Conference PaperPDF Available

Unlocking Potentials of Building Energy Systems’ Operational Efficiency: Application of Digital Twin Design for HVAC systems


Abstract and Figures

Building energy systems are complex because of their nonlinear behaviour and stochastic environmental interactions. Thus, their efficiency depends on both component design and control strategy. Aiming for high efficiencies over the whole lifecycle, we apply Product Lifecycle Management and Digital Twin Design to HVAC systems, initially using an energy recovery ventilation (ERV) simulation model. We develop a Digital Twin Prototype in Modelica of the ERV unit to predict physical system behaviour. Verifying functionality and suitability, we set up a Digital Twin Instance (DTI) by calibrating the model against measurement data of the physical twin. DTI and physical twin communicate via MQTT interfaces with each other. The approach enables DTIs to calculate predictions for different scenarios per instance, which increases systems’ efficiency significantly. In conclusion, this framework allows an application of innovative and predictive control and maintenance strategies to HVAC systems.
Content may be subject to copyright.
Unlocking Potentials of Building Energy Systems’ Operational Efficiency: Application of
Digital Twin Design for HVAC systems
Christian Vering1, Philipp Mehrfeld1, Markus Nürenberg1, Daniel Coakly2,
Moritz Lauster1, Dirk Müller1
1Institute for Energy Efficient Buildings and Indoor Climate, Aachen, Germany
2Mitsubishi Electric, Edinburgh, Scotland
Building energy systems are complex because of their
nonlinear behaviour and stochastic environmental
interactions. Thus, their efficiency depends on both
component design and control strategy. Aiming for high
efficiencies over the whole lifecycle, we apply Product
Lifecycle Management and Digital Twin Design to
HVAC systems, initially using an energy recovery
ventilation (ERV) simulation model.
We develop a Digital Twin Prototype in Modelica of the
ERV unit to predict physical system behaviour. Verifying
functionality and suitability, we set up a Digital Twin
Instance (DTI) by calibrating the model against
measurement data of the physical twin. DTI and physical
twin communicate via MQTT interfaces with each other.
The approach enables DTIs to calculate predictions for
different scenarios per instance, which increases systems’
efficiency significantly. In conclusion, this framework
allows an application of innovative and predictive control
and maintenance strategies to HVAC systems.
Digitalization of energy systems enables the application
of sophisticated control strategies increasing operational
efficiencies significantly (Afram, 2014). Unlocking the
entire potential, there is an urgent need to design systems
carefully by simultaneous consideration of functional
requirements and control strategies with digital interfaces
in early development stages (Dongellini 2017; Poppi,
2016). However, integral descriptions that face these
concepts for HVAC systems do not exist, yet. On the one
hand, an integral product development approach
regarding the whole lifecycle is required. We propose
Product Lifecycle Management to be suitable (Stark,
2015). On the other hand, the digital twin approach has
proven great suitability for complex system designs and
their operation (Grieves, 2017). A combination of both
concepts for HVAC Systems has currently not been found
in literature.
Building energy systems are complex because of their
nonlinear behaviour and stochastic environmental
interactions (Staffel, 2012). Furthermore, HVAC systems
make up the majority of commercial building energy
consumption, and are directly responsible for maintaining
the health and wellbeing of occupants in Germany
(Arbeitsgemeinschaft Energiebilanzen e.V., 2017).
However, their efficiency strongly depends on both
component design and control strategy (Afram, 2014).
Aiming for high efficiencies, we apply systematically
Digital Twin Design (DTD) to HVAC systems, initially
using a detailed model of an energy recovery ventilation
(ERV) unit.
Thereby, this work contributes to the application of the
digital twin approach to HVAC systems by example of an
energy recovery ventilation unit considering Product
Lifecycle Management:
We combine the approaches of Product Lifecycle
Management and Digital Twin Design (DTD) to
demonstrate its potential for the use in realistic,
manufacturing applications in Chapter 2.
In Chapter 3, we show the modelling approach of the
ERV unit in Modelica and a room model to apply it
to a use case. We calibrate the ERV model using
measurement data.
A use case of predictive maintenance is evaluated in
Chapter 4 were we evaluate different energy
consumptions due to different maintenance
Finally, we summarize and conclude our findings.
Digital Twins in Product Lifecycle
Product Lifecycle Management (PLM) is a production
engineering approach for the holistic control and
administration of product-related information along the
entire products’ lifecycle. Methods, models and tools
(especially IT systems) support the product development
process and the product usage phase in order to increase
productivity and efficiency. PLM integrates all data and
information collected along the product lifecycle. (Stark,
The first use of the term in literature dates back to the
beginning of the 21st century. A uniform definition of the
exact components that a PLM system must present in
order to be declared as such does not exist. Previous
applications refer mainly to physical products and
processes (Tao, 2018). In the context of this work, PLM
is to be additionally implemented for virtual models in the
context of digital twins. (Stark, 2015)
PLM distinguishes between fife phases of products’
lifecycle. In the planning phase, a product is defined on
the basis of customer requirements, market analyses,
innovative ideas and knowledge from former products. A
preliminary design is carried out in which technical
parameters and functions are defined. These are
concretized in the create-phase by first models or
prototypes. In the build phase, production is finally
started based on the defined product design. The
following use of the product is described in the sustain
phase. This also includes services such as maintenance
and repair. Ultimately, the disposal phase after the end of
use includes processes for disposal or recycling. (Yoga
Mule, 2012)
At the moment, there is no standardized definition of what
exactly a digital twin is. A widespread understanding
describes the term digital twin as an intelligent, digital
image of a real product or process. Against this
background, a separate approach to the implementation of
a digital twin will be developed within the presented
framework. The first approaches to the implementation of
a digital twin come from the field of aeronautics and space
technology. In this context, NASA in particular should be
mentioned as a "pioneer" in the development and use of
digital twins. Their intention to use digital twins is to
monitor the functional condition of an aircraft or space
shuttles as an overall system as well as individual
components. (Schleich 2017; Tuegel, 2011)
NASA defines a digital twin e.g. as follows: "A Digital
Twin is an integrated multiphysical simulation of a
vehicle or system actually built, using the best available
physical models, sensor technologies, fleet history, etc.,
to reflect the life of the corresponding flying twin
(Schleich, 2017). Grieves, who worked in an advisory
capacity for NASA, is considered the founder of the term
digital twin, which he describes in the context of PLM
(Grieves, 2014). It defines a digital twin as "a set of virtual
information that completely describes a potential or
actually manufactured physical product from the micro
atomic plane to the macro geometric plane". (Grieves,
Combining both approaches, PLM and digital twins, and
applying them to HVAC systems, we propose Digital
Twin Design.
