Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Energy & Buildings 179 (2018) 333–343

Contents lists available at ScienceDirect

Energy & Buildings


journal homepage: www.elsevier.com/locate/enbuild

Comparison of dynamic urban building energy models (UBEM):


Sigmoid energy signature and physical modelling approach
P. Nageler a,∗, A. Koch b, F. Mauthner c, I. Leusbrock c, T. Mach a, C. Hochenauer a, R. Heimrath a
a
Institute of Thermal Engineering, Graz University of Technology, Inffeldgasse 25/B A-8010 Graz, Austria
b
European Institute for Energy Research, Emmy-Noether-Strasse 11 D-76131 Karlsruhe, Germany
c
AEE – Institute for Sustainable Technologies, Feldgasse 19 8200 Gleisdorf, Austria

a r t i c l e i n f o a b s t r a c t

Article history: Thermal energy demand represents over 30% of overall energy consumption and thus is a vital tar-
Received 13 April 2018 get for the decarbonisation of our energy supply. This fact is the reason why dynamic thermal mod-
Revised 10 August 2018
elling of buildings, components and infrastructure has been gaining increasing importance in the con-
Accepted 23 September 2018
text of the transformation of existing energy systems into smart energy systems. For the first time,
Available online 26 September 2018
this study presents a comparison of two major bottom-up approaches to modelling urban neighbour-
Keywords: hoods: a physical modelling approach and a data-driven approach. The physical method is represented
Dynamic urban building energy modelling by a GIS-based automated modelling approach with detailed dynamic building simulation in IDA ICE. In
Data-driven approach the data-driven approach, an energy signature is applied, which uses a non-linear data-driven method.
Physical approach These two methods were validated on the basis of a multi-family house, an office building and a res-
Heating demand forecasting idential area with 34 buildings by means of detailed multi-zone building simulation and measure-
ment data. The simulation results show that both applied approaches are applicable in these cases
within good agreement compared to the measurement data (physical/data-driven: multi-family house
(RN_RMSE(%) = 14.7/9.3; R2 = 0.68/0.87), office building (RN_RMSE(%) = 7.6/5.4; R2 = 0.92/0.96) and resi-
dential area (RN_RMSE(%) = 8.2/4.8; R2 = 0.92/0.97)). Finally, the fields of application of both approaches
are discussed. A major finding here is that the energy signature shows slightly better results at deriving
the load profile when the measured heating demand is present from a previous heating season. Fur-
thermore, the number of buildings does not affect the duration of a simulation because cumulative user
profiles are used. Whereas the simulation duration of the physical approach depends essentially on the
size of the investigated area. The physical approaches have the advantage of being able to include den-
sification or renovation scenarios, demand forecasting and coupled simulations of buildings and smart
energy systems of neighbourhoods.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction An analysis of the building energy demand is thus essential for


city planners to analyse the energy demand of the actual building
International energy and climate policies are aiming at a re- stock or evaluate restructuring measures, densifications of urban
duction of global energy demand and carbon dioxide emissions development areas or the expansion potential of district heating
to limit global warming [1]. Two-thirds of the world’s energy de- networks (DHN). The underlying approaches, challenges, and op-
mand and 70 % of global CO2 emissions occur in cities [2], and en- portunities for urban energy system models have been presented
ergy needs continue to increase through urbanization (from 54 % by Keirstead et al. [5] and Kavgic et al. [6]. Reinhart and Davila
in 2014 to 66 % in 2050 [3]). Buildings need a large portion of the [7] published a review of urban building energy modelling (UBEM)
world‘s energy consumption. For example, European buildings ac- tools. Two major bottom-up UBEM approaches exist in the litera-
count for approximately 37 % of the total final energy demand [4]. ture.
Retrofitting the existing building stock and integrating the build- The first group are UBEM with physical modelling approaches,
ings in future energy systems are therefore key strategies for cities which can be used to evaluate the current building stock as well as
to reduce energy demand and greenhouse gas emissions. the retrofit or densification scenarios. Most of the physical UBEM
use model simplification procedures, such as resistance and capac-
ity models [8–12], model order reduction methods [13], or simpli-

Corresponding author. fications regarding their temporal resolution, such as steady-state
E-mail address: peter.nageler@tugraz.at (P. Nageler).

https://doi.org/10.1016/j.enbuild.2018.09.034
0378-7788/© 2018 Elsevier B.V. All rights reserved.
334 P. Nageler et al. / Energy & Buildings 179 (2018) 333–343

