Energy and Buildings 129 (2016) 367377

Contents lists available at ScienceDirect

Energy and Buildings

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

Adaptive predictive control of thermo-active building systems (TABS)

based on a multiple regression algorithm: First practical test
Martin Schmelas a, , Thomas Feldmann a , Patrick Wellnitz b , Elmar Bollin a
a
b

Institute of Energy System Technology (INES), Offenburg University of Applied Sciences, Badstrasse 24, 77652 Offenburg, Germany
Industrielle Werke Basel (IWB), Margarethenstrasse 40, 4002 Basel, Switzerland

a r t i c l e

i n f o

Article history:
Received 6 June 2016
Received in revised form 2 August 2016
Accepted 3 August 2016
Available online 3 August 2016
Keywords:
Thermo-activate building system (TABS)
Concrete core activation
Adaptive predictive control
Multiple regression
Thermal comfort
Energy savings

a b s t r a c t
There is a growing trend for the use of thermo-active building systems (TABS) for the heating and cooling
of buildings, because these systems are known to be very economical and efcient. However, their control is complicated due to the large thermal inertia, and their parameterization is time-consuming. With
conventional TABS-control strategies, the required thermal comfort in buildings can often not be maintained, particularly if the internal heat sources are suddenly changed. This paper shows measurement
results and evaluations of the operation of a novel adaptive and predictive calculation method, based on
a multiple linear regression (AMLR) for the control of TABS. The measurement results are compared with
the standard TABS strategy. The results show that the electrical pump energy could be reduced by more
than 86%. Including the weather adjustment, it could be demonstrated that thermal energy savings of
over 41% could be reached. In addition, the thermal comfort could be improved due to the possibility to
specify mean room set-point temperatures. With the AMLR, comfort category I of the comfort norms ISO
7730 and DIN EN 15251 are observed in about 95% of occasions. With the standard TABS strategy, only
about 24% are within category I.
2016 Elsevier B.V. All rights reserved.

1. Introduction
From 1990 to 2010, the global primary energy demand increased
by 50% [1]. The building sector is a major consumer of energy. In
2012, the share of nal energy consumption in the European Union
was at 42.6% in the tertiary sector, at 25.6% in the industry sector,
and at 31.8% in the transportation sector [2]. A large part of the tertiary sector accounts for the energy consumption of buildings. Most
of it is needed for their heating and cooling. Therefore, heating and
cooling concepts for renewable energy sources become increasingly important. For this purpose, low-temperature systems such
as thermo-active building systems (TABS) are particularly suitable.
There are different system congurations and designations for
TABS. A very popular and often-used one is known as concrete
core conditioning and concrete core activation. One of the rst
buildings to be equipped with concrete-core activation was built
in Switzerland in 1990/1991 [3]. According to a 2001 market analysis [4], there were a total of more than 60 buildings with TABS in
Germany at that time. In 2007, 508 new buildings were equipped
with 483,000 m2 of concrete-core-activated areas in Germany.

Corresponding author.
E-mail address: martin.schmelas@hs-offenburg.de (M. Schmelas).
URL: http://www.ines.hs-offenburg.de (M. Schmelas).
http://dx.doi.org/10.1016/j.enbuild.2016.08.013
0378-7788/ 2016 Elsevier B.V. All rights reserved.

General guidelines for the use of TABS in buildings can be found

in [5].
Reasons for the rapid spread are mainly the low investment
costs for new buildings (about 20 euro/m2 according to [6]) and
high economic efciency due to the energy efciency of these
systems. Low over- and under-temperatures as well as large heattransfer areas allow for the use of environmental energy (soil,
groundwater, ambient air) with the help of heat pumps. Of particular interest to architects is the factor that TABS are not visible,
because they are placed inside the concrete. By natural convection,
there is no high-velocity air movement in a room, and therefore the
problem of draft effects can be excluded. Due to the possibility of
actively loading and unloading the thermal mass of a building, load
peaks can also be attened and loads can be shifted within one day.
Nevertheless, there are also some disadvantages. The large thermal inertia of these systems does not allow for fast reactions to
room-temperature changes. For conventional control strategies, it
can be very difcult to handle this thermal inertia. So there is often
an initial operation phase over a period of one to two years because
the parameterization of the conventional operating strategy must
be adjusted by experts during the operation of the building. In addition, after the completion of a building, there is not yet full occupation, and residual moisture still is present in the concrete. By changing internal loads, for example the occupation of rooms, parameters
must be readjusted. Another disadvantage are the low heating and