Digital Twin Design has PLMs’ fife phases. These phases
are combined with digital twin definitions. Regarding
(Grieves, 2017), we combine three PLM phases with
existing digital twin definitions:
Create phase Digital Twin Prototype (DTP)
Build phase Digital Twin Instance (DTI)
Sustain phase Digital Twin Aggregate (DTA)
In addition, we propose two further Digital Twins to be
on PLM phases:
Dispose phase Digital Twin Knowledge (DTK)
Plan phase Digital Twin Concept (DTC)
The first step in product development is the creation of a
Digital Twin Concept (DTC). The conception takes place
in the PLM plan phase. It comprises the entire planning
and design of a concept for the implementation of a digital
twin for HVAC systems. Based on customer needs,
market analyses and individual ideas, goals are derived
which are pursued by the implementation of a digital twin.
Once these objectives have been defined, a requirements
analysis can be carried out to define product
characteristics, functions and operating strategies.
Besides former typical requirements such as physical
concepts, in this phase requirements for the virtual model
need to be set up.
Figure 1: Applying Product Lifecycle Management and
digital twins to Digital Twin Design for HVAC systems.
The first requirement of the virtual model is the definition
of model’s inputs and outputs. Applied to HVAC systems,
typical inputs are measurement data points like volume
flows, temperatures and pressure levels. Based on this
inputs our model needs to predict outputs. Typically these
are power consumptions e.g. of fans, pumps or heaters.
If inputs and outputs are defined, a second requirement
needs to be met. Regarding the application it has to be
defined, which level of accuracy the model should has. In
general, we can distinguish between
black box,
grey box and
white box model.
Black box models are mathematical models, whereas grey
box models use physical equations and some simplifying
assumptions to represent physical system behaviour.
White box models are a perfect description of the physical
system. With an increasing level of detail, calculation
durations of model predictions increase as well.
Consequently, the trade-off between accuracy and
simulation speed needs to be clarified in an early design
stage. (Mueller, 2016)
Creating the Digital Twin Prototype (DTP) in the PLM
create phase, a simulation model of the HVAC system
must be created. In addition to modelling, the subsequent
parameterization of the model is also carried out in the
DTP phase. In our use case of an ERV system, we need
simulation models of heat exchangers, ventilators, air
filters, heater, cooler, temperature control, sources, sinks
and the target room. Therefore, we use Modelica because
of both modelling of dynamic systems as well as a high
degree of modularity and reusability.
After modelling all components, a parametrisations needs
to be done. For this purpose, safe, known parameters are
transferred into the model. In particular, geometric
boundary conditions must be taken into account, such as
heat-transferring surfaces, wall thicknesses, room
dimensions, etc. The materials used for the individual
components are also among the parameters known in
many applications.
Assumptions can be made for parameters that are not
known or are subject to uncertainties. However, it must be
borne in mind that this creates uncertainties in the overall
system, which can lead to a deviation between simulation
and measured values. Due to uncertain parameters, a
subsequent calibration is of enormous importance when
modelling real systems. With the completion of the
simulation model, the prototype of the digital twin exists.
The mentioned calibration of the simulation model is
intended as a transition to the build phase according to
The creation of a Digital Twin Instance (DTI) in the PLM
build phase requires a calibrated simulation model and a
communication structure between physical and digital
twins. The calibration of the simulation model serves to
map selected output variables of the real system through
the simulation. Any deviations between measured values
and simulation results should be as small as defined in the
planning phase.
Several methods exist for the calibration of HVAC
systems. Bayesian calibration has proven to be suitable
for dynamic systems. This is mainly due to the good
predictability of output variables, so that there is a high
probability that measured values will be well represented
by the simulation after calibration (Li, 2017).
The disadvantages of the high complexity and numerous
required simulations are accepted. There is a promising
approach in the literature for the implementation of
Bayesian calibration (Chong, 2018):
1) Generate/procure measurement data
2) Sensitivity analysis/parameter testing
3) Generate simulation results
4) Combining measurement data and simulation results in
Gaussian process
5) Distribution determination (Markov Chain Monte
Carlo method)
6) Check convergence and assess the accuracy of the
calibrated model.
In the course of the development of the DTD approach, a
(partially) automated implementation of this
methodology for digital twins is considered to be
promising. A practical implementation for the simulation
model of the ventilation system has not yet taken place.
The existing data basis was not sufficient in this respect.
Instead, a manual calibration of the model is carried out.
The target values of the model and therefore for
calibration are power consumption of the fan (Pel) on the
one hand and the effectiveness of the heat transfer in the
cross-flow heat exchanger () on the other hand.
Measured values from the manufacturer were available in
this respect. For several operating points, different
volume flows per stage and related power consumptions
are available. The calibration was conducted fitting the
heat transition coefficient of supply air, heat exchanger
material and exhaust air in order to match the both target
The Root Mean Square Error (RMSE) is used as the
criterion for assessing the quality of the calibration. The
RMSE is calculated as follows (Mehdiyev, 2016):
 
. (1)
First, simulation results  are generated with the
non-calibrated model. By comparison with
measured  values, a deviation for the target values
can be determined. The lower the RMSE the better a
calibration was conducted. With this calibrated model,
Digital Twin Design is ready to use.
In order to exploit the benefit of a Digital Twin Design,
digital and physical twins are constantly exchanging data.
