Professional Documents
Culture Documents
A Numerical Model To Evaluate The Thermal Behaviour of PCM Glazing System Configurations
A Numerical Model To Evaluate The Thermal Behaviour of PCM Glazing System Configurations
A Numerical Model To Evaluate The Thermal Behaviour of PCM Glazing System Configurations
a r t i c l e i n f o a b s t r a c t
Article history: The adoption of Phase Change Materials (PCMs) in building components is an up-to-date topic and a
Received 11 June 2012 relevant number of research activities on this issue is currently on the way. A particular application
Received in revised form 20 July 2012 of PCMs in the building envelope focuses on the integration of such a kind of material into transpar-
Accepted 28 July 2012
ent envelope components. A numerical model that describes the thermo-physical behaviour of a PCM
layer in combination with other transparent materials (i.e. glass panes) is developed to perform numer-
Keywords:
ical analyses on various PCM glazing systems configurations. The paper illustrates the structure of the
Building envelope
model, the main equations implemented and the hypotheses adopted for the model development. The
Numerical model
Thermal simulation
comparison between numerical simulations and experimental data of a simple PCM glazing configura-
Responsive building element tion is also presented to show the potentials and the limitations of the numerical model. While a good
Advanced glazing agreement between simulations and experimental data can be shown for the surface temperature of the
PCM glazing, the comparison between simulated and measured transmitted irradiances and heat fluxes does
not always reach the desired accuracy. However, the numerical tool seems to predict well the thermo-
physical behaviour of the system and may therefore represent a good starting point for simulations on
different configurations of PCM glazing systems.
© 2012 Elsevier B.V. All rights reserved.
0378-7788/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.enbuild.2012.07.036
142 F. Goia et al. / Energy and Buildings 54 (2012) 141–153
Fig. 1. Scheme of the computational grid (glass panes and PCM layer).
The model herewith illustrated is expected to be used to perform heat capacity of the components (glass panes and PCM layer) is
preliminary and sensitivity analyses on the application of PCM glaz- taken into account, so that the model can simulate the behaviour
ing systems in different building types and climates. The final aim of the system under dynamic conditions.
is to evaluate the effectiveness of such a class of technologies in The following hypotheses have been adopted for the develop-
reducing the energy consumption in buildings. Furthermore, due ment of the model:
to the structure of the model and its implementation in the Mat-
lab/Simulink environment, combinations of PCM layers with other • each node represents a layers that is supposed to be homoge-
glazing-related technologies (e.g. solar shading devices, smart win- neous;
dows, airflow in window cavities) can be investigated in the next • the glass surface is considered a grey body in the IR range;
future. • the surfaces of the outdoor and indoor environment are consid-
The aim of the model is therefore not to exactly replicate the ered as black and/or grey body;
physical phenomena, taking place at micro-, meso- and macro- • the convection within the PCM layer (when in liquid state) is
scale inside the PCM. In fact, the exact replication would be quite neglected, due to the cavity geometry (thickness: 15 mm);
complex, because of the nature of the material, that is highly non • the radiative exchange between the two glass surfaces facing the
homogeneous, and of the different phenomena that occur at very cavity filled with PCM is neglected too, being the PCM, both when
different scales. Most of the these phenomena own their com- in liquid state and in solid state, highly non-transparent to the
plexity to the non-linearity of many of the thermal and optical long-wave radiation;
properties. The non-linear behaviour makes the modelling of the • the optical properties of the glass panes are only function of the
system extremely complicated. Some of the most complex aspects incidence angle of solar radiation;
that should be taken into account to reach the fully replication • the thermal properties of the glass are temperature independent;
of the thermo-physical and optical phenomena are: the change • the optical and thermal properties of the PCM depend on the
in the state of aggregation with some super-cooling effects, the temperature (state of the PCM).
possible creation of macroscopic crystals after the re-solidification,
the non homogeneity when in solid state, convective mass move-
In Fig. 1, the schematic representation of the node grid is shown.
ment, the absorption of solar radiation characterized by scattering
For the “basic” element of a PCM glazing system (i.e. two panes of
effects, and the shortwave radiative transfer through a random
glass and a cavity filled with PCM), a total of 11 nodes have been
medium. Quite complex calculation tools, e.g. CFD/FEM codes,
used. Seven nodes (1 each glass pane, 5 for the PCM layer) have
would be necessary to reach the target of the “exact modelling”.
