The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
UTILIZING BUILDINGS AS SHORT-TERM THERMAL ENERGY STORAGE
1
2
J. Kensby , A. Trüschel , and J.-O. Dalenbäck
1
2
1
Chalmers University of Technology, Gothenburg, Sweden
Department of Energy and Environment, Division of Building Services Engineering
ABSTRACT
Heat demand in a district heating system can exhibit
significant variation within one day, which sets
problematic conditions for efficient heat generation.
Short-term thermal energy storage can decrease this
daily variation and make the conditions for generating
heat more favourable. By periodically overheating and
under-heating buildings, causing small variations in the
indoor temperature, their thermal inertia can be utilized
as short-term thermal energy storage. This study
presents the results from a pilot test where the potential
to function as short-term thermal energy storage was
tested in five multifamily residential buildings in
Gothenburg, Sweden. These results are then up-scaled
to study the consequences for a whole-district heating
system from a large-scale implementation. The signal
from the outdoor temperature sensors in the test
buildings were adjusted in different cycles over a total
of 52 weeks. The delivered heat and indoor
temperature were measured during the test. The
results show that heavy buildings with a structural core
of concrete can tolerate relatively large variations in
heat delivery while still maintaining a good indoor
climate. Storing 0.1 kWh/m2floor area of heat will very
rarely cause variations in indoor temperature greater
than ±0.5°C in a heavy building. Utilizing about 500
substations for short-term thermal energy storage in
large residential buildings would provide capacity for
storing heat equivalent to that of a hot water storage
tank with a volume of 14,200 m3 for the city of
Gothenburg. This would decrease the daily variations
in heat load by 50%, reduce the need for peak heat
generation, and reduce the number of starts and stops
of heat-generation units.
INTRODUCTION
The heat generation in district heating (DH) systems is
mainly demand driven. When customers increase their
heat consumption, the heat supplier must increase the
heat generation or the temperatures in the distribution
network will drop. The heat demand can exhibit
significant variation within short periods and, hence,
heat generation, causing many starts and stops of heat
sources. This gives the heat supplier limited freedom to
control the heat generation. These conditions can
drastically reduce the ability to plan and control heat
generation, hence reducing efficiency. With a larger
variation in heat load comes a larger need for peak
heat generation, which often runs on fossil fuels with
large operational costs and high environmental impact.
Short-term thermal energy storage (TES) can decouple
heat demand and heat generation in district heating
systems and have many positive effects, such as the
following:
•
•
•
•
•
•
•
Reduced load variation
Better fuel economy
Fewer starts and stops in heat generation
Increased security of supply
Less need to invest in peak load heat generation
Operate combined heat and power plants (CHP)
according to electrical price
Operate heat pumps and direct electrical heaters in
DH systems according to electrical price
As presented in [1], there are mainly four different
strategies for short-term TES in district heating
systems: hot water storage tanks; phase change
materials (PCM); varying temperature in the DH
network; and utilizing building thermal inertia. This
study has focused on utilizing building thermal inertia
for short-term TES in district heating systems. Such
strategies are commonly included in the term “demand
side management” (DSM).
The purpose of this study is to evaluate the potential for
load shifting in a large-scale implementation of
buildings as short-term TES in a DH system. “Short
term” can, in this case, be defined as normally a few
hours, but a period of up to a few days is also possible.
A few studies known to the authors have treated this
subject, but this is the first study to base a full-scale
simulation on the actual behavior of buildings in a pilot
test. Earlier work has either assumed the full-scale
effects [2], focused on nighttime setback [3] or on
energy-saving potential [4]. The energy-saving
potential in buildings utilized as short-term TES
probably comes from a reduction of excessive
temperatures due to the implementation of indoor
temperature measurements for control of the building
heating system. This study is also the first study known
to the authors to make a comparison of utilizing
building thermal inertia and hot water storage tanks for
short-term TES purposes.
The study can be divided into mainly two parts. The
first part focuses on determining the potential for shortterm TES in individual buildings. A pilot test has been
carried out to determine what quantities of thermal
energy can be stored and how the indoor climate is
affected. A more in-depth study of this pilot test can be
found in [1]. The second part scales the results from
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
the first part and studies how an overall DH system
could benefit from utilizing buildings as short-term TES.
STATE OF THE ART
For the purposes of this study, it is most relevant to
study residential buildings with a large thermal mass,
e.g. concrete structures, that are heated by radiator
systems connected to DH. The reasons are that such
buildings are common in many large DH systems and
that they are well suited for utilization as short-term
TES. A large thermal mass has been shown to
increase a building’s suitability as short-term TES in [1,
5]. Buildings with radiator heating systems are better
suited for short-term TES than buildings with airborne
heating systems, according to [6].
The most common way of describing how the indoor
temperature is affected by an increase or decrease in
the heat delivered is to assume a correlation based on
a time constant. It was shown in [1] that this measure
does not reflect reality in an accurate way. It is,
however, true that the indoor temperature in a building
with a higher measured time constant will be less
affected then in a building with a lower measured time
constant for the same relative change in delivered heat.
