Sei sulla pagina 1di 19

Energy and Buildings 150 (2017) 447–465

Contents lists available at ScienceDirect

Energy and Buildings


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

An evaluation of the suitability of lumped-capacitance models in


calculating energy needs and thermal behaviour of buildings
Jacopo Vivian ∗ , Angelo Zarrella, Giuseppe Emmi, Michele De Carli
Department of Industrial Engineering, Applied Physics Section, University of Padova, Via Venezia 1, 35131, Italy

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

Article history: Simple, reliable building models have been receiving quite a bit of attention recently particularly with
Received 3 November 2016 regard to diverse applications, such as building design for inexpert energy modellers, simulation of neigh-
Received in revised form 10 April 2017 bourhood or city districts and model predictive control. The International Standard ISO 13790 and the
Accepted 10 June 2017
German Guideline VDI 6007 use two different lumped-capacitance models (5R1C and 7R2C, respectively)
Available online 12 June 2017
based on deterministic, analytical procedures to identify their parameters. The current work investigates
the suitability of these models in calculating peak loads and seasonal energy needs and their accuracy in
Keywords:
estimating buildings’ dynamic behaviour. A room and an apartment were thus simulated using simplified
Building simulation
Simplified model
models and with the benchmarked software TRNSYS. Four reference envelopes with different thermal
Lumped-capacitance model insulation and heat capacities were examined in four climatic conditions. Each of the models was able to
ISO 13790 estimate quite precisely energy needs in both the heating and cooling modes, although the 7R2C model
VDI 6007 was slightly more accurate. The 5R1C model was, however, unable to follow the thermal response of the
TRNSYS buildings during the cooling season, which in turn implied a systematic underestimation of the cooling
peak load. The 7R2C model identified a significant reduction in the root mean squared error (RMSE) both
in the indoor air temperature and in the heating/cooling loads with respect to the reference profiles. That
model would seem then more suitable for the dynamic simulation of single thermal zones with hourly
time steps in both heating and cooling modes.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction The heat dynamics of a building can be modelled following two


different approaches. The first consists in explicitly modelling the
In view of the fact that households are responsible for approxi- physical phenomena occurring within the building, i.e. all physical
mately 25% of the total energy consumption in the European Union processes that significantly affect the energy balance of the ther-
[1], reducing the energy demand of buildings has become one of mal zone(s) considered. Models following this approach are termed
the key points of its energy policy [2]. white-box models. The second approach relies on statistical methods
Appropriate dynamic models are necessary both during early that are applied to available data (heat consumption and tempera-
design stages of buildings and neighbourhoods as well as when tures) without any explicit reference to equations governing their
retrofitting solutions for existing ones are being evaluated. energy balance and are thus called black-box models [4].
Besides retrofitting and modernising Heating Ventilation and Air- In most cases, such as during the design of the building, data are
Conditioning (HVAC) systems, smart controls can significantly not available and white-box models must be used. Since buildings
contribute to reduce the energy demands of existing buildings. are complex energy systems that are affected by multiple, overlap-
Model predictive control (MPC) is an advanced control technique ping physical processes such as transient heat conduction, parallel
that calculates, at each time step, the optimal control inputs for heat exchange by convection and radiation, radiative and convec-
an HVAC system based on the predicted demand of conditioned tive heat generation by internal heat sources etc., energy needs for
spaces over a finite time horizon [3]. This requires a reliable and space heating/cooling can be calculated in a more or less detailed
computationally light building mode.. manner depending on the objective of the analysis. Steady-state,
quasi-steady-state or dynamic methods can be used for this pur-
pose [5]. Quasi-steady-state methods calculate the heat balance of
the building zone over sufficiently long periods of time (typically
∗ Corresponding author. one month or a whole season) and then take dynamic effects into
E-mail address: jacopo.vivian@dii.unipd.it (J. Vivian). account using empirically determined gain/loss utilization factors

http://dx.doi.org/10.1016/j.enbuild.2017.06.021
0378-7788/© 2017 Elsevier B.V. All rights reserved.
448 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

unsuitable for model predictive control (MPC) applications [15] or


Nomenclature the simulations of neighbourhoods/city districts [16,17]. Due to the
large number of buildings being simulated, Urban Building Energy
a Absorption coefficient (−) Models (UBEM) such as SimStadt [18] and CitySIM [19] rely on
A Surface area (m2 ) either steady-state methods [18] or simplified dynamic models [19]
C Thermal capacitance (J/K) to simulate the energy consumption of individual buildings.
cp Heat capacity (J/m2 K) Simplified dynamic models of buildings have been receiving
Fso Shading reduction factor for external obstacles (−) growing attention. The system domain of these models is dis-
H Heat transfer coefficient (W/K) cretized into a set of nodes connected by thermal resistances and
Isol Solar irradiance (W/m2 ) capacitances (i.e. the parameters of the model); their main advan-
mf Frontal mass (kg/m2 ) tage is that once their parameters have been identified, the system
R Thermal resistance (K/W) can be solved analytically, thus avoiding problems of convergence
RMSE Root mean squared error and stability that typically arise with numerical schemes. Since each
t Time of simulation (s) thermal capacitance of the equivalent circuit corresponds to a state
U Thermal transmittance (W/m2 K) variable of the model, the number of thermal capacitances leads
to the order of the model. Reducing the model order curtails, in
Greek symbols
turn, the number of parameters to be identified, thus achieving
 Temperature (K)
the objective of describing the building dynamics with a simple
␾ Heat flow rate (W)
model. The parameters of the equivalent thermal network can be
␣ Heat transfer coefficient (W/(m2 K))
identified analytically, by model order reduction (for a review of
ε Relative error (−)
the techniques see [20]) or by tuning the model to the data set of
␭ Thermal conductivity (W/(m K))
temperature and energy consumption available. In the latter case,
 Density (kg/m3 )
these models, which can be considered a hybrid between black-and
εlw Emissivity (−)
white-box models, are often referred to in the literature as grey-box
t Simulation time-step (s)
models [21].
AW Non-adiabatic building elements
It is clear that calculating parameters analytically can deviate
conv Convective
the results of building simulation from real world conditions as the
e External air
model receives no feedback from the conditioned space. On the
fl Floor
other hand, parameter estimation from measured datasets is influ-
gnd Ground
enced by the case study being considered and thus lacks generality.
hc Heating/cooling load
Coherently with the approach used in the present work, the fol-
hg Heat gains
lowing review mainly focuses on simplified models based on the
i Indoor air
analytical identification of the parameters. For a comprehensive
int Internal (heat gains)
review on simplified thermal models, see [22].
IW Adiabatic building elements
The first attempt to describe the dynamic behaviour of building
lw Long-wave
elements using the electrical analogy dates to 1936 when Beuken
m Thermal mass
[23] derived the equations for a generic network with n thermal
op Opaque building elements
capacitances. That study formed the basis for the n-capacitances
rad Radiative
room model proposed by Rouvel [24]. The latter describes each
se External surface
building element using either a first- or a second-order model, thus
si Internal surface
obtaining a model with at least n thermal capacitances, where n is
sky Apparent sky
the number of the building elements of the room. The latter, which
sol Solar (heat gains)
has been implemented in the GEBSIMU software [25], has been
sup Supply air
extensively used and tested since the 70s.
sw Short-wave
A few years later, Laret [26] proposed a simple analytical method
tr Transmission
to represent each construction element with a one-node model
ve Ventilation
consisting of two resistances and one capacitance. The subdivision
of the thermal resistance was carried out by calculating a vari-
able referred to as the ‘accessibility factor’. Based on this approach,
[6]. Conversely, dynamic methods use short time steps (typically Lorenz and Masy [27] lumped all the construction elements of the
one hour long) to calculate heat balance taking into consideration thermal zone considered into a simple model with two time con-
thermal transmission, heat flow by ventilation, thermal storage, stants: one for the air volume and one for the building structure.
and internal and solar heat gains in the building zone [5]. Crabb et al. [28] used a first- and a second-order model based on
The most frequently used softwares for building simulation (e.g. the prescriptions of Lorenz and Masy to predict the indoor air tem-
TRNSYS [7], Energy Plus [8], or ESP-r [9]) rely on models that require perature in a school. The study showed good agreement between
a detailed description of the building in terms of both its geometry the predictions of internal air temperature and the temperatures
and physical properties. Given the high number of input parameters registered. The authors thus concluded that the dynamic thermal
and the detailed information that are requested, they are unsuitable response of intermittently occupied buildings can be described by
for inexpert users. The underlying models find an explicit solu- first- or second-order lumped-capacitance models.
tion to the heat conduction equation by finite difference [10] or Tindale [29] argued that a second-order model is over-
by analytical methods, namely the transfer function method [11] responsive when used on medium and heavyweight buildings
(an improvement with respect to the frequency response factor which have high-gain daily heat cycles or, equivalently, large diur-
method [12]) and the admittance matrix methods [13]. A synthetic nal temperature swings. He thus proposed a third-order model
description of these methods has been provided by Underwood and obtained by adding a third capacitance connected to the sol-air
Yik [14]. As these methods require a high computational effort and node. Comparisons against a benchmarked finite difference model
cannot readily be interfaced with optimization solvers, they are
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 449

Fig. 1. The single room.