Ensuring this requirement, an infrastructure is required on
which communication can take place. In the following,
we will show how such a communication interface can be
implemented to enable data transfer between digital and
physical instances as well as data storage in a cloud.
Based on the results of extensive literature research,
MQTT (Message Queuing Telemetry Transport) has
proven to be suitable for implementing communication
structures for IoT applications. (HiveMQ, 2019)
Notably, low demands on the digital infrastructure, such
as network and bandwidth, make the almost universal use
of the digital twin possible. In addition, an enormous
number of end devices can participate in this
communication structure. The intention of executing
several instances of the digital twin in parallel, MQTT has
many advantages. Hirmer et al. (Hirmer, 2016) present an
approach for the implementation of an architecture for
automated communication for IoT environments. The
easily scalable communication protocol MQTT is used.
Hence, the access to sensor data of physical systems is
realized. Communication with digital instances takes
place on a publish/subscribe architecture. Haag and
Anderl (Haag, 2018) describe the implementation of a
digital twin based on an MQTT architecture.
In addition to the communication between the physical
and digital instance, a link is provided to display data and
results in real time via a graphical user interface. Based
on this implementation, potentials with regard to
monitoring and the resulting ease of use for users can be
The MQTT-Broker as central instance for communication
manages the data transfer between physical and digital
instances as well as the storage of data in a cloud. All
subscribers (terminals) send requests to the MQTT broker
using the commands publish and subscribe, which ensures
their implementation. Data is made available by publish,
while data is requested by subscribe.
Meeting requirements of communication services, the
Quality of Service (QoS) level should be at least 1, as this
ensures that data arrives. In order to ensure that the data
arrive exactly once, QoS level 2 must be selected, which
however results in increased data traffic due to increased
communication. If the resource requirement plays a minor
role due to increased data traffic, this disadvantage can be
ignored and level 2 can be implemented. (HiveQM, 2019)
With regard to communication with the physical instance,
an approach to implementation has been developed as part
of further work. A transfer of the developed approach to
the DTD is planned, but must still be tested in practice. To
implement communication on the part of the digital
instances, an MQTT-Dymola interface is currently under
development at the Institute for Energy Efficient
Buildings and Indoor Climate (EBC) at RWTH Aachen
University. This is to serve as an interface to read
simulation results from Dymola (Modelica) and to
introduce parameters into Dymola.
The implementation of the communication interface
completes the build phase. A functional instance of the
Digital Twin (DTI) is available. In the following, the
developed DTD approach provides for the aggregation
and execution of several instances the sustain phase.
The simultaneous execution of several DTIs is necessary
to achieve some of the objectives of the DTC. In the PLM
sustain phase several DTIs are aggregated to Digital Twin
Aggregate (DTA). To enable simultaneous, stable
execution of the aggregates, a suitable software
environment is required. In this respect, known
approaches were examined within the scope of this work
and a theoretical implementation possibility was
integrated into the DTD approach developed.
Virtual machines are a well-known advocate for creating
an environment in which software can be run in isolation.
Due to some disadvantages of virtual machines,
especially in the areas of efficiency and performance,
alternatives have gained in importance in recent years. An
example of this is Docker (Bauer, 2018). Docker uses
containers to create an environment in which software can
be run in isolation. Hence, Docker is suitable for the
implementation of the DTD approach in the DTA phase
due to simple possibilities for instantiating processes.
Individual instances in the Docker containers can simulate
different scenarios. These can vary in many ways.
Various weather data are just as conceivable as simple
parameter variations. These include, for example, a
decreasing mass flow due to leakages or additional
internal heat gains due to deviating user interaction. This
corresponds to carrying out various case studies. In
addition, more extensive instances can also be executed,
such as complete annual simulations. As mentioned in the
description of the DTC phase, the implementation of the
DTD approach pursues several objectives like operational
optimization or predictive maintenance.
On the one hand, the digital twin can be used to optimize
the operation of the physical Twin. Different scenarios of
the DTIs are of central importance here. Due to the
simulations of the DTIs carried out parallel to the real
operation, simulation results are available for various
parameters under the currently prevailing boundary
conditions. Based on these results, the digital twin can
determine the optimal scenario for the operation and
transfer the parameters used to the real system via the
developed communication structure. The implementation
of the ERV system looks as follows:
Parallel to the operation of the real system, several
instances are executed that simulate different scenarios.
An instance recreates the current operation of the real
system. The other instances vary single or multiple
parameters, such as air mass flow or pressure drop due to
the use of a new air filter. Different scenarios lead to the
fact that the instances deliver different simulation results.
Within the framework of operational optimisation, the
efficiency of the operation can be determined based on
predefined assessment variables (e.g. consumption of
electrical energy).
Through the communicated comparison between the
individual instances, the most efficient operating point
can be determined. The instance that simulated it,
transfers the operating parameters to the real system in
real time, so that it is executed at the optimum operating
point. The required valuation parameters can be very
complex in this context. Accordingly, the realization of a
digital twin for operational optimization is a complex
process. Nevertheless, the idea in connection with
resource-saving energy supply in the HVAC sector seems
to be promising for the creation of more efficient systems.
Another possible use of the digital twin is predictive
maintenance. This is closely related to the ability of the
digital twin to monitor the operation of the real system.
The digital twin as a virtual image of the real system has
the current operating data of the system at its disposal.
This makes it possible to monitor the functionality and
ensure operation in real time. Based on the evaluation of
current operating data, the digital twin can draw
predictive attention to maintenance and repair measures.
Problems can thus be recognised in advance and ideally
solved before they become acute.
By implementing predictive maintenance, maintenance
routines can be planned and downtimes are avoided
consequently. Both technical and personnel resources can
thus be used in an optimised manner. As a result, an
increase in the life cycle efficiency of products is
achieved. (Schreiner, 2018)
The prerequisite for implementing predictive
maintenance is simulation models that provide results in
real time and predictively. In addition, the simulation
models must have the ability to predict malfunctions
based on characteristic values. (Schreiner, 2018)
Large amounts of data are generated during the entire
sustain phase of the digital twin. Following the DTD, all
operating data of the digital twin are stored. After the end
of the use in the DTA phase, a data collection is available
which is finally used in the last phase of the life cycle.