The progression of the indoor temperature in such a
model is, however, inaccurate and may lead to overusage or under-usage of a building’s capacity for
storing heat if implemented in the control method.
Earlier pilot tests
Utilizing building thermal inertia as short-term TES in a
district heating system is not a new concept. The oldest
pilot test known to the authors is from 1982 [7]. The
main aim of this test was to increase the supply
security for the heat customers located farthest away
from a heating plant in case of a shortage. Eighty
residential and office buildings located in Stockholm,
Sweden participated, and their heat deliveries were
remotely reduced by a control system. The magnitude
and durations of the reduced heat deliveries were
based on assumed time constants for the buildings and
a maximum accepted drop in indoor temperature of
3°C. The indoor temperature was measured in two of
the buildings. The variations were at a normal level
except during the test with the longest duration (48 h).
Another pilot test was conducted during the winter of
2002–2003 in two Finnish buildings with concrete
structures and radiator heating systems [6]. The test
revealed that the heat load could be reduced by 20–
25% over 2–3 h, causing a drop in indoor temperature
of up to 2°C. These tests were performed at outdoor
temperatures of -10°C to 0°C. The same study
demonstrated a smaller potential for load shifting in a
building complex consisting of offices and facilities for
streetcar maintenance in Mannheim, Germany. The
peak demand for heating was reduced by 4.1% during
the tests. The main reason for the lower potential was
that the heating system was mainly airborne. The main
aim of these tests was to evaluate the potential for the
reduction of peak load production in the district heating
system. This was also the main focus of the
subsequent studies presented here.
A residential area in Karlshamn, Sweden, was the
subject of a pilot test where DSM was implemented in
the form of agent-based load control [8, 9]. The control
was distributed among agents on the production side,
on a cluster level, and on a customer level. These
agents monitored and controlled the local systems.
They also communicated with each other to achieve
system-wide peak reduction and optimization. The
system displayed the potential for reducing peaks as
well as reducing the energy consumption by 4%, even
though the thermal storage capacity was only partly
utilized in this test. The average return temperatures to
the district heating system were also reduced by 2°C
while the system was in operation [10]. A subsequent
larger test of this technology was performed in three
major Swedish district heating systems [11]. A total of
58 substations serving one to several buildings each
were included in this test. Peak load reductions of
approximately 15–20% and energy savings of 7.5%
were achieved.
The effect of the utilization of buildings for short-term
TES on the indoor temperature was studied in [12]. The
test was performed in an office building with a light
construction and concrete slabs. The heat load was
reduced during short periods of up to 1 h and longer
periods of 4 to 8 h. Both single and frequently recurring
heat load reductions were tested. The average
deviation was chosen as the measurement for the
variation in indoor temperature. During periods with
load reductions, the average deviation increased to
0.29°C from the normal 0.19°C.
A study with the aim of estimating the possible heat
storage potential of different building types was
conducted in Gothenburg, Sweden [5]. The heat
deliveries to the different buildings were reduced over
periods of 24 h, and the heat deliveries and indoor
temperatures were measured. Time constants for each
building
were
calculated
based
on
these
measurements. Wooden buildings reported time
constants of 102 h, stone buildings 155 h, and tower
blocks 218 to 330 h.
Large-scale implementation
The effects of the large-scale implementation of
buildings’ thermal inertia as short-term TES in district
heating systems has been studied in a few
publications. They have adopted very different
approaches.
A case study of how the implementation of DSM would
affect the fuel and operational costs of the DH system
in Næstved, Denmark was included in [2]. Two cases
were considered where the heat load was assumed to
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
be adjusted by 20% and 80%, respectively, toward the
mean heat load. They resulted in total savings of 1%
and 2.6%.
District heating systems where a considerable number
of the buildings utilize nighttime setback can have large
peaks in heat demand in the morning hours. A
simulation study regarding the DH network of
Altenmarkt im Pongau, Austria studied the effects of
applying DSM strategies to buildings utilizing nighttime
setback [3]. The buildings were controlled so that they
recovered from their nighttime setback at different
hours. Up to 35% peak shaving would be achieved if
applied to the overall district heating network.
The effects of three energy conservation measures on
the local energy system in Linköping, Sweden were
compared in [4]. The compared measures were heat
load control (utilizing buildings’ thermal inertia), attic
insulation, and electricity savings. Heat load control
showed a potential for energy savings primarily in the
spring and autumn. It would also be economically
profitable for both the DH provider and the residents.
The analyzed installation for heat load control was
described in [11].
METHODOLOGY
Since this study consists of several parts, their
methodologies are described here separately, starting
with the pilot test. This is followed by a description of
how the results from the pilot test are up-scaled to
describe a large-scale implementation of buildings as
short-term TES. Last follows a description of how the
large-scale implementation is simulated and evaluated.