368

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

Nomenclature
Q
V
I glob
T amb
T Rs et
TR
a, b, c
d
e
fHC
H
hdd
m
n
P
p
Qmc
Qua
t
Tamb
Ted
Trm
TRW
TR
TSW
u
W
w
y
z
AMLR
BMBF
COP
DIN
EN
F
GF
hist
INES
ISO
IWB
MPC
MWK
OLS
PID
PLC
pred
SIA
TABS
TCP
TF
UBB
VDI
VIP

energy ow
volume ow
mean daily global irradiation
mean daily ambient temperature
set-point mean daily room temperature
mean daily room temperature
variables of the multiple linear regression
time of actual integrated and mean daily values
errors
heating and cooling curve
hysteresis
heating-degree days
measured values
time of historical values
status signal of the control valve, electrical energy
time of predicted values
daily energy of a slab
daily energy demand of a zone
time of actual values
ambient temperature
mean daily ambient temperature
moving average ambient air temperature
return-water temperature
room temperature
supply-water temperature
control signal
weighting function
set-point
control variables
disturbances
adaptive predictive control of TABS based on a multiple linear regression
German Federal Ministry of Education and Research
coefcient of performance
Deutsches Institut fr Normung
Europische Norm
oor
ground oor
historical
Institute of Energy System Technology
International Organization for Standardization
Industrielle Werke Basel
model-based predictive control
Ministry of Science, Research and Arts of BadenWrttemberg, Germany
ordinary least squares
proportional-integral-derivative
programmable logic controller
prediction
Schweizerischer Ingenieur- und Architektenverein
thermo-active building systems
transmission control protocol
top oor
unknown-but-bounded
Verein Deutscher Ingenieure
vacuum insulation panels

cooling powers that move depending on the system in heating

between 30 and 50 W/m2 and in cooling mode between 50 and
70 W/m2 [6]. This is justied by the supply-water temperatures of
these systems, which are limited due to condensation and discomfort, e.g. in ofce buildings to a minimum of 16 C and a maximum

of 30 C. Due to the use of historical ambient temperatures and no

radiation data, it is neither possible to react to changes in weather.
This work focuses on measurement results and evaluations
of the operation of a novel adaptive and predictive calculation
method, based on a multiple linear regression (AMLR) for the control of TABS. The measurement results are compared with the
standard TABS strategy related to thermal comfort, energy efciency and adaptability.
2. Control strategies for TABS
In this section, the most frequently used control strategies are
presented for heating and cooling with TABS. Some of these nonpredictive strategies are analyzed in more detail and compared to
each other in [79]. The strategies from Sections 2.1 and 2.2 belong
to these control strategies.
2.1. Three-step control in dependency of the room temperature
The three-step control in dependency of the room temperature
is comparatively the simplest method to control TABS. In [10,11],
this type of control was investigated with the aid of experiments
and simulations. It does not contain any information about the
dynamics of the building. The supply-water temperature of TABS
corresponds with the undercut of the rst set point to the maximum possible supply-water temperature in the heating mode or
by overstepping the second set point to the minimum possible
supply-water temperature in the cooling mode. Here, the reference
variable is the room temperature.
2.2. Outside-temperature-compensated supply- and
return-water temperature control
The outside-temperature-compensated supply-water temperature control has been presented in [12,7]. There are different
variants of this control; however, the supply-water temperature
of TABS is always a function of the measured ambient temperature
and is dened in so-called heating and cooling curves. There are
several variations in the form of the ambient temperature included
in the strategy. This includes the current ambient temperature, the
ambient temperature averaged over a period from several hours
to days, and moving average values of the ambient temperature.
It is recommended to dene different parameters for heating and
cooling as well as (for ofce buildings) for workdays and weekends
(because the internal loads differ signicantly during these days). In
[13], this control strategy was used for demand-side management
mechanisms and realized through parameter adjustments of the
heating and cooling curve. In advance, these heating and cooling
curves can be adjusted for the building through appropriate planning tools as well as dynamic simulation programs and should be
optimized based on experiences during operation. The planning
tool TABSDesign (downloadable at http://www.faktor.ch/tabstool.html) is a Microsoft Excel sheet designed to simplify the planning and design process of TABS as well as to nd the rst parameters for this control strategy without taking demand-side management mechanisms into account. It is specically adapted to the socalled unknown-but-bounded (UBB) method which is described in
detail in [1416]. This is an integral approach that uses the outsidetemperature-compensated supply-water temperature control.
If there is a sensor for the return-water temperature (instead of
the supply-water temperature) in the heating and cooling circuits
of a building, the return-water temperature can be dened by the
heating and cooling curves. It includes information on the energy
demand of a building. Furthermore, the difference between the
supply- and return-water temperatures can be considered to be a
control variable. The larger this difference is, the higher the power

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

of the TABS. If the difference of the supply- and return-water

temperature undercuts a predened value, the TABS pump can
be switched off because there is almost no more energy transfer.
If the TABS pump is switched off, there is no indication for the
return-water temperature and thus the energy requirement of
the zone. Therefore, a circulation of the water has to take place in
regular (or continuous) intervals.