(APACHE) proved that the third-order model was superior to the gate energy demand and transient behaviour and that the accuracy
previous one. depended on the type of structure that was considered.
Antonopoulos and Koronaki [30] found the effective (dynamic) Gouda et al. [20] used model order reduction to identify suit-
thermal capacitance of a first-order model by fitting the latter to the able parameters for each construction element of a well-known
numerical solution obtained by a detailed model solved with the unoccupied building of the University of Northumbria (UK) with
finite-difference method. The latter was compared to the apparent two external walls, two adiabatic partitions (ceilings and floors). A
(static) thermal capacitance (sum of the products of density, thick- trial-and-error comparison with a benchmarked 20th-order model
ness and specific heat capacity of all the building components) for revealed that a 3R2C model was the best choice to reproduce
12 reference buildings with different sizes and thermal insulations. the dynamic behaviour of single construction elements. These ele-
It was found that the effective thermal capacitance normalized ments were then lumped into a single 11th-order model to predict
by the floor area was always higher than the apparent one with the free floating indoor air temperature measured in the build-
a ratio ranging from 2.2 (one-storey) to 3.1 (six-storey block) in ing. The model was found to improve the prediction with respect
cases of well-insulated buildings and even higher for uninsulated to a 6th −order model based on Lorenz and Masy’s prescriptions
envelopes. although both models significantly under-estimated the indoor air
Starting with a second-order model, Fraisse et al. [31] proposed temperature, probably due to the unknown infiltration rate of air
adding two further capacities corresponding to 5% of the internal flowing into the space considered.
heat capacities of the building structure to consider the response of In accordance with the same approach used in the afore-
the wall surface to rapid indoor temperature changes. The response mentioned n-capacitance model [25], Rouvel and Zimmermann
to high frequency solicitations (heat gains or temperature changes) [34] separated the building elements into those under symmet-
becomes important when simulation time step is shorter than one ric and asymmetric loads, which are typically internal partitions
hour, as necessarily occurs in control applications. and external walls, respectively. These groups of building elements
Nielsen [32] developed a simple two-node model to be used were then aggregated (by parallel connection of the correspond-
in the early stages of building design. The simplified model found ing resistance-capacitance circuits), thus making it possible to
good agreement with the detailed simulation software BSim [33] in find an analytical solution for all the building elements having
estimating the energy demand of an office building in Copenhagen the same boundary conditions. The result of this procedure is a
(Denmark). The transient behaviour during the heating and cooling lumped-capacitance model with two similar time constants: one
seasons was evaluated by calculating the indoor air temperature for adiabatic walls and another for non-adiabatic ones. Both mod-
during, respectively, the first week of January as well as that of els (2C and n-C) are now implemented in the GEBSIMU software
July. It was found that eliminating the differences in the calculation [25].
of the transmitted solar radiation, the two-node model produced More recently, Rodriguez Jara et al. [35] proposed a method
reliable results for the aforementioned purpose. to analytically assess the parameters of a first-order model based
Kämpf and Robinson [16] slightly modified the structure of on the hypothesis that the position of the capacitance varies at
Nielsen’s model and calculated the heat load of eight reference every time step in response to changes in the excitation value.
rooms using the test reference year of Geneva (Switzerland). They The method showed promising results as far as the accuracy in the
then compared the resulting load to that obtained with the detailed thermal behaviour of single construction elements was concerned.
ESP-r simulation software [9]. The model was then extended to However, a discussion on its application for whole room simulation
reproduce a multi-zone building. Comparisons with ESP-r showed has yet to be carried out.
that the second-order model accurately reproduced both aggre- The International ISO 13790 Standard [5] fully prescribes a
quasi-steady-state calculation method (monthly method) and a
450 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

simple dynamic method (simple hourly method) based on a first- Table 1


Characteristics of the considered reference climates.
order lumped-capacitance model with five thermal resistances and
one thermal capacitance. Despite the fact that the latter produces Palermo Venice Vienna Helsinki
hourly results, it has been validated, in accordance with the ISO Maximum temperature [◦ C] 34.6 33.6 31.7 28.7
15265 Standard, only at the aggregated monthly heat demand level Minimum temperature [◦ C] 5.9 −5.8 −18.3 −21.7
[36]. Individual hourly values may thus contain important rela- Heating degree days 801 2267 3180 4856
tive errors [5]. The International ISO 52016 Standard [37] that is Cooling degree days 1002 474 212 39
Annual Solar irradiance [kWh/m2 ] 1458 1102 1123 947
about to be released will replace the ISO 13790 and will be suit-
able for the calculation of hourly heating and cooling load including
dehumidification; the new Standard does not lump all the build-
ing components in one single element as the previous does, but
discretizes each of them into five or two nodes depending on the
component type, opaque and glazed respectively.
The ISO 13791 [38] and ISO 13792 [39] Standards, which have
recently been released, outline the procedure to validate simpli-
fied models to calculate the internal air temperature of a single
room under transient hourly conditions. The models are suitable
for warm periods, but they have been limited, until now, to sit-
uations characterized by free floating temperatures without any
cooling system.
Rouvel’s n-C model was validated and achieved the best class
according to ISO 13792. This implies deviations of the internal and
operative temperature falling between −1 K and +1 K with respect
to the reference values [39]. Rouvel and Zimmermann’s 2C model
[34] was implemented in the German Guideline VDI 6007-1 [40].
Despite the fact that several studies have investigated these
models, the current review uncovered a gap in the discussion on
their accuracy. Indeed, until now no studies have set out to examine
the accuracy of lumped-capacitance models for different climate
conditions and diverse types of building structures that explicitly
separate the results into heating and cooling seasons, thus produc-
ing a meaningful guideline for potential and/or inexpert users.
A single room and an apartment were thus simulated with the Fig. 2. Floor plan of the apartment.

one-capacitance (1C) model described in the ISO 13790 [5] and with
the two-capacitances (2C) model described in VDI 6007-1 [40], tak-
using both resistance-capacitance (RC) models and TRNSYS com-
ing into consideration four diverse building structures under four
mercial software under the same boundary conditions that are
different climate conditions (Palermo, Venice, Vienna and Helsinki).
described in detail here below.
The results of the simplified models were then compared with
those obtained from the well-established TRNSYS software (used
as benchmark), in order to quantify the improvement of the 7R2C 2.1. Weather conditions
model over the 5R1C one in both heating and cooling seasons in
terms of seasonal energy needs, peak loads, and transient thermal One-hour time steps were used over a one-year simulation
behaviours. period including a heating season (from 15 October to 15 April), a
The most recent studies on the accuracy of lumped-capacitance cooling season (from 15 May to 15 September); the remaining peri-
models focus on grey-box models with parameters identified by ods of the year had free-floating indoor air temperatures. Although
historical data series (e.g. [41–43]). We decided to use an analyt- this methodology obviously leads to unrealistic values with regard
ical approach here in order to obtain results that do not depend to overall energy needs, particularly in northern climates where the
on the quality of the dataset and on the parameter identification heating season is longer, it nevertheless allowed us to compare the
procedure [42]. Using models from well-known Standards makes indoor air temperature and the heating and cooling load profiles
the results reproducible and allows to generalize results pertaining over equal periods of time.
to different climate conditions and building structures. In addition, The test reference year (TRY) files of four reference European
while most studies use free floating indoor air temperature as the locations were used to examine a wide range of weather conditions
only output variable, the current work used both the indoor air tem- according to an updated version of the Köppen-Geiger classification
perature and the heating/cooling load to assess the model accuracy. [44]:
This approach makes it possible to compare the peak loads, the
overall energy needs, and the mean distance values in both heating • Palermo: warm temperate climate with dry and hot summer;
and cooling seasons obtained using simplified models, thus pro- • Venice: fully humid with hot summer;
ducing useful reference values for those who intend to use one of • Vienna: fully humid with warm summer;
these models. • Helsinki: fully humid snow-dominated climate with warm sum-
mer.
2. Methodology
Table 1 summarizes the main characteristics of the climates
In order to learn more about the extent to which the lumped- mentioned. Both heating and cooling degree days are in reference
capacitance models are able to gauge the thermal behaviour of to the baseline of 18.3 ◦ C [45]. Solar irradiance refers to the annual
buildings, a single room and an entire apartment were simulated amount of global radiation on the horizontal surface.
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 451

Table 2
Opaque building components of heavyweight building structures.

Building component Construction material Thickness [m] ␭ [W/(m K)] ␳ [kg/m3 ] cp [J/(kg K)]
2 2
External wall EW H1 (U = 1.06 W/m K;mf = 410 kg/m ) Indoor plaster 0.015 0.700 1700 1010
Lightweight brick 0.080 0.360 1200 840
Airhole 0.040 0.260* 1.2 1000
Cavity block 0.200 0.560 1300 1050
Outdoor plaster 0.015 0.900 1800 910
External wall EW H2 (U = 0.26 W/m2 K;mf = 410 kg/m2 ) Indoor plaster 0.015 0.700 1700 1010
Lightweight brick 0.080 0.360 1200 840
Polystirene 0.120 0.040 30 1300
Cavity block 0.200 0.560 1300 1050
Outdoor plaster 0.015 0.900 1800 910
Internal wall IW H Indoor plaster 0.015 0.700 1700 1010
Lightweight brick 0.080 0.360 1200 840
Indoor plaster 0.015 0.700 1700 1010
Ceiling CE H, floor FL H Ceramic tile 0.015 1.200 2300 1000
Screed 0.060 1.400 2200 1050
Rockwool 0.040 0.047 75 840
Floor slab 0.220 0.740 1800 920
Indoor plaster 0.020 0.700 1700 1010

Table 3
Opaque building components of lightweight building structures.

