations within each
country. Next, assuming the heat demand at dierent locations to be proportional to the population, the spatial
time series are weighted with population geodata from Eurostat, aggregated to national time series, and normal-
ized to one TWh average yearly demand. Finally, for the years 2008 to 2013, these are scaled with data on the
annual nal energy consumption for space and water heating from the EU Building Database and corrected for
nal-energy-to-heat conversion losses. Figure2 provides an overview of the methodology applied.
e COP time series are based on quadratic COP curves, which take the dierence between the heat source
and heat sink temperatures as an input. e curves are parametrized on manufacturer data9 and distinguish
between three heat pump types: air-source heat pumps (ASHP), ground-source heat pumps (GSHP), and
groundwater-source heat pumps (WSHP). Heat source temperature time series for air and soil are retrieved from
the ERA-Interim archive, and a constant groundwater temperature is used. Heat sink temperature time series are
computed with literature-based heating curves for oor heating and radiators, and the hot water supply tempera-
ture is set constant. Assuming a homogenous adoption of dierent heat pump technologies across dierent loca-
tions within each country and across various building types, the spatial COP proles are nationally aggregated
with respect to the heat demand. Finally, the time series are corrected based on eld measurements10. Figure3
summarizes the methodology applied.
Methods
is section describes the methodology behind the When2Heat dataset. At rst, the data that serve as an input to
the calculation of both the heat demand and the COP time series are introduced. Subsequently, the procedures
applied for the preparation of the heat demand time series and of the COP time series are presented in detail,
respectively. Finally, the code availability is outlined.
 e time series of the present dataset are based on weather data from the ERA-Interim archive,
a global atmospheric reanalysis from the European Centre for Medium-Range Weather Forecasts (ECMWF)11.
e following parameters are used:
• 2 metre temperature: e ambient air temperature at 2 m above ground
• Soil temperature level 4: e ground temperature at 1.00–2.89 m below ground (https://conuence.ecmwf.int/
pages/viewpage.action?pageId=56660259)
• 10 metre wind speed: e wind speed at 10 m above ground.
Fig. 1 Countries included in the dataset. In alphabetic order: Austria, Belgium, Bulgaria, Czech Republic,
Germany, France, Great Britain, Croatia, Hungary, Ireland, Netherlands, Poland, Romania, Slovenia, and
Slovakia.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