Description of test buildings
During 2010 and 2011, the ability of five buildings to
function as thermal energy storage in a district heating
system was tested in Gothenburg, Sweden. The five
buildings that were included in the analysis are all
residential buildings with 3 to 5 stories. A summary of
the building data is presented in Table 1. There are
some differences in the buildings, and they can be
grouped into two categories: light and heavy. This
classification is based on the thermal mass of the
building. A light building typically has a core of steel or
wood, which results in a low capacity for storing heat. A
heavy building typically has a core of concrete, which
results in a higher capacity for storing heat. One of the
buildings can be classified in the light category. All of
the buildings were constructed between 1939 and 1950
and have a yearly heating demand of approximately
150 kWh/m2floor area per year. This is a normal energy
performance for these types of buildings in the city of
Gothenburg, Sweden, which has a yearly average
temperature of 8°C. A major portion of the large public
housing stock that was built in the 1960s and 1970s is
similar to the buildings tested in this study regarding
energy performance [13]. More recently constructed
buildings generally have superior energy performance.
Table 1. Building Data.
Building
A
B
C
D
E
Year of
construction
1950
1939
1934
1939
No info
Living area
2
[m ]
1,178
904
900
904
No info
Stories
3
5
3
5
3
Apartments
20
24
19
24
25
Heavy
Light
Heavy
Heavy
Plastered
Wood,
brick
Brick
Brick
Estimated
Heavy
thermal mass
Facade
Plastered
The heat deliveries to the buildings were increased and
reduced during specified periods, and the indoor
temperature, T, was measured in two apartments in
each building. Temperature meters were placed on a
wall in the hall in each apartment. All buildings were
connected to district heating and had a radiator heating
system.
Control of test building
All of the buildings in the pilot test adjusted the heating
power by controlling the supply temperature to the
radiator system using a conventional feedback
controller. The supply temperature was set based on
the outdoor temperature and a control curve. Fine
adjustment of the heating power within each individual
apartment was performed via thermostats on the
radiators. To control the heating power delivered to the
buildings in this test, the signal from the outdoor
temperature sensor, u, was adjusted in different cycles
as shown in Fig. 1. This affected the set point for the
water supply to the radiators in the feedback controller.
For example, to discharge a building, 7°C was added to
the outdoor temperature signal. The real outdoor
temperature was 3°C, but the control system receive
the signal 10°C (3°C + 7°C). According to the control
curve, this resulted in a lower supply temperature to the
radiator system. The apartments then received radiator
water with a lower temperature than they needed to
maintain their indoor temperature, T, at the current
outdoor temperature. The indoor temperature, T, slowly
started to drop in the apartments, and the building
affected the district heating system, similar to
discharging a hot water storage tank. This test setup
was similar to the one used in [12].
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
15
10
5
0
-5
-10
u
15
10
5
0
-5
-10
u+ u
T
Cycle I
8
6
4
2
0
-2
-4
-6
-8
∆u [°C]
u
∆u [°C]
15
10
5
0
-5
-10
0
9 12 15 18 21
t [h]
0
Building heating system
Fig. 1. Schematic sketch of how the control was
implemented in the pilot test.
Five different cycles of charging and discharging were
tested; they are shown in Fig. 2. The following
notations are used to describe them:
CP—Charge period; the building receives more heat
than it normally would at the current outdoor
temperature.
DP—Discharge period; the building receives less heat
than it normally would at the current outdoor
temperature.
NOP—Normal operation period; the building heating
system operates as it normally would.
Δu—Adjustment to outdoor temperature signal
Cycle II was the most extensively tested. It was tested
in all five buildings and produced 19 complete weeks of
measurement data without any obvious measurement
errors. Cycle II is also the cycle with the largest
variation in Δu and therefore should be the cycle that
provided the largest utilization of the building’s thermal
energy storage capacity and produced the largest
variations in indoor temperatures.
0
∆u [°C]
In this test, the adjustments to the outdoor temperature
signal were performed in 21-h cycles. Most of the
tested control cycles contained one 9-h period of
discharging, one 9-h period of charging, and one 3-h
period of normal operation. The reason to use a test
cycle that was 21 h (and not 24 h) was that this caused
the charging and discharging to occur at different times
each day. This made it possible to separate variations
in indoor temperature caused by the test from normal
variations caused by, e.g., sunlight and the tenants’
behavior. Eight cycles of 21 h make one full week.
Cycle III
8
6
4
2
0
-2
-4
-6
-8
∆u [°C]
Indoor
temperature
sensors
Q
District heating system
6
3
6
9 12 15 18 21
t [h]
Controller
∆u [°C]
Outdoor
temperature
sensor
3
T
Cycle II
8
6
4
2
0
-2
-4
-6
-8
3
6
9 12 15 18 21
t [h]
Cycle IV
8
6
4
2
0
-2
-4
-6
-8
0
3
6
9 12 15 18 21
t [h]
Cycle V
8
6
4
2
0
-2
-4
-6
-8
0
3
6
9 12 15 18 21
t [h]
Fig. 2. The five test cycles from the pilot test.
Large-scale implementation
To study a large-scale implementation of short-term
TES in buildings, a group of buildings suitable for
implementation needs to be analyzed. For this
purpose, Västra Gårdsten, a residential area in
Gothenburg, was selected. The area has 13
substations, each supplying heat to a group of 2–3
buildings. There is a total of 1,000 apartments in the
area with an average living area of 76 m². The average
annual heat consumption for the area is 12.1 GWh.