2.3. Model-based predictive control (MPC) for TABS

In the late 1970s, MPC came up in the application process
industry. Under this concept, a class of regulatory procedures is
summarized which use a model and minimize a cost function (also
called objective function) in order to predict an optimized control
signal. This method can be applied to linear and non-linear processes. MPC for TABS is increasingly being investigated in research
projects such as [1719]. High costs are associated with the implementation of an MPC in building automation through the enormous
modeling effort and the non-trivial mathematical calculations [20].

2.4. Adaptive and predictive control of TABS (AMLR)

The adaptive and predictive algorithm for the control of TABS
(AMLR) is based on both a dynamic and a steady-state resistance
capacity model for TABS as well as a multiple linear regression. The
applied TABS models as well as all mathematical calculations, the
comparison to other standard TABS strategies, and several weeks
of laboratory testing can be found in [21]. Simulation results for a
passive house with the algorithm are presented in [22].
The AMLR utilizes the knowledge that the daily energy demand
of a zone (Qua ) to obtain a mean daily room temperature (T R )
depends on the internal loads, the mean daily ambient temperature (T amb ), and the mean daily global irradiation (I glob ) according
to Eq. (1), which implies the steady-state model.
Qua = a + b (T R T amb ) + c I glob

(1)

The variable coefcients a, b, and c are determined by the ordinary least squares (OLS) method. The OLS method is the reason
why the AMLR strategy can adapt to different buildings and usage
proles. In the OLS method, it is possible to involve a weighting function W in the calculation of the variable coefcients. The
dependent variables are weighted by a function. Particularly important values are weighted with 1 and unimportant values with 0. A
linear function is chosen in dependency of the age of the historical
days for the weighting function.
Eq. (1) only use historical values in order to determine the
plane coefcients a, b, and c. With weather forecasts of the next

369

24 h averaging an ambient temperature T amb,pred as well as the

next 24 h averaging global irradiation I glob,pred , it is possible to
estimate the energy demand of a zone for a predened average
room temperature T R,set in advance.
The functional diagram of the AMLR controller is shown in Fig. 1
with its inputs and outputs. The variables n, t, p and d describe the
instant of time of the data. The time of the historical daily data
are denoted by n. t is the time of the current time step. d is the
time of the current integrated and mean daily value, respectively. p
describes the prediction time. Averaged and integrated daily values
are stored in a database.
For the steady-state TABS and room model as well as the
dynamic TABS model TABS-specic parameters must be dened
once. These include the location of pipes in the concrete, the activated TABS area, and the structural material data of the TAB system.
This data can be taken from the construction plans of the building.
The TABS-specic parameters are necessary for the calculation of
the resistances and the capacity of the models.
The controller with the steady-state model is needed once
a day and in case of change of the predicted disturbances z(p).
Disturbances, which are taken into account at the steady-state
model, are the measured (T amb,hist ) and forecasted mean daily
ambient temperatures (T amb,pred ) as well as global irradiation
(I glob,hist , I glob,pred ) and internal loads. Due to the different internal
loads, it has to be dened which days are weekdays or belong to a
weekend (or holiday).
The historical measured errors e(n) between the historical setpoints w(n) and the measured control variables y(n) (here T R,set )
are used as an input for a PID (proportional-integral-derivative)
controller, which is integrated inside the controller with steadystate model. It shifts the plane of the multiple linear regression up
and down, so it affects the coefcient a of Eq. (1).
With help of the dynamic model and the actual measured data
m(t) of the supply-water temperature (TSW ), the status signal of the
control valve P, and the room temperatures above (TR,2 ) and below
(TR,1 ) the TABS, energy ows can be calculated to (Q mc ) and out of
the slab (Q ua ) at every time step.
Output of the steady-state model is a predicted energy package
(Qmc,pred ) that has to be supplied to the zone within the next day.
Furthermore, Q mc is output of the dynamic model, which is integrated within a day (Qmc,day ). Once the energy package has been
supplied to the TABS zone, the control valve is closed for the rest of
the day. The operation mode (which determines the time the TABS
are loaded) can thus be selected freely. Short-term load shifts can
thus be realized.
With the help of the AMLR and one control valve per zone, it is
possible to perform a zone-specic loading of the TABS in contrast to the standard strategy. This means that only those zones

Fig. 1. Functional diagram of the AMLR algorithm for one TABS zone with its inputs and outputs.