family of models and their application is provided by Day [37] and


Nomenclature ASHRAE’s inverse modelling toolkit [38]. Energy signatures can be
further divided into linear [33,38,39] and non-linear [40,41] re-
Abbreviations gression models. At the scale of neighbourhoods, energy signatures
DB database were more recently used by Raffio et al. [42] as well as Mazzarella
DHN district heating network et al. [43].
DHW domestic hot water To the best of the authors’ knowledge, this study presents a
GIS geographic information system comparison of two major UBEM approaches for the first time. The
IDA ICE simulation tool from EQUA: IDA Indoor Climate study points out the advantages and disadvantages of a data-driven
and Energy non-linear energy signature in comparison to a GIS-based physi-
IP inflection point cal approach using the detailed dynamic building simulation tool
MFH multi-family house IDA ICE [44]. By statistical validation with measured data the er-
NSB night setback of the heating system rors of both approaches are quantified. Furthermore, the simula-
SFH single family house tion results are compared against a detailed multi-zone building
RC model resistance and capacity model simulation in IDA ICE. This comparison and validation helps other
UBEM urban building energy model scientists to choose the right approach for the present task and to
assess possible uncertainties. The first part of the study introduces
Symbols, unit
the case study, which consists of a multi-family house (MFH), an
A area, m2
office building and a residential area with 34 buildings. The sec-
CVRMSE coefficient of variance of root mean squared error,
ond part presents the two UBEM approaches in detail. In the sec-
%
ond part, the simulation results of the case study are presented.
FD weekday factor
Finally, the fields of application of both approaches are discussed.
GFA gross floor area, m2
h normalised daily energy demand, kWh
MBE mean bias error, % 2. Materials and methods
n total number of data points in the dataset
Qd de-normalized daily heating demand, kWh This section introduces the case study with the study area and
QMD measured heating demand, kWh provides relevant information concerning the measurement data
RN_RMSE range normalized root mean squared error, % acquisition and applied metrics for validation purposes. Subse-
R2 coefficient of determination quently, the section presents a physical method and a data-driven
V volume, m3 modelling approach to model UBEM. First, a GIS-based physical ap-
Y experimental data proach is presented with detailed dynamic building simulation in
Y
ˆ model output IDA ICE. Afterwards, a non-linear energy signature is introduced.
ϑa calculated temperature of the day, °C
ϑ0 40 °C (point of discontinuity), °C 2.1. Study area
ϑt measured outdoor temperature of the current day,
°C The two UBEM approaches are tested on the basis of a case
μ mean value of measurement data, kWh study. The investigated area is a part of the building stock of a
small town in the Eastern Styria, Austria. The town is a typical Cen-
tral European small town with terraced houses in the town centre
models [14]. In the literature, few UBEM use detailed building sim-
and many detached houses on the outskirts. The case study is di-
ulation tools (EnergyPlus [15–18], Dymola [19] and IDA ICE [20]) to
vided into two cases.
ensure a temporal and spatial resolution in the desired level of de-
tail. Building archetypes [21,22] are used in these methods to ob-
tain the mostly unknown data about the building envelope. Tools 2.1.1. Individual buildings
and methods for automated modelling are therefore essential to In the first case, two buildings with different building usage,
shorten the time consuming modelling effort [23–26]. which are a MFH and an office building (see Fig. 1), are examined.
The second group are data-driven UBEM, which use a mathe- The MFH was built in 2011 and consists of 11 residential units with
matical description of the system based on measurement data [27]. a total gross floor area (GFA) of 1020 m2 . The office building was
Data-driven models can be applied to predict operational changes built in 1966 and has 1083 m2 GFA. The two methods were val-
or operational problems [18], to derive temporal courses of the idated on the basis of these buildings with the energy supplier’s
heating load from annual energy billing or to predict the energy measurement data and a detailed building simulation in IDA ICE
demand of whole cities based on training data [28,29]. However, it (see Section 2.2).
is difficult to evaluate retrofit or densification scenarios using data-
driven models [18]. The big advantage is that data-driven mod- 2.1.2. Residential area
els are less complicated in design and more robust in simulation In the second case, the two UBEM were applied to a typically
than physical models [30]. The empirical or black box approach residential area with 34 buildings, which consists of 30 single fam-
is widespread among the data driven modelling approaches [31], ily houses (SFH) and four MFH (see Fig. 2). The SFH of the residen-
which defines the input and output relation without describing tial area can be characterized by the following mean values: the
the physical properties. A more sophisticated approach is that of mean building heated gross floor area is 275 m2 , the mean build-
grey box models, also known as energy signatures, which consider ing height is 7.1 m and the average building age is 38 years. The
building archetypes and the dependency of heating demand and buildings are typically aligned to the southwest. A DHN supplies
environmental variables such as the outdoor temperature [32,33]. the residential area with heat for domestic hot water (DHW) and
Energy signatures can be applied to predict the temporal course of space heating. The DHN water supply temperature at the substa-
the heating and cooling load based on climate data and the annual tions is set in the simulation to 85 C and the heat exchanger at the
heating demand [34,35]. In this field of application, energy signa- substation is dimensioned with a temperature decrease with 30 K
ture models were first used by Jacobsen [36]. An overview of this (mean supply/return temperature at the supply unit is 88/60 °C).
P. Nageler et al. / Energy & Buildings 179 (2018) 333–343 335

Fig. 1. Study area (detailed multi-zone building model in IDA ICE) of the first case: (a) multi-family house and (b) office building.

tion plans and wall structures were achieved from the official en-
ergy certificates of the individual buildings. The evaluation interval
is set to 1.1.2015 till 31.12.2015 with hourly resolution. Both UBEM
approaches and the detailed multi-zone building simulation were
calibrated to the annual heating demand in order to achieve better
comparability between the approaches, but the dynamic heating
load was not calibrated in any approach. Only the existing control
strategies or heating parameters were implemented, such as night
reduction of the heating. The physical models were calibrated by
variation of the infiltration rate. This means that only the heat-
ing load prediction and not the annual heating demand forecast is
evaluated in the physical approach. In the second case, the cumu-
lative load profile of an entire residential area is compared with
measured data from the energy supply unit. This load profile in-
cludes consequently the heat load from the buildings and the heat
losses of the piping to the environment. The residential area is as-
sembled of 30 SFH and 4 MFH. The building usage and the year
of construction are provided by the utilities services. To reflect a
realistic application case, climate corrected measurements of the
previous year were used for the residential area and compared to
the actual measurements. In this case, the UBEM approaches use
the climate corrected heating demand from the last heating season
(1.1.2016 to 31.12.2016) to calibrate the model and the evaluation
interval is for both UBEM approaches from 1.1.2017 to 31.12.2017
Fig. 2. Study area of the residential area refers to the second case; 30 single family with hourly resolution. The building height is obtained from an
houses are depicted orange and four multi-family houses blue (For interpretation elevation model by airborne laser scanning [47] and the weather
of the references to color in this figure legend, the reader is referred to the web
data from the ZAMG are used [48].
version of this article.).
It is common practice to use metrics to validate simulation
models in order to better evaluate failures in the simulation [49–
51]. In all metrics described below, the following nomenclature
2.2. Measurement data, validation and applied metrics applies: n is the total number of data points in the dataset, Yk
indicates experimental data and Y ˆ k indicates model output and
Data on buildings and district heating infrastructure for this μ is the mean value of the measurement data. In a recent re-
case study were supplied by the utility services [45]. The UBEM search, Chakraborty and Elzarka [51] suggested that range normal-
approaches are validated with real measurement data derived from ized root mean squared error (RN _RMSE) should be used as the
the utilities’ database [45], which operate the DHN and collect the primary metric for validation of energy models instead of variance
measurement data for heat demand at the substations. The vali- of root mean squared error (CVRMSE). The RN _RMSE normalizes
dation relies on the evaluation of heating load at the DHN sub- the root mean squared error with the range of the dataset allowing
stations. Unfortunately, the case study provides no measurement the comparison by using the RN _RMSE for different data sets. This
data about cooling, which limits the evaluation to the heating per- metric is defined according to Eq. (1) and evaluates the accuracy
formance. The heat is measured with gauged heat meters. The of models, whereas a lower value is desirable. Coefficient of deter-
first case validates simulation results of both UBEM approaches mination (R2 ) should be used as a secondary metric for validation
with measurement data. Furthermore, the results are also com- of energy models. R2 specifies how well the regression estimate
pared with detailed multi-zone building simulation in IDA ICE, be- fits the data, see Eq. (2). Due to more frequent usage, more com-
cause such comparison is also relevant in the case of individual mon metrics such as CVRMSE and mean bias error (MBE) are also
buildings (MFH and office building). The detailed building models used, see Eqs. (3) and (4). All metrics are applied to the results for
in IDA ICE consist of 51 thermal zones for the MFH and 37 ther- daily heating demand in all calculations. Furthermore, the tempo-
mal zones for the office building, see Fig. 1. Internal loads and set ral course of the daily heating demand is displayed over a year;
points were chosen from room usage data for energy and building the hourly heating load is presented over one week in the winter
technology [46]. Geometry data was obtained from the construc-
336 P. Nageler et al. / Energy & Buildings 179 (2018) 333–343