The buildings are all residential except for one small
dental practice and one office for about 20 persons. All
buildings are 3 to 5 stories and have a core of
concrete. They are very similar to the heavy buildings
in the pilot test. This building type is also very common
in Sweden, as many large residential areas similar to
Västra Gårdsten were built in the 1960s and 1970s.
Due to their similarities, it is assumed in this study that
the buildings in Västra Gårdsten will perform identically
to the heavy buildings in the pilot test with regard to the
ability to function as short-term TES. To scale the
results from pilot test to Västra Gårdsten, the heating
power signature is used. The heating power signature
is the heat demand dependency of the outdoor
temperature. It is determined by finding the linear
dependency with the smallest squared error based on
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
three years of measurements of the delivered heat and
the outdoor temperature.
For the full-scale study of the potential of buildings as
short-term TES in DH networks, the city of Gothenburg
is studied. The study consists of four cases: buildings
consuming 10% of the yearly heat generation in the
Gothenburg DH network, the same case but with 20%,
30%, and a reference case with no thermal storage. It
is assumed that there are enough residential areas
similar to Väsra Gårdsten that can be utilized for shortterm TES in Gothenburg to cover the four cases. This
evaluation is based on a comparison of the actual heat
generation (the reference case) and fictive heat
generation that would be possible with access to
thermal storage. The heat generation data are from the
DH system in Gothenburg from 2010–2012.
Properties of a short-term TES
Based on the findings from the pilot test and the
heating power signature of Västra Gårdsten, two main
parameters defining the short-term TES can be
established for each case:
•
•
Power limitation [MW]—The maximum power that
the short-term TES can be charged/discharged
with.
Storage capacity limitation [MWh]—The amount of
heat that the short-term TES can store.
The power limitation of the storage is valid when the
outdoor temperature is 8°C or less. When the outdoor
temperature is higher, the heating power in the
buildings is less than the limitation. Since negative
heating power cannot be delivered to the buildings, the
power limitation for discharging decreases linearly from
8°C to 15°C, where it reaches zero. Charging of the
buildings can still be done at higher temperatures, and
the power limitation then starts to decrease linearly at
15°C and reaches zero at 22°C.
Simulation model
An optimization problem was formulated with the aim of
minimizing the variation in heat load. The parameter to
be optimized is the maximum peak load reduction, and
the limitations for adjustments to the heat load are the
power limitation and the storage capacity limitation.
With a resolution in time of 1 h and a heating power of
1 MW, the number of solutions is small enough to be
solved with a brute force iteration approach, testing all
possible solutions.
The progression for the iterative solution is first to split
the data set into periods of 200 h each to speed up the
simulation. For each time period, the highest hourly
heat load Ph(t) is reduced by one step (1 MW) and the
lowest hourly heat load Ph(t) is increased by one step
(1 MW). A check is performed to see whether any of
the two limitations was violated. If the check passed the
iteration, the test was started over, and if not, the
program proceeded to test all combinations of
decreasing points in descending order where Ph(t) >
Ph(t-1) and/or Ph(t) > Ph(t+1) and increasing points in
ascending order where Ph(t) < Ph(t-1) and/or Ph(t) <
Ph(t+1). The iteration continued until no further
improvements could be made. The method is quite
computational heavy (solving at about 10,000 times
real time) but guarantees a solution with the maximum
possible peak reductions. To avoid boundary
constraints from the 200-h periods influencing the
results, the full iteration was performed a second time
with overlapping time periods.
Evaluation of results
To evaluate the results, the relative daily variation was
studied for each case. Relative daily variation is
defined in [14] as follows:
“The relative daily variation is the accumulated positive
difference between the hourly average heat load and
the daily average heat load divided by the annual
average heat load and the number of hours during a
day. The relative daily variation is expressed with 365
values per system and year.”
1
Gd = 2
∑24
h=1|Ph −Pd |
Pa ∙24
∙ 100 [%]
(1)
Ph—hourly heat load
Pd—daily heat load
Pa—annual heat load
Gd—relative daily variation
RESULTS
All the heavy buildings in the test showed that it is
possible to utilize them as short-term TES with the
restrictions from Cycle II and still maintain a good
indoor climate. Cycle II is the cycle with the largest
variation in Δu and therefore should be the cycle that
provided the largest utilization of the building thermal
energy storage capacity and produced the largest
variations in indoor temperatures.
Effect on heat delivery in the pilot test
The relation between the heat delivered to the
buildings, Q, and the adjustment to the outdoor
temperature signal, Δu, was studied. Cycles I, II, and V
in Building A were selected for this study, as these
tests have the most measurement data available. To
separate the variation in Q caused by the test from the
normal variations, an average profile for each week
was created based on the eight cycles of 21 h each. To
make different periods of time comparable, the average
heating power for the present week, Qmean, was
subtracted from each weekly profile. Graphs showing
these heating profiles are presented in Fig. 3.
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
Table 2. Ratio between heat deliveries and adjustment to
the outdoor temperature signal for Building A.