buffer (i.e. heat capacity) and absorption (in the shortwave range)
However, these tools are mostly suitable for simulations under
properties. The nodes at the interfaces between the glass panes
steady state conditions, due to the considerable calculation load
and the indoor/outdoor environment, as well as the nodes at the
that they require. Furthermore, they are not suitable to perform
interface between the glass panes and the PCM layer, do not possess
system analyses (that is the coupled behaviour of the component
heat capacity and do not absorb solar energy. If a more complex
with the building), as well as sensitivity analyses, since they are
glazing configuration needs to be simulated (e.g. a triple glazing
not trivial and quick to use. Analyses of the component’s perfor-
with a cavity filled with PCM and the another cavity filled with
mance under transient conditions, or under real climatic conditions
air/gas), no nodes are associated to the air/gas layer and the heat
and/or the operative conditions would be extremely complex and
transfer process (radiation and convection) is computed between
time-consuming.
the two surface nodes that face each other through the cavity itself
(node Teg,i and Tig,o in Fig. 2). The two schemes (Figs. 1 and 2) can
2. Physical–mathematical model be combined together to create multi-layered PCM glazing systems
that contain both PCM layer and non-ventilated cavities.
2.1. Calculation scheme and hypotheses Each PCM node (nodes Tpcm,1 to Tpcm,5 in Fig. 1) represents a
slice of PCM. Each slice of PCM can be in a solid state, in the melt-
The developed numerical procedure is based on a 1D nodal ing phase, or in liquid state, depending on the temperature of the
model (heat and shortwave radiation transmission along x and node itself. This approach, which is different from the moving grid
y axis are not considered). In the standard model, three nodes procedure, allows the PCM layer to melt and re-solidify without
are associated to each glass pane, while the PCM layer is repre- selecting a priori the direction of the melting/re-solidification front.
sented by means of five nodes. The number of nodes for both the In fact, during an experimental campaign [7], it was observed that
glass pane and the PCM layer can be increased, but the discretized the melting process of the PCM layer does not necessarily occur
scheme herewith presented (Fig. 1) is a good compromise between from the outer region and proceeds towards the inner ones (or
accuracy and computational time. Heat and shortwave radiation vice versa). The dynamics of the melting process (as well as the
transport equations are implemented for the glass pane and the one of the re-solidification process) is more complex and depends
PCM layer. Energy conservation equations are written for each node on the boundary conditions. A grid moving procedure is not able
and numerically solved to obtain the thermal field. Obviously, the to replicate this phenomenon. The adopted discretization allows a
144 F. Goia et al. / Energy and Buildings 54 (2012) 141–153
Fig. 2. Scheme of the cavity modelling in the case of more complex PCM glazing systems.
higher degree of freedom with respect to the melting process, but is underway, making use of a large integrating sphere (diameter:
still it presents some limitations in replicating the behaviour of the 0.75 m) because of the elevate scatter phenomenon caused by the
system (e.g. patches of solid parts casually dispersed in the liquid). material when in solid state. The optical properties of the PCM are
At each time step, the algorithm reads the input data (some of analysed for either the solid state or liquid state [12], but not for a
them are time-dependent, some of them are time independent) and PCM that is undergoing a phase change – i.e. when the PCM layer is
iteratively solves the energy conservation equations until conver- not completely solid or completely liquid. The preliminary results
gence is assured. show that the solar and light transmittance of glazing prototype
The model is divided in two sub-models. The main one is dedi- when the PCM is in liquid state (0.75 and 0.85, respectively) are
cated to solve the heat transfer process and the other one is aimed considerably higher than the solar and light transmittance when
at computing the solar irradiance transmitted and absorbed by the PCM in solid state (0.46 and 0.55, respectively). This reduction of
glazing. The absorbed energy calculated by the second sub-model the solar and light transmittance is due to the increase of both the
is fed to the first sub-model – i.e. it becomes an input data for the absorptance and reflectance.