The last step of the product life cycle (PLM dispose
phase) is the disposal of the product. Since the digital twin
has processed, generated and stored an amount of data and
information in the course of its life cycle, Digital Twin
Knowledge (DTK) phase provides evaluation and
utilization of the data. In the following, some ideas are
shown, how and for which purpose the evaluation of the
data generated and stored over the entire life cycle can
take place. During the operation of the Digital Twin
Aggregate, generated operational data is stored in a cloud
via the implemented communication structure. The data
thus generated over the entire life cycle can be analysed
after active use in the DTK phase. Various software
fundamentals are required for the entire data evaluation
process. (Meier, 2018)
At its core is an architecture that provides an environment
for cloud computing a database that can manage
generated time series data from simulations and sensor
readings, and a user interface that enables real-time
monitoring of operational data. Such an architecture
ensures that data generated over the lifecycle of the digital
twin can be meaningfully stored, managed and used.
PLM is a closed loop process. After the last phase, the
cycle starts again from the beginning. In the DTD, the
entire available data basis is consequently transferred to a
new life cycle via the evaluation step. This integration
takes place in the form of information, experience and
knowledge. This step accounts for a large part of the
added value of using a digital twin. Based on the
transferred information, a second generation of the real
product or the corresponding digital twin can be
produced, for example. With the help of evaluation as part
of the DTK phase, both positive and negative
characteristics of the first generation can be evaluated.
The aim for the second generation should be to eliminate
negative aspects and to strengthen positive aspects.
The DTK phase thus leads to improved products and
associated digital twins. To illustrate the DTD, an ERV
system will be evaluated using dynamic simulation
models, in order to be able to simulate various scenarios
for the operation and thus demonstrate the applicability of
the DTD. Ultimately, this approach is intended to identify
previously hidden potentials with regard to energy and
resource efficiency in the life cycle of HVAC systems.
Modelling and Calibration
Within this work, DTD is applied to an ERV system in
order to evaluate its potential of predictive maintenance.
Therefore, dynamic simulation models are necessary.
These are developed in Modelica and Dymola as
compiler. The ERV system is shown in Figure 2.
Figure 2: ERV system in Modelica consisting of a
ventilator, cross flow heat exchanger, filter, heater,
cooler and a room model.
The core component of the ventilation system is the heat
exchanger. This is designed as a cross-flow heat
exchanger. To create the simulation model for the
ventilation system, a heat transfer model is integrated for
cross-flow design. According to the Association of
German Engineers (VDI), the following equation applies
to the effectiveness of heat transfer in cross-flow heat
exchanger with fluid mixed on both sides (VDI, 2013):
 (2)
This effectiveness calculation  is implemented as a
function in Modelica using the number of transfer units
(NTU) and heat capacity flow R. In addition, the
manufacturer's instructions state that the heat-transferring
surface is made of paper. In this respect, a new record is
implemented for paper, which converts the material
properties into Modelica.
An existing room from the Modelica Buildings model
library is used as the room model (Wetter, 2019). This
model offers various possibilities for parameterization.
These include: geometric dimensions of the room wall,
floor and ceiling materials external and internal heat gains
port for reading in weather data. The current room
temperature is implemented by means of an energy
balance over the room.
The filter is modelled as pressure drop  (in Pa) between
inlet and outlet of the filter. Assuming particles to be in
the air, the filter is getting defiled over time by dust feed.
Hence, it pressure drop increases time-dependently
described as follows (Trox, 2012)
     . (3)
For the other components simple energy and mass
balances are introduced to describe the physical behaviour
of the system.
Regarding the calibration of the system, five ventilation
stages are used. In order to meet these target values for all
five ventilation stages and the associated volume flows, a
third-degree polynomial is stored, which takes into
account the volume flow dependence of the pressure
losses in the model. This results in an acceptable
calibration quality of 0.861 W based on a determined
RMSE. Table 1 summarises the associated electrical
Table 1: Calibration of electric output power.
In the second step, the effectiveness of the heat transfer is
calibrated against the measured values given for the five
ventilation stages. As before, the generated simulation
results are compared with the manufacturer data. It can be
stated that the effectiveness of the simulation is
underestimated. This is due to the calculation of (VDI,
2013). Real efficiencies are therefore better than the
calculated ones. For the model, the effectiveness must be
corrected upwards in the course of calibration. Based on
the individual deviations between simulated and
measured efficiencies, a mean deviation is determined,
with which the correction of the calculation is made in the
model. This allows a quality of 0.021 to be achieved on
the basis of the RMSE. Table 2 shows the associated
Table 2: Calibration of effectiveness.
In all cases, we have a well modelled and calibrated
Digital Twin Instance that can be studied regarding
predictive maintenance.
Predictive Maintenance of Filter systems
As shown in Figure 2, the ventilation system has a filter
to cut off particles from the air that is delivered in the
room. During operation, the filter is filled with particles.
The pressure drop increases. To ensure the adjusted mass
flow, the ventilation systems’ power consumption
A typical maintenance action would be to exchanger the
filter. Using DTD we study different schedules to find the
best time for an exchange. During operation, we consider
four different scenarios for annual simulations. Herein,
different criterions must be met for an exchange:
1) Exchange every year (Baseline scenario)
2) Exchange at 150 Pa pressure drop
3) Exchange at 200 Pa pressure drop
4) Exchange every 2.5 years
Assuming a price for electrical power of 29.5 ct/kWh and
55 € per exchanger, we calculate for a lifecycle of ten
years total operation costs and energy consumption.
Referred to the baseline scenario, the results are shown in
Figure 3. (Farr, 2006; Statistische Bundesamt, 2018)
Exchanging the filter once a year, the lowest electric
consumption is archived with 3827 kWh. The total costs
are 1129 €.