Relative heat deliveries to Building A (heavy)
20
Q - Qmean [kW]
Period
10
Δu–Δumean
Q–Qmean
(Q–Qmean)/
(Δu–Δumean)
[°C]
[kW/°C]
[kW/°C]
0
Cycle I
-10
DP
-20
NOP
4.00
CP
0
3
6
9
12
15
18
n/a
n/a
n/a
5.49
-1.83
21
Cycle II
DP
7.00
-11.84
-1.69
CP
-7.00
12.71
-1.82
NOP
0.00
-2.62
NaN
Adjustment to outdoor temperature signal
8
Cycle V
4
2
DP
5.29
-7.63
-1.44
0
CP
-4.71
7.10
-1.51
NOP
-1.71
1.58
-0.92
-2
Cycle I
-4
Cycle II
-6
Cycle V
Indoor temperature variation in the pilot test
-8
0
3
6
9
12
15
18
21
Time within cycle [h]
Fig. 3. Heat delivery profiles relative to control cycle
average for Building A. Each profile is based on 5–8
weeks of measurements.
Based on the same dataset used to create Fig. 3, a
more qualitative analysis of how the heat delivered to
the buildings, Q, relates to the adjustment to the
outdoor temperature signal, Δu, was performed. The
ratio between relative heat deliveries, Q–Qmean, and the
relative adjustment of the outdoor temperature signal,
Δu–Δumean, was studied. The average adjustment to the
outdoor temperature signal during one control cycle,
Δumean, is defined in equation 2 The reason for studying
Δu—Δumean instead of just Δu is that Cycles I and V are
not symmetrical like Cycle II, and their Δumean ≠ 0.
Δumean =
ΔuDP ×tDP +ΔuCP ×tCP +ΔuNOP ×tNOP
tDP +tCP +tNOP
[°C]
(2)
Δumean = 0°C for Cycle II (because it is symmetric),
1.79°C for Cycle I, and 3°C for Cycle V. The results
from the study are presented in Table 2. As observed
in this table, the ratio is between -1.44 kW/°C and -1.83
kW/°C for all but two cases. The NOP for Cycle V
differs from the other results. This is most likely
because the length of this period was only 3 h and
approximately half of that time is the time it takes for
the water in the radiator system to circulate once and
reach steady supply and return temperatures. The fact
that all other values have only a small variation in their
ratios implies that the change in heat deliveries and the
adjustment of the outdoor temperature signal, Δu, have
close to a linear relation.
To separate the variations in indoor temperature, T,
caused by the test from the normal variations, an
average indoor temperature profile for each week was
created based on the eight cycles. These profiles were
created in the same manner as the profiles for the
relative heating power in Fig. 3. An example of these
profiles in one of the heavy buildings is presented in
Fig. 4. As only the variation in temperature is of
interest, the temperature profiles have been made
more easily comparable by centering them on the Yaxis. Thus, Tmin + (Tmax – Tmin)/2 is subtracted from each
weekly profile.
Relative indoor temperature in Building A (heavy)
0,4
0,3
Indoor temperature difference [°C]
∆u [°C]
-1.83
-3.00
Time within cycle [h]
6
-7.32
0,2
0,1
0
-0,1
Cycle I
Apartment 1
Cycle II
Apartment 1
Cycle II
Apartment 2
Cycle V
Apartment 2
-0,2
-0,3
-0,4
0
3
6
9
12
15
Time within cycle [h]
18
21
Fig. 4. Indoor temperature variations caused by the pilot
test. Each curve is based on the average values over a
period of 4–6 weeks.
From Fig. 4, it can be observed that the effect on
indoor temperature, T, from Cycle II is, as expected,
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
larger than the effect from Cycle I and Cycle V. It
should also be noted that there is a time delay from
when the mode of operation changes and when the
effect on the indoor temperature, T, starts to be
displayed. This is due to the circulation time in the
radiator heating system; it takes some time before the
temperature front reaches the radiators and affects the
indoor temperature. Thus, dead time is added to the
system.
For each week in each apartment in each building, the
indoor temperature variation, Tvar.21h, caused by the
pilot test was calculated. Tvar.21h is defined as the
difference between the maximum and minimum
temperature for a weekly 21-h profile divided by two. A
summary of the variations is presented in Table 3.
from Cycle II and three important relations between the
adjustment to the outdoor temperature signal, Δu, heat
delivery, Q, and indoor temperature, T, that were
presented in [1]. With the restriction of |Δu| < 7°C, the
thermal energy storage capacity can then be measured
in degree hours [°Ch]. The thermal energy storage
capacity of the heavy building in this case is then
simplified to 63°Ch (7°C × 9 h).
To find the power limitation [MW] and storage capacity
limitation [MWh] for a large-scale, short-term TES, we
need to combine the results from the pilot test with the
heating power signature for the intended building stock.
The power signature for Västra Gårdsten, which has an
average annual heat consumption of 12.1 GWh, is
shown in Fig. 5.
Table 3. Average variation in indoor temperature caused
by the pilot test.