main sub-model. It is important to mention the high coupling that Due to the structure of the model, each PCM sub-layer makes use
occurs between the two sub-modules in the case of a PCM glaz- of a reflection , transmission and absorption ˛ coefficient that
ing. In fact, in the case of a conventional glazing system, the two represent the optical behaviour of a PCM layer with a thickness
sub-modules can be solved in sequence (the second one first, and that is equal to the one of the sub-layer (i.e. 3 mm, when the total
then the main one), since the optical properties of the glazing (that PCM thickness is 15 mm). Because of the non-linearity of the opti-
affect the amount of energy absorbed by the fenestration) can be cal phenomena, it was not possible to obtain these data from the
considered temperature independent. On the contrary, the opti- experimental campaign on a 15 mm thick PCM layer. Instead, the
cal properties of the PCM layer depend on the temperature of the calibration of the physical–mathematical model on a set of inde-
PCM itself (solid or liquid). Therefore, the two sub-models must be pendent experimental data was eventually preferred. Additional
solved simultaneously since they are mutually influenced. information on this activity can be found Section 3.
As far as the glass pane is concerned, reflection , transmis-
sion and absorption ˛ coefficients are computed by making use
2.2. Short-wave (UV–VIS–NIR) transport modelling
of parametric curves (obtained with the software WIS) that pro-
vide the , and ˛ coefficients as functions of the glass thickness
The solar radiation plays a crucial role in the behaviour of a
and the incident angle of the solar radiation. The parametric curves
glazed system and its importance is even more enhanced in case of
implemented in the models are not here reported for the sake of
a PCM glazing system. In this last case, the energy storage effect
brevity.
obtained within the PCM layer is mostly due to the interaction
The total amount of solar energy that enters the indoor envi-
between the PCM and the solar radiation. Direct and diffuse radia-
ronment, that is absorbed in each layer and that is exchanged with
tion must be separately processed, because their angles of incidence
the outdoor environment depends on the properties of the layers
are different (and thus it is different their interaction with the glass
that compose the glazing system and on their interaction. These
panes and PCM layer). In case of a PCM glazing system, an additional
amounts can be found, once the impinging solar radiation, , and
complexity arises: the PCM layer (when in solid state) “converts”
˛ coefficients of each layer and the amount of shortwave radiation
a direct incident solar radiation in a transmitted/reflected diffuse
that is reflected by the indoor environment are known, by solving
radiation.
iteratively the energy balance of the radiative heat fluxes [13] for
The conventional modelling approach for the short-wave reflec-
each layer (node):
tion in semitransparent media makes use of the Fresnel and Snell
laws (to assess the reflection coefficient), while the solar absorp-
tion and transmission coefficients can be modelled by making use of q̇sol,n = (In−1→n + In+1→n ) − (In→n+1 + In→n−1 ) (1)
the Bouguer–Beer Law, cf. [10,11]. As mentioned, the correct optical
modelling of a PCM layer presents additional difficulties due to the where q̇sol,n is the absorbed solar radiation in the nth layer [W m−2 ]
scattering effects that occur when the PCM layer is in solid state. The and the Ix→y variables are the short-wave radiative fluxes [W m−2 ]
incorporation of a numerical method that computes the short-wave that enter/leave the nth layer, as shown in Fig. 3.
radiation transport of such a material within the main heat trans- As far as the , and ˛ coefficients of each PCM sub-layer are
fer model is a non-trivial task. As shown by different authors [3,6], concerned, two values for each coefficient are determined: a value
it is highly preferable to independently determine, by means of a that describes the behaviour of the PCM sub-layer when in solid
separate model (e.g. Monte Carlo simulations of photons and their state and a value that is associated to the liquid state. A sub-layer in
interaction with the matter) or by experimental measurements, the phase transition (Tmelt,low < T(t) < Tmelt,high ) is still considered within
reflection, transmission and absorption coefficients of a PCM layer the solid state and the coefficients used in the calculation are
as a function of its thickness and state of aggregation. those for the solid state. Only when the PCM sub-layer completes
In the present paper the latter approach is chosen. A detailed the phase transition (T(t) > Tmelt,high ) the values that describe the
experimental campaign on the optical properties of the PCM layer behaviour of the PCM sub-layer in liquid state are used instead.
F. Goia et al. / Energy and Buildings 54 (2012) 141–153 145
Fig. 3. Scheme of the energy balance of radiative short-wave heat fluxes in multi-layered glazing systems.