3
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
e temperature parameters are retrieved for the years 2008 to 2018 in six-hourly temporal resolution, and
the wind speed data are retrieved for all available years (1979–2018) in monthly resolution. All parameters fea-
ture a 0.75 × 0.75° spatial grid, which is equivalent to approx. 28 × 17 km. For the wind speed, the average of all
October-to-April heating periods from 1979 to 2018 is determined for every location, which serves to classify
them into “normal” and “windy” locations in the following.
For their spatial aggregation, local time series are weighted with population geodata from the
Eurostat GEOSTAT dataset (http://ec.europa.eu/eurostat/web/gisco/geodata/reference-data/
population-distribution-demography/geostat). ese data originally feature a 1 km² resolution and are thus ini-
tially mapped to the 0.75 × 0.75° grid of the ERA-Interim data. For the nal scaling of the demand proles, yearly
data on the nal energy consumption for space and water heating in residential and non-residential buildings are
retrieved from the EU Building Database (http://ec.europa.eu/energy/en/eu-buildings-database).
 Temporal heat demand proles are determined by three factors: weather con-
ditions, building properties, and occupant behavior. Its calculation can follow either statistical methods, includ-
ing standard and reference load proles, or physical approaches (for an overview, see Fischer et al.12). For the
When2Heat dataset, the German statistical methodology for calculating gas standard load proles has been cho-
sen, which is permanently used by gas suppliers for non-daily metered customers. e proles explicitly refer to
National time series
For building typesand
applications as above
Spatialtime series
0,75 x 0,75°
Fordifferent buildingtypes
•single-familyhouses
•multi-family houses
•commercialbuildings
For different applications
•Space heating
•Waterheating
Spatialtime series
0,75 x 0,75°
Weighted mean temperature, mean wind speed (BGW)
Profile functions(BDEW)
Temperature-dependent hourly profiles (BGW)
Weightingwithpopulation(Eurostat)
Aggregation by countries & normalization(1 TW h/a)
Scalingwithfinal energyfor heating(EU Building Dat abase
)
Multiplyingwithfinal-energy-to-heatconversionefficiency
Temperature andwind speed
ERA-Interim dataset
Daily reference temperatureand
Oct-Apr wind speedaverages
Daily demand
Hourly demand
Hourly profiles
Hourly demand
Fig. 2 Methodology applied for calculating the heat demand time series. Starting from spatial temperature and
wind speed time series from the ECMWF ERA-Interim dataset, gas standard load proles are derived according
to BGW7 and BDEW8. e time series are weighted with population data from Eurostat GEOSTAT dataset,
spatially aggregated, normalized and scaled using the EU Building Database.
Spatialtime series
0,75 x 0,75°
For different heat sources
•Air
•Ground
•Groundwater
For different heat sinks
•Floor heating
•Radiators
•Waterheating
National time series
For heat sourcesasabove
For different heating systems
•Floor & water heating
•Radiators & water heating
Spatialtime series
0,75 x 0,75°
Heatingcurves(own literature-based assumptions)
QuadraticCOP curves (regression on manufacturerd
ata)
Aggregation with respecttothe heat demand
Correction factor basedon fieldmeasurements
Ambient airand soil temperature
ERA-Interim dataset
Hourlytemperature difference
Hourly COP
uncorrected
HourlyCOP
corrected
Fig. 3 Methodology applied for calculating the COP time series. Spatial temperature time series from the
ECMWF ERA-Interim dataset are combined with COP curves based on manufacturer data9 to calculate the
temperature dierence between various heat sources and sinks. Aer spatial aggregation, the time series are
corrected with respect to eld measurements10.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

4
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
space and water heating, and it is assumed that (1) gas boiler operation follows the original heat demand and (2)
gas heated buildings are representative for the whole building stock.
e gas standard load prole methodology has been presented by BGW7 and updated BDEW8. While the
calculation of daily reference temperatures is equally included in both references, the calculation of the daily
demand has been rened in BDEW8, and the calculation of wind speed averages (for the assignment of dierent
proles) and the calculation of hourly demand is exclusively described in BGW7. Here, both references are used
and extended by a spatial dimension to calculate national time series.
Daily reference temperature time series are calculated for each of the ERA-Interim locations. To capture the
thermal inertia of buildings, the daily reference temperature,
Tdl
ref
,
, is dened as the weighted mean of the daily
average ambient air temperature of the actual day,
Tdl
amb
,
, and the corresponding temperatures of the three preced-
ing days,
…
−−
T T
dl
ambdl
amb
1, 3,
, for every day, d, and every location, l7,8:
=+. +. +.
+.+. +.
−− −
TTT TT05 0250125
105025 0125 (1)
dl
refdl dl
amb
dl
amb
dl
amb
,,1,2,3,
Daily demand factors are derived from the reference temperatures using prole functions. ese demand fac-
tors, fd,l, can be interpreted as unscaled daily demand, which is normalized in the following. e prole function
is dened by a combination of a sigmoid function and two linear functions8:
=
+