370

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

are opened which have an energy demand, and other zones remain
closed. For the standard strategy, no control valves are usually used.
Therefore, each zone is loaded with energy, even if it does not
require any. The neutral zone describes the energy range within
the control valve remains closed. The deviation between the mean
room temperature and its set-point is that low, that no energy must
be supplied to the zone. On weekends, the mean room set-point
temperature does not have to be met. Therefore, a weekend temperature reduction will be applied in addition to a neutral zone. For
this purpose, Qmc,pred will be reduced to a predened value.
The mean room set-point temperature has been set for the
period shown here at 22 C. The neutral zone was dened with
150 Wh/m2 on workdays and with 300 Wh/m2 on weekends. The
weekend temperature reduction is 300 Wh/m2 . The calculation of
the predicted energy packages as well as the loading of the TABS
starts daily at 6:00 p.m.
3. IWB CityCenter building description

On the other oors, general working hours are on weekdays from

8 a.m. to 5 p.m.
The standard strategy for controlling the TABS in IWB CityCenter
was implemented by means of a central return-water temperature sensor and a predened heating curve fHC . The set-point of the
return-water temperature TRW,set is based on the current ambient
temperature Tamb by Eq. (2). The return-water temperature sensor
measures an average temperature of all zones and is located inside
the heat pump. The standard TABS control is also located in the heat
pump. An admixture does not occur here, which enables avoiding
energy losses [26]. Instead, the return-water temperature is kept
within a certain range over a dened hysteresis H. The heat pump
will be enabled according to Eq. (3) and disabled according to Eq.
(4). In order to achieve a representative return-water temperature,
the TABS pump operates continuously.
TRW,set = fHC (Tamb )

(2)

IWB (Industrielle Werke Basel) is the company for energy, water,

and telecommunications in the Swiss Basel region. The IWB CityCenter is the advisory centre for clients of IWB to all matters relating
to the issue of energy (ground and 1st oor). The remaining oors
26 serve as ofces for employees of IWB. The building is located
at Steinenvorstadt 14, 4051 Basel, Switzerland and represented in
Fig. 2. It was completed in 2008 in the MINERGIE-P standard [23].
Therefore, it is nearly a zero-energy building. The key facts are:

Vacuum insulation panels (VIP) insulation

Triple and quadruple glazing
Building volume according to SIA 416/1 [24]: 3583 m2
Net oor area according to SIA 416 [24]: 821 m2
Energy reference area: 1177 m2
Central ventilation system with heat recovery on the ground and
1st oors and decentralized ventilation systems on the other projectiles
Reversible brine/water heat pump for heating and cooling with a
nominal electrical power of 2.85 kW and a heat output of 11.4 kW
(equivalent to a COP of 4 for B0/W35 in accordance with DIN EN
14511 [25])
TABS and TABS distribution on each oor
TABS pump with a maximum power of 180 W
Public trafc on the ground and the 1st oors: Wednesday to
Friday 11 a.m. to 7 p.m. and Saturdays 11 a.m. to 4 p.m.

Fig. 2. Exterior view of the IWB CityCenter with advisory center for clients of IWB
on the ground and 1st oor as well as ofces for employees of IWB on the other
oors.

Fig. 3. Hydraulic diagram of the IWB CityCenter with sensors, actuators, the TABS,
the TABS pump, and the heat pump.

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

371

Fig. 4. Schematic diagram of the integration of the AMLR into the IWB CityCenter.

HP on = TRW,set H

(3)

HP off = TRW,set + H

(4)

The hydraulic diagram of the building after including the AMLR as

well as the various sensors, actuators, the TABS, the TABS pump,
and the heat pump is shown in Fig. 3. It shows that there are a total
of seven zones that can be opened and closed via a control valve in
each case. The top oor hydraulically depends on the 5th oor, as

there is only one control valve here. In addition, it can be seen that
this is a two-pipe system, i.e. it can only be either heated or cooled.
Other hydraulic variants for the integration of TABS are shown in
[27]. As part of the integration of the AMLR to the building, the
following sensors and actuators were retrotted:
Room-temperature sensors on each oor (previously, they were
only located on the ground, 2nd and top oors),

Fig. 5. Part of the measured data from 10.12.2014 until 12.12.2014 for the room temperatures for three oors, supply temperature, ambient temperature and the thermal
power of the TABS from the IWB CityCenter with the standard TABS strategy.