However we calculated the lowest electric energy
consumption in the baseline scenario, it is not the scenario
with lowest total costs. Scenario 2 and 3 show lower total
costs by about 5 to 6 % savings. Simultaneously, the
energy consumption increases by about 6 to 10 %. Both
the highest electric consumption as well as the highest
total costs is archived with scenario 4. Here, total costs
increase by about 1 % and electric energy consumption
increases about 30 %.
Figure 3: Total costs and electric energy consumption of
four different Digital Twin Aggregates.
These fictive scenarios indicate the potential of DTD. If
we can provide the digital twin with measurement data
and apply fault detection, filter exchanges can be
predicted and both total costs and electric consumption
can be decreased significantly.
Using these results, Digital Twin Design shows potential
for the whole systems’ lifecycle efficiency. Applying this
to other, more sophisticated control and maintenance
scenarios higher savings can be reached, respectively.
Hence, we showed that DTD, consisting of PLM and
digital twins, was successfully applied to HVAV systems
and the lifecycle efficiency could be increased due to
predictive maintenance.
This work shows that Digital Twin Design can be applied
to HVAC systems using Product Lifecycle Management
and digital twins for life cycle analysis. On the basis of
five life cycle phases, Digital Twin Concept, Digital Twin
Prototype, Digital Twin Instance, Digital Twin Aggregate
and Digital Twin Knowledge, the presented approach
serves as a detailed framework for the implementation of
a digital twin to a whole product lifecycle.
To demonstrate the feasibility of the developed approach,
a grey box model of a ventilation system was created and
manually calibrated. In addition, a communication
structure based on MQTT was presented.
The potential of predictive maintenance with various
routines with regard to air filter replacement for the
ventilation system was successfully demonstrated. These
are based on the existence of an optimum of energy
consumption and total costs. Savings of total costs up to
10 % are predicted using filter exchanges at a pressure
loss of 200 Pa. By integrating a digital twin for HVAC
systems, it is therefore possible to increase lifecycle
efficiency in terms of both energy and total costs.
0 2000 4000 6000
2,5 years
200 Pa
150 Pa
1 year
Total costs [€] Energy consumption [kWh]
Due to the complexity and interdisciplinary of the topic of
digital twin, not all phases of the DTD approach could be
tested for applicability and practicability in this work.
Consequently, there is a need for continued research
work, particularly in the areas of machine communication
and software technology.
Overall, the full potential of digital twins is not exhausted,
yet. Rather, the technology is still in its infancy. In the
future, further developments in and with digital twins are
to be expected. Their use is highly likely to multiply in the
coming years, as the basic infrastructure for
implementation is already in place.
In addition, some interesting approaches and ideas for the
use of digital twins have already emerged. Only the
implementation in practice is still open in many cases. It
will also be exciting to observe the extent to which new
business models with digital twins will occur. The market
for applications with digital twins is, in any case, almost
endlessly large in all sectors as digitalization progresses.
Afram, A. ; Janabi-Sharifi, F.: Theory and applications of
HVAC control systems A review of model
predictive control (MPC). In: Building and
Environment 72 (2014), S. 343355.
Dongellini, M. et al.: Sizing effects on the energy
performance of reversible air-source heat pumps for
office buildings. In: Applied Thermal Engineering
(2017), S. 10731081.
Poppi, S., et al.: Influence of boundary conditions and
component size on electricity demand in solar thermal
and heat pump combisystems. In: Applied Energy 162
(2016), S. 10621073.
Stark, J.: Product Lifecycle Management. Cham :
Springer International Publishing, 2015 (Decision
Staffel, I., et al.: A review of domestic heat pumps. In:
Energy Environ. Sci. (2012), Nr. 5, S. 92919306.
Arbeitsgemeinschaft Energiebilanzen e.V..
Auswertungstabellen zur Energiebilanz Deutschland:
1990 bis 2016,
me=ausw_24juli2017_ov.pdf, 2017.
Tao, F., et al.: Digital twin-driven product design
framework. In: International Journal of Production
Research 12 (2018), Nr. 1, S. 119.
Yoga Mule, J.: Concept and Evolution of PLM. In:
International Journal of Applied Information Systems
4 (2012), Nr. 3, S. 2528.
Schleich, B., et al.: Shaping the digital twin for design and
production engineering. In: CIRP Annals 66 (2017),
Nr. 1, S. 141144.
Tuegel, E. J, et al.: Reengineering Aircraft Structural Life
Prediction Using a Digital Twin. In: International
Journal of Aerospace Engineering 2011 (2011), Nr. 3,
S. 114.
Grieves, M.: Digital Twin:Manufacturing Excellence
through Virtual Factory Replication (2014).
Mueller, D.: Simulationsmodelle für die Heiz- und
Raumlufttechnik: Heizflächen. Lecture, RWTH
Aachen University, Aachen, 2016.
Li, C., et al.: A dynamic Bayesian network approach for
digital twin. In: 19th AIAA Non-Deterministic
Approaches Conference 2017.
Chong, A., Menberg, K.: Guidelines for the Bayesian
calibration of building energy models. In: Energy and
Buildings 174 (2018), S. 527547.
Mehdiyev, N., et al.: Evaluating Forecasting Methods by
Considering Different Accuracy Measures. In:
Procedia Computer Science 95 (2016), S. 264271.
How to Get Started with the lightweight IoT
started-with-mqtt, 2019.
Hirmer, P., et al.: Automating the Provisioning and
Configuration of Devices in the Internet of Things. In:
Complex Systems Informatics and Modeling
Quarterly (2016), Nr. 9, S. 2843.
Haag, S. ; Anderl, R.: Digital twin Proof of concept. In:
Manufacturing Letters 15 (2018), S. 6466.
Bauer, R.: What’s the Diff: VMs vs Containers,,
Schreiner, J., Mossner, J.: IoT-Basics: Was ist Predictive
Meier, A.: Werkzeuge der digitalen Wirtschaft: Big Data,
NoSQL & Co. Wiesbaden: Springer Fachmedien
Wiesbaden, 2018.