Fig. 3. Workflow of the physical UBEM approach.

period for detailed comparison and evaluation (see Chapter 3). TRNSYS or MATLAB. The DB system consists of a project-related
 DB and a reusable typology DB, which are both object-relational
(Yk −Yˆ k )
n 2
k=1 PostgreSQL DB [53]. The typology database stores non-geometrical
n
RN _RMSE (% ) = × 100 (1) project independent data, which are for example wall construc-
maximum(Y ) − minimum(Y )
tions for typical buildings or load profiles for DHW [52] and inter-
n  2 nal loads [46]. These data are reusable for the next project (D). All
2 k=1 Yk − Y
ˆk
R =1− n (2) non-geometric data of a building is linked in the so-called build-
k=1 (Yk − μ )2 ing type, which are data of the wall- and window constructions,
 the building usage, the percentage of glazing, the specific heat de-
k=1 (Yk −Yk )
n 2
ˆ
mand and peak power for space heating. The desired building type
n
CV RMSE (% ) = × 100 (3) can be selected from a list of building types. The building age and
μ
the building usage is an indicator for the energy demand in build-
n   ings. The building types were thus created according to both the
k=1 Yk − Y
ˆk
MBE (% ) = n × 100 (4) building usage and the building age. So-called archetypes were cre-
k=1 Yk ated, which can be subdivided, for example, for Austrian residential
buildings into the following eight building age classes [21]. Data
2.3. UBEM approaches that has not yet been implemented in the database can be added
and new building types can be created from these data (E). The
This section introduces the two major UBEM approaches in de- project database stores the relevant project data with a spatial ref-
tail. First, a GIS-based physical approach is presented with dynamic erence, which are the required input data (A), the assigned data
building simulation using IDA ICE. Thereafter, an energy signature (B) by means of a categorization/dimensioning process (2) and the
is introduced, which is a non-linear data-driven approach. dynamic simulation results (C). This means that the same typology
database is reusable while the project database is created for each
2.3.1. Physical approach: GIS-based automated dynamic thermal city new project or project version. In addition to the DB, a file system
modelling using IDA ICE is provided, in which the source code of the models is saved as
The physical modelling of entire neighbourhoods proves to be text files in order to make the results reproducible.
extremely time-consuming, because every single building has to The workflow of the UBEM approach is divided into the follow-
be geometrically modelled and parameterized with wall construc- ing five process steps (see Fig. 3). The first step 1) deals with data
tions, heating systems and internal gains. An automation process acquisition. This UBEM approach uses publicly available data. The
is required to shorten the considerable time needed for manual method is based on only three kinds of data, which are the build-
input. This study thus uses a validated method for automated ing geometry, the land utilization plan and the building age. The
building modelling, which has been developed by Nageler et al. building geometry is composed of the ground plan and the height.
[20] in order to minimize the considerable time required to model The ground plan can be obtained through Open Data such as Open-
the building stock. The approach can be integrated as a planning StreetMap [54] or basemap.at [55] or through cadastral plans. Laser
tool into an urban development process, see dashed/dotted line in scanning data [56,57] provide an opportunity to obtain the build-
Fig. 3. The planning development process begins with the develop- ing height. QGIS [58] is chosen as an open source GIS tool to ma-
ment of possible action or planning options, see Fig. 3. The energy- nipulate, visualize and analyse the collected data. In the second
technical definition of a planning option is called a scenario. step (2) the buildings are categorized and the heating system is
First, a database (DB) system is created, which allows the user dimensioned. A categorization process assigns the buildings a use,
to manage multiple sources for these data sets and make them a building type and a heating type. This assignment is made by
accessible to various simulation tools such as DYMOLA, IDA ICE,
P. Nageler et al. / Energy & Buildings 179 (2018) 333–343 337