Regardless the layer which is crossed by the income short-wave is far larger than the one of the façade, the radiative exchange, per
radiation (i.e. glass pane or PCM sub-layer), the first step of the unit of surface q̇rad,in , can be written as (10):
calculation procedure is aimed at assessing the component that is
4 4
reflected by the layer. The second step calculates the amount of q̇rad,in = · εig,i · (Tig,i − Tsur,i ) (10)
shortwave radiation that is absorbed by the layer. The amount of
where εs,i is the emissivity of the inner glass surface of the PCM
shortwave energy that is eventually transmitted by the layer is then
glazing system.
obtained by difference between the total impinging radiation and
The IR radiative exchange, per unit of surface q̇rad,cav , between
the reflected and absorbed parts.
the two glass slabs that face a cavity is given by (11), being εeg,i and
εig,o the emissivity of the outer and inner glass surfaces, respec-
2.3. Long-wave radiative heat exchange
tively.
where seg is the thickness of the external glass, eg is the den- hpcm
∗
sity of the external glass, ceg is the specific heat capacity of the cpcm,m = (20)
Th
external glass, eg is the density of the external glass, q̇sol,eg is
the absorbed solar radiation in the external glass, Teg,c is the tem-
Th = Tpcm,fi − Tpcm,in (21)
perature of the node that represents the external glass, Teg,o and
Teg,i are the temperature at the outdoor and indoor surface of the
glass pane respectively, and t is the time. The ODE (16) is calcu- This approach allows the conductive heat transfer problem to be
lated in a continuous time domain by Matlab/Simulink, by means transformed into a “single-phase”, non-linear conduction problem
of the Dormand–Prince method, obtaining the temperature Tpcm,n in the entire calculation range (T1 to T2 ), where the non-linearity
by directly solving the integral equation (17). is given by the different value that the specific heat capacity can
−1 assume, according to the state of aggregation of the PCM, or in other
1 s
n 1 words, according to the temperature of the PCM (22).
Teg,c = · · · (Teg,o − Teg,c )
dt
(seg · eg · ceg ) 2 n
−1 ∗ (T
cpcm pcm ) = cpcm,s if Tpcm ≤ Tpcm,in
s n 1
− · · (Teg,c − Teg,i ) + q̇sol,eg dt (17) ∗ (T ∗
2 n cpcm pcm ) = cpcm,m if Tpcm,in < Tpcm < Tpcm,fi (22)
∗ (T
cpcm pcm ) = cpcm,l if Tpcm ≥ Tpcm,fi
∗ (T
cpcm pcm ) = cpcm,s if Tpcm ≤ Tpcm,in
∗
cpcm,m − cpcm,s ∗
cpcm,m − cpcm,s
∗ (T
cpcm pcm ) = · Tpcm +cpcm,s − · Tpcm,i if Tpcm,in < Tpcm ≤ Tpcm,m
Tpcm,m − Tpcm,in Tpcm,m − Tpcm,in
∗ ∗
(23)
∗ (T
cpcm,l − cpcm,m cpcm,l − cpcm,m
cpcm pcm ) = − · Tpcm +cpcm,l + · Tpcm,f if Tpcm,m < Tpcm < Tpcm,fi
Tpcm,fi − Tpcm,m Tpcm,fi − Tpcm,m
∗ (T
cpcm = cpcm,l if Tpcm ≥ Tpcm,fi
pcm )
∗
where cpcm,m is the maximum value of equivalent specific heat
capacity that can be reached. This value is reached when the
melting/re-solidification process reaches its peak (i.e. when Tpcm =
∗ • the time profiles of the temperature of each layer (glass and PCM,
Tpcm,m ). cpcm,m is automatically calculated by the code as a function
inner and surface temperatures);
of the PCM properties (cpcm,s , cpcm,l , Tpcm,in , Tpcm,fi , Tpcm,m , hpcm ),
• the time profile of transmitted shortwave heat fluxes (i.e. trans-
according to Eq. (24).
mitted irradiance);
• the time profile of the “longwave” heat fluxes exchanged between
∗
Tpcm,in − Tpcm,m Tpcm,m − Tpcm,fi 2hpcm
cpcm,m = · cpcm,s + · cpcm,l + (24) the various surfaces of the glazing system, and of the surface heat
Th Th Th
fluxes (that is radiative IR heat exchange plus convective heat
exchange) with the indoor and outdoor environment.