Verein Deutscher Ingenieure (VDI): VDI-Wärmeatlas:
Mit 320 Tabellen. 11. Aufl. //11. bearb. und erw. Aufl.
Berlin : Springer Berlin and Springer Vieweg, 2013.
Wetter, M.: Modelica Buildings Library., 2019.
Trox Technik (Trox): Filtertechnologie,
OX, 2012.
Farr, Camfil (2006): Saubere Luft mit geringen Kosten,, 2018.
Statistische Bundesamt (Destatis) (2018): Daten zur
Energiepreisentwicklung - Lange Reihen von Januar
2005 bis Oktober 2018, 2018.
... Although the concept first appeared in manufacturing, it can be applied to various domains including building operation [8]. The potential application of digital twins in building operation covers such diverse applications as energy [9], smart grids [10], construction monitoring and project management [11], sustainability assessment [12], urban planning [13], maintenance [14], logistics and facilities management [9]. Similarly, the scale of a Digital Twin can also vary considerably, reaching from an apartment up to complete districts [10]. ...
... While machine learning and data analytics are prominently used to create digital twins [8], simulation-based digital twins are also described in the literature [9]. Vering et al. [14] show how a simulation-based digital twin can be used to determine the optimum of energy consumption and total costs for air filter replacements. Zaballos et al. [15] combined wireless sensor networks with building information modelling (BIM) to capture and monitor thermal, visual, acoustic, and air quality comfort. ...
... Zaballos et al. also highlight the research need in the domain of digital twins applied to building engineering, especially when it comes to testing and smartness of the digital twin. This is in line with other work pointing out research needs for the potential use of digital twins in optimisation [10], digital twin business models [14] and general hurdles and success factors for future digital twin application in building engineering [9]. ...
Full-text available
Increasing demands on indoor comfort in buildings and urgently needed energy efficiency measures require optimised HVAC systems in buildings. To achieve this, more extensive and accurate input data are required. This is difficult or impossible to accomplish with physical sensors. Virtual sensors, in turn, can provide these data; however, current virtual sensors are either too slow or too inaccurate to do so. The aim of our research was to develop a novel digital-twin workflow providing fast and accurate virtual sensors to solve this problem. To achieve a short calculation time and accurate virtual measurement results, we coupled a fast building energy simulation and an accurate computational fluid dynamics simulation. We used measurement data from a test facility as boundary conditions for the models and managed the coupling workflow with a customised simulation and data management interface. The corresponding simulation results were extracted for the defined virtual sensors and validated with measurement data from the test facility. In summary, the results showed that the total computation time of the coupled simulation was less than 10 minutes, compared to 20 hours of the corresponding CFD models. At the same time, the accuracy of the simulation over five consecutive days was at a mean absolute error of 0.35 K for the indoor air temperature and at 1.2% for the relative humidity. This shows that the novel coupled digital-twin workflow for virtual sensors is fast and accurate enough to optimise HVAC control systems in buildings.
... Here, some studies and research work have focused on the generation of simulation models that serve this purpose. For example, Vering et al. (2019) [50] developed an energy recovery ventilation (ERV) simulation model to predict physical system behaviour of a ventilation system as a part of a digital twin. The model is used to achieve greater efficiency in the operation of HVAC systems. ...
... Here, some studies and research work have focused on the generation of simulation models that serve this purpose. For example, Vering et al. (2019) [50] developed an energy recovery ventilation (ERV) simulation model to predict physical system behaviour of a ventilation system as a part of a digital twin. The model is used to achieve greater efficiency in the operation of HVAC systems. ...
Full-text available
The COVID-19 pandemic has generated new needs due to the associated health risks and, more specifically, its rapid infection rate. Prevention measures to avoid contagions in indoor spaces, especially in office and public buildings (e.g., hospitals, public administration, educational centres, etc.), have led to the need for adequate ventilation to dilute the possible concentration of the virus. This article presents our contribution to this new challenge, namely the Ventilation Early Warning System (VEWS) which has aims to adapt the operation of the current Heating, Ventilating and Air Conditioning (HVAC) systems to the ventilation needs of diaphanous workspaces, based on a Smart Campus Digital Twin (SCDT) framework approach, while maintaining sustainability. Different technologies such as the Internet of Things (IoT), Building Information Modelling (BIM) and Artificial Intelligence (AI) algorithms are combined to collect and integrate monitoring data (historical records, real-time information, and location-related patterns) to carry out forecasting simulations in this digital twin. The generated outputs serve to assist facility managers in their building governance, considering the appropriate application of health measures to reduce the risk of coronavirus contagion in combination with sustainability criteria. The article also provides the results of the implementation of the VEWS in a university workspace as a case study. Its application has made it possible to detect and warn of inadequate ventilation situations for the daily flow of people in the different controlled zones.
... Based on the Modelica language, they developed their digital twin model to predict the behavior of the physical system. They communicated with the physical model through the Message Queuing Telemetry Transport (MQTT) interface to achieve the virtual reality mapping of the physical model and the digital twin model, which can improve the efficiency of energy recovery devices [11]. In order to monitor the operation status of the space station in real time, Xing Tao et al. established seven subsystem models of dynamics and control, energy, environmental and thermal control, propulsion, information, data management, and measurement and control of the core cabin, test cabin I and test cabin II of the space station based on the Modelica language and the MWorks system simulation platform, covering the general system, subsystems, and key stand-alone equipment, and conducted simulation analyses of the typical working conditions of subsystems, respectively. ...
... Eng. 2023, 11, 429 ...
Full-text available
The marine steam power system includes a large amount of thermal equipment; meanwhile, the marine environment is harsh and the working conditions change frequently. Operation management involves many disciplines, such as heat, machinery, control, electricity, etc. It is a complex multi-discipline physical system with typical nonlinear, multi-parameter, strong coupling characteristics. In order to realize the health management of a marine steam power system, based on digital twin technology combined with the Modelica language, modular modeling, etc., this paper conducts in-depth research on the multi-domain modeling of the marine steam power system, characteristic analysis of variable working conditions, fault simulation, etc. The analysis results show that the dynamic response trend of the model is consistent with the actual operation, the error of the main steam flow at 1800 s is the largest and is −4.9%, and the error of the main steam flow, steam turbine output power, cooling water outlet temperature and other key parameters is within ±5%. Virtual reality mapping between the digital model and the physical equipment is realized, which lays a foundation for mastering the dynamic characteristics of the marine steam power system.