means of the building geometry, the land utilization plan and the Table 1
Parameters sets (A, B, C and D) for office buildings and
building age. Thereafter, the heating system is dimensioned due
multi-family houses (MFH) / residential areas.
to specific heating loads, which are set in the building types. The
heat is provided by one central heat generation unit or with a so- Office building [35] MFH / residential area [27]
lar thermal plant combination. In this approach, statistical assign- A 2.717 2.059
ments are made on various levels in order to make basic assump- B −35.141 −39.762
tions, e.g. about the refurbishment rate. This kind of data can thus C 7.130 5.384
D 0.0 0 0 0.070
be assigned through a compound interest calculation using a char-
acteristic refurbishment rate of the investigated area. This means
older buildings are more likely to be considered refurbished. It is
ing. The normalized daily energy demand h is calculated by using
important to note, that every assumption can be overridden manu-
Eq. (5) and depends on the calculated temperature of the day ϑa .
ally for each building, if more information is available. In the third
The temperature ϑ0 describes the point of discontinuity with tem-
step (3), a building model is automatically generated from the in-
perature equals 40 °C.
put data and the assigned data. This method uses a tool that gener-
ates detailed dynamic building models automatically [23]. The tool A
h ( ϑa ) =  C + D (5)
is an interface between the PostgreSQL DB and the dynamic build- 1 + ϑa −B ϑ
0
ing simulation environment IDA ICE [44]. Each building consists of
a 2.5D geometry based on the ground plan and the mean build- Fig. 4 shows variations of each individual parameter based on
ing height. The zoning of the building is carried out by up to three a typical SFH (Fig. 4: blue line). The parameter A shifts the curve
vertically layered thermal zones and depends on the number of (Fig. 4: broken orange line) vertically at low outdoor temperatures,
different uses. The building model generation tool also generates whereas the parameter B shifts the curve (Fig. 4: dashed/dotted
and parameterizes a detailed heating system for each building. The yellow line) horizontally at the inflection point (IP). The param-
data used for parameterization is stored in the DB. The method eter C changes the slope of the inflection tangent (Fig. 4: dotted
uses predefined and scalable heating systems for each heating type magenta line). The curve flattens off at low outside temperatures
(e.g. water tanks, boilers, ambient air and ground heat pumps, so- due to a changed user behavior. The window ventilation at cold
lar thermal power, photovoltaic and wind turbine power, district temperatures decreases sharply, thus the heating demand does not
heating connection or heat distribution systems). This avoids errors always increase linearly. These parameters can be seen as corre-
during the modelling process. In the fourth step (4) the automat- sponding to the performance and the building type while param-
ically created model is simulated with the dynamic building sim- eter D represents the non-temperature related part of energy de-
ulation environment IDA ICE, which allows automated modelling mand such as the DHW and shifts the curve (Fig. 4: green line)
through scripting and the simulation process is parallelizable. IDA vertically.
ICE, which uses balance equations for CO2, humidity and energy in Table 1 shows a parameter comparison for an office building
each thermal zones, is a whole-year detailed and dynamic multi- and a MFH /residential area, which area also applied in the case
zone simulation tool for study of thermal indoor climate as well study. The parameters for the office building were extracted for dif-
as the energy consumption of the entire building. The simulation ferent building uses using sampling data by minimizing error vari-
tool considers the mutual shading between buildings and also the ance [35]. Parameter sets were developed by Koch [27] for the ap-
shading of the terrain, but the microclimate is not considered. In plication at the scale of urban neighbourhoods and MFH. The non-
the fifth step ((5), the results are graphically prepared for further temperature related part of the energy demand (DHW) increases
analysis and evaluation or as input for discussion with stakeholders in MFH and residential areas due to an increasing D parameter.
and decision makers, which are for example municipalities, energy In order to represent dynamic effects in the building‘s heating
suppliers, urban planners and investors. QGIS is used to create heat demand such as the thermal inertia of the building, a calculated
maps, 3D visualizations or spatio-temporal animations of building temperature smooths the temporal course of the daily heating de-
(thermal) energy demands. mand. This calculated temperature ϑa is defined as the measured
Finally, the results are evaluated, e.g. by comparing the results outdoor temperature of the current day ϑt as well as the last three
with previously set benchmarks like CO2 reduction goals or in days ϑt−1 , ϑt−2 and ϑt−3 with weighting factors, as can be seen in
discussion with respective decision makers. The evaluation deter- Eq. (6) [35].
mines whether the results for this scenario meet the requirements 1 · ϑt + 0.5 · ϑt−1 + 0.25 · ϑt−2 + 0.125 · ϑt−3
ϑa = (6)
and benchmarks. In this case the process is finished. If the result 1 + 0.5 + 0.25 + 0.125
is considered insufficient, the process starts again. Eq. (7) provides the de-normalized daily heating demand Qd ,
which is calculated by using the measured heating demand QMD
2.3.2. Data driven approach: non-linear energy signature and the normalized daily heating demand. The measured heating
Energy signature models are regression functions that are used demand of a specific period is divided by the cumulated normal-
in various forms. For the assessment of energy performance of ized daily heating demand of this period. The cyclical behaviour
buildings, linear functions are used with single or multiple inflec- of the building user is taken into account by means of a weekday
tion points as well as non-linear functions such as the sigmoid factor FD . Geiger and Hellwig [35] described these factors, which
model described here. For the simulation of heating or cooling especially for non-domestic users, show a distinct pattern with the
needs, outdoor temperature measurements are often used as a sin- highest modifications on weekends.
gle regressor. The energy signature model used in this compari-
QMD
son was initially developed at the TU Munich by Geiger and Hell- Q d ( ϑ a ) = n · h(ϑa ) · FD (7)
wig [35] for the gas load prediction at the scale of gas distribu- 1=1 h (ϑa ) · FD
tion zones, Eichlseder [59] adapted the sigmoid function for the To generate an hourly load profile from the calculated daily
Austrian context. At the scale of groups of buildings and urban ar- heating demand, statistical hourly load profiles can be used. In the
eas the model was evaluated and adapted by Koch [27]. The sin- case study these profiles depend on the outdoor temperature and
gle variant model is based on a geometric series of the outdoor the building use. At a larger scale, site-specific load profiles can
temperature and uses a set of four parameters (A, B, C and D), be used for small building clusters or aggregated load profiles for
which are defined depending on the use and the age of the build- neighbourhoods [60].
338 P. Nageler et al. / Energy & Buildings 179 (2018) 333–343

Fig. 4. Sigmoid function with different parameter variations.

In this case study, the model was applied for the MFH as also ulation and energy signature, but the physical UBEM approach is
the case of the office building using the annual sum of the mea- with 40.2 % above this value. Only the heating load peak at the
sured heating demand of the actual period and of the previous beginning of October is not recorded in the detailed building sim-
year for the residential area. Furthermore, not every building in the ulation. The reason is an overestimation of the storage mass in the
settlement area is simulated individually, but the entire residential building that increases the capacity of the building, which has a
area is calculated with the cumulated heating demand. room temperature-compensating effect. This also explains the re-
duced alternating part of the heating load compared to the mea-
sured data. This effect is also evident in the physical UBEM ap-
3. Results proach (see Fig. 5: orange line). Another effect shortens the heat-
ing period, because the simulation model only uses one thermal
This section presents the results for the two cases as described zone for the building. This approximation results in an underes-
in Section 2.1. It is important to consider that UBEM approaches timation of annual heating demand, because the simulation tool
and the detailed multi-zone building simulation were calibrated treats a thermal zone as a node within a network with perfectly
to the annual heating demand in order to achieve better compa- mixed air volumes [26]. Therefore, possible local energy surpluses
rability between the approaches. In the first case, the individual due to solar irradiation in south-facing rooms that may occur dur-
buildings are examined, which are the MFH and the office building ing winter are absorbed with underserved north facing rooms. This
(see Fig. 1). In the second case, the two UBEM were applied to a aspect becomes more important with larger buildings [20]. This ef-
residential area with 34 buildings (see Fig. 2). The graphical rep- fect can amount to 10 % of the annual heating demand in multi-
resentation of the simulation results is always carried out in sub- family houses. The calibration with the air change causes a vertical
plots in the same way in all cases: (a) First, the temporal course of offset in the heating demand curve with an approximately constant
the daily heating demand is displayed over a year, (b) Thereafter, heating period. This explains the higher peak demands compared
the daily heating demand is displayed as a function of the mean to the measurement data. The summer base load, which is caused
outdoor air temperature, (c) Then, the hourly heating load of one by the domestic hot water demand, is properly covered by both
week is shown and d) Finally, the applied metrics are illustrated in methods.
a bar chart. Subplot Fig. 5b shows that the fluctuation range of the heating
demand as a function of the outside air temperature in the en-
3.1. Result of the individual buildings ergy signature is significantly lower and in the detailed multi-zone
building simulation is slightly too low. This means that the thermal
The heating demand of the two UBEM approaches were vali- inertias have not been sufficiently considered. The opposite effect
dated and compared with measurement data and detailed multi- is evident in the physical UBEM approach.
zone building simulation in IDA ICE (see Section 2.2). Fig. 5 shows The weekly profile of the hourly heating load at the beginning
the analysis of the simulation results of the MFH. The trajectories of January is illustrated in the subplot of Fig. 5c. Here it can be
of the daily heating demand is in a good agreement with the mea- seen that the heating load profile of the measured data agrees well
surement data for both approaches and the detailed building sim- with the detailed building simulation. The night setback and the
ulation, see subplot Fig. 5.a. The figure shows a very good agree- subsequent start of the heating phase and also a reduction in the
ment of the energy signature (see Fig. 5: yellow line) with the de- room air temperature during noon hours can be well represented
tailed building simulation (see Fig. 5: blue line) and the measured with the detailed building simulation. The simulation results of the
data (see Fig. 5: violet line). The simulation results of all applied energy signature show a smoothed profile, in which too much heat
methods are below a RN _RMSE value of 14.7 %, a CVRMSE value of demand during the night and too little during the day is predicted.
40.2 %, above a R2 of 0.68 and a MBE value of 0 %, because of cali- This can be explained by the larger number of users, for which the
bration with the annual heating demand. A R2 of 0.68 means that hourly load curves were initially developed. The opposite effect is
68 percent of the variance in the measurement data set can be evident in the physical UBEM approach. The reason for this also
explained by the modelled values. The ASHRAE guideline 14 sug- lies in the assumption that one thermal zone represents a large
gests that the CVRMSE value for calibrated energy models must building. As a result, the load profile is not smoothed by switching
be less than 30 % and the MBE must be within ± 10 % [50]. This on/off of heating at different times and in individual rooms.
CVRMSE limit can be met by the detailed multi-zone IDA ICE sim-
P. Nageler et al. / Energy & Buildings 179 (2018) 333–343 339