++









⋅+
⋅+









⋅°
−


fADmax mTCb
mTCb
1
/
/,
(2)
dl BC
TT
C
spacedl
refspace
waterdl
refwater
,,
,
dl
ref
,0
with T0 = 40 °C. BDEW8 presents sets of prole function parameters, A, B, C, D, mspace, bspace, mwater, bwater, for
various building types, namely single-family houses, multi-family houses, and commercial buildings. Parameters
for more and less temperature-sensitive proles are provided for dierent regional weather conditions, which
are related to the local wind speed7. erefore, all locations are clustered based on the averaged ERA-Interim
wind speed data: for averages above 4.4 m/s, the sigmoid functions for “windy” locations is applied. Otherwise,
the locations are assigned to the “normal” category. Figure4 depicts a selection of the resulting prole functions.
Hourly demand time series are derived for each location from the daily values by means of hourly demand
factors. BGW7 presents these factors for the dierent building types, ten dierent temperature ranges, and – in
the case of commercial buildings – various weekdays (See page 55 for single- and multi-family houses and pages
85–86 for commercial buildings). Note that dierent classes are distinguished by the share of old buildings and
the type of commerce, but here the German average is considered. ese demand factors can be interpreted as
hourly shares of the daily demand, i.e. they sum up to 100% per day. For commercial buildings, BGW7 addition-
ally derived weekday factors, which scale the daily demand according to the day of the week. Figure5 depicts a
selection of the hourly demand factors, where the weekday factors are already included, i.e. the hourly factors of
each day sum up to the weekday factor in the case of commercial buildings.
Separate time series for space and water heating are of interest, e.g., to allow for considering their dierent
temperature levels for the COP calculation. In BDEW8, the temperature independent component of the sigmoid
function, parameter D, and the linear function for water heating, ⋅+
mTb
waterdl
refwater
,
, are associated with the gas
consumption for water heating. However, the linear water heating function only applies to temperatures at which
its value is higher than this of the linear space heating function (see Eq. (2)). is is the case for temperatures
above the average heating threshold of approximately 15 °C. For higher temperatures, it is assumed here that the
demand for water heating remains constant. us, daily water heating demand factors,
fdl
water
,
, are determined at
each location according to the following equation:
0
1
2
3
4
-20-10 0102
03
0
[srotcafdnamedyliaD-]
Reference temperature [°C]
SFH
MFH
COM
SFH windy
SFH_water
Fig. 4 Daily heat demand factors as a function of the reference temperature. Exemplary prole functions for
single-family houses (SFH), multi-family houses (MFH), and commercial buildings (COM) and for single-
family houses at windy locations (SFH_windy). In addition, daily water heating demand factors for single-
family houses are displayed (SFH_water).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

5
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
=









+⋅ +>
+⋅+≤


fDm TCbT C
Dm bTC
/,15
15 ,15
(3)
dl
waterwaterdl
ref
waterdl
ref
waterwater dl
ref
,,,
,
Concerning the hourly demand factors, there is no such explicit distinction between space and water heating
in BGW7. However, assuming there is no space heating at high ambient air temperatures, the hourly demand fac-
tors for the highest temperature range (higher than 25 °C) are related to water heating. Hence, daily water heating
factors are multiplied with high-temperature hourly demand factors (including weekday factors for commercial
buildings) to compute water heating demand time series for each building type. e space heating demand is
calculated as the dierence between the total heat demand and the water heating demand. ereby, some negative
values occur at the hourly resolution during summer, which are set to zero.
Finally, the resulting spatial demand time series are weighted with the population geodata from Eurostat,
aggregated by countries, and normalized to one TWh average yearly demand. us, weather year variations
cause the exact yearly sum of the normalized time series to uctuate around one TWh. For the years 2008 to
2013, for which data are available from the EU Building Database, proles are additionally scaled with the
annual nal energy consumption for heating. For the residential sector, the demand time series of single- and
multi-family houses are aggregated assuming a ratio of 70:30. Aer scaling, the time series for the residential and
non-residential sectors are aggregated for space heating and for water heating, separately. en, the nal energy
consumption for heating is transformed to the useful heat demand assuming an average conversion eciency of
0.9, and the time series are corrected for daylight saving time and dierent time zones. Space and water heating
time series are aggregated in the end, but separate time series are likewise included in the dataset.
 e COP of heat pumps generally depends on the temperatures and heat transfer condi-
tions at the heat source and at the heat sink, which are in turn linked to technical properties and varying weather
conditions.
e temperature dependence of the COP for the thermodynamically ideal process is described by the Carnot
eciency, which can be scaled down with a quality factor to model real heat pump processes13. As an alternative,
more generic option for modelling COP variance, quadratic regression has been proposed by Fischer et al.14.
Here, the latter approach is applied to manufacturer data on the COP under dierent temperature conditions9. As
shown in Fig.6, a regression is performed for each of the heat pump types with the following results:
=