... Dave, Buda [120] developed a web-based platform integrating IoT sensors with energy consumption and occupant comfort campus data to enhance facility management efficiency through informed decision-making. Vering, Nürenberg [121] developed a digital twin prototype that communicates prediction set for different scenarios via MQTT interfaces for increased system efficiency. Machado, Dezotti [122] integrated BIM and IoT for real-time building energy performance monitoring using a visual programming language. ...
... Vering et al. [60] used Product Lifecycle Management and Digital Twin Design for HVAC systems, firstly utilizing an energy recovery ventilation (ERV) simulation model in order to achieve high efficiency for the equipment. To forecast physical system behavior, they created a DT prototype of the ERV unit to test functionality and applicability by calibrating the model against physical twin measurement data. ...
Full-text available
Over the last few decades, energy efficiency has received increasing attention from the Architecture, Engineering, Construction and Operation (AECO) industry. Digital Twins have the potential to advance the Operation and Maintenance (O&M) phase in different application fields. With the increasing industry interest, there is a need to review the current status of research developments in Digital Twins for building energy efficiency. This paper aims to provide a comprehensive review of the applications of digital twins for building energy efficiency, analyze research trends and identify research gaps and potential future research directions. In this review, Sustainability and Energy and Buildings are among the most frequently cited sources of publications. Literature reviewed was classified into four different topics: topic 1. Optimization design; topic 2. Occupants’ comfort; topic 3. Building operation and maintenance; and topic 4. Energy consumption simulation.
... nX which provides models for both building physics and HVAC system simulation. The use of Modelica's multidomain simulation approach allows the substitution of different model parts with measured data and vice versa. It is thus predestined to analyse and optimize building, control and HVAC system behaviour using a Digital Twin model approach (c.f. Vering et. al., 2019). ...
Full-text available
Open-source modeling libraries facilitate the standardization and harmonization of model development. In the context of building energy systems, Modelica is a suitable modeling language as it is equation-based and object-oriented. As an outcome of the IBPSA project 1 cooperation, four open-source modeling libraries have been successfully deployed which all share the core library Modelica IBPSA. One of them is the AixLib modeling library. AixLib supports different modeling depths ranging from component to district level and covers all relevant domains in the context of building energy systems. To ensure high-quality simulations , continuous integration has been successfully added to automatically compare simulation results with existing validation data. This paper presents AixLib's key features, its scope, and associated tools. We present three use cases that highlight the broad application range of AixLib models. Furthermore, an overview of relevant research and industry projects is provided. Finally, we give an outlook on future development goals.
The purpose of a digital twin (DT) is to gain insight into and predict the performance of a physical product, process, or piece of infrastructure. Numerous advantages accrue from the energy industry's adoption of DT technology, such as improved asset performance, higher profits and efficiencies, and less harmful effects on the environment. This paper's goal is to present a literature evaluation that classifies DT principles, usage patterns, and benefits in the energy sector. A thorough literature review covering the past decade of studies on DT in the energy sector was conducted. The originality of this study is in-depth examination of DT's use across the whole energy value chain from power generation and storage to energy usage in buildings, transportation, and industrial applications. From this analysis, it was clear that there is a growing interest in using DT in the energy industry and minimizing energy use is the primary focus of the literature on digital twins. Growth of DT technologies will be aided by recent developments in machine learning and artificial intelligence, as well as the development of more sophisticated control systems, allowing for the enhancement of energy system efficiency and effectiveness, thereby fostering the clean energy transition, and reshaping the future of energy.
Full-text available
Digital twins can transform agricultural production systems and supply chains, curbing greenhouse gas emissions, food waste and malnutrition. However, the potential of these advanced virtualization technologies is yet to be realized. Here, we consider the promise of digital twins across six typical agrifood supply chain steps and emphasize key implementation barriers.
Full-text available
With the advent of new generation information technologies in industry and product design, the big data-driven product design era has arrived. However, the big data-driven product design mainly places emphasis on the analysis of physical data rather than the virtual models, in other words, the convergence between product physical and virtual space is usually absent. Digital twin, a new emerging and fast growing technology which connects the physical and virtual world, has attracted much attention worldwide recently. This paper presents a new method for product design based on the digital twin approach. The development of product design is briefly introduced first. The framework of digital twin-driven product design (DTPD) is then proposed and analysed. A case is presented to illustrate the application of the proposed DTPD method.
Full-text available
Miniaturization and price decline enable the integration of information, communication and sensor technologies into virtually any product. Products become able to sense their own state as well as the state of their environment. Paired with the ability to process and communicate this data allows for the creation of digital twins. The digital twin is a comprehensive digital representation of an individual product that will play an integral role in a fully digitalized product life cycle. To prove the digital twin concept a cyber-physical bending beam test bench was developed at DiK research lab.
Full-text available
The digitalization of manufacturing fuels the application of sophisticated virtual product models, which are referred to as digital twins, throughout all stages of product realization. Particularly, more realistic virtual models of manufactured products are essential to bridge the gap between design and manufacturing and to mirror the real and virtual worlds. In this paper, we propose a comprehensive reference model based on the concept of Skin Model Shapes, which serves as a digital twin of the physical product in design and manufacturing. In this regard, model conceptualization, representation, and implementation as well as applications along the product life-cycle are addressed.
Full-text available
The Internet of Things benefits from an increasing number of interconnected technical devices. This has led to the existence of so-called smart environments, which encompass one or more devices sensing, acting, and automatically performing different tasks to enable their self-organization. Smart environments are divided into two parts: the physical environment and its digital representation, oftentimes referred to as digital twin. However, the automated binding and monitoring of devices of smart environments are still major issues. In this article we present a method and system architecture to cope with these challenges by enabling (i) easy modeling of sensors, actuators, devices, and their attributes, (ii) dynamic device binding based on their type, (iii) the access to devices using different paradigms, and (iv) the monitoring of smart environments in regard to failures or changes. We furthermore provide a prototypical implementation of the introduced approach.