Therefore, the final differential equation describing the “single-
phase”, non-linear conduction problem, regardless the state of
aggregation of the PCM is Eq. (25) and the temperature Tpcm,n of
the node n of the PCM layer is obtained by Eq. (26): 3. Model calibration
−1
∂Tpcm,n s n
The optical properties (, and ˛ coefficients) of each PCM sub-
∗
(sn · n · cpcm,n (Tpmc,n )) · = · (Tpcm,n−1 − Tpcm,n ) layer are needed to perform simulations. However, due to the fact
∂t n
−1 that optical phenomena are not linear (e.g. the solar transmission
s n of a 6 mm thick PCM layer is not double the one of two 3 mm thick
− · (Tpcm,n − Tpcm,n+1 ) + q̇sol,n (25) PCM layers, and not the half of a 12 mm thick PCM layer), the deter-
n
mination of the values of these coefficients is not a trivial task.
Furthermore, highly non-linear phenomena are registered when
−1 the PCM is in phase transition. Therefore, as far as the , and ˛
1 s n coefficients are concerned, the physical–mathematical model was
Tpcm,n = ∗ · · (Tpcm,n−1 − Tpcm,n )
dt
(sn · n · cpcm,n (Tpcm,n )) n calibrated on a set of experimental data (more information on the
−1 experimental campaign are given in Section 4.1 and in Goia et al.
s n [7]). Data used to calibrate the model belongs to a different period
− · (Tpcm,n − Tpcm,n+1 ) + q̇sol,n dt (26)
n that the ones that are used to validate the model. Data used for
the calibration represent periods when the indoor air temperature
∗
where cpcm,n (Tpcm,n ) is then the equivalent specific heat capacity of was kept at the desired set-point, and periods when the indoor air
the PCM in the node n, as defined in Eq. (23), and sn is the thickness temperature was not controlled by means of a HVAC system and
of the PCM slice – the spatial discretization of the PCM layer is the test cell was in free running mode. Given this assumptions, the
obtained by dividing the total thickness of the PCM layer in 5 slices; independence between the calibration process and the validation
therefore, the PCM layer is represented by 5 nodes, which have both process is assured.
heat capacity and internal heat source (cf. Fig. 1). In particular, the calibration of the model was carried out con-
∗
It is worth mentioning that cpcm,n (Tpcm,n ) can assume different sidering a period of 14 days. , and ˛ coefficients of the PCM
values in the different computational nodes, at the same time step, sub-layers were searched ‘til a good agreement between the model
since each n node can be at a different temperature. Furthermore, and the experimental data concerning the transmitted solar irra-
it must be highlight that hysteresis phenomena occurring between diance was found. Fig. 5 shows the good agreement between the
the melting process and the re-solidification process are not taken measured values of transmitted solar irradiance and the numeri-
into account in this model. Therefore, cpcm ∗ (T cal simulation. A very satisfactory agreement was reached both in
pcm ) is independent
from the direction of the phase change process (i.e. from solid state days with high solar irradiation (when the PCM layer completes the
to liquid state or vice versa). melting process) and in days with low solar irradiation (when the
PCM layer remains in solid state). To assess the reliability of the cal-
ibrated optical properties, the comparison between measured and
2.7. Model input/output data simulated indoor surface temperature of the PCM glazing systems
was also analysed during the same period. The very good agreement
The model inputs are: the thermophysical properties of the between the measured and the simulated values of indoor surface
materials (glass and PCM) and the boundary conditions. As far as the temperature (Tig.i , in Fig. 1) is illustrated in Fig. 6. Therefore, it is pos-
boundary conditions are concerned, the indoor/outdoor air temper- sible to state that, after the good replication of the transmitted solar
ature time profiles, the profile of the incident solar irradiance and irradiance was achieved (first stage of the calibration), the model
of the incident angle of the solar radiation and the profile of the was also able to correctly simulate, using the optical properties that
wind speed are needed. were determined, the indoor surface temperature (second stage of
The model provides, as output data: calibration, validation of the calibration).
148 F. Goia et al. / Energy and Buildings 54 (2012) 141–153
Fig. 5. Calibration of the model: comparison between simulated and measured transmitted solar irradiance.