Heating power signature for Västra Gårdsten
5
Building
I
A—heavy
±0.26
-
8
C—light
±0.23
±0.39
18
A—heavy
±0.40
±0.40
6
B—heavy
±0.29
±0.29
6
D—heavy
±0.09
±0.19
5
E—heavy
±0.06
±0.27
1
III
E—heavy
±0.11
±0.22
2
IV
E—heavy
±0.06
±0.10
1
V
A—heavy
-
±0.30
5
II
Tvar.21h
Tvar.21h
Number of
Apartment 1 Apartment 2 test weeks
[°C]
[°C]
As shown in Table 3, all four heavy buildings have
average values for the indoor temperature variation
caused by Cycle II of ±0.40°C or less. If we look at
each individual week, there is only one week in one of
the apartments in one of the buildings that caused
variations in indoor temperature larger than ±0.50°C. In
that case, the variation was ±0.53°C. It is unlikely that
the variation caused by Cycle II combined with the
normal variations will cause a total variation in indoor
temperature larger than ±1.0°C on any given day.
Therefore, most heavy buildings similar to those
studied in the test should be able to utilize their thermal
inertia with restrictions similar to those in Cycle II
without jeopardizing the quality of service provided by
the heating system. Hence, the restrictions from Cycle
II are used to decide the limiting parameters for the fullscale simulation.
Limiting parameters of a building’s short-term TES
When utilizing buildings as thermal energy storage, it is
beneficial to have more freedom in the control than
what is entailed by Cycle II. It might be beneficial to
have several shorter DPs on one day or store heat
from one day to utilize the next day. A simple yet
accurate enough model for utilizing buildings as shortterm TES can be established based on the restrictions
Data points
4
Heat demand [MW]
Test
cycle
2.28
- 0.13
Linear
( ) × Tout
3
2
1
0
-20
-10
0
10
Outdoor temperature [°C]
20
30
Fig. 5. The inclination of the trend line is the heating power
signature for the residential area Västra Gårdsten,
Gothenburg.
Fig. 5 shows that an increase in the outdoor
temperature of 1°C would result in a decrease in the
heat delivered to the area of 0.13 MW. Hence, an
increase in Δu of 1°C would result in a decrease in the
heat delivered to the area of 0.13 MW. With the
limitations of |Δu| < 7°C and 63°Ch of thermal storage
capacity, this area could be utilized as thermal storage
with a power limitation of 0.13 MW/°C × 7°C = 0.91 MW
and a storage limitation of 0.13 MW/°C × 63°Ch = 8.19
MWh. This corresponds to a storage limitation of about
0.1 kWh/m2floor area. These substations, like many others,
already have a data connection that sends
consumption data each hour. All that is required to
utilize this area as thermal storage is adjusting the 13
substations so that the adjustment to the outdoor
temperature signal, Δu, can be controlled remotely.
There are many areas similar to Västra Gårdsten in
Gothenburg (and in other cities), so it is possible to
scale these results for a city-wide implementation.
Full-scale implementation
Since the cost of implementing building short-term TES
is proportional to the number of substations that need
adjustments, it is better to utilize the substations with
the largest yearly heat demand first. From 2010–2012,
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
the DH system in Gothenburg had an average annual
heat generation of 4.26 TWh. The total amount of
delivered heat to customers was 4.04 TWh, of which
2.12 TWh was delivered to the 4,457 substations in
multifamily residential buildings. The heat deliveries to
these substations are sorted in Fig. 6 to find the
required number of substations for each case
presented in Table 4.
Total heat generation in DH system [MW]
800
700
600
10%
500
30%
Adjustment to outdoor temperature signal, Δu [°C]
8
2,5
50
2
40
1,5
30
1
10%
20%
0,5
20
10
30%
0
4
Part of total heat genertaion [%]
Cumulative yearly heat delivery [TWh]
20%
400
Cumulative yearly heat deliveries
in multifamily residential buildings
0
-4
-8
0,25
500 1000 1500 2000 2500 3000 3500 4000
0
Number of substations
Fig. 6. Cumulative yearly heat deliveries to all substations
in multifamily residential buildings in Gothenburg, Sweden.
-0,25
Based on the data from Fig. 6 and the parameters
found from the study of Västra Gårdsten, the five
simulation cases can now be summarized in Table 4.
-0,5
Table 4. Summary of the four simulation cases.
Case
0% (ref)
Effect on indoor temp in most affected buildings [°C]
0,5
0
0
10%
0% (ref)
Yearly heat
delivery to
utilized
substations
[GWh]
Number of
utilized
substations
Power
limitation
Storage
capacity
limitation
[MW]
[MWh]
0
12
24
36
48
60
72
84
96
108
120
Time [h]
Fig. 7. Utilizing buildings as short-term TES in Gothenburg
DH system; effect on heat generation, outdoor
temperature signal and indoor temperature in utilized
buildings.
Relative daily variation
0
0
0
0
426
165
32
285
20%
852
507
63
571
30%
1,279
1046
95
856
Based on the data from Table 4 and the hourly heat
generation data for Gothenburg from 2010–2012, a fullscale simulation was performed. An example showing
a five day period of the results can be found in Fig. 7.
Around the 24 h mark in Fig. 7 the short-term TES is
discharging for the 10%, 20% and 30% cases. This can
be seen from the reduced heat generation in the top
graph, the positive Δu in the middle graph and the
falling indoor temperature in the bottom graph. The
effect on indoor temperature in the bottom graph is
estimated for the most affected buildings, those similar
to Building A in the pilot test. The bottom graph can
also be seen as the charging level of the TES, where at
+0.5°C, the storage is charged to its full storage
capacity limitation.