Building component Construction material Thickness [m] ␭ [W/(m2 K)] ␳ [kg/m3 ] cp [J/(kg K)]

External wall EW L1 Board formwork 0.020 0.140 500 2520


(U = 1.04 W/m2 K; Plasterboard 0.050 0.100 625 1000
mf = 50 kg/m2 ) Board formwork 0.020 0.140 500 2520
External wall EW L2 Board formwork 0.020 0.140 500 2520
(U = 0.28 W/m2 K; Plasterboard 0.050 0.100 625 1000
mf = 70 kg/m2 ) Wood fiber 0.100 0.038 1200 1000
Board formwork 0.020 0.140 500 2520
Internal wall IW L Plasterboard 0.050 0.100 625 1000
Boundary wall BO L same as EW L1
Ceiling CE L, Ceramic tile 0.015 1.200 2300 1000
floor FL L Lightweight screed 0.030 0.140 800 1050
Wood fiber 0.100 0.038 1200 1000
Screed 0.040 1.400 2200 1050
Indoor plaster 0.015 0.700 1700 1010

many authors have indicated that solar gains are one of the main
sources of uncertainty [41,42]. Solar gains were calculated using
Type 56 of TRNSYS [46]. At each time step, the solar heat entering
the building was given by the sum of shortwave transmission and
secondary heat flux through external windows and secondary heat
flux of the external window. The solar radiation absorbed by the
exterior walls and the longwave radiation emitted by both glazed
and opaque building components were calculated separately (see
Sections 3.1 and 3.2 for further details).
The apparent sky temperature  sky and ground temperature  gnd
were calculated according to the equations outlined in the Standard
Fig. 3. Reference node. VDI 6007 Part 1 [40] and Part 3 [47].
Table 5 outlines all of the building components composing the
Table 4 four building structures.
Glazed building components. Other boundary conditions for both simplified and detailed
building models are:
Building component U[W/(m2 K)] g [–] Ff [–] Fsh [–]

Single pane external window SP 5.68 0.855 0.150 1.0


Double pane external window DP 2.83 0.755 0.150 1.0 • emissivity of surfaces for long-wave thermal radiation (εlw ): 0.9;
• absorption coefficient of exterior surfaces (ase ): 0.6;
• convection heat transfer coefficient of internal surfaces (␣si,conv ):
2.2. Building structures 2.7 W/(m2 K) for horizontal heat flow direction, 1.7 W/(m2 K) for
vertical heat flow direction;
• convection heat transfer coefficient of external surfaces (␣se,conv ):
Four building structures representing different combinations of
weight and thermal insulations [heavyweight not insulated (named 20 W/(m2 K).
H1), heavyweight insulated (H2), lightweight not insulated (L1) and
lightweight insulated (L2)] were considered. The building compo- 2.2.1. The room
nents are outlined in Tables 2–4. The first case study was a single 5.0 × 5.0 × 3.0 m room with
Both simplified models used the solar heat gain obtained from only one external wall containing a 7 m2 glazed surface which was
the detailed simulation as the input signal. This was done because oriented southwards or westwards depending on the simulation
452 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

Table 5
Building envelopes considered in the analysis.

Envelope H1 H2 L1 L2

Description Heavyweight Heavyweight well Lightweight Lightweight


uninsulated insulated uninsulated well insulated

Non-adiabatic building components EW H1 EW H2 EW L1 EW L2


SP DP SP DP
Adiabatic building components IW H, CE H, FL H IW H, CE H, FL H IW L, BO L, CE L, IW L, BO L, CE L,
FL L FL L

considered. All the other walls (internal walls, floor and ceiling) • Case A: the apartment was simulated with the same boundary
were considered as adiabatic. The room is shown in Fig. 1. conditions as the room, i.e. no ventilation and no internal heat
A dynamic simulation was carried out for each combination of gains. This type of simulation helps us to understand if the results
building envelope and orientation: 16 simulation runs with a south- obtained for the single room are also valid for larger thermal
oriented external wall (S) and another 16 runs with a west-oriented zones.
external wall (W). For a given orientation of the external wall, the • Case B: the dimensions and orientation of the apartment were not
16 cases refer to the combination of 4 reference weather files (the subject to change with respect to those of Case A, but the venti-
test reference year of Palermo, Venice, Vienna and Helsinki) with lation and internal heat gains were included for a more realistic
the 4 building envelopes described above (H1, H2, L1 and L2). description of the apartment. Both the air mass flow rate due to
infiltration and the natural ventilation and the heat flow rate due
to internal heat gains were assumed constant in time and uniform
within the thermal zone because the analysis mainly looks at the
2.2.2. The apartment evaluation of the peak load and energy needs. Both the heat flow
The second building was a 93.6 m2 single-storey residential rate from internal sources and the air change rate due to nat-
building (see Fig. 2). The apartment has two external walls oriented ural ventilation/infiltration were calculated in accordance with
along north-south direction −or east-west, depending on the con- the Italian Standard UNI TS 11300-1 [48]. The former was calcu-
sidered simulation- and the construction elements in all the other lated according to Eq. (1); the air change rate was assumed to be
directions were considered adiabatic. The latter include internal 0.5 vols per hour, as is generally assumed for domestic dwellings
partitions, boundary walls towards conditioned spaces (adjacent [48].
apartments), floor and ceiling. All glazed components on the north
(or east) side are windows of height 1.3 m (overall glazed area of int = 5.294Afl − 0, 01557Afl 2 [W ] (1)
5 m2 ), whereas all glazed components on the opposite side are
doors of height 2.3 m (overall glazed area of 9.4 m2 ). The height 2.3. Evaluation of the models’accuracy
of the apartment was 3 m and the overall surface of the internal
partitions was 95 m2 . The different load profiles obtained by the RC models were com-
Two types of simulations of the apartment were carried out: pared to assess their accuracy with respect to the TRNSYS model

Fig. 4. Five resistances one capacitance (5R1C) model of EN ISO 13790.


J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 453

as far as the dynamic response and overall energy needs were The simple hourly method is based on the equivalent resistance-
concerned. The dynamic response is evaluated by calculating the capacitance (RC) circuit shown in Fig. 4. The parameters of the
distance between the heat load profiles of the RC models and the model are the thermal transmission coefficients of the glazed and
heat load profile obtained with TRNSYS. The metric used is the root opaque building components (Htr,w and Htr,op respectively), the
mean squared error, calculated as in Eq. (2). Cm ventilation coefficient Hve , the coupling conductance Htr,is and
 the thermal capacitance
n
(y
i=1 i
− yi∗ )2 The final variable was calculated in accordance with the ISO
RMSEy = (2)
n 13786 [50]. The thermal transmittance of opaque building com-
ponents Htr,op is divided into Htr,em and Htr,ms as shown in Fig. 4.
In order to compare the RMSE obtained with reference to dif-
The nodes represent the temperatures of the outdoor air  e , the
ferent building structures and under different weather conditions,
supply air  su , the indoor air  m  i , the internal wall surface  s and
the latter was normalized with respect to the mean heat load of the
the thermal mass
case considered, thus obtaining the relative error ␧, as shown in Eq.
Heat gains (int and sol ) are distributed to nodes  i ,  s and  m
(3). The mean heating/cooling load was obtained by dividing the
according to the three semi-analytical correlations that depend on
energy needed for space heating/cooling by the number of hours
the ratio between effective mass area and area of the internal build-
with heating/cooling systems turned on.
ing components and on the thermal transmittance of the glazed
RMSEy Htr,w building components
εy = (3)
ȳ The effective mass area Am represents the thermal capacitance
of the building in terms of surface area, as described in the Standard
3. Simplified models [5]. The solar heat gain sol includes here not only the solar radi-
ation entering through external windows but also the short-wave
Lumped-capacitance models assume that the distributed ther- radiation absorbed by the external walls and the long-wave radia-
mal mass of the dwelling is lumped into a discrete number of tion emitted by the external surfaces to the outdoor environment.
thermal capacitances, depending on the model type [42]. Therefore In the 1C model, all these terms sum up to give an additional heat
they are often referred to as xRyC models, where x is the number of gain, as shown in Equations (5) and (6). The first term is the short-
thermal resistances and y the number of thermal capacitances of wave radiation absorbed by opaque building components, where
the equivalent electrical circuit. These models rely on the following Fso is the shading reduction factor for the external obstacles, Isol
assumptions: linearity of the heat transfer mechanism, represen- the solar irradiance, Uop and Aop the thermal transmittance and the
tation of multilayer wall characteristics by lumped parameters and projected area of the opaque building components, as and Rse their
single zone approximation [28]. absorption coefficient and the surface heat resistance, respectively.
The lumped-capacitance model was solved by a linear system All these variables refer to the external surface of the exterior walls
composed of n heat balance equations, where n is the number of and must be considered separately for each orientation (the sub-
nodes of the corresponding thermal network. Fig. 3 shows, by way script of the orientation is omitted for the sake of simplicity). The
of example, a temperature node of an equivalent circuit. second term is the extra heat flow due to the thermal radiation to
The following heat balance holds true for the reference node. the sky, where Fr is the form factor between the building element
    and the sky (0.5 for vertical walls), ␣rad is the radiative heat transfer
1,t − 1,t−t
H0 0,t − 1,t + H1 2,t − 1,t + 1,t = C1 (4) coefficient, and  er is the difference between the external air tem-
t
perature and the apparent sky temperature. The interested reader
As is usual in building simulations, the system has one degree of will find additional details on each parameter in the reference [5].
freedom unless one variable is fixed by the user. This leads to two  
possible model uses: sol = Fso as Rse Uop Aop Isol − Fr r (5)
r = Rse Uop Aop ˛rad er (6)
• Calculation of the heat load. The indoor air temperature  i is set by
the user and the output of the model is the heat load hc , i.e. the 3.2. The 7R2C model
energy needed by the building to maintain the specified set-point
temperature  set ; The model distinguishes between adiabatic and non-adiabatic
• Calculation of the indoor air temperature. The heat load hc is set by building components and assigns an equivalent thermal capacity
the user, and the output of the model is the indoor air temperature to each of the two groups. This is because under asymmetrical load
i . the heat flux through the building component must be taken into
consideration in addition to its heat storage capacity [40]. In the
The models considered do not include the balance of water 5R1C model [5], the entire thermal mass of the building is lumped
vapour in the indoor ambient which means that the calculation into one single element, thereby making this distinction impossible.
of the latent heat load (to be delivered to or extracted from con- Instead, in the second-order model of Lorenz and Masy [27], the
ditioned spaces) is not included. The water vapour balance can, two thermal capacitances represent the whole thermal mass of the
however, be included simply as a post-processing calculation after building and the indoor air volume, rather than the adiabatic and
the thermal simulation. The models were developed in the MATLAB non-adiabatic building elements, respectively. This, in turn, implies
[49] environment; their main features are briefly described in the a less accurate thermal response at high frequencies. The model
following section. intends to simulate the building with 1 h time step.
The AW and IW subscripts respectively indicate non-adiabatic
3.1. The 5R1C model and adiabatic building components. Thus, R1;AW and R1;IW are the
dynamic thermal transmittances (without surface resistances) and
As reported above, the ISO 13790 International Standard [5] C1;AW and C1;IW are the dynamic thermal capacitances of the non-
presents two methods that rely on the same inputs to calculate adiabatic and adiabatic building components. The combination
the building’s energy use at different levels of detail: the monthly of different building components to achieve equivalent thermal
method with one month time intervals and the simple hourly method resistances and capacitances (for example, an adiabatic roof, an adi-
with one hour time intervals. abatic ceiling and internal walls with reference to the single room)
454 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