.−.⋅∆+.⋅∆
.−.⋅∆+.⋅∆
.−.⋅∆+.⋅∆
COPTTASHP
TTGSHP
TTWSHP
608009 00005 ,
10 29 02100012 ,
997020 00012 ,(4)
2
2
2
For simplicity, variable-speed ASHP have not been considered in the regression, i.e. only on-o modulating heat
pumps are included. Note that this laboratory-based COP parametrization is adjusted for real-world ineciencies
in the following.
As an input to the COP curves, temperature dierences between all possible combinations of heat sources and
heat sinks,
∆Thl
sink source
,,
, are computed from the heat sink temperatures,
Thl
sink
,
, and the heat source temperatures,
Thl
source
,
, for each hour, h, at every location, l, using
∆=−.TTT
(5)
hl
sink source hl
sink hl
source
,,,,
Regarding the source temperature, dierent heat pump types are distinguished. For ASHP, the ambient air
temperature from the ERA-Interim dataset is directly used. For GSHP, the manufacturer data refer to the brine
temperature rather than the ground temperature. To account for the heat transfer from the ground to the brine,
0%
2%
4%
6%
8%
10%
06:00
12:00
18:00
06:00
12:00
18:00
06:00
12:00
18:00
06:00
12:00
18:00
06:00
12:00
18:00
06:00
12:00
18:00
06:00
12:00
18:00
06:00
12:00
18:00
06:00
12:00
18:00
MonTue WedThu FriSat Sun
SFHMFH COM
srotcafdnamedylruoH
Cold (-15 to -10°C) Medium (5 to 10°C) Warm (more than 25°C)
Fig. 5 Hourly demand factors at dierent temperature ranges. Exemplary functions for single-family houses
(SFH), multi-family houses (MFH), and commercial buildings (COM). Note that only the factors of commercial
buildings depend on the weekday.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

6
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
a temperature dierence of 5 K is subtracted from the ERA-Interim ground temperature. For WSHP, a constant
temperature of 10 °C and a temperature dierence of 5 K for possible intermediate heat exchangers are consid-
ered. Hourly values are linearly interpolated between the six-hourly data from ERA-Interim.
e heat sink temperatures are calculated for oor heating, radiator heating, and water heating. For oor
heating and radiators, the temperature time series are derived from the (hourly interpolated) ambient air temper-
ature,
Thl
amb
,
, as described by the following heating curves, which are averages of the existing literature14,15, as
depicted in Fig.7:
=









−.⋅
−.⋅


TCTradiator heating
CTfloor heating
40 10 ,
30 05 ,
(6)
hl
sink hl
amb
hl
amb
,,
,
In the case of water heating, a constant heat sink temperature of 50 °C is assumed according to German eld
measurements10.
By using these average heating curves, unrealistically small temperature dierences are computed at relatively
warm outdoor temperatures. To avoid these, a minimum temperature dierence of 15 K is introduced, which is
in line with the manufacturer data (Fig.6).
e resulting spatial COP time series, COPh,l, are aggregated to national time series, COPh,c, for every country,
c, using the following equation:
∑
==













−



COPQ
PQQ
COP,
(7)
hc hc
hc hc l
hl
hl
,,
,,,
,
1
where

Qhl,
and

Qhc,
denote the spatial and national heat demand time series, which are calculated as described
above. Ph,c is the national electricity consumption by the heat pumps. For simplicity, the COP time series do not
distinguish between dierent building types, and the sum of the normalized heat demand time series for the dif-
ferent building types is used here. e COP time series for oor and radiator heating systems are spatially aggre-
gated with respect to the space heating demand time series, whereas the COP time series for water heating are
spatially aggregated using the water heating demand time series.
A constant correction factor is applied to all COP time series to account for such real-world eects. As shown
in the Technical Validation section, the resulting COP time series dier signicantly from eld measurements.
0
1
2
3
4
5
6
7
8
9
10 20 30 40 50 60 70 80
COP
Delta_T [K]
WSHP
GSHP
ASHP
Fig. 6 Estimation of COP curves. Quadratic regressions are performed on manufacturer data9 distinguishing
between air-source heat pumps (ASHP), ground-source heat pumps (GSHP), and groundwater-source heat
pumps (WSHP).
0
10
20
30
40
50
60
-15-10 -5 0510 15
[erutarepmetknistaeH °C]
Ambient air temperature [°C]
HT Radiator (Fischer et al.)
LT Radiator (Fischer et al.)
Radiator (Nabe et al.)
Radiator (Own Assumption)
HT Floor Heating (Fischer et al.)
LT Floor Heating (Fischer et al.)
Floor Heating (Nabe et al.)
Floor Heating (Own Assumption
)
Fig. 7 Estimation of heating curves. Own assumptions are compared to literature values form Fischer et al.14
and Nabe et al.15, distinguishing between radiators and oor heating systems. HT: high-temperature; LT: low-
temperature.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