4. Simulation and model validation value was adopted, conservatively, for all the thermocouples. The
heat flux meter sensors presented an accuracy of ±5%, with a confi-
In order to carry out a validation of the numerical model, a dence interval of 95%. Three hemispherical pyranometes were used
simple PCM glazing system is simulated. The obtained theoreti- for the experimental campaign. Their sensitivities were provided
cal results are compared against the experimental data collected by the producers and periodically verified. The hourly total mea-
during a test cell campaign [7]. surement uncertainty, declared by the producers, lies in the range
of 2–5%, including all the factors of influence that contribute to
4.1. Experimental data the measurement error (such as non-linearity sensitivity variation,
temperature dependency, directional error).
The PCM glazing prototype used for these tests was made of two
panes of clear glass (6 mm and 8 mm thickness) and a layer of paraf-
4.2. Numerical simulation and data analysis
fin wax (15 mm thickness) – cf. Fig. 1. The PCM glazing system was
1.40 m long and 1.15 m high. The used paraffin wax had a nominal
The same PCM glazing configuration tested during the experi-
melting temperature of 35 ◦ C and a melting range of about 10 ◦ C.
mental campaign [7] is then simulated by means on the numerical
The sample was installed on an outside test cell located in a humid
model herewith described. The boundary conditions (e.g. both
subtropical climate (Cfa – Köppen climate classification, cf. [20]).
indoor and outdoor air temperature, solar irradiance) used as input
The indoor air temperature of the test cell was maintained at the
data for the simulation are the ones measured during the exper-
desired set-point (tolerance ± 1 ◦ C) by means of an all air recircula-
imental campaign. In Fig. 7 the hourly profile of the boundary
tion system. The set-point temperature was set to 26 ◦ C in summer,
conditions used to validate the numerical model (7 summer days
20 ◦ C in winter and 23 ◦ C in mid-season. In addition, experiments
and 7 winter days) are shown. A total of 169 couple of values (mea-
were also carried out by letting the indoor air temperature to freely
sured data and simulated values) are used to validate the model,
float (free running mode). The measurement apparatus consisted of
each season. In Figs. 8–10 the comparisons between the experimen-
40 sensors (thermocouples, heat flux meters, pyranometers) con-
tal and numerical hourly profiles of the main physical quantities
nected to a data logger. It was designed to monitor both the physical
(indoor surface temperature, surface heat flux and transmitted
quantities of two glazed units (a conventional DGU was also instru-
solar irradiance, respectively) are shown, both for the summer and
mented for reference purpose) and the boundary conditions inside
the winter season. In Table 1 the material properties used during
and outside the test cell.
the simulation are summarized.
Data acquisition was performed on a 5 min basis and lasted for
Alongside with the comparison of the hourly profile of the
about 1 year. Data collected during the experimental campaign
indoor surface temperature, of the transmitted irradiance and of
were therefore representative of the behaviour of the PCM glaz-
the indoor surface heat flux, the simulated and measured daily
ing system under different boundary conditions and in different + −
energy (E24 h ) and the disaggregated daily energies (E24h , E24 ) are
seasons. h
calculated and compared. In particular, it holds:
The measurement accuracy of each thermocouple was assessed
after they were calibrated. As a result of this procedure, the highest
likely uncertainty, using the 95% confidence limit, was ±0.3 ◦ C. This q̇tot (t) = Itr (t) + q̇s,i (t) (27)
Fig. 6. Calibration of the model: comparison between simulated and measured indoor surface temperature of the PCM glazing.
F. Goia et al. / Energy and Buildings 54 (2012) 141–153 149
Fig. 7. Boundary conditions (outdoor air temperature and solar irradiance on the outdoor vertical surface) used for the validation of the numerical model.
Fig. 8. Simulated and measured indoor surface temperature of the PCM glazing.
Fig. 9. Simulated and measured indoor surface heat flux of the PCM glazing.
150 F. Goia et al. / Energy and Buildings 54 (2012) 141–153
Fig. 10. Simulated and measured transmitted solar irradiance of the PCM glazing.
where is given by Eq. (34);
j
1
Short-wave and long-wave radiative heat fluxes plus convective heat flux, = sj − ej (34)
exchanged between the indoor surface of the glazing and the indoor environment. j
F. Goia et al. / Energy and Buildings 54 (2012) 141–153 151
+ −
Fig. 11. Simulated and measured disaggregated (E24 h
, E24 h
) daily energy.