It can be clearly seen in Fig. 8 that the variation in heat
generation has decreased and that the conditions for
generating heat are more favourable with a building’s
short-term TES.
Relative daily variation cumulative distribution function [%]
14
0% (ref)
12
10%
10
20%
8
30%
6
4
2
0
0
200
400
600
Time [days]
800
1000
Fig. 8 Relative daily variation cumulative distribution
function for three years (2010–2012).
It can be seen in Fig. 8 that the relative daily variation
has significantly decreased when some of the buildings
are utilized as thermal storage. The decrease from no
storage to 10% is larger than the decrease from 10% to
20%. This is because, in some cases, 10% is enough
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
to cut a peak, and there is then no need for larger
storage. The utilization is then larger for the smaller
storage of 10%. This can also be seen in the middle
graph in Fig. 7 where the buildings more often receive
a stronger control signal in the 10% case than in the
20% and 30% cases.
If we look at the average values of the relative daily
variation, we get a simple measurement for comparing
the four cases:
0% (reference):
10%:
20%:
30%:
3.63%
2.44%
1.74%
1.38%
In the 20% case, the average relative daily variation is
reduced by 50% compared to the reference case. This
comes at the cost of increasing the variation in indoor
temperature in the customers’ buildings in most cases
by less than ±0.5°C and the investment in adjusting the
substations. This should be compared with the value of
reducing the variation in the heat generation and other
storage options.
For the specific case of Gothenburg, the 20% case
would require adjustments in about 500 substations.
This can be compared with investing in a storage tank
with a storage capacity of 576 MWh and a power
limitation of 64 MW. With a supply temperature of 80°C
and a return temperature of 45°C, this would result in a
hot water storage tank of 14,200 m3 with a maximum
flow of 0.44 m3/s. Such a storage tank would have an
investment cost of roughly 3–6 M€. This estimation
includes all related costs required to get the storage
tank in operation and is based on interviews with three
Swedish district heating companies. Assuming that the
required adjustments to the substations can be made
cheaper than 6,000–12,000€ per substation, utilizing
buildings as short-term TES can be a more economical
alternative than hot water storage tanks. It should,
however, be noted that this is a very rough economic
comparison that only includes the investment cost.
DISCUSSION
The capacity for storing heat in buildings is highly
dependent on the restrictions on the indoor
temperature variation. No study of how the tenants
experienced the indoor climate during the pilot test was
carried out, but the landlords all stated that the
complaints were at a normal level. For this study, the
restriction of not increasing the variation in indoor
temperature more than ±0.50°C was used. These
variations are small compared to the normal variation
caused by variations in sunlight and tenant activity.
Allowing a larger variation in the indoor temperature
would increase the thermal storage capacity in the
buildings. What limits the potential to utilize buildings
as short-term TES is how the tenants experience the
indoor climate. Here, the operative temperature and
sociological factors are of importance. The operative
temperature in an apartment is mainly affected by the
indoor air temperature and surface temperatures. This
puts a power limitation on the charging/discharging of
buildings to avoid radiators that are too warm or cold.
What sets the power limitation for the utilization as
short-term TES might not even be the thermal comfort
of the tenants. It could be tenants experiencing
unreasonable hot or cold radiators, believing that
something is wrong with the heating system, and
complaining or adjusting their thermostats in an
unfavorable way. In the simulation, especially the 20%
and 30% cases, it was the power limitation rather than
the storage capacity limitation that restricted the
capacity.
The potential for a large-scale implementation might be
underestimated since it is based on results from the
worst performing building in the pilot test. The benefit is
that buildings could be utilized to the calculated
potential without risking large variations in indoor
temperatures. For a higher degree of utilization, it might
be recommended to measure the indoor temperature in
the buildings continually and implement it in the control.
OUTLOOK
Utilizing buildings as short-term TES can have great
potential as a cost-effective method for storing heat,
but there are two main potential obstacles that need to
be studied further before this technology can take the
step from pilots to large-scale applications.
First is the method for controlling such storage. Several
practical methods for controlling thermal energy
storage utilizing buildings’ thermal inertia have been
studied in earlier works. They include direct load
control [1, 6, 7, 12, 15], control through price incentives
[6, 15-17], and other more indirect DSM strategies [4,
6, 8-11, 15, 17, 18].
Second is what type of business model is to be used
and what the contract between the DH providers and
the customers should contain. This can be difficult
since there might be three parties involved: the DH
provider, the customer/landlord, and the tenants.
Questions that need to be addressed include the
following:
•
•
•
•
Who is responsible for the indoor climate when
load control is active?
How should the investment cost be financed?
Is it beneficial to combine installations of short-term
TES with energy-efficiency measures?
How (if at all) should customers be compensated
for allowing the DH provider to utilize their buildings
as short-term TES?
Both these obstacles should be easy to overcome if a
major DH provider is willing to invest in a large-scale
building short-term TES. It is hoped that such an
investment in a technical solution and business mode
can be exported to other DH providers.