7
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
is can be explained by the assumption that the manufacturer data that are used for the COP curve regression
are obtained under ideal operating conditions and additional losses will occur in the real world. For example,
ideal conditions would imply a steady-state operation at full load, whereas in the real world, the adjustment of
heat pump operation to the current demand will be subject to losses. Further ineciencies may occur from the
pumping of groundwater for WSHP and brine for GSHP. e magnitude of the correction factor is set to 0.85,
corresponding to eld measurements from Günther et al.10.

All data are available on the Open Power System Data platform5. For every country, 24 time series are included in
the dataset (Table1). Nine time series report the absolute heat demand in MW for the dierent heat applications
and building types as well as their aggregation. Note that these time series are available only for the years 2008
to 2013, where data for scaling are available from the EU Building Database. Six proles cover the normalized
heat demand in MW per TWh average yearly demand for the dierent heat applications and building types. e
remaining nine time series represent the COP for dierent heat pump congurations (three heat sources times
three heat sinks). e COP values have no unit or, to emphasize the energy conversion, the unit MWth/MWel.
e whole dataset is provided in three dierent le types (csv, xlsx, and sqlite) and three dierent shapes
(singleindex, multiindex, and stacked), following the Open Power System Data time series standard. All formats
have pros and cons. In contrast to the stacked format, both the singleindex and the multiindex are easy to read
for humans and small in le size. e singleindex and stacked les are compatible with the datapackage standard.
e multiindex is easy to read into GAMS, which is a widely used technology for modelling electricity markets.
All les are indexed by the Coordinated Universal Time (UTC) and Central European (Summer-) Time (CE(S)T)
timestamps, indicating the start of the hourly period that the values of each row correspond to.