372

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

Control valves in front of each TABS distributor that can be opened

and closed via thermal actuator.
In addition to the heat pump, TABS provides a hot-water storage
tank and an air curtain with heat. If one of these storage demands
energy, they are priority-provided by the heat pump with heat.
Fig. 4 shows how the AMLR has been integrated into the existing system. On a PC, all communication channels converge. It has
a database for the recording of measured values and was equipped
with a LabView Runtime Engine. Via a website, the PC downloads
weather forecasts and historical weather data twice a day and
writes it into the database. Through Modbus TCP, it can communicate with the heat pump and the programmable logic controller
(PLC). All measurement data of the temperature sensors and actuators are detected by the PLC and are made available to the network
via Modbus TCP. In order to be able to cool down and heat up at
a predened time, the internal TABS control of the heat pump is
handled through the adjustment of heating and cooling curves. The
AMLR itself was programmed on LabView and runs as an executable
le on the PC. An additional web server visualizes all forecasts and
measurements so that the user of the building is informed about
the current status at all times.
4. Results and evaluation
The month of December was chosen as the period for comparing
the standard strategy with the AMLR. From 1st to 26th December
2014, the standard strategy described in Section 2.2 was active.

From 1st to 26th December 2015, the AMLR algorithm in Section

2.4 described was active. These periods were chosen because they
have similar ambient temperatures and the same amount of workand non-workdays.
In Fig. 5, the most important measured values are shown for two
days of the standard operation of TABS. In Fig. 9, the overall considered period of the standard operation of TABS is shown. At this
time, only the room temperatures for the ground, the 2nd, and the
top oor were recorded. In addition, no return-water temperature
was recorded, and therefore, there was no indication of the power
of TABS Q mc . The change in the supply-water temperature TSW can
allow for a conclusion of when energy was supplied to the TABS
circuit. Since the heat pump only allows the states on or off, it
can be assumed with a change of a TSW that Q mc equals 13 kW. It is
good to see the short-term interruptions of Q mc resulting from the
priority control of the hot-water tank and the air curtain.
In Fig. 6, the measured data of two days are shown that have
been stored during the TABS operation with the AMLR algorithm.
The ambient temperatures Tamb in Fig. 6 are similar to Fig. 5. The
overall considered period with its measurements of the AMLR algorithm can be found in Fig. 10. Room temperature measurements are
available for each oor. The return-water temperature TRW and the
were also recorded, whereby Q mc can be calculated.
mass ow m
In order to make a correct and valid comparison between the two
periods, the constant power TABS Q mc,constant was also assumed at
13 kW. It can be seen that most of the time, the Q mc values are identical to the Q mc,constant ones. However, short-term peaks can be seen
which probably arise due to measurement inaccuracies.

Fig. 6. Part of the measured data from 10.12.2015 until 12.12.2015 for the room temperatures for every oor, supply and return temperatures, ambient temperature and the
thermal power of the TABS from the IWB CityCenter with the AMLR TABS strategy.

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

In contrast to the supply-water temperature of Fig. 5, the one in

Fig. 6 goes up to 60 C. Due to the highly resistant plastic pipes that
are used for TABS, there are no material problems. The high supplywater temperatures result from the temporary opening of only a
single zone. Here, the total power of the heat pump is supplied to
this particular zone. Due to the high resistance and the capacitance
of the slab, this temperature is, however, passed on to the surface
of the slab only in a very attenuated condition. Thus, this is not
in conict with the maximum supply-water temperature of 30 C
specied in Section 1. This limitation applies only if the supplywater temperature is applied to the slab for a prolonged period.
In order to evaluate the thermal comfort, the European DIN EN
15251:2012-12 [28] standard was used. It differentiates between
naturally ventilated and air-conditioned buildings. The comfort is
divided into four categories, out of which three of the international
DIN EN ISO 7730:2006 [29] standard can be assigned. Category
IV describes any value outside of the other three categories. In
contrast to DIN EN ISO 7730:2006, the winter case is determined
by a moving average ambient air temperature of Trm < 10 C, while
the summer case with Trm > 15 C is calculated by Eq. (5) according
to DIN EN 15251:2012-12. In order to avoid a jump between the
summer and winter cases, linearization was performed in the
interval 10 C Trm 15 C.
Trm =

(Ted1 + 0.8Ted2 + 0.6Ted3 + 0.5Ted4 + 0.4Ted5 + 0.3Ted6 + 0.2Ted7 )

,
3.8

(5)

in which Ted is the mean daily ambient temperature.

373

Table 1
Percentage of the distribution of the room temperatures during occupancy in accordance with ISO 7730 and DIN EN 15251 for the standard control strategy of TABS
for December 2014.