The 14th International Symposium on District Heating and Cooling,
September 7th to September 9th, 2014, Stockholm, Sweden
CONCLUSIONS
The pilot test in this study has shown that heavy
buildings with a structural core of concrete can tolerate
relatively large variations in heat deliveries while still
maintaining a good indoor climate. Storing 0.1
kWh/m2floor area of heat will very rarely cause variations
in indoor temperature larger than ±0.5°C in the most
affected heavy buildings. This corresponds to adjusting
the outdoor temperature signal, Δu, by 7°C over 9 h.
Most heavy buildings will experience even smaller
variations in indoor temperature and could possibly be
utilized to a larger extent than the parameters found in
the pilot test.
Utilizing about 500 substations for short-term thermal
energy storage in large residential buildings would
provide a capacity for storing heat equivalent to
constructing a hot water storage tank with a volume of
14,200 m3 for the city of Gothenburg, Sweden. This
would decrease the daily variations in heat load by
50%, reduce the need for peak heat generation, and
reduce the number of starts and stops of heat
generation units. Assuming that the required
adjustments to the substations can be made cheaper
than 6,000–12,000€ per substation, utilizing buildings
as short-term TES can be a more economical
alternative than hot water storage tanks.
ACKNOWLEDGEMENT
8.
9.
10.
11.
12.
13.
This work was financially supported by Göteborg
Energi AB (Gothenburg Energy).
REFERENCES
1.
2.
3.
4.
5.
6.
7.
J. Kensby, A. Trüschel and J.-O. Dalenbäck,
Potential of residential buildings as thermal
energy storage in district heating systems —
Results from a pilot test (accepted for
publication), Applied Energy, 2014.
M. Wigbels, B. Bohm and K. Sipilaek, Dynamic
heat storage optimisation and demand side
management, IEA DHC CHP, 2005.
D. Basciotti and R.R. Schmidt, Demand side
management in district heating networks:
Simulation case study on load shifting,
Euroheat and Power (English Edition), Vol. 10,
pp. 43–46, 2013.
K. Difs, et al., Energy conservation measures
in buildings heated by district heating — A
local energy system perspective, Energy, Vol.
35, pp. 3194-3203, 2010.
L.C. Olsson Ingvarson and S. Werner, Building
mass used as short term heat storage, in the
11th International Symposium on District
Heating and Cooling, Reykjavik, Iceland, 2008.
S. Kärkkäinen, et al., Demand side
management of the district heating system,
VTT research notes 2247, Espoo, Finland,
2003.
B. Österlind, Effektbegränsning av fjärrvärme:
Försök med centraliserad styrning av
14.
15.
16.
17.
18.
abonnenternas effektuttag (in Swedish),
Byggforskningsrådet, report no. R63:1982,
1982.
F. Wernstedt, P. Davidsson, and C.
Johansson, Demand side management in
district heating systems, in the Sixth Intl. Joint
Conf. on Autonomous Agents and Multi-Agent
Systems (AAMAS 07), Honolulu, HI, US, 2007.
F. Wernstedt and C. Johansson, Intelligent
distributed load control, in the 11th
International Symposium on District Heating
and Cooling, Reykjavik, Iceland, 2008.
F. Wernstedt, C. Johansson and J
Wollerstrand, Sänkning av returtemperaturer
genom
laststyrning
(Decreased
returntemperatures through heat load control) (in
Swedish),
Swedish
District
Heating
Association, report no. 2008:2, 2008.
C. Johansson, F. Wernstedt, and P. Davidson,
deployment of agent based load control in
district heating systems, in First Agent
Technology for Energy Systems Workshop,
Ninth
International
Conference
on
Autonomous Agents and Multiagent Systems,
Toronto, Canada, 2010.
C. Johansson, and F. Wernstedt, Heat load
reductions and their effect on energy
consumption, in The 12th International
Symposium on District Heating and Cooling,
pp. 244-249, Tallin, Estonia, 2010.
Energimyndigheten,
Energistatesik
för
flerbostadshus 2012 (Energy statistics for
multi-dwelling buildings in 2012) (in Swedish),
report no. ES 2013:03, Eskilstuna, Sweden,
2013.
H. Gadd, and S. Werner, Daily heat load
variations in Swedish district heating systems,
Applied Energy, Vol. 106, pp. 47-55, 2013.
M. Wigbels, B. Bøhm, and K. Sipilä,
Operational optimisation: Dynamic heat
storage and demand side management
strategies, Euroheat and Power (English
Edition), Issue. 2, p. 58-61, 2005.
J. Van Deventer, J. Gustafsson and J. Delsing,
Controlling district heating load through prices,
5th IEEE International Systems Conference,
SysCon, pp. 461-465, Montreal, QC, Canada,
2011.
S.F. Van Der Meulen, Load management in
district heating systems, Energy and Buildings,
Vol. 12, pp. 179-189, 1988.
C. Molitor, et al., New energy concepts and
related information technologies: Dual demand
side management, 2012 IEEE PES Innovative
Smart Grid Technologies, ISGT 2012,
Washington, DC, US, 2012.