Fig. 5. Evaluation of the simulation results in the multi-family house: (a) temporal course of the daily energy demand, (b) daily energy demand depending on the outdoor
temperature, (c) heating load profile of one week and (d) evaluation metrics.

Fig. 6 illustrates that the simulation results of the office build- surement data for both approaches and the detailed building sim-
ing are in very good agreement for both UBEM approaches with ulation, see subplot Fig. 7a. The simulation results of both applied
the measurement data and the results of the detailed multi-zone UBEM approaches are below a RN _RMSE value of 8.4 %, a CVRMSE
building simulation. Fig. 6a shows that all temporal courses are value of 21.4 % and above a R2 of 0.92. The MBE value is 2.1 %, be-
very similar and are in consensus with the mean outdoor tem- cause the UBEM approaches are calibrated to the climate-corrected
perature. The reason for this is that the user behaviour in office annual heat demand from the last heating season.
buildings is better predictable and structured, leading to a better Fig. 7b shows that the physical UBEM approach (Fig. 7b: orange
match between the load profile of the simulation results and the line) at low outside air temperatures overestimates the heating de-
measurement results than in the MFH. The simulation results of all mand and underestimates the heating demand at higher tempera-
applied methods are below 7.6 % RN _RMSE, below 24.9 % CVRMSE, tures in the range of about 3 to 15 °C. The reason is again that the
a MBE value of 0.0 % (calibration) and above 0.92 R2 . The energy simulation model only uses one thermal zone for the whole build-
signature lies closest to the measurement data with a RN _RMSE of ing.
5.4 %, a CVRMSE of 17.6 % and a R2 of 0.96. The summer base load Fig. 7c shows a good agreement of the hourly heating load pro-
drops to almost zero except for a few cold days, since office build- file. The load peaks of the day profile, which are present in SFH,
ings require almost no hot water. are smoothed out in residential areas by time-shifted switching on
The fluctuation ranges of the heating demand as a function / off of heating.
of the outside air temperature is in both UBEM approaches and The physical UBEM approach uses archetypes that depend on
the detailed multi-zone building simulation in agreement with the the refurbishment status of the building. This explains a relatively
measurement data, which is shown in Fig. 6b. This means that the large spread of the simulation results of a single building. Thus,
dynamic effects in all approaches are correctly integrated. some buildings are simulated with modelled refurbishment, but in
Subplot c in Fig. 6 also shows a good agreement of the hourly reality they are not refurbished and vice versa. Furthermore, the
heating load profile. The heating reduction at the weekend (3 and extent of a refurbishment is always different and detailed data on
4 January) due to setting down of temperature is also apparent in the state frequently not available. In the proposed physical method,
the plot. Only the load profile of the energy signature (Fig. 6c: yel- an advanced refurbishment of a building with a year of construc-
low line) shows a spread between day and night that is too low. tion before 1919 causes a decrease of the overall heat transfer co-
Furthermore, the energy signature shows a temporal offset of the efficient of walls from 1.4 to 0.12 W (m2 K)−1 due to additional
heating system switch-on points in the morning. insulation [21].

3.2. Results for the residential area 4. Discussion

Fig. 7 shows the evaluation of the simulation results for the This section discusses relevant results and presents the applica-
residential area. In this case, the simulation results of the UBEM tion areas and import features of physical and data-driven UBEM
approaches are not compared to detailed multi-zone building sim- approaches. The results from Chapter 3 show that the accuracy of
ulation due to the enormous modelling effort. The trajectories of the estate level tends to be better than at the building level. The
the daily heating demand are in good agreement with the mea- biggest uncertainty factor in the building simulation is the user
340 P. Nageler et al. / Energy & Buildings 179 (2018) 333–343

Fig. 6. Evaluation of the simulation results in the office building: (a) temporal course of the daily energy demand, (b) daily energy demand depending on the outdoor
temperature, (c) heating load profile of one week and (d) evaluation metrics.

Fig. 7. Evaluation of the simulation results in the residential area: (a) temporal course of the daily energy demand, (b) daily energy demand depending on the outdoor
temperature, (c) heating load profile of one week and (d) evaluation metrics.
P. Nageler et al. / Energy & Buildings 179 (2018) 333–343 341

Table 2
Fields of application of physical and data-driven UBEM approaches.

Fields of application physical approach data-driven approach



Densification scenarios ✗

Retrofit scenarios ✗

Demand forecasting ✗
√ √
Temporal course of heating or cooling load

Derive temporal courses from annual demands ✗

Energy demand of whole cities based on training data ✗
model the dynamic interactions between buildings and

smart energy systems in districts ✗

Building modelling at any level of detail ✗
√ √
Modelling of urban development areas

Modelling of building‘s heating system ✗
√ √
Climate change scenarios
√ √
Operational changes