is achieved by parallel connection of complex thermal resistances. with the 1C model and to +7% with the 2C model). On the other
Parallel connection is possible given that all these components are hand, the peak load for space cooling is systematically underesti-
identically loaded. Rges;AW is the steady-state thermal resistance of mated (from −15% to −27%) by the 1C model, and the 2C model
the exterior wall (including surface thermal resistances on both overestimates it by up to +10%. The energy needs for heating and
sides) of the exterior building components. Thus, RRest;AW results cooling clearly depend on both the climate and the building enve-
as: lope. The ratio between heating and cooling needs ranges from 0.25
1 (Palermo) to 5.9 (Helsinki) for a room with a low thermal insulation
RRest;AW = Rges;AW − R1;AW − 1 1
(7) and from 0.03 (Palermo) to 1.2 (Helsinki) for a room with high ther-
R˛;conv;AW
+ R˛;conv;AW/IW mal insulation. Table 7 seems to extend the patterns found for the
peak load to the energy needs calculation. In fact, the energy needs
Rve is the thermal resistance due to ventilation that connects the
for space heating is slightly underestimated by the RC models for
indoor air temperature  i with the node of the supply air temper-
buildings with low thermal insulation, while it is overestimated for
ature  sup .
highly insulated building envelopes. The trends in the space cooling
The heat exchange between the indoor air and the internal wall
mode remained almost entirely unaltered, although in this case the
surfaces are accounted for by three surface thermal resistances: the
error in the lumped-capacitance models was not as significant as
radiative heat exchange between the surfaces of the external and
that in the peak load calculation; this would suggest that although
internal walls R␣;rad,AW/IW , the convective heat exchange between
they do not accurately follow the fluctuations of the heat load, sim-
the indoor air and the internal wall surfaces and R␣;conv;IW and
plified models are better able to estimate the energy needs of the
between the indoor air and the exterior wall surfaces and R␣;conv;AW .
building over the entire heating/cooling season (Fig. 5).
The delta connection that links these thermal resistances at sur-
faces is converted into an equivalent star connection (thus giving
4.1.2. The evaluation of transient behaviour
rise to R˛∗ ;AW , R˛∗ ;IW and R˛∗ ;i ) in order to facilitate the resolution
Fig. 6 shows the RMSE of the simplified models and the peak
of the loop and node equations. The model assumes that all room
load in the corresponding season for the west oriented room. The
surfaces contribute to the radiation exchange in proportion to their
peak load, which was extracted from the heat flow profile produced
respective surface areas.
by TRNSYS, serves as a valid reference for both simplified models.
The effects of solar radiation absorbed by the external walls
The peak loads and the energy needs for space heating were
and the radiation emitted by the external surfaces to the external
obviously increased when we turned to examine cold climates, as
environment (sky and ground) are included in the equivalent air
can be seen in Fig. 6 and in Table 7. The RMSE of the simplified
temperature  A;eq;gew . The following holds true for the equivalent
models increased from Palermo to Vienna but showed a smaller
outdoor temperature of each exterior surface:
increase with reference to the snow-dominated climate of Helsinki.
e;eq = e + e;eq;lw + e;eq;sw (8) This trend could be linked to the different patterns of temperature
differences between the indoors and outdoors, which represents
The temperature drop  e;eq;lw is due to the thermal radiation
the main driving force for space heating load in winter months.
emitted by external surfaces to the sky and to the ground, and the
Indeed, temperatures are less prone to register significant diurnal
temperature gain  e;eq;sw is due to the solar energy absorbed by
fluctuations in the Nordic climate of Helsinki than in the other loca-
the opaque walls.
tions considered in this study. This trend seems to be valid for both
Radiative and convective heat gains (including the heat-
simplified models regardless of the building envelope. Fig. 6(a),
ing/cooling load) are divided between three different nodes: the
for example, shows that while the peak load increases almost two
indoor air temperature  i , the surface temperature of internal and
times from Palermo to Vienna (from 500 W to 1383 W) and the
external walls  s;IW and  s;AW . This is another significant improve-
RMSE1C increases four times (from 20 W to 102 W), this ratio is no
ment over the 1C model.
longer valid when we turn our attention from Vienna to Helsinki
(+25% of peak load and only +9% of the RMSE1C ).
4. Results and discussion The accuracy of the 7R2C model is greater than that of the
5R1C one, as is noted in Fig. 6 which compares the RMSE2C to the
Sections 4.1 and 4.2 respectively present the simulation results RMSE1C . The improvement of the 7R2C model over the 5R1C one
for the single room and for the apartment and report the peak in terms of RMSE reduction rises from −28% to −53% depending
loads and the overall energy needs for space heating and cool- on the boundary conditions. The trivial reason for the improve-
ing that were produced by the simplified models as well as by the ment of the 2C model over the 1C one is linked to the presence
benchmarking TRNSYS model. Transient behaviour is described by of the second thermal capacitance that makes it possible to dis-
means of the distance of the heating/cooling load profile from that tinguish between adiabatic and asymmetrically loaded building
obtained by the reference model. components, as explained in Section 3.2. Moreover, the presence
of two internal surface temperature nodes ( IW and  AW in Fig. 5)
4.1. The room may lead to further improvement with respect to the 1C model
due to both a more coherent distribution of heat gains throughout
The ventilation and the internal heat gains were omitted in the the wall surfaces and the introduction of a radiative heat exchange
calculations made for the single room to more accurately mea- between the inside surface of external walls and the surfaces of
sure the response of the building structures to fluctuating weather internal building components.
conditions, i.e. the outdoor air temperature and solar radiation. The peak load and energy needs are not as clearly seen as a
function of the weather variable in the cooling mode as it is in the
4.1.1. The evaluation of peak loads and energy needs heating one. Indeed, the cooling load is determined by the combi-
The peak loads and the overall energy needs for both space heat- nation of two driving forces: the external air temperature and solar
ing and cooling are outlined in Tables 6 and 7, respectively. Table 6 radiation. This may also be the underlying reason for the lower
shows that lumped-capacitance models tend to slightly underes- accuracy of lumped-capacitance models in the cooling mode with
timate the peak load for the space heating of uninsulated building respect to the heating one. When RMSE1C and the RMSE2C are com-
structures (down to −8% with the 1C model and to −4% with the 2C pared (see Fig. 6), it becomes immediately clear that the error of the
model) and to overestimate it for insulated buildings (up to +10% simplified models in the cooling mode is significantly higher than
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 455

Table 6
Peak load of the west-oriented single room.

Envelope Weather Peak load for space heating Peak load for space cooling [W]

TRNSYS[W] 1C[%] 2C[%] TRNSYS[W] 1C [%] 2C[%]

H1 PALERMO 500 −6.0 −4.6 1230 −20.6 7.0


VENEZIA 983 −8.4 −4.1 1163 −16.5 4.4
VIENNA 1383 −7.9 −3.6 2015 −16.2 5.9
HELSINKI 1735 −7.9 −4.0 1513 −15.7 8.7
H2 PALERMO 184 10.1 5.3 945 −23.6 4.7
VENEZIA 396 5.7 6.6 882 −19.8 0.6
VIENNA 542 3.5 4.8 1608 −20.0 1.3
HELSINKI 725 4.1 4.9 1339 −19.9 1.8
L1 PALERMO 561 −6.3 0.2 1357 −25.2 7.1
VENEZIA 1052 −8.2 −1.3 1282 −22.6 5.6
VIENNA 1520 −7.5 −1.3 2197 −21.5 5.7
HELSINKI 1780 −6.7 −1.9 1622 −21.2 10.8
L2 PALERMO 210 7.7 5.1 1279 −22.1 10.3
VENEZIA 426 2.4 6.3 1167 −18.7 8.2
VIENNA 594 1.4 3.8 2160 −17.8 7.6
HELSINKI 773 2.0 2.2 1813 −17.3 10.3

Table 7
Heating and cooling needs of the west-oriented single room.