Table 2
Values of the relevant material properties used during the simulation.
Physical quantity Season PRMSE RMSE ¯ STD
¯
Surface temperature of the PCM Summer & winter 7.1% 1.7 ◦ C 1.1 ◦ C 1.3 ◦ C
glazing system (ig,i) Summer 6.0% 1.9 ◦ C 1.4 ◦ C 1.2 ◦ C
Winter 8.1% 1.5 ◦ C 0.9 ◦ C 1.2 ◦ C
Transmitted surface heat flux (q̇s,i ) Summer & winter – 14.9 W m−2 11.4 W m−2 9.5 W m−2
Summer – 17.9 W m−2 13.3 W m−2 11.9 W m−2
Winter – 11.1 W m−2 9.6 W m−2 5.5 W m−2
Transmitted solar irradiance (Htr ) Summer & winter – 20.3 W m−2 6.5 W m−2 19.3 W m−2
Summer – 27.6 W m−2 10.4 W m−2 25.6 W m−2
Winter – 7.9 W m−2 2.7 W m−2 7.5 W m−2
n 5. Discussion
2
1
¯= ·
STD −
¯ (35)
n j j The comparison between the numerical results and the experi-
j=1
mental data demonstrates the good ability of the numerical model
to simulate the phenomena that occur in a PCM glazing system,
despite the several assumptions and simplifications that have been
In Table 2, the PRMSE, RMSE, ¯ and STD
¯ for the hourly done.
profile of the indoor surface temperature, surface heat flux and In particular, as far as the indoor surface temperatures of
transmitted irradiance are reported, for the summer season, the the glazing are concerned (Fig. 8), a good agreement between
winter season and summer and winter seasons together. the experimental data and the numerical simulation is found.
The model is capable of correctly predicting highest and lowest A less satisfactory agreement between numerical and experi-
peak values of the hourly profile, even if a little over-estimation of mental results is found as far as the transmitted solar irradiance
the highest values and an under-estimation of the lowest values through the PCM glazing is concerned. This is particularly note-
are recorded, both in summer and in winter. A more detailed anal- worthy in sunny summer days, when the PCM layer undergoes a
ysis of the time profiles reveals that the accuracy is slightly worst complete phase change and becomes highly transparent to the solar
during the charge phase of the PCM, while a better agreement is radiation (cf. Goia et al. [7]). In such case, see e.g. Fig. 10 (summer),
reached when the paraffin layer re-solidifies after having experi- during the third and the forth day, the simulated peak values of
enced a complete fusion. As clearly shown by Fig. 4a, the model has the transmitted solar irradiance are about 250 W m−2 , while the
the tendency to over-estimate the surface temperature during the measured values are about 100–125 W m−2 . These are two sum-
melting process, that is, it predicts a quicker charge phase than the mer days with high outdoor air temperature and (above all) high
one which actually occurs. On the contrary, during the discharge solar irradiance, which cause the complete melting of the PCM
phase, even in case of a complete melting of the PCM layer, a bet- layer. On the contrary, a different behaviour is registered in the fifth
ter agreement between the experimental data and the numerical day. In this case the boundary conditions are quite similar to the
output is seen. The model reliability in predicting the surface tem- previous day (cf. Fig. 7), but surprisingly the agreement between
perature of the glazing during the winter (Fig. 4b) is better than in measured and predicted quantities is very good. Such characteris-
summer. Such behaviour can be explained by the fact that the PCM tic may give rise to the question whether the experimental data or
in this season never completes the phase change and remains in the numerical data (or none of them) are the ones that represent
the solid state (or in a mix state where the solid phase is predom- with the best reliability the physical phenomenon. The measure-
inant) for most of the time. A slight over-estimation of the surface ment of the transmitted solar irradiance, in fact, is not a trivial task,
temperature of the glazing is shown during the winter time when since the PCM glazing shows a very complex and non homoge-
the temperature of the PCM layer is relatively high; a little under- neous behaviour. Therefore, it may happen that the actual readings
estimation occurs when the PCM layer is significantly lower than of the pyranometer are negatively influenced by shades caused by
the transition temperature (e.g. during the night time or in cold small fractions of solid paraffin dispersed in an almost completely
cloudy days). liquid layer. A good agreement between experiment and simula-
The PRMSE and RSME testify the good agreement between the tion is, instead, always achieved when the PCM layer does not melt
simulated values and the measured too. It is noticeable that the completely (i.e. in cloudy days in summer and during the winter
PRMSE of the indoor surface temperature is, over the entire period, season).