Thermal comfort analysis ✗

behaviour. On the one hand, user behaviour influences the inter- requires the simulation of the building‘s heating system, which is
nal loads such as the heat input by people, lighting and equip- coupled with smart energy systems at urban scale such as a com-
ment. On the other hand, the ventilation behaviour and the set bination of a low temperature DHN with the integration of fluc-
point adjustment of the heating and cooling system is also influ- tuating renewables and seasonal storage [65]. The dynamic inter-
enced by the user behaviour. This aspect coincides with the find- actions between buildings and smart energy systems can thus be
ings from the literature. Schnieders et al. [61] compared measure- evaluated. As a result, it is possible to investigate operational opti-
ment data of four low-energy residential areas. The results show mizations of power supply units or special operating cases such as
a log-normal distribution of the specific heating demand of build- failure or demand undersupply scenarios or the starting behaviour
ings with same construction, which can only be explained by the of power supply units with temperature waves in the DHN. In ad-
user behaviour. This influence smoothes with increasing number of dition, thermal comfort can be assessed by physical approaches.
users. This smoothing effect is in agreement with Koch [27], who This means that, for example, analyses can be carried out with re-
identified that the correlation between simulation and measure- gard to the room air temperature or the air quality.
ments increases with the number of aggregated buildings. With the availability of more intelligent metering devices, smart
In the case study, only the heating demand was considered, home applications and improved individual billing solutions, en-
as there were no measurements for the cooling demand. Nev- ergy demand data is becoming ever more fully and extensively
ertheless, both methods can be applied to simulate the heat- available for a large part of the building stock. The family of sta-
ing and cooling performance, but the results include more un- tistical models described can be used to predict the energy de-
certainty factors. Here, the user behaviour also has a significant mand of whole cities or smaller urban areas using historic en-
influence on the cooling requirements. Furthermore, in contrast ergy demand data in combination with readily available environ-
to the heating system, it is not clear which building or parts of mental data (i.e. outdoor temperature) and limited information on
the building are cooled. The UBEM approaches are thus only valid the building use. The presented method in Section 2.3.2 has not
within a wider observation horizon for cooling, because the above- been developed for this kind of application, but the results support
mentioned problems are smoothed through statistics. Energy sig- the assumed scalability from large regional areas to neighbourhood
natures are applied to simulate the cooling demand in communi- scale. The application was based on a parameter set developed
ties, which mostly calculates the electricity demand of the cooling based on building samples in Germany, local training data would
device [62,63]. thus not always appear to be necessary for the applications shown.
The case study includes only residential buildings and one of- A big advantage of energy signatures is that the temporal course
fice building. However, the methods can also be applied to build- of heating and cooling loads of buildings or neighbourhoods can
ings with other uses by adaption of the parameters of the energy be easily derived from an annual energy billing. If annual energy
signature (see Section 2.3.2) [27] and choosing other profiles for billing is not available, e.g. in urban development areas, the spe-
internal loads in the physical building model [20]. cific heating demand by building age group and building use can
Table 2. shows the application areas of physical and data-driven be taken from the statistics. Physical approaches are more complex
UBEM approaches, which are commonly used at urban scale. Those and thus require a detailed modelling of the buildings and suit-
fields of application, which cannot be covered by means of the able modelling tools. Heating loads may be of interest to energy
two presented approaches in Chapter 2, are explicitly pointed out. suppliers, for example to design DHN according to the maximum
The strength of physical approaches is that building physical sce- load and the base load, or to decide whether a DHN is economical.
narios can be simulated. This allows the evaluation of retrofit sce- Both approaches show a good agreement with the measurement
narios by changing the material properties and thicknesses of the data with respect to the mean daily energy demand and the load
wall constructions. Densification scenarios can also be simulated profile during one day, see Sections 3.1 and 3.2. A further appli-
by adapting the building geometry of existing buildings or by inte- cation case can be the development of continuous benchmarks for
grating new buildings into the building stock. Furthermore, energy monitoring purposes, as the calculation requires little knowledge
demand forecasts can be simulated using planning materials from in energy modelling.
buildings. This allows the evaluation of urban development areas, Operational changes can be considered in both approaches,
in which new buildings are added to the building stock. Another which are e.g. changes in the building usage or changes regard-
advantage is that the new challenges of the transformation of city ing the heating system control such as a reduction of the heating
energy systems, as described by Lund et al. [64], can be assessed. demand because of a night set back in a residential building.
Such simulation tools can be used to serve as a decision-support After the fields of application have been described, several im-
tool for city planners or investors during the planning phase. This portant features of the UBEM approaches are discussed, such as the
342 P. Nageler et al. / Energy & Buildings 179 (2018) 333–343

Table 3
Properties and abilities of physical and data-driven UBEM approaches.