Envelope Weather Energy need for space heating [kWh] Energy need for space cooling [kWh]

TRNSYS [kWh] 1C[%] 2C [%] TRNSYS [kWh] 1C[%] 2C[%]

H1 PALERMO 366 −9.0 −13.9 1434 −4.1 2.4


VENEZIA 2171 −8.1 −7.2 1052 −2.9 4.3
VIENNA 2537 −9.1 −7.5 1426 −0.2 5.2
HELSINKI 3901 −8.2 −6.0 865 3.6 10.2
H2 PALERMO 35 60.0 31.4 1287 −6.5 −2.7
VENEZIA 715 11.5 8.1 1057 −7.3 −3.1
VIENNA 771 13.4 10.2 1548 −6.7 −3.5
HELSINKI 1460 8.8 7.3 1250 −8.3 −4.8
L1 PALERMO 455 −7.9 −9.5 1186 −4.8 2.3
VENEZIA 2218 −5.8 −4.5 861 −6.2 3.3
VIENNA 2628 −6.8 −4.6 1147 −3.7 3.9
HELSINKI 3877 −5.7 −3.6 662 −5.4 6.8
L2 PALERMO 33 12.1 −3.0 1424 −0.2 2.8
VENEZIA 712 4.6 2.0 1195 −0.3 3.3
VIENNA 778 4.4 3.0 1735 1.0 3.4
HELSINKI 1468 4.0 2.8 1428 1.4 3.9

Fig. 5. Seven resistances two capacitances (7R2C) model of VDI 6007.

that in the heating mode. Ventilation, however, will significantly mer (free cooling). Thus, this observation needs to be corroborated
augment the mean load for space heating in the simulations of the in the apartment (case B).
whole apartment while it could have an opposite effect in the sum-
456 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

Fig. 6. RMSE of lumped-capacitance models and peak load for the west-oriented single room with envelopes: (a) H1, (b) H2, (c) L1 and (d) L2.

The improvement of the 7R2C model over the 5R1C one is sub- 4.2.1. The evaluation of the peak loads and energy needs
stantial with regard to the cooling load: the reduction of the RMSE The tendency of lumped-capacitance models to underesti-
ranges from −38% to −54% depending on the boundary conditions. mate/overestimate the peak load for space heating for low/high
Fig. 7 shows that heating (during the third week of January) thermal insulation also holds true for the apartment (case A) −see
and cooling (during the third week of July) load profiles calculated Table 8. With the exception of the case in Palermo with insulated
by TRNSYS are better approximated using the 7R2C model with walls where a very low heating demand was found, the overestima-
respect to the 5R1C one. This is particularly evident in the cool- tion of the 5R1C models ranges from +6% to +11%, and it is slightly
ing mode (Fig. 7(b)) in which the single capacitance model has a attenuated in the 7R2C model. The underestimation reaches −7%
damping effect on the heat load profile, making it difficult to per- with the 1C model (in Vienna with envelope L1) and almost dis-
ceive rapid fluctuations that can, instead, be almost entirely caught appears with the 2C model. In the cooling mode, the 1C model is
by the 2C model. still responsible for underestimating the peak load (down to −27%),
Analogous results were obtained for the south-oriented room, while the 2C model does not show a clear trend as its error is
as reported in App. 1. much lower and not uniformly distributed throughout the building
envelope.
Introducing internal heat gains and natural ventilation (case B)
does not significantly change the trends found for case A. Indeed,
the presence of a constant inflow of cold air introduces a fur-
4.2. The apartment ther heat load in the heating mode that can be reproduced by the
resistance-capacitance models, resulting in a peak load that runs
The aim of this section was to verify whether the behaviour of from −6% to +8% with respect to that resulting from the TRNSYS
the simplified models discussed above is the same when a real simulation. Results regarding case B confirm that the 1C model is
residential building is simulated. The conditioned space is first inappropriate for the thermal simulation of buildings in the cool-
upgraded from a single room to an apartment (case A) and nat- ing mode. In fact, the underestimation of the peak load ranges from
ural ventilation and internal heat gains are then introduced (case −7% to −17%. Conversely, the peak load calculated by the 2C model
B). This will make it possible to verify how the separate effects of is very close to that obtained by TRNSYS. Indeed, it is approximated
the building size and of the additional boundary conditions (natu- by an error ranging between −4% and +4%, the only exception being
ral ventilation and internal heat gains) eventually affect the trends the apartments with low thermal insulation oriented on the north-
outlined for the single room. The peak load and the overall energy south direction in Palermo. In that case, the peak load occurs at noon
needs (for space heating and cooling) for both Cases A and B are and the 2C model deviates significantly from the values obtained by
outlined in Tables 8 and 9, respectively. TRNSYS. The peak load is over-estimated by +10% for a lightweight
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 457

Fig. 7. The third week of (a) January and (b) July for the west-oriented single room in Palermo with envelope H2.

Table 8
Peak load of the apartment (case A and B) oriented east-west.

Env. Weather Peak load for space heating Peak load for space cooling

TRNSYS [W] 1C [%] 2C [%] TRNSYS [W] 1C [%] 2C [%]

A B A B A B A B A B A B

H1 PALERMO 1231 1564 1.9 −0.5 −1.5 −2.2 2084 2720 −8.9 −10.5 6.0 −1.5
VENEZIA 2494 3321 −2.8 −3.6 −1.7 −1.7 2030 2678 −8.4 −6.3 −1.6 −0.8
VIENNA 3463 4882 0.3 −1.2 0.6 −0.3 3264 3810 −6.9 −7.7 0.5 0.4
HELSINKI 4442 6037 −2.6 −2.6 −1.7 −1.7 2260 2582 −5.9 −5.7 3.0 2.5
H2 PALERMO 399 717 21.8 8.1 11.0 4.0 1731 2346 −14.0 −11.1 4.0 −1.2
VENEZIA 924 1745 9.1 2.0 7.1 2.5 1600 2307 −9.1 −8.1 0.8 −0.4
VIENNA 1244 2629 8.1 2.5 7.7 3.4 3149 3390 −18.6 −9.9 −1.6 1.4
HELSINKI 1696 3277 6.0 2.0 5.5 2.0 2282 2571 −9.0 −9.1 4.1 1.8
L1 PALERMO 1426 1761 −4.1 −4.8 1.0 0.1 2506 3164 −15.4 −14.5 2.2 1.3
VENEZIA 2696 3560 −6.0 −5.5 −0.4 −0.7 2494 3184 −15.9 −12.9 2.6 1.7
VIENNA 3928 5355 −4.9 −4.6 −0.5 −0.9 4057 4602 −15.0 −13.8 2.3 2.0
HELSINKI 4626 6223 −4.3 −3.8 −1.4 −1.5 2929 3290 −15.2 −13.1 4.7 4.1
L2 PALERMO 489 803 15.7 6.6 7.0 2.9 1964 2637 −15.6 −13.8 6.4 −2.7
VENEZIA 987 1831 7.6 2.9 5.8 2.0 1866 2572 −13.3 −10.8 −1.7 −2.1
VIENNA 1384 2796 9.5 2.7 9.5 3.4 3334 3895 −14.1 −13.4 −0.3 −0.8
HELSINKI 1796 3378 5.4 1.7 5.1 1.7 2826 3003 −17.4 −12.6 −1.0 −0.4

structure and by +7% for a heavyweight one. Since the value of the son lightweight structures perform better than heavyweight ones
peak load calculated by the 7R2C model is conservative, the latter because of their ability to dissipate heat more rapidly. In fact, in
can be safely used for this purpose. case A, a rather small reduction in the cooling demand (from −1.3%
As Table 9 shows, when there is poor thermal insulation to −4% depending on the climate considered) occurs from H2 to
(envelopes H1 and L1), the seasonal energy needs for space heat- L2, while it is barely observable (less than −2%) in the presence of
ing in Palermo is approximately 40% of the energy needs for space ventilation and internal heat gains.
cooling. In that climate, the heat demand in the cold season is null
with high thermal insulation even when ventilation is present. In
the Nordic climate of Helsinki the heating demand of the apart-
ment with low thermal insulation is 11 times and 17 times larger 4.2.2. The evaluation of transient behaviour
than the cooling demand in cases A and B, respectively. In the same Figs. 8 and 9 show the RMSE of the simplified models and the
climate, the insulated apartment presents a ratio between heating peak load (according to TRNSYS) in the corresponding season for
and cooling demand of 1.45 (case A) that rises to 3.6 in case B due to the apartment facing east- and westwards in Case A and B, respec-
natural ventilation. Qualitatively speaking, during the cooling sea- tively. Analogous results were obtained for the apartment facing
the north-south direction, as reported in Appendix A.
458 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

Table 9
Heating and cooling needs of the apartment (case A and B) oriented east-west.