Ground oor
2. oor
Top oor

Category I

Category II

Category III

53.64%
2.23%
17.19%

37.64%
51.03%
54.35%

8.72%
44.94%
28.45%

Category IV
0%
1.8%
0%

In Figs. 7 and 8, the available room temperatures were assigned

to categories of comfort. On the Y-axis, the standard for thermal comfort has an operative room temperature. Since there was
no sensor available for it, the measured room temperature was
used. Each dot represents a measured value during working hours.
The overheating of rooms with the standard strategy is obvious, whereas category I could be maintained in 95% of the time
on average with the AMLR on all oors. The remaining 5% are
within category II, which should be maintained for ofce buildings
according to the standard. With the standard strategy, most room
temperatures are within category II, which still leaves about 27%
within and 2% outside of category III. Here only about 24% are within
category I. The percentage distribution of room temperatures in the
four categories of the two periods is presented in Tables 1 and 2.
Thus, it is apparent that a signicantly higher level of thermal comfort is provided by the AMLR as compared to the standard strategy.
In addition to thermal comfort, especially a saving in energy consumption is the target of a new control strategy. Therefore, both
periods were evaluated in terms of the energy demand of TABS
Qmc and the electrical energy demand of the TABS pump Ppump . The

Fig. 7. Room temperatures plotted against the moving average ambient temperature during occupancy, according to ISO 7730/DIN EN 15251 for Dec 2014 (standard TABS
strategy).

374

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

Fig. 8. Room temperatures plotted against the moving average ambient temperature during occupancy, according to ISO 7730/DIN EN15251 for Dec 2015 (AMLR TABS
strategy).

results can be found in Table 3. As the ambient temperature deviates at around 0.7 K on average, a heating-degree day adjustment
of TABS energy demand was performed according to VDI 3807 [30].
For this purpose, the heating-degree days hdd are calculated with
Eq. (6):
hdd =

z


(T R T amb ),

(6)

in which z is the number of heating-degree days, T R is a mean daily

room temperature (here 20 C according to VDI 3807), and T amb is
the mean daily ambient temperature of each heating day.
Comparing the adjusted TABS energy demand Qmc,hdd of the two
periods with each other, a thermal energy saving of about 41%
could be reached with the AMLR. This heating-degree day adjustment cannot be transferred to Ppump . Without a heating-degree day
Table 2
Percentage of the distribution of the room temperatures during occupancy in accordance with ISO 7730 and DIN EN 15251 for the AMLR control strategy of TABS for
December 2015

Ground oor
1. oor
2. oor
3. oor
4. oor
5. oor
Top oor

Category I

Category II

Category III

Category IV

98.51%
100%
96.19%
86.93%
99.95%
93.19%
92.48%

1.49%
0%
3.81%
13.07%
0.05%
6.81%
7.52%

0%
0%
0%
0%
0%
0%
0%

0%
0%
0%
0%
0%
0%
0%

Table 3
Energy consumption, mean ambient temperature, heating-degree days and savings for December 2014 (standard TABS strategy) and December 2015 (AMLR TABS
strategy) at the IWB Citycenter.

Thermal TABS energy

Electrical energy of
the pump
Mean ambient
temperature
Heating-degree days
Heating-degree day
adjustment of
thermal TABS
energy

Dec. 2014

Dec. 2015

Savings/
difference

3067 kWh
102.06 kWh

1695 kWh
13.95 kWh

44.73%
86.33%

7.8 C

8.5 C

0.7 K

304.98 Kd
3067 kWh

286.22 Kd
1807 kWh

18.76 Kd
41.08%

adjustment, the lower operation time of the pump results in energy

savings of over 86%. Ppump was not measured directly, but was calculated due to the pump data and the measurement of the volume
ow V according to Eq. (7).
P1
=
P2

V 1
V 2

3
(7)

5. Conclusion and outlook

In this publication, the use of the newly developed adaptive
predictive algorithm for controlling TABS based on a multiple

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

linear regression (AMLR) is reported for a building under real

conditions for the rst time. It has been shown which sensors
and actuators are required for this purpose and how the AMLR
could be integrated into the existing automation system. The test
period shown here was the month of December 2015. The test
results could be directly compared with the measurement results
from the December 2014. Here, the standard TABS strategy was
active.
In both test periods, the number of non-working and working
days are equal and the mean ambient temperature differed only
slightly at around 0.7 K. Without the heating-degree day adjustment, a saving of thermal energy by more than 44% could be
detected. The demand for electrical pump energy could be reduced
by more than 86% at the same time. With the heating-degree
day adjustment, a reduction of thermal energy use by over 41%
could still be demonstrated. The savings could be explained by
the zone-specic loading of TABS and the presence overheating in
the standard strategy of the building. With the ability to specify
mean room set-point temperatures, the thermal comfort within
the building could be greatly improved. The AMLR matched category I in 95% of cases, whereas an overheating took place with the
standard strategy, with which only about 24% of cases were within
category I. Another advantage of the AMLR is its adaptability to