Properties and abilities Physical approach Data-driven approach



Co-simulation capable ✗

Fast resource-saving simulation ✗

No annual energy billing necessary ✗
No climate data necessary ✗ ✗

Physical properties of buildings are not required ✗

Various types of heating systems considered ✗

simulation duration, the necessary input data or interfaces to other measurement data. Finally, the fields of application of both ap-
simulation tools. The properties and abilities of the two UBEM proaches are discussed. Energy signatures are more suitable for
approaches are listed in Table 3, in which a green check mark calculating the load profile when the measured heating demand is
indicates a positive result. The major drawback of the physical present during a heating season and the physical UBEM approaches
UBEM approaches is the computationally expensive simulation and have the advantage of simulating densification or renovation sce-
the time-consuming collection of building data such as geometry, narios, demand forecasting and coupled simulations of buildings
physical properties, uses, etc. Physical approaches offer the possi- and smart energy systems of neighbourhoods are possible.
bility that they can be coupled with other simulation tools. Co-
simulation provides the opportunity to couple the best and most Acknowledgments
computationally efficient tool for each subtask. In this context, co-
simulation is an application that employs at least two simulation This research was supported / funded by the Austrian Fed-
tools to solve differential equations that are coupled to exchange eral Ministry of Transport, Innovation and Technology (bmvit) and
data depending on state variables [66,67]. In addition, the model is associated with the project "EnergyCityConcepts - method and
libraries of individual tools (e.g., IDA ICE, EnergyPlus, and Dymola) concept development for the implementation of sustainable energy
can be linked and so a large repertoire of models can be obtained systems in cities and communities” (project duration from Febru-
without even having to create new models. Co-simulation can help ary 2016 to February 2019, FFG project number 850129).
to model smart energy systems in cities, taking into account vari- This research was also supported by the Stadtwerke Gleisdorf,
ous heating system types in buildings [68]. the city of Salzburg and the Salzburg AG.
In general it can be stated that the data acquisition is a ma-
jor problem in UBEM. To simplify data acquisition, statistics and References
archetypes were used in the physical approach to derive all phys-
ical building properties based on 3 data sources, which are gener- [1] UN Climate Change Conference Paris, 2015. Online]. Available Accessed 31-
ally well-available. The data-driven approaches, on the other hand, August -2017 http://www.un.org/sustainabledevelopment/cop21/.
[2] International Energy Agency, World Energy, Outlook, 2008 2008.
only require the annual energy billing and the outdoor air temper- [3] United Nations, Department of Economic and Social Affairs, World Urbaniza-
ature as the environment variable. If this data is available, a fast tion Prospects, 2014.
resource-saving simulation is possible, which requires no physical [4] L Pérez-Lombard, J Ortiz, C Pout, A review on buildings energy consumption
information, Energy Build 40 (3) (2008) 394–398.
properties of the building. [5] J Keirstead, M Jennings, A Sivakumar, A review of urban energy system mod-
els: approaches, challenges and opportunities, Renew Sustain. Energy Rev. 16
(6) (2012) 3847–3866.
5. Conclusion [6] M Kavgic, A Mavrogianni, D Mumovic, A Summerfield, Z Stevanovic,
M Djurovic-Petrovic, A review of bottom-up building stock models for energy
consumption in the residential sector, Build Env. 45 (7) (2010) 1683–1697.
This study presents a comparison of two models represent- [7] CF Reinhart, CC Davila, Urban building energy modeling – a review of a
ing the major UBEM approaches, which are a data-driven non- nascent field, Build Env. 97 (2016) 196–202.
linear energy signature and a GIS-based physical approach using [8] JA Fonseca, A Schlueter, Integrated model for characterization of spatiotempo-
ral building energy consumption patterns in neighborhoods and city districts,
the dynamic building simulation tool IDA ICE. The two presented
Appl. Energy 142 (2015) 247–265.
UBEM approaches differ in that energy signatures use annual en- [9] B Morille, N Lauzet, M Musy, SOLENE-microclimate: a tool to evaluate en-
ergy billings and physical approaches use physical building proper- velopes efficiency on energy consumption at district scale, Energy Procedia 78
ties as their basis. This comparison helps other scientists to choose (2015) 1165–1170.
[10] JH Kämpf, D Robinson, A simplified thermal model to support analysis of ur-
the right approach for the present task. In principle, physical ap- ban resource flows, Energy Build 39 (4) (2007) 445–453.
proaches can be used more extensively, e.g. for densification or [11] S Coccolo, J Kaempf, JL Scartezzini, The EPFL campus in Lausanne: new energy
retrofit scenarios, demand forecasting of planned urban develop- strategies for 2050, Energy Procedia 78 (2015) 3174–3179.
[12] Walter E, Kämpf JH. A verification of CitySim results using the BESTEST and
ment areas or interactions between buildings and smart energy monitored consumption values. In: Proceedings of the 2nd Building Simulation
systems. Although, in overlapping fields of application and in the Applications conference, Bozen -Bolzano, Italy; 2015:215–222.
case of existing annual energy billings, the energy signature is to [13] EJ Kim, G Plessis, JL Hubert, JJ Roux, Urban energy simulation: simplification
and reduction of building envelope models, Energy Build 84 (2014) 193–202.
be preferred due to the lower computational and modelling effort. [14] R Nouvel, A Mastrucci, U Leopold, O Baume, V Coors, U Eicker, Combin-
These are applications that examine different climatic conditions ing GIS-based statistical and engineering urban heat consumption models:
or illustrate the heating and cooling load profile when annual en- towards a new framework for multi-scale policy support, Energy Build 107
(2015) 204–212.
ergy billings are available. A further application case is the devel-
[15] G Zucker, F Judex, M Blöchle, M Köstl, E Widl, S Hauer, A Bres, J Zeilinger,
opment of daily benchmarks for continuous monitoring of urban A new method for optimizing operation of large neighborhoods of buildings
areas. using thermal simulation, Energy Build 125 (2016) 153–160.
[16] A Bres, K Eder, S Hauer, F Judex, Case study of energy performance analyses
The presented approaches are applied to case studies, which
on different scales, Energy Procedia 78 (2015) 1847–1852.
consist of a multi-family house, an office building and a residen- [17] H Samuelson, S Claussnitzer, A Goyal, Y Chen, A Romo-Castillo, Parametric en-
tial area with 34 buildings. The UBEM approaches are validated ergy simulation in early design: high-rise residential buildings in urban con-
by means of detailed building simulation and measurement data. texts, Build Env. 101 (2016) 19–31.
[18] Y Chen, T Hong, MA Piette, Automatic generation and simulation of urban
The simulation results show that both UBEM approaches are appli- building energy models based on city datasets for city-scale building retrofit
cable in such cases within reasonable deviation compared to the analysis, Appl Energy 205 (2017) 323–335.
P. Nageler et al. / Energy & Buildings 179 (2018) 333–343 343