Env. Weather Energy need for space heating Energy need for space cooling

TRNSYS [kWh] 1C [%] 2C [%] TRNSYS [kWh] 1C [%] 2C [%]

A B A B A B A B A B A B

H1 PALERMO 1276 1293 −0.5 −2.5 −5.8 −6.5 2689 3353 −4.9 −3.7 −1.8 −1.3
VENEZIA 6020 7377 −4.0 −4.0 −4.5 −4.2 1730 2170 −4.3 −3.0 −0.9 −0.4
VIENNA 7213 9304 −4.7 −4.3 −4.4 −3.9 2232 2282 −0.3 −0.1 −0.2 0.1
HELSINKI 10565 13557 −4.7 −4.2 −4.0 −3.6 995 809 4.1 4.2 2.3 3.4
H2 PALERMO 93 137 65.6 32.6 34.4 16.4 2761 3402 −4.2 −3.0 −2.1 −1.3
VENEZIA 1761 3055 11.8 5.6 7.6 3.2 2171 2589 −4.5 −3.0 −1.9 −1.0
VIENNA 1916 3890 13.5 5.7 9.7 3.8 3207 3071 −4.0 −3.2 −2.7 −1.9
HELSINKI 3575 6528 9.1 3.7 6.6 2.4 2462 1811 −5.1 −5.0 −3.7 −2.7
L1 PALERMO 1296 1336 −2.2 −5.2 −5.2 −6.3 2652 3313 −3.6 −2.4 −0.2 0.1
VENEZIA 5926 7300 −3.2 −3.5 −3.4 −3.5 1783 2228 −4.8 −3.5 0.8 1.2
VIENNA 7105 9203 −4.0 −3.9 −3.4 −3.3 2306 2409 −1.4 −2.1 1.3 1.6
HELSINKI 10283 13291 −3.3 −3.2 −2.9 −2.8 1088 952 −0.6 3.7 4.1 4.7
L2 PALERMO 131 179 34.4 14.2 18.3 7.5 2671 3337 −2.4 −2.9 −1.0 −1.4
VENEZIA 1823 3129 10.8 4.8 7.2 2.8 2143 2578 −3.5 −3.2 −1.2 −1.1
VIENNA 2024 3989 11.5 4.6 8.9 3.2 3090 3017 −2.1 −2.8 −1.6 −1.5
HELSINKI 3186 6531 9.9 4.0 7.6 2.5 2375 1794 −3.9 −5.1 −3.2 −2.5

Fig. 8. RMSE of lumped-capacitance models and peak loads for the apartment (case A) with east and west-oriented external walls with envelopes: (a) H1, (b) H2, (c) L1 and
(d) L2.

The positive effect of a cold climate on the accuracy of lumped- 6% from Palermo to Vienna, and drops to 4.5% for Helsinki. Similar
capacitance models in the heating mode is also noted in the trends can be noted for the apartment with ventilation and internal
apartment without ventilation and internal gains, as can be seen heat gains.
in Fig. 8. The ratio between RMSE and peak load for the envelopes The relative errors (defined in Eq. (3)) of both simplified mod-
H1 and H2 was analysed at different latitudes. With reference to els are presented graphically in Fig. 10 for the 32 simulations of the
the 2C model, the ratio for envelope is equivalent to 3% in Palermo apartment (case B). In this case, it is evident that the type of building
and Venice and drops to 2.5% in Vienna and Helsinki. For an enve- structure does not significantly affect the accuracy of the simpli-
lope with high thermal insulation, the ratio ranges between 5 and fied models in the heating mode, as the effect of climate conditions
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 459

Fig. 9. RMSE of lumped-capacitance models and peak loads for the apartment (case B) with east and west-oriented external walls with envelopes: (a) H1, (b) H2, (c) L1 and
(d) L2.

Fig. 10. The relative error ε for heating and cooling for the apartment (case B).

prevails. In fact, the blue indicators that represent the climate of greater for the apartment (case A) than that for the single room.
Helsinki are always lower than the red ones that represent the cli- In fact, the reduction in the RMSE ranges from −30% to −60% in
mate of Palermo, while the green and the orange indicators are the heating mode and from −45% to −75% in the cooling one. The
always somewhere in between. This holds true both for the 1C introduction of ventilation and internal heat gains (case B) does not
and 2C models, although the error of the former presents a higher modify the average RMSE reduction that remains close to approx-
dispersion. The improvement of the 2C model over the 1C one is imately 45% in the heating mode and to 60% in the cooling one.
460 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

Fig. 11. RMSE of the free floating indoor air temperature  i during off-seasons.

Fig. 10 shows that while the error ε1C of the 1C model shoots to very benchmark, it does not provide any information about the error
high values (20–30%) when the transition is made from the heating distribution.
to cooling mode, the error ε2C of the 2C model does not show such Fig. 12 shows the heating/cooling load calculated with the 7R2C
a sharp increase and always remains below the threshold of 12.5%. model as a function of the heating/cooling load calculated with
Contrary to what takes place in the heating mode, the type of TRNSYS for the apartment (case B) in Venice. Upper and lower
building envelope does seem to affect the accuracy of the models boundaries of +10% and −10% make it possible to see where
during the warm season. Indeed, lightweight building structures the simplified models tend to deviate from the reference values
present a higher error than heavyweight ones. This is particularly provided by the TRNSYS simulation. It was found that the under-
evident when the error of the 2C model is compared, since ε2C estimation and the overestimation detected for the buildings with,
increases from 6–7% to 12% when there is low thermal insulation respectively, low and high thermal insulations, are not uniformly
(i.e. from H1 to L1) and from 7–11% to 10–12% when it is high (i.e., distributed along the x axis. Indeed, it was apparent that the ther-
from H2 to L2). As reported in the literature review, Tindale [29] mal load for space heating went beyond the mentioned bounds for
argued that a second-order model is over-responsive for medium all the building envelopes especially when the heating load itself
and heavyweight building structures in presence of significant heat was low, i.e. during the middle seasons. This was particularly evi-
gain daily cycles −which typically occur during summer due to dent for the 1C model for envelopes H2 and L2 (Figs. 11(b) and
the effect of solar radiation. In our case, indeed, the second-order (d)). For buildings with high thermal transmittance (H1 and L1),
model is under-responsive with lightweight building structures. the deviation from the reference model was still high. The values of
It is the same effect (the low accuracy of the second-order model the cooling load obtained using the 1C model fell mostly outside of
during the cooling season) but obtained with two models that use the bounds traced confirming that the 1C model is not suitable for
different thermal capacitances. In Tindale’s model, the second ther- building simulation in the cooling mode since it is unable to detect
mal capacitance represented the volume of heated/cooled air. Thus, the fluctuations caused by the high heat gains occurring during the
due to its low value, the model was over-responsive when used to summer season.
reproduce the thermal behaviour of heavyweight buildings. In the Fig. 13 shows how the error of the indoor air temperature is dis-
VDI 6007 model both thermal capacitances represent the thermal tributed during the periods of free floating temperature, i.e. when
mass of building components (internal and external ones). Thus, the HVAC systems are turned off. What emerges is that the disper-
the model is not able to reproduce effects that have higher fre- sion of the 2C model is always lower than that of the 1C one and that
quency, such as the response of lightweight building structures to the error distribution seems to be affected to a greater extent by the
rapid fluctuations of solar heat gains. thermal insulation than by the thermal mass of the building since
Fig. 11 shows the RMSE of the simplified models in calculating envelopes H1-L1 (a-c) and H2-L2 (b-d) have similar distributions.
the indoor air temperature during periods when the HVAC systems The mean value of the air temperature (on the x-axis) in Fig. 13(b-d)
are turned off. The trend of the error roughly follows that of the is higher due to the greater thermal insulation with respect to the
relative error of the cooling load shown on the right-hand side of H1 (a) and L1 (c) envelopes. The error of the 2C model seems to be
Fig. 10. This probably takes place because during these periods the uniformly distributed around a positive value (bias) roughly falling
dynamic behaviour of the building (in this case the apartment) is between 0 and 0.2 K for the envelopes with low thermal insulation.
significantly affected by solar radiation. This leads to high diurnal
fluctuations of the indoor air temperature especially for climates
with significant differences in the day and night time external air 5. Conclusions
temperature, i.e. Vienna and Helsinki.
Although the RMSE indicates the approximation error made The 5R1C and 7R2C lumped-capacitance models described in
by the lumped-capacitance models compared to the TRNSYS the International Standard ISO 13790 and in the German Guide-
line VDI 6007 were used to simulate the thermal behaviour of a
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 461

Fig. 12. Distribution of heating and cooling load by the lumped-capacitance models for apartment (case B) in Venice with envelope (a) H1, (b) H2, (c) L1, (d) L2.

single room and an entire apartment using four reference building software. The accuracy of the simplified models was measured by
envelopes in four different climates. Those same reference build- comparing their results with those produced by TRNSYS as far as
ings were also simulated using the TRNSYS detailed simulation peak load, seasonal energy need, and dynamic thermal behaviour
462 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

Fig. 13. Error distribution of the indoor air temperature calculated by the lumped-capacitance models for apartment (case B) in Venice with envelope (a) H1, (b) H2, (c) L1,
(d) L2.

in both heating and cooling seasons were concerned. The latter was and between the indoor air temperature during the two periods of
measured by calculating the RMSE between the hourly load profiles free floating indoor air temperature.
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 463