375

different buildings, which enormously simplied its commissioning. The elaborate parameterization of heating and cooling curves
is not necessary anymore. The algorithm automatically reacts to
changes in internal loads, which would entail a new parameterization of heating and cooling curves in the standard strategy.
In further steps, measured data are to be collected from other
periods so that a statement about the energy-saving potential of an
entire year can be made. In addition, the AMLR will be implemented
in other buildings. By calculating energy packages, the AMLR also
offers the option of load shifts without violating the thermal comfort within the building. This potential is to be quantied in other
investigations.
Acknowledgments
The AMLR has been developed within the scope of a PhD work.
The work has been supported by the Ministry of Science, Research
and Arts (MWK) of Baden-Wrttemberg, Germany, the German
Federal Ministry of Education and Research (BMBF) and the company IWB in the framework of an industry project.
Appendix

Fig. 9. Measured data from 01.12.2014 until 26.12.2014 for the room temperatures for three oors, supply temperature, ambient temperature and the thermal power of the
TABS from the IWB CityCenter with the standard TABS strategy.

376

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

Fig. 10. Measured data from 01.12.2015 until 26.12.2015 for the room temperatures for every oor, supply and return temperatures, ambient temperature and the thermal
power of the TABS from the IWB CityCenter with the AMLR TABS strategy.

References
[1] U.S. Department of Energys (DOE), Ofce of Energy Efciency and Renewable
Energy, 2011 Buildings Energy Data Book, 2012.
[2] European Union, Statistical Pocketbook 2014, 2014, http://dx.doi.org/10.
2833/24150, URL http://ec.europa.eu/energy/publications/doc/2014
pocketbook.pdf.
[3] G.P. Henze, C. Felsmann, D.E. Kalz, S. Herkel, Primary energy and comfort
performance of ventilation assisted thermo-active building systems in
continental climates, Energy Build. 40 (2) (2008) 99111, http://dx.doi.org/10.
1016/j.enbuild.2007.01.014, URL http://www.sciencedirect.com/science/
article/pii/S0378778807000333.
[4] S. Lohmller, C. Koob, F. Leisering, Das Marktvolumen fr Flchenkhlung in
Deutschland, HLH. Heizung, Lftung/Klima, Haustechnik (60) (2009) 5860.
[5] J.-H. Lim, J.-H. Song, S.-Y. Song, Development of operational guidelines for
thermally activated building system according to heating and cooling load
characteristics, Appl. Energy 126 (0) (2014) 123135, http://dx.doi.org/10.
1016/j.apenergy.2014.03.087, URL http://www.sciencedirect.com/science/
article/pii/S0306261914003274.
[6] J. Pfafferott, D. Kalz, R. Koenigsdorff, Bauteilaktivierung: Einsatz
Praxiserfahrungen Anforderungen, Fraunhofer IRB Verlag, Stuttgart, 2015.
[7] B.W. Olesen, Radiant oor-heating in theory and practice, ASHRAE J. (44)
(2002) 1926.
[8] B.W. Olesen, F.C. Dossi, Operation and control of activated slab heating and
cooling systems, in: Building for the Future: The 16th CIB World Building
Congress, 2004, URL http://www.irbnet.de/daten/iconda/CIB1743.pdf.
[9] B. Olesen, K. Sommer, B. Dchting, Control of slab heating and cooling
systems studied by dynamic computer simulations, ASHRAE Trans. 108 (2)
(2002) 110.
[10] S.H. Cho, M. Zaheer-uddin, An experimental study of multiple parameter
switching control for radiant oor heating systems, Energy 24 (5) (1999)

[11]

[12]
[13]

[14]

[15]

[16]

[17]

[18]