Full-text available
Choosing the appropriate forecasting technique to employ is a challenging issue and requires a comprehensive analysis of empirical results. Recent research findings reveal that the performance evaluation of forecasting models depends on the accuracy measures adopted. Some methods indicate superior performance when error based metrics are used, while others perform better when precision values are adopted as accuracy measures. As scholars tend to use a smaller subset of accuracy metrics to assess the performance of forecasting models, there is a need for a concept of multiple accuracy dimensions to assure the robustness of evaluation. Therefore, the main purpose of this paper is to propose a decision making model that allows researchers to identify the superiority of a forecasting technique over another by considering several accuracy metrics concurrently. A multi-criteria decision analysis approach, namely the preference ranking organization method for enrichment evaluation (PROMETHEE), was adopted to solve this problem. Bayesian Networks, Artificial Neural Networks, SVMs, Logistic Regression, and several Rule and Tree-based forecasting approaches were included in the analysis. After introducing a detailed description of accuracy measures, the performance of the prediction models are evaluated using a chosen dataset from the UCI Machine Learning Repository.
This paper provides practical guidelines to the Bayesian calibration of building energy models using the probabilistic programming language Stan. While previous studies showed the applicability of the calibration method to building simulation, its practicality is still impeded by its complexity and the need to specify a whole range of information due to its Bayesian nature. We ease the reader into the practical application of Bayesian calibration to building energy models by providing the corresponding code and user guidelines with this paper. Using a case study, we demonstrate the application of Kennedy and O’Hagan’s (KOH) [1] Bayesian calibration framework to an EnergyPlus whole building energy model. The case study is used to analyze the sensitivity of the posterior distributions to the number of calibration parameters. The study also looks into the influence of prior specification on the resulting (1) posterior distributions; (2) calibrated predictions; and (3) model inadequacy that is revealed by a discrepancy between the observed data and the model predictions. Results from the case study suggest that over-parameterization can result in a significant loss of posterior precision. Additionally, using strong prior information for the calibration parameters may dominate any influence from the data leading to poor posterior inference of the calibration parameters. Lastly, this study shows that it may be misleading to assume that the posteriors of the calibration parameters are representative of their true values and their associated uncertainty simply because the calibrated predictions matches the measured output well.
Das essential führt in Big Data ein und erläutert die wichtigsten Werkzeuge zur Nutzung von SQL- wie NoSQL-Technologien. Neben semantischer Datenmodellierung, Abfragesprachen, Konsistenzgewährung mit ACID oder BASE werden NoSQL-Datenbanken vorgestellt und organisatorische Aspekte des Datenmanagements erläutert. Der Leser erhält mit diesem Kompendium die wichtigsten Grundlagen sowohl zu SQL- wie auch zu NoSQL-Datenbanken. Der Inhalt • Big Data, SQL- und NoSQL-Datenbanken • Semantische Modellbildung • Relationenorientierte und graphbasierte Abfragesprachen • Konsistenzsicherung mit ACID oder BASE und Aspekte der Systemarchitektur Die Zielgruppen • Führungskräfte und Projektleiter aus der Wirtschaft, Entscheidungsträger in der öffentlichen Verwaltung • Dozierende und Studierende der Informatik/Wirtschaftsinformatik Der Autor Andreas Meier ist Professor für Wirtschaftsinformatik an der Universität Fribourg/Schweiz und beschäftigt sich mit eBusiness, eGovernment und Big Data. Er verfügt über breite Erfahrungen im Dienstleistungssektor und hat diverse Kooperationsprojekte mit der Privatwirtschaft sowie mit der öffentlichen Verwaltung erfolgreich abgeschlossen.
This paper deals with the numerical analysis of the energy performance of HVAC systems for heating and cooling, based on a reversible electrical air-source heat pump. The aim of this study is to highlight in which way the energy consumption of these systems are influenced by the heat pump modulating capability, the heat pump sizing and the climate. A numerical model based on the bin method and the energy signature procedure, able to take into account the performance of the heat pump at partial load, has been followed in order to calculate the seasonal and annual performance of the HVAC system coupled to a typical office building, located in three different European locations (Frankfurt, Istanbul, Lisbon). Different kinds of heat pumps (mono-compressor on-off, multi-compressor and inverter-driven heat pumps) have been considered. The numerical results presented in this paper point out that the heat pump sizing strongly affects seasonal and annual HVAC energy consumption: the ratio between the full thermal capacity of the heat pump and the building peak load strongly influences the energy performance of the system. It has been demonstrated that, depending on the season, the seasonal performance of the above-mentioned kinds of heat pumps are influenced in a different way by their sizing.
Solar thermal and heat pump combisystems are used to produce domestic hot water (DHW) and space heating (SH) in dwellings. Many systems are available on the market. For an impartial comparison, a definite level of thermal comfort should be defined and ensured in all systems. This work studied the influence of component size on electricity demand for a state of the art solar thermal and heat pump system. A systematic series of parametric studies was carried out by using TRNSYS to show the impact of climate, load and size of main components as well as heat source for the heat pump. Penalty functions were used to ensure that all variations provided the same comfort requirements. Two reference systems were defined and modelled based on products on the market, one with ambient air and the other with borehole as heat source for the heat pump. The results show that changes in collector area from 5 to 15 m2 result in a decrease in system electricity of between 305 and 552 kW h/year. Changes in heat exchanger size for DHW preparation were shown to give nearly as large changes in electricity use due to the fact that the set temperature in the store was changed to give the same thermal comfort in all cases. Decrease in heat pump size was shown to give a decrease in electricity use for the ASHP in the building with larger heat demand while it increased or had only a small change for other boundary conditions. Heat pump losses were shown to be an important factor highlighting the importance of modelling this factor explicitly.