The first-order model underestimated the peak load for space firmed the higher accuracy of the 2C model with respect to the 1C
heating of the apartment with low insulation (falling as low as one due to a lower dispersion of the hourly values of indoor air tem-
−6%) and overestimated it for the high thermal insulation situation perature (average RMSE of 0.57 K for the 1C model and 0.35 K for
(reaching up to +8%). This pattern was repeated in the second-order the 2C) and confirmed the influence of climate conditions linked to
model, although it was less evident. Furthermore, while the first- the effect of solar heat gains at different latitudes.
order model systematically underestimated the peak load for space In conclusion, both lumped-capacitance models appear to reli-
cooling (-11% on average), the second-order model made a fairly ably calculate the overall energy needs of buildings in both heating
accurate calculation. and cooling seasons. As far as transient behaviour is concerned,
The analysis of the energy needs for space heating calculated the first-order model of ISO 13790 model seems inappropriate to
by the lumped-capacitance models uncovered the same trend calculate neither the hourly cooling load profile nor the cooling
found for the peak load calculation. The heating needs of the peak load. The second-order model proposed by VDI 6007 is more
apartment with low/high thermal insulation were slightly under- accurate in both the heating and cooling modes.
/overestimated (approximately +/−5%) by both simplified models. The current study hopes to be useful to building designers who
The energy needs for space cooling were, instead, calculated with must choose between simplified simulation tools based on the stan-
sufficient accuracy by both models. The second capacitance led to a dards mentioned and to researchers who intend to integrate the
slight improvement with respect to the first-order model, reducing lumped-capacitance models presented here into city districts sim-
the average error from −3% to almost zero. ulation tools.
The 2C model brought a significant improvement over the 1C
model in terms of accuracy as far as the dynamic behaviour was Acknowledgements
concerned. The ratio between the RMSE and the mean load (termed
relative error, ε) for the simulation of the apartment was approxi- These research activities were supported by the Interdepart-
mately 6% and 9% in the heating and cooling seasons, respectively. mental Centre for Energy Technology and Economics “G. Levi Cases”
These values corresponded to an average decrease in the relative of University of Padua.
error by almost 50% in the heating mode and by more than 60% in
Appendix A.
the cooling one when we switched from the first-order ISO 13790
model to the second-order VDI 6007 model. The two months with See Table A1–A4
free floating temperature (15 Apr-15 May and 15 Sep-15 Oct) con-

Table A1
Peak load of the south-oriented single room.

Envelope Weather Peak load for space heating Peak load for space cooling [W]

TRNSYS[W] 1C[%] 2C [%] TRNSYS[W] 1C [%] 2C[%]

H1 PALERMO 409 −4.3 −10.2 1287 −19.4 7.4


VENEZIA 923 −6.4 −5.3 1104 −16.5 2.8
VIENNA 1200 −7.1 −5.1 1853 −16.6 1.1
HELSINKI 1652 −5.8 −3.1 1845 −14.5 3.7
H2 PALERMO 0 0 0 1054 −24.0 2.2
VENEZIA 360 10.5 7.4 891 −25.7 0.1
VIENNA 478 9.4 7.4 1570 −26.7 −3.8
HELSINKI 716 5.3 5.0 1518 −18.7 −1.0
L1 PALERMO 533 −5.2 1.1 1449 −26.4 7.5
VENEZIA 983 −8.4 −3.1 1189 −23.4 7.2
VIENNA 1352 −8.3 −1.8 1982 −21.7 2.5
HELSINKI 1694 −5.3 −1.2 1980 −21.0 4.4
L2 PALERMO 0 0 0 1474 −22.2 9.6
VENEZIA 390 7.3 7.3 1434 −15.6 8.1
VIENNA 524 5.5 5.9 1945 −19.0 5.9
HELSINKI 759 3.3 3.7 2030 −17.0 5.9

Table A2
Heating and cooling needs of the south-oriented single room.

Envelope Weather Energy need for space heating [kWh] Energy need for space cooling [kWh]

TRNSYS [kWh] 1C [%] 2C [%] TRNSYS [kWh] 1C [%] 2C[%]

H1 PALERMO 42 −27.8 −39.6 1175 −4.6 2.7


VENEZIA 1672 −9.8 −8.8 1119 −2.8 4.1
VIENNA 1983 −10.6 −9.2 1468 −0.2 5.1
HELSINKI 3414 −8.8 −6.8 1167 2.6 8.4
H2 PALERMO 0 0.0 0.0 1017 −7.7 −3.2
VENEZIA 422 17.6 12.9 1076 −7.4 −3.3
VIENNA 458 17.9 14.0 1551 −6.7 −3.5
HELSINKI 1142 10.2 8.3 1537 −7.5 −4.3
L1 PALERMO 147 −31.5 −21.3 983 −12.0 1.8
VENEZIA 1798 −7.9 −5.7 922 −6.1 2.9
VIENNA 2182 −8.9 −5.6 1188 −3.9 3.6
HELSINKI 3476 −6.6 −4.1 904 −5.0 −5.7
L2 PALERMO 0 0.0 0.0 1122 0.6 3.0
VENEZIA 420 4.4 2.4 1209 −0.1 3.3
VIENNA 470 3.4 3.7 1733 1.1 3.5
HELSINKI 1163 3.7 2.9 1731 1.7 3.8
464 J. Vivian et al. / Energy and Buildings 150 (2017) 447–465

Table A3
Peak load of the apartment (case A and B) oriented north-south.

Env. Weather Peak load for space heating Peak load for space cooling

TRNSYS [W] 1C [%] 2C [%] TRNSYS [W] 1C [%] 2C [%]

A B A B A B A B A B A B

H1 PALERMO 1170 1493 3.4 0.7 −1.6 −2.5 1701 2255 −13.6 −12.1 9.3 7.0
VENEZIA 2465 3170 −4.2 −3.2 −4.4 −1.9 1817 2412 −10.2 −7.1 −5.0 −4.4
VIENNA 3257 4565 −2.8 −3.2 −2.5 −0.4 2722 3222 −10.5 −8.1 −3.8 −3.9
HELSINKI 4206 5764 −2.5 −3.3 −2.4 −1.8 2431 2760 −7.2 −6.7 −1.0 −1.6
H2 PALERMO 0 51 0 435.3 0 337.3 1463 1928 −18.4 −12.3 5.5 0.5
VENEZIA 883 1600 10.8 2.1 7.8 2.6 1401 2014 −11.6 −7.1 −1.4 −2.0
VIENNA 1140 2395 10.3 1.5 7.1 1.1 2824 2837 −26.6 −9.6 −3.6 −1.1
HELSINKI 1641 3112 9.0 3.4 6.4 2.1 2401 2698 −14.3 −9.2 −1.5 0.3
L1 PALERMO 1399 1721 −3.8 −4.4 0.9 0.3 2200 2655 −22.7 −17.4 7.0 9.6
VENEZIA 2586 3443 −6.4 −6.5 −0.4 −0.6 2180 2778 −16.9 −12.6 2.3 1.1
VIENNA 3683 5102 −7.1 −6.3 −0.7 −1.0 3383 3884 −17.6 −14.6 2.2 1.3
HELSINKI 4423 6020 −4.3 −4.4 −1.3 −1.5 3188 3505 −17.4 −15.2 3.3 2.7
L2 PALERMO 0 606 0 7.8 0 7.3 1860 2189 −24.8 −16.1 2.3 1.3
VENEZIA 948 1719 8.8 1.4 8.5 2.2 1650 2247 −17.0 −11.1 −3.4 −3.5
VIENNA 1238 2568 8.8 1.3 5.6 2.8 2814 3304 −18.3 −14.1 −3.3 −2.8
HELSINKI 1713 3208 9.3 3.1 6.7 2.0 2826 3217 −17.4 −14.5 −0.9 −2.4

Table A4
Heating and cooling needs of the apartment (case A and B) oriented north-south.

Env. Weather Energy need for space heating Energy need for space cooling

TRNSYS [kWh] 1C [%] 2C [%] TRNSYS [kWh] 1C [%] 2C [%]

A B A B A B A B A B A B

H1 PALERMO 729 772 −2.9 −2.8 −9.7 −7.5 1696 2377 −7.9 −5.3 −1.9 −1.4
VENEZIA 5518 6823 −5.3 −4.1 −5.7 −4.2 1361 1812 −6.0 −4.6 −0.8 −0.6
VIENNA 6624 8637 −6.0 −4.4 −5.7 −4.0 1794 1886 0.0 −0.7 0.6 0.0
HELSINKI 10122 13013 −5.8 −4.2 −5.0 −3.6 930 778 4.6 2.9 3.9 2.6
H2 PALERMO 0 0 0.0 0.0 0.0 0.0 1802 2463 −6.3 −3.9 −2.7 −1.4
VENEZIA 1356 2623 14.6 6.2 9.8 3.7 1792 2211 −5.6 −3.6 −2.4 −1.2
VIENNA 1448 3325 15.4 6.8 11.3 4.8 2789 2661 −4.7 −3.8 −3.1 −2.1
HELSINKI 3090 6028 10.7 4.2 7.8 2.8 2474 1788 −5.6 −5.6 −4.1 −3.0
L1 PALERMO 785 875 −6.6 −10.1 −6.3 −7.4 1716 2384 −7.6 −4.8 −0.6 0.0
VENEZIA 5389 6775 −3.7 −3.9 −3.6 −3.6 1436 1888 −7.9 −5.7 0.6 1.1
VIENNA 6487 8584 −4.5 −4.3 −3.5 −3.4 1916 2042 −3.1 −3.8 1.3 1.5
HELSINKI 9770 12775 −3.6 −3.4 −3.0 −2.9 1070 941 −3.6 −6.3 3.9 4.5
L2 PALERMO 0 18 0.0 42.0 0.0 58.0 1744 2429 −4.5 −3.0 −1.5 −0.7
VENEZIA 1426 2696 12.7 5.4 9.0 3.4 1769 2211 −4.7 −3.3 −1.6 −0.7
VIENNA 1554 3474 13.0 5.0 11.1 4.0 2682 2623 −2.8 −2.9 −2.0 −1.2
HELSINKI 3186 6067 9.9 4.2 7.6 2.9 2375 1778 −3.9 −5.3 −3.2 −2.3