433444, http://dx.doi.org/10.1016/S0360-5442(98)00101-7, URL http://

www.sciencedirect.com/science/article/pii/S0360544298001017.
M. Sourbron, R.D. Herdt, T. van Reet, W. van Passel, M. Baelmans, L. Helsen,
Efciently produced heat and cold is squandered by inappropriate control
strategies: a case study, Energy Build. 41 (10) (2009) 10911098, http://dx.
doi.org/10.1016/j.enbuild.2009.05.015, URL http://www.sciencedirect.com/
science/article/pii/S0378778809001145.
R. Meierhans, Room air-conditioning by means of overnight cooling of the
concrete ceiling, ASHRAE Trans. (102-1) (1996) 693697.
A. Arteconi, D. Costola, P. Hoes, J.L.M. Hensen, Analysis of control strategies for
thermally activated building systems under demand side management
mechanisms, Energy Build. 80 (0) (2014) 384393, http://dx.doi.org/10.1016/
j.enbuild.2014.05.053, URL http://www.sciencedirect.com/science/article/pii/
S0378778814004770.
J. Tdtli, TABS control: Steuerung und Regelung von thermoaktiven
Bauteilsystemen, in: Handbuch fr Planung, Auslegung und Betrieb,
Schriftenreihe Technik, Faktor-Verl, Zrich, 2009.
J. Tdtli, M. Gwerder, F. Renggli, W. Gntensperger, B. Lehmann, V. Dorer, K.
Hildebrand, Regelung und Steuerung von thermoaktiven Bauteilsystemen
(TABS), Bauphysik 31 (5) (2009) 319325, http://dx.doi.org/10.1002/bapi.
200910042.
M. Gwerder, B. Lehmann, J. Tdtli, V. Dorer, F. Renggli, Control of
thermally-activated building systems (TABS), Appl. Energy 85 (7) (2008)
565581, http://dx.doi.org/10.1016/j.apenergy.2007.08.001, URL http://www.
sciencedirect.com/science/article/pii/S0306261907001328.
M. Sourbron, Dynamic thermal behaviour of buildings with Concrete Core
Activation, Katholieke Universiteit Leuven Faculty of Engineering, 2012,
Dissertation.
F. Oldewurtel, C. Sagerschnig, E. Z cekov,
S. Prvara, J. Cigler, Z. Vna,
Building modeling as a crucial part for building predictive control, Energy
Build. (2013) 822, http://dx.doi.org/10.1016/j.enbuild.2012.10.024.

M. Schmelas et al. / Energy and Buildings 129 (2016) 367377

L. Ferkl, J. Cigler, Model predictive control of a building

[19] S. Prvara, J. Sirok
y,
heating system: the rst experience, Energy Build. 43 (2) (2011) 564572,
http://dx.doi.org/10.1016/j.enbuild.2010.10.022.

F. Oldewurtel, J. Cigler, S. Prvara, Experimental analysis of model

[20] J. Sirok
y,
predictive control for an energy efcient building heating system, Appl.
Energy 88 (9) (2011) 30793087, http://dx.doi.org/10.1016/j.apenergy.2011.
03.009.
[21] M. Schmelas, T. Feldmann, E. Bollin, Adaptive predictive control of
thermo-active building systems (TABS) based on a multiple regression
algorithm, Energy Build. (103) (2015) 1428, http://dx.doi.org/10.1016/j.
enbuild.2015.06.012, URL http://www.sciencedirect.com/science/article/pii/
S0378778815300402.
[22] M. Schmelas, PrTABS Controller am Beispiel des Passiv-Seminargebudes der
Hochschule Offenburg, 37. Internationaler Uponor Kongress TGA Alles
geregelt?!, 2015, pp. 2736.
[23] MINERGIE, Der MINERGIE-Standard fr Gebude: Planung und
Projektierung, 2010, URL http://www.minergie.ch/tl les/download/
Publikationen/Praxisbeispiel/Der-Minergie-Standard-fuer-Gebaeude 2010.
pdf.
[24] SIA 416:2003, Flchen und Volumen von Gebuden.

377

[25] DIN EN 14511-1:2015-12, Luftkonditionierer, Flssigkeitskhlstze und

Wrmepumpen fr die Raumbeheizung und -khlung und Prozess-Khler
mit elektrisch angetriebenen Verdichtern Teil 1: Begriffe.
[26] B. Lehmann, V. Dorer, M. Gwerder, F. Renggli, J. Tdtli, Thermally activated
building systems (TABS): energy efciency as a function of control strategy,
hydronic circuit topology and (cold) generation system, Appl. Energy 88 (1)
(2011) 180191, http://dx.doi.org/10.1016/j.apenergy.2010.08.010, URL
http://www.sciencedirect.com/science/article/pii/S0306261910003260.
[27] A.K.d. Wit, C.J. Wisse, Hydronic circuit topologies for thermally activated
building systems - design questions and case study, Energy Build. 52 (0)
(2012) 5667, http://dx.doi.org/10.1016/j.enbuild.2012.05.010, URL http://
www.sciencedirect.com/science/article/pii/S0378778812002654.
[28] DIN EN 15251:2012-12, Eingangsparameter fr das Raumklima zur Auslegung
und Bewertung der Energieefzienz von Gebuden - Raumluftqualitt,
Temperatur, Licht und Akustik; Deutsche Fassung EN 15251:2007.
[29] DIN EN ISO 7730:2006, Ergonomie der thermischen Umgebung - Analytische
Bestimmung und Interpretation der thermischen Behaglichkeit durch
Berechnung des PMV- und des PPD-Indexes und Kriterien der lokalen
thermischen Behaglichkeit.
[30] VDI 3807, Verbrauchskennwerte fr Gebude.