e following two subsections validate the heat demand and COP time series, respectively.
 e gas standard load prole methodology applied for the When2Heat data-
set is permanently used by German gas suppliers, which conrms its validity for the case of Germany. When
transferring the approach to other countries, dierent local weather conditions are explicitly considered through
the spatial ECMWF data. Building properties and occupant behavior, however, are implicitly considered in the
parameters of the standard load prole approach, and their possible cross-country variations are not modelled.
is section provides evidence that the approach is nevertheless valid for all countries included in the dataset.
e motivation for the creation of the present dataset is the absence of national heat demand time series,
which impedes direct validation. However, as already done above, actual data of the non-daily metered gas con-
sumption can be interpreted as a proxy for the temporal prole of the heat demand. Following this interpre-
tation, an exemplary validation is performed for the UK, where such data are available in a daily resolution.
Subsequently, the validity of the time series for the other countries included in the dataset is substantiated by
induction based on national building insulation statistics.
For the exemplary UK validation of the dataset, actual non-daily metered gas consumption data from UK
National Grid (https://mip-prod-web.azurewebsites.net/DataItemExplorer) serve as the benchmark. Daily time
series are retrieved for the year 2013, which is the rst year fully covered by published data. e yearly sum of
the time series amounts to 491 TWh/a, which is assumed to break-down into 365 TWh/a of consumption for
space heating, 89 TWh/a of consumption for water heating, and 37 TWh/a of consumption for other purposes,
according to DECC16. As the present dataset presents heat time series only, the non-heat component is deducted
from the National Grid data, assuming a constant temporal distribution. e resulting time series is compared to
the sum of the modelled space and water heating demand time series, which are aggregated by days and scaled
according to the gas consumption for space and water heating as reported by DECC16. Figure8 compares the tem-
poral prole of gas consumption for space and water heating as reported by National Grid to the modelled tempo-
ral prole of the present dataset. Both the chronological time series (le plot) and the load duration curve (right
plot) show high consistency (R² = 0.95). is conrms the applicability of the standard load prole approach not
only for Germany but also for the UK, even though cross-country variations of building properties and occupant
behavior have not been modelled.
For the other countries included, building characteristics are compared. The most important build-
ing characteristic is insulation, which can best be quantied by the heat transfer coecient. Average national
heat transfer coefficients can be retrieved from the EU Building Database (http://ec.europa.eu/energy/en/
eu-buildings-database) and are depicted in Fig.9. Apparently, the insulation of buildings in Germany and UK
diers quite substantially but this dataset’s methodology has shown to be robust to this dierence and to further
dierences in building properties and occupant behavior. By induction, it is argued that the methodology will be
an adequate estimate for all countries of which the building insulation is in the range of German and UK insula-
tion. is is the case for cold-temperate climate EU countries, which are thus included in the When2Heat dataset.
Scandinavian features much better insulation than Germany whereas Mediterranean features much worse insu-
lation compared to the UK, so that these countries are excluded from the dataset.
 To validate the COP time series, the averages are compared to the results of a large-scale
monitoring project from July 2012 to June 2013 in Germany10. Note that averages are calculated with respect to
the demand series and oen referred to as performance factors. As the eld measurements relate to a portfolio
of around 20% water heating and 80% space heating, 85% of which is equipped with oor heating and 15% with
radiators, the time series from the When2Heat dataset are aggregated accordingly (analogously to Eq. (7)). As
already discussed in the Methods section, the modelled COP is signicantly higher than the led measurements,
Content courtesy of Springer Nature, terms of use apply. Rights reserved

8
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
which is why a correction factor has been applied to increase the validity of the dataset (Fig.10). For WSHP,
Günther et al.10 have examined only four systems, of which the individual results range from 3.6 to 4.2. e
corrected average of the When2Heat dataset is at the top end of this interval, which is satisfactory as the small
measurement sample features exceptionally high heat sink temperatures.

e following subsections discuss the applicability and the limitations of the dataset, in turn.
 e When2Heat dataset is tailored to electricity market models. For a review of power-to-heat
modelling approaches, the reader might refer to Bloess et al.17. Technically, the multiindex format is most conven-
ient for the use in GAMS, which is a widely used technology for modelling electricity markets, but other soware
tools might be used with the various output formats.
e normalized heat proles correspond to one TWh average yearly demand. In other words, the exact yearly
sum of the normalized time series varies from year to year (around one TWh). In case inter-yearly changes of the
yearly heat demand are of interest, the time series should simply be multiplied with assumptions on the average
yearly demand. Otherwise, scaling should consider the specic yearly value of the normalized time series.
Var i able Attribute Description
heat_demand
total Heat demand for space and water heating
space Heat demand for space heating
water Heat demand for water heating
space_SFH Heat demand for space heating in single-family houses
space_MFH Heat demand for space heating in multi-family houses
space_COM Heat demand for space heating in commercial buildings
water_SFH Heat demand for water heating in single-family houses
water_MFH Heat demand for water heating in multi-family houses
water_COM Heat demand for water heating in commercial buildings
heat_prole
space_SFH Normalized heat demand for space heating in single-family houses
space_MFH Normalized heat demand for space heating in multi-family houses
space_COM Normalized heat demand for space heating in commercial buildings
water_SFH Normalized heat demand for water heating in single-family houses
water_MFH Normalized heat demand for water heating in multi-family houses
water_COM Normalized heat demand for water heating in commercial buildings
COP
ASHP_oor COP of air-source heat pumps with oor heating
ASHP_radiator COP of air-source heat pumps with radiator heating
ASHP_water COP of air-source heat pumps with water heating
GSHP_oor COP of ground-source heat pumps with oor heating
GSHP_radiator COP of ground-source heat pumps with radiator heating
GSHP_water COP of ground-source heat pumps with water heating
WSHP_oor COP of groundwater-source heat pumps with oor heating
WSHP_radiator COP of groundwater-source heat pumps with radiator heating
WSHP_water COP of groundwater-source heat pumps with water heating
Tab le 1. Overview of the 24 time series that are included in the dataset for every country.
0,0
0,5
1,0
1,5
2,0
2,5
3,0
0100 200300
Gas consumption [TWh/d]
Time [d]
Actual (from UK National Grid)
0,0
0,5
1,0
1,5
2,0
2,5
3,0
0100 20
03
00
Gas consumption [TWh/d]
Time, sorted [d]
Modelled (from the When2Heat dataset)
Fig. 8 Validation of the heat demand time series for the UK. Modelled UK gas consumption for space and
water heating in 2013 are compared to published data from UK National Grid. Daily time series are displayed
chronologically (le) and sorted (right).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