[19] M Fuchs, J Teichmann, M Lauster, P Remmen, R Streblow, D Müller, Workflow [42] G Raffio, O Isambert, G Mertz, C Schreier, K Kissock, Targeting residential en-
automation for combined modeling of buildings and district energy systems, ergy assistance, in: Proceedings of the Energy Sustainability Conference, Long
Energy 117 (2) (2016) 478–484. Beach, California, 2007.
[20] P Nageler, G Zahrer, R Heimrath, T Mach, F Mauthner, I Leusbrock, [43] L Mazzarella, M Liziero, C Neumann, D Jacob, Description of European Proto-
H Schranzhofer, C Hochenauer, Novel validated method for GIS based au- type Tool for Evaluation of Building Performance and the National tools. Po-
tomated dynamic urban building energy simulations, Energy 139 (2017) litecnico Milano, Fraunhofer Institute for Solar Energy Systems ISE, 2009.
142–154. [44] EQUA, http://www.equa.se/en/ [Accessed 16 September 16].
[21] T Loga, B Stein, N. Diefenbach, TABULA building typologies in 20 European [45] Stadtwerke Gleisdorf. Wärme, http://www.stadtwerke-gleisdorf.at/992_DEU_
countries—making energy-related features of residential building stocks com- HTML.htm [Accessed 4 October 16].
parable, Energy Build 132 (2016) 4–12. [46] Swiss Society of Engineers and Architects, SIA 2024, Standard-
[22] M Aksoezen, M Daniel, U Hassler, N Kohler, Building age as an indicator for Nutzungsbedingungen für die Energie- und Gebäudetechnik (2006).
energy consumption, Energy Build 87 (2015) 74–86. [47] Digitaler Atlas Steiermark:: Höhen- & Reliefkarte, http://gis2.stmk.gv.at/
[23] P Nageler, T Mach, R Heimrath, H Schranzhofer, C Hochenauer, Generation tool atlas/(S(kfkazafhk5p4fksmbm1afwwv))/init.aspx?karte=gel&ks=das&cms=
for automated thermal city modelling, Appl. Mech. Mater. (2018) (in press). da&redliningid=kgbyngmhkqpn131mqgpr33z4&box=359853.849598163;
[24] S Leal, S Hauer, F Judex, S Gahr, Implementation of an automated building 5162155;610146.150401837;5297226&srs=32633 [Accessed 23 January 18].
model generation tool, in: Proceedings of 12th IEEE International Conference [48] meteorological and geophysical service of Austria (ZAMG), http://www.zamg.
on Industrial Informatics (INDIN), Porto Alegre , Brazil, 2014, pp. 457–462. ac.at/cms/de/topmenu/kontakt [Accessed 14 October 16].
[25] T Dogan, C Reinhart, P Michalatos, Autozoner: an algorithm for automatic ther- [49] L Ljung, System Identification: Theory for the User, 2nd ed., Prentice-Hall PTR,
mal zoning of buildings with unknown interior space definitions, J. Build Per- 1999.
form Simul. 9 (2) (2016) 176–189. [50] ASHRAE, Guideline 14-2014, Measurement of Energy, Demand and Water Sav-
[26] L Smith, K Bernhardt, M Jezyk, Automated energy model creation for concep- ings, American Society of Heating, Ventilating, and Air Conditioning Engineers,
tual design, in: Proceedings of the Symposium on Simulation for Architecture Atlanta, GA, 2014.
and Urban Design, San Diego, CA, USA, 2011, pp. 13–20. [51] Debaditya Chakraborty, Hazem Elzarka, Performance testing of energy mod-
[27] EA Koch, Continuous Simulation for Urban Energy Planning Based on a Non– els: are we using the right statistical metrics? J. Build. Perform. Simul. (2017),
Linear Data-Driven Modelling Approach, FBTA, Karlsruhe Institute of Technol- doi:10.1080/19401493.2017.1387607.
ogy (KIT), Karlsruhe, 2016 Thesis. [52] U Jordan, K Vajen, DHWcalc: program to generate domestic hot water profiles
[28] J Ma, JCP Cheng, Estimation of the building energy use intensity in the ur- with statistical means for user defined conditions, in: Proceedings of the ISES
ban scale by integrating GIS and big data technology, Appl. Energy 183 (2016) Solar World Congress, Orlando, FL, USA, 2005.
182–192. [53] PostgreSQL, https://www.postgresql.org/ [Accessed 17 October 2016].
[29] CE Kontokosta, C Tull, A data-driven predictive model of city-scale energy use [54] OpenStreetMap, https://www.openstreetmap.org/ [Accessed 14 October 2016].
in buildings, Appl. Energy 197 (2017) 303–317. [55] Basemap at:, http://basemap.at/ [Accessed 14 October 2016].
[30] ASHRAE. Energy Estimating and Modelling Methods, ASHRAE Handbook - Fun- [56] M Kedzierski, A Fryskowska, Methods of laser scanning point clouds integra-
damentals, American Society of Heating, Refrigerating and Air-Conditioning tion in precise 3D building modelling, Measurement 74 (2015) 221–232.
Engineers, Atlanta, GA, USA, 2005 2005. [57] L Mahdjoubi, C Moobela, R Laing, Providing real-estate services through the
[31] AJ Heller, Heat-load modelling for large systems, Appl. Energy 72 (1) (2002) integration of 3D laser scanning and building information modelling, Comput.
371–387. Ind. 64 (9) (2013) 1272–1281.
[32] M Bauer, JL Scartezzini, A simplified correlation method accounting for heat- [58] QGIS. A free and Open Geographical Information System, http://www.qgis.org/
ing and cooling loads in energy-efficient buildings, Energy Build 27 (2) (1998) en/site/ [Accessed 17 October 16].
147–154. [59] H. Eichlseder, Lastprofile Nicht-Leistungsgemessener Kunden (HE, HM, HG, PG,
[33] A Rabl, A Rialhe, Energy signature models for commercial buildings: test with PK Und PW) Der Gasnetzbetreiber Österreichs – Überarbeitung 2008. Graz,
measured data and interpretation, Energy Build 19 (2) (1992) 143–154. Technical University of Graz, 2008.
[34] A Koch, Heat Demand Modeling at the Scale of Urban Neighborhoods, in: Pro- [60] J Grohmann, Lastanalyse und –synthese zur Optimierung leitungsgebundener
ceedings of the Third International Conference On Infrastructure Systems And Energiesysteme, Herrsching, E und M Energie und Management Verl.-Ges,
Services: Next Generation Infrastructure Systems For Eco-Cities (INFRA), Shen- 20 0 0.
zhen, China, 2010. [61] J Schnieders, W Feist, R Pflug, O Kah, CEPHEUS - Wissenschaftliche Begleitung
[35] B Geiger, M Hellwig, Entwicklung von Lastprofilen für die Gaswirtschaft im und Auswertung, Endbericht. Retrieved from Darmstadt, Passive House Insti-
Auftrag des Bundesverbandes der Deutschen Gas- und Wasserwirtschaft e.V. tute, 2001.
und des Verbandes kommunaler Unternehmen e.V.-Haushalte, Institute for En- [62] MT Ali, M Mokhtar, M Chiesa, P Armstrong, A cooling change-point model of
ergy Economy and Application Technology, Technical University of Munich, community-aggregate electrical load, Energy Build 43 (1) (2011) 28–37.
Munich, 2002. [63] C Ghiaus, Equivalence between the load curve and the free-running tempera-
[36] FR Jacobsen, Energy signature and energy monitoring in building energy man- ture in energy estimating methods, Energy Build 38 (5) (2006) 429–435.
agement systems, in: Proceedings of CLIMA 20 0 0 World Congress, Copen- [64] H Lund, S Werner, R Wiltshire, S Svendsen, JE Thorsen, F Hvelplund, BV Math-
hagen, 1985. iesen, 4th generation district heating (4GDH): integrating smart thermal grids
[37] T Day, Degree-days: Theory and Application, The Chartered Institution of into future sustainable energy systems, Energy 68 (2014) 1–11.
Building Services Engineers (CIBSE), London, 2006 TM41. [65] G Schweiger, J Rantzer, K Ericsson, P Lauenburg, The potential of power—
[38] JK Kissock, J Haberl, DE Claridge, Inverse modeling toolkit (1050RP): numerical to–heat in Swedish district heating systems, Energy 137 (2017) 661–669.
algorithms, ASHRAE Trans. 109 (part 2) (2003). [66] M Wetter, Co-simulation of building energy and control systems with the
[39] L Girardin, F Marechal, M Dubuis, N Calame-Darbellay, D Favrat, EnerGis: a building controls virtual test bed, J. Build. Perform. Simul. 4 (3) (2011)
geographical information based system for the evaluation of integrated energy 185–203.
conversion systems in urban areas, Energy 35 (2) (2010) 830–840. [67] Engel G, Chakkaravarthy A, Schweiger G. A methodology to compare different
[40] HA Nielsen, H Madsen, Modelling the heat consumption in district heating sys- co-simulation interfaces: a thermal engineering case study. In: Proceedings of
tems using a grey-box approach, Energy Build 38 (1) (2006) 63–71. the 7th International Conference on Simulation and Modeling Methodologies,
[41] P Woods, ETI Macro Distributed Energy Project - Summary Report Loughbor- Technologies and Applications, Madrid, Spain; 2017.
ough, Energy Technologies Institute, 2012. [68] P Nageler, G Schweiger, H Schranzhofer, T Mach, R Heimrath, C Hochenauer,
Novel method to simulate large-scale thermal city models, Energy 157 (2018)
633–646.

You might also like