around 7%. This value slightly increases when the winter season The analysis of the RMSE, ¯ and STD ¯ reveals once more
alone is analysed. However, it must be stated that this increase is
that a quite satisfactory agreement is anyway reached. The RMSE
mostly due to the fact that the average surface temperature during is about 20 W m−2 and the ¯ is about 6 W m−2 , with a range of
the winter is lower than in summer, and therefore the percentage values that can vary from 0 W m−2 up to almost 250 W m−2 . Once
error is inherently higher. The good agreement
in winter time is more, the best agreement is reached between the measured and
demonstrated by the fact that the ¯ is lower than the one dur- simulated values in winter time.
ing the summer. As previously mentioned, the closer the measured The analysis of both the total daily energy E24 h (cf. Fig. 12) and
datum to 0, the less reliable the PRMSE indicator. The ¯ along the disaggregated daily energies, E24 + −
and E24 (cf. Fig. 11), reveals
h h
the entire year is around 1 ◦ C, with a quite large range of values of that, in most of the case, the numerical tool is able to estimate the
surface temperature that vary from 6 ◦ C to more than 40 ◦ C (and to amount of energy that enters (E24+
) or leaves (E24−
) the indoor envi-
h h
more than 52 ◦ C, when the system is in free running mode). ronment through the PCM glazing, with a satisfactory agreement –
The comparison between the experimental and numerical sur- + −
quite often the calculated E24 h , E24 h
, and E24 h
are within the mea-
face heat fluxes is resumed in Fig. 9. The model is able to predict surement uncertainty range. However, it can be observed that in the
the exchanged heat fluxes under most of the situations with a rel- summer, in Day 3 and Day 4, the simulated value is quite far from
atively good reliability. The picture is similar to the prediction of the measured one. This can be explained considering the differ-
the hourly profiles of the surface temperatures, but a worse agree- ent value of the transmitted irradiance, as already explained above.
ment is revealed between simulations and measurements during Furthermore, it is necessary to state that the numerical model has
the discharge phase of the PCM. In sunny summer days, the model the tendency to over-estimate both positive and the negative daily
under-estimates the heat flux during the heat release process, energies, even if the values are very often in the range of the mea-
which mostly occurs in the late afternoon, evening and night, and sured ones.
over-estimates them during the charge phase. In winter, the heat
fluxes provided by the simulation tool appear to be quite similar
to the ones recorded during the experimental campaign, though 6. Conclusion
their values are systematically slightly lower. This phenomenon is
emphasized during the central hours of cold nights. The discrep- Advanced glazing systems are an important element of the
ancy between measurement and simulation may be caused by the construction to achieve the nZEB target. The adoption of PCM in
assumption done in the code (e.g. calculation of the convective heat transparent components has been considered a promising strat-
transfer coefficient, geometry of the room), and/or by the errors egy to improve the poor thermal inertia of conventional glazing
during the measurements of the heat fluxes – being the measure- systems. PCM glazing systems were originally developed for cold
ment of the heat flux in transparent/translucent components not climates, but they seem to be promising also for moderate and
straightforward. In fact, the sensor may suffer from the effect of warmer climates. However, to fully investigate their potentials
the solar radiation and then need to be shielded: the presence of and drawbacks, the availability of robust and reliable numerical
the shields can affect the actual readings of the sensor and provide tools is fundamental. Thanks to these tools it would be possible
lower or higher values [21]. to test different configurations of PCM glazing systems (e.g. with
The analysis of the RMSE, ¯ and STD ¯ shows the satisfac- various PCMs, layered-structures) under different operative con-
tory level of accuracy that is achieved, being the RMSE and ¯ ditions without the need of performing extensive and expensive
always one order of magnitude less that the maximum value of experimental analyses. It would also be possible to assess the effec-
the variable (the surface heat flux can vary from −60 W m−2 up to tiveness of this technology in different climates and for different
120 W m−2 ). building types (e.g. dwellings, office buildings).
F. Goia et al. / Energy and Buildings 54 (2012) 141–153 153