9
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
 Critical assumptions were made concerning the spatial aggregation. First, using the population
as a proxy for the spatial distribution of the heat demand ignores, e.g., regionally dierent insulation standards
and inhomogeneous per capita living areas. We argue that deriving national time series from population-weighted
spatial time series is nevertheless an improvement over using only one representative temperature time series.
However, the user might integrate diering weighting assumptions with our publicly available code. Note that an
aggregation by other regions, e.g. the European NUTS regions, is also possible through including the correspond-
ing shapeles. Second, the calculation of the COP proles assumes homogeneous diusion within each country
and across building types, which neglects the fact that some heat pumps might match certain applications better
than others.
Concerning the COP time series, the major simplication is the consideration of average values and constant
correction factors. Indeed, the eciency of individual heat pumps depends on various factors and may deviate
signicantly from the mean10,18. Furthermore, note that the COP is related to the operation strategy of the heat
pumps, which might be endogenously determined in the electricity market models. For instance, if heat pumps
are optimized with respect to the electricity price by charging a thermal buer storage during low-price periods,
their eciency will drop due to the higher heat sink temperatures in the buer storage19.

All code is implemented in Python and published at https://github.com/oruhnau/when2heat under the open MIT
license. e entire processing workow is documented in a single Jupyter Notebook (processing.ipynb), which
draws on custom functions that are structured in dierent Python scripts (download.py, read.py, preprocess.py,
demand.py, cop.py, write.py, metadata.py, misc.py).
While the download of weather and population data is automated, the input data from the EU Building
Database, BGW7, and BDEW8, as well as the COP curve parameters are included in the code repository.

1. Pfenninger, S., DeCarolis, J., Hirth, L., Quoilin, S. & Staell, I. e importance of open data and soware: Is energy research lagging
behind? Energy Policy 101, 211–215 (2017).
2. uhnau, O., Banni, S., Otten, S., Pratinjo, A. & obinius, M. Direct or indirect electrication? A review of heat generation and
road transport decarbonisation scenarios for Germany 2050. Energy 166, 989–999 (2019).
3. Quiggin, D. & Buswell, . e implications of heat electrication on national electrical supply-demand balance under published
2050 energy scenarios. Energy 98, 253–270 (2016).
4. Wilson, I. A. G. et al. Historical daily gas and electrical energy flows through Great Britain’s transmission networs and the
decarbonisation of domestic heat. Energy Policy 61, 301–305 (2013).
5. uhnau, O. When2Heat Heating Proles. Open Power System Data, https://doi.org/10.25832/when2heat/2019-08-06 (2019).
6. Wiese, F. et al. Open Power System Data – Frictionless data for electricity system modelling. Applied Energy 236, 401–409 (2019).
7. German Federation of the Gas- and Water Industry (BGW) & German Association of Local Utilities (VU). Anwendung von
Standardlastprolen zur Belieferung nicht-leistungsgemessener unden [Application of Standard Load Proles for Supplying Non-
Daily-Metered Customers]. Practice Information P 2006/8 (BGW, 2006).
0
0,5
1
1,5
2
2,5
3
Estonia
Finland
Sweden
Denmark
Lithuania
Germany
Slovenia
Austria
Czech Republic
Hungary
Bulgaria
Poland
Luxembourg
Romania
France
Ireland
Netherlands
Slovakia
Croatia
Un
ited Kingdom
Belgium
Greece
Italy
Spain
Portugal
Cyprus
Malta
U-value [W/m²K]
Countries explicitly validate
d
Further countries included
Countries excluded
Fig. 9 National average heat transfer coecients (U-values). Dierent EU countries are compared to assess
the transferability of the German gas standard load prole methodology. Countries with values in between or
similar to Germany or UK are included in the dataset.
0
1
2
3
4
5
6
ASHP GSHP WSHP
COP
When2Heat,
uncorrected
When2Heat,
corrected
Günther et al
.,
measured
Fig. 10 Validation of COP time series. Modelled COP average from July 2012 to June 2013 are compared to
eld measurements from Günther et al.10. Air-source heat pumps (ASHP), ground-source heat pumps (GSHP),
and water-source heat pumps (WSHP) are distinguished.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