References [13] N.O. Milbank, J. Harrington-Lynn, Thermal response and the admittance
procedure, Build. Serv. Eng. Res Technol. 42 (1974) 38–47.
[1] European Commission, Eurostat, http://ec.europa.eu/eurostat/statistics- [14] C.P. Underwood, F.W.H. Yik, Modelling Methods for Energy in Buildings,
explained/index.php/Consumption of energy (Last access: 30 October 2016). Blackwell, Oxford, 2004.
[2] European Parliament, Directive 2010/31/EU of the European Parliament and [15] S. Privara, J. Cigler, Z. Vaňa, F. Oldewurtel, et al., Building modeling as a crucial
of the Council of 19 May 2010 on the energy performance of buildings part for building predictive control, Energy Build. 56 (2013) 8–22.
(recast). Official Journal of the European Union n. 153 of the 18/06/2010, 2010. [16] J.H. Kämpf, D. Robinson, A simplified thermal model to support analysis of
[3] J. Siroky, F. Oldewurtel, J. Cigler, S. Privara, Experimental analysis of model urban resource flows, Energy Build. 39 (2007) 445–453.
predictive control for an energy efficient building heating system, Appl. [17] M. Lauster, J. Teichmann, M. Fuchs, R. Streblow, D. Mueller, Low order thermal
Energ. 88 (9) (2011) 3079–3087. network models for dynamic simulations of buildings on city district scale,
[4] H. Madsen, J. Holst, Estimation of continuous-time models for the heat Build. Environ. 73 (2014) 223–231.
dynamics of a building, Energy and buildings 22 (1995) 67–79. [18] R. Nouvel, A. Mastrucci, U. Leopold, O. Baume, V. Coors, U. Eicker, Combining
[5] International Organization for Standardization ISO 13790: Energy GIS-based statistical and engineering urban heat consumption models:
performance of buildings: calculation of energy use for space heating and towards a new framework for multi-scale policy support, Energy Build. 107
cooling. 2nd ed. Geneva, Switzerland, 2008. (2015) 204–212.
[6] V. Corrado, E. Fabrizio, Assessment of building cooling energy need through a [19] D. Robinson, F. Haldi, J. Kämpf, P. Leroux, D. Perez, A. Rasheed, U. Wilke,
quasi-steady state model: simplified correlation for gain-loss mismatch, Citysim: comprehensive micro-simulation of resource flows for sustainable
Energy Build. 39 (2007) 569–579. urban planning, in: Proc. of the 11th Intern. IBPSA Conf. Building Simulation
[7] S.A. Klein, et al., TRNSYS 17: A Transient System Simulation Program, Solar 2009, Glasgow, 2009.
Energy Laboratory, University of Wisconsin, Madison, USA, 2010 http://sel. [20] M.M. Gouda, S. Danaher, C.P. Underwood, Building thermal model reduction
me.wisc.edu/trnsys. using nonlinear constrained optimization, Build. Environ. 37 (2002)
[8] US Department of Energy Efficiency and Renewable Energy, Washington DC, 1255–1265.
2003. http://www.eere.energy.gov/buildings/energyplus/. [21] F. Deque, F. Ollivier, A. Poblador, Grey boxes used to represent buildings with
[9] University of Strathclyde, Energy Systems Research Unit, Glasgow, 2003. a minimum number of geometric and thermal parameters, Energy Build. 31
http://www.esru.strath.ac.uk/Programs/ESP-r.htm. (1) (2000) 29–35.
[10] J. Clarke, Energy Simulation in Building Design, Butterworth-Heinemann, [22] R. Kramer, J. van Schijndel, H. Schellen, Simplified thermal and hygric building
Oxford, 2001. models: a literature review, Front. Arch. Res. 1 (2012) 318–325.
[11] D.G. Stephenson, G.P. Mitalas, Calculation of heat conduction transfer [23] K.L. Beuken, Wärmeverluste bei periodisch betriebenen Öfen, Dissertation
functions for multilayer slabs, ASHRAE Transactions 77 (2) (1971) 117–126. Freiburg (1936).
[12] D.G. Stephenson, G.P. Mitalas, Room thermal response factors, ASHRAE Trans.
73 (2) (1967), III.2.1–III.2.10.
J. Vivian et al. / Energy and Buildings 150 (2017) 447–465 465

[24] L. Rouvel, Berechnung des wärmetechnischen Verhaltens von Räumen bei [38] International Organization for Standardization ISO 13791: Thermal
dynamischen Wärmelasten, Brennst-Wärme-Kraft 24 (1972), Nr. 6 (in performance of buildings — Calculation of internal temperatures of a room in
German). summer without mechanical cooling — General criteria and validation
[25] Rouvel L. Thermische Gebäudesimulation GEBSIMU- Berechnungsverfahren procedures. Geneva, Switzerland, 2012.
zum instationären thermischen Gebäudeverhalten, 2015. http://www. [39] International Organization for Standardization ISO 13792: Thermal
gebsimu.de. performance of buildings − Calculation of internal temperatures of a room in
[26] L. Laret, Use of general models with a small number of parameters: part summer without mechanical cooling − Simplified methods. Geneva,
1—theoretical analysis, in: Proc. of 7th Int.Congress of Heating and Air Switzerland, 2012.
Conditioning CLIMA 2000, Budapest, 1980. [40] German Association of Engineers, Calculation of transient thermal response of
[27] F. Lorenz, G. Masy, Mèthode d’évaluation de l’économie d’énergie apportée rooms and buildings − modelling of rooms. 91.140.10 (VDI 6007-1), Beuth
par l’intermittence de chauffage dans les bâtiments. Traîtement par Verlag GmbH, Düsseldorf, 2012.
differences finies d’un model a deux constantes de temps, in: Report No. [41] P. Bacher, H. Madsen, Identifying suitable models for the heat dynamics of
GM820130-01. Faculte des Sciences Appliquees, University de Liege, Liege, buildings, Energy Build. 43 (2011) 1511–1522.
Belgium, 1982 (in French). [42] G. Reynders, J. Diriken, D. Saelens, Quality of grey-box models and identified
[28] J.A. Crabb, N. Murdoch, J.M. Penman, A simplified thermal response model, parameters as function of the accuracy of input and observation signals,
Build. Serv. Eng. Res. Technol. 8 (1987) 13–19. Energy Build. 82 (2014) 263–274.
[29] A. Tindale, Third-order lumped-parameter simulation method, Build. Serv. [43] H. Harb, N. Boyanov, L. Hernandez, R. Streblow, D. Müller, Development and
Eng. Res. Technol. 14 (3) (2003) 87–97. validation of grey-box models for forecasting the thermal response of
[30] K.A. Antonopoulos, E. Koronaki, Apparent and effective thermal capacitance of occupied buildings, Energy Build. 117 (2016) 199–207.
buildings, Energy 23 (3) (1998) 183–192. [44] F. Rubel, M. Kottek, Observed and projected climate shifts 1901–2100
[31] G. Fraisse, C. Viardot, O. Lafabrie, G. Achard, Development of a simplified and depicted by world maps of the Köppen-Geiger climate classification,
accurate building model based on electrical analogy, Energy Build. 34 (2002) Meteorol. Z. 19 (2010) 135–141.
1017–1031. [45] U.S. Department of Energy Energy Efficiency and Renewable Energy D.C.
[32] T.R. Nielsen, Simple tool to evaluate energy demand and indoor environment Washington, 2003 https://energyplus.net/weather/simulation (Last Access:
in the early stages of building design, Sol. Energy 78 (1) (2005) 73–83. 30 October 2016).
[33] Danish Building, Urban Research BSim, Danish Building and Urban Research, [46] Solar Energy Laboratory University of Wisconsin, Madison, 2007. TRNSYS 16,
Hørsholm, Denmark, 2003, pp. 2002 http://www.bsim.dk. a TRaNsient SYstem Simulation program,Vol. 6 (Multizone Building modeling.
[34] L. Rouvel, F. Zimmermann, Berechnung des instationären thermischen with Type56 and TRNBuild) http://web.mit.edu/parmstr/Public/
Gebäudeverhaltens. LH 55 3 2004 39–46; LH 55 (4) (2004) 24-30. Documentation/06-MultizoneBuilding.pdf (Last Access: 30 October 2016).
[35] E.A. Rodriguez, F.J. Jara Sanchez de la Flor, S.A. Dominguez, et al., A new [47] German Association of Engineers, Calculation of transient thermal response of
analytical approach for simplified thermal modelling of buildings: rooms and buildings—modelling of solar radiation. 91.140.10 (VDI 6007-3),
self-Adjusting RC-network model, Energy Build. 130 (2016) 85–97. Beuth Verlag GmbH, Düsseldorf, 2015.
[36] European Committee for Standardization BS EN 15265: Energy performance [48] Ente Nazionale Italiano di Unificazione UNI TS 11300-1. Determinazione del
of buildings. Calculation of energy needs for space heating and cooling using fabbisogno di energia termica dell’edificio per la climatizzazione estiva ed
dynamic methods. General criteria and validation procedures. Brussel, invernale (Evaluation of energy need for space heating and cooling) (in
Belgium, 2007. Italian). UNI, Milan, Italy, 2014.
[37] International Organization for Standardization ISO/DIS 52016-1: Energy [49] MATLAB version 7.10.0. Natick, Massachusetts : The MathWorks Inc., 2010.
performance of buildings − Calculation of the energy needs for heating and [50] International Organization for Standardization ISO 13786: Thermal
cooling, internal temperatures and heating and cooling load in a building or performance of building components −Dynamic thermal
building zone − Part 1: Calculation procedures. Geneva, Switzerland, 2014. characteristics—Calculation methods. Geneva, Switzerland, 1999.

Potrebbero piacerti anche