10
Scientific DATA | (2019) 6:189 | https://doi.org/10.1038/s41597-019-0199-y
www.nature.com/scientificdata
www.nature.com/scientificdata/
8. German Association of Energy and Water Industries (BDEW), German Association of Local Utilities (VU) & European
Association of Loacal Energy Distributors (GEODE). Abwiclung von Standardlastprolen Gas [Execution of Gas Standard Load
Proles]. Guidelines (BDEW, 2015).
9. Stiebel-Eltron. Planung und Installation von Wärmepumpen [Planning and installation of heat pumps]. Planning Manual (Stiebel-
Eltron, 2016).
10. Günther, D., Miara, M., Langnner, ., Helmling, S. & Madani, H. WP Monitor - Feldmessung von Wärmepumpenanlagen [WP
Monitor - Field Measurements of Heat Pumps]. Final Project eport (Fraunhofer ISE, 2014).
11. Berrisford, P. et al . e EA-Interim archive Version 2.0. EA eport Series (European Centre for Medium-ange Weather Forecasts,
ECMWF, 2011).
12. Fischer, D., Wolf, T., Scherer, J. & Wille-Haussmann, B. A stochastic bottom-up model for space heating and domestic hot water load
proles for German households. Energy and Buildings 124, 120–128 (2016).
13. Zottl, A., Nordman, . & Miara, M. Benchmaring method of seasonal performance. SEPEMO-Build Project eport (2012).
14. Fischer, D., Wolf, T., Wapler, J., Hollinger, . & Madani, H. Model-based exibility assessment of a residential heat pump pool.
Energy 118, 853–864 (2017).
15. Nabe, C. et al. Potenziale der Wärmepumpe zum Lastmanagement im Strommart und zur Netzintegration erneuerbarer Energien
[Potential of Heat Pumps for Load Management in the Electricity Maret and for the Grid Integration of enewable Energies].
eport for the German Federal Ministry of Economic Aairs and Energy (Ecofys and Prognos, 2011).
16. Department of Energy & Climate Change (DECC). Estimates of heat use in the U 2013. Energy Trends (DECC, 2014).
17. Bloess, A., Schill, W.-P. & Zerrahn, A. Power-to-heat for renewable energy integration: A review of technologies, modeling
approaches, and exibility potentials. Applied Energy 212, 1611–1626 (2018).
18. Nolting, L., Steiger, S. & Pratinjo, A. Assessing the validity of European labels for energy eciency of heat pumps. Journal of
Building Engineering 18, 476–486 (2018).
19. Nolting, L. & Pratinjo, A. Techno-economic analysis of exible heat pump controls. Applied Energy 238, 1417–1433 (2019).

e authors gratefully acknowledge valuable comments from Jonathan Mühlenpfordt and Robin Beer on the
scripts and the dataset.

O.R. designed the approach, developed the scripts and the dataset, and wrote the manuscript. L.H. and A.P.
supervised the project and reviewed the manuscript.

Competing Interests: e authors declare no competing interests.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
e Creative Commons Public Domain Dedication waiver http://creativecommons.org/publicdomain/zero/1.0/
applies to the metadata les associated with this article.
© e Author(s) 2019
Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com