Energy and Buildings 94 (2015) 10–20

Contents lists available at ScienceDirect

Energy and Buildings

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

Fener: A Radiance-based modelling approach to assess the thermal

and daylighting performance of complex fenestration systems in
office spaces
Bruno Bueno a,∗ , Jan Wienold b , Angelina Katsifaraki a , Tilmann E. Kuhn a
a
Fraunhofer Institute for Solar Energy Systems ISE, Freiburg, Germany
b
Ecole Polytechnique Fdrale de Lausanne EPFL, Lausanne, Switzerland

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

Article history: A new Radiance-based modelling approach called Fener is presented. The motivation is to be able to
Received 11 December 2014 perform detailed analyses of complex fenestration systems (CFS) from the energetic and daylighting
Received in revised form 13 February 2015 points of view in a computationally efficient manner, so the benefits of innovative products can be easily
Accepted 14 February 2015
quantified. The model couples daylighting and thermal simulations in a time-step basis, so that shading
Available online 21 February 2015
control strategies that depend on thermal variables, such as indoor air temperature and energy load,
can be simulated without iterating between full-year simulations of a thermal model and a daylighting
Keywords:
model. Fener is a single-zone energy model that uses the three-phase method and bi-directional scattering
Daylighting
Complex fenestration systems distribution functions (BSDF) to predict the transmitted solar irradiance and indoor illuminance of office
Bi-directional scattering distribution spaces with CFS. An evaluation of the model is presented. Fener is tested against EnergyPlus and classic
function Radiance for different fenestration systems and sky conditions. Cooling and heating energy demand,
Three phase method transmitted solar irradiance and indoor illuminance are compared. As an exemplary application, Fener
Building simulation is used to assess the performance of an innovative perforated lamella system together with a control
Solar control strategy that depends on indoor air temperature.
EnergyPlus © 2015 Elsevier B.V. All rights reserved.
Radiance

1. Introduction In recent years, a number of modelling strategies have been pro-

Complex fenestration systems (CFS) refer to any light transmit-
ting window technology that incorporates at least one non-clear
(non-transparent) layer or one layer with swichable properties.
Examples include translucent insulating panels, shading devices,
such as venetian blinds and roller shutters, and electrochromic
glazing. CFS, combined with a suitable control strategy, have
the potential to improve the thermal and visual comfort of
indoor spaces as well as to save energy for lighting, cooling and
heating. Recent energy reduction requirements are forcing the
architectural-engineering community to integrate daylighting and
shading technologies in their designs. However, the prediction of
the light scattering properties of CFS remains an open research
field. There is still a need for the integration of advanced meth-
ods for thermal and daylighting simulation in order to facilitate the
development of new products and the adoption of commercially
available technologies.
∗ Corresponding author. Tel.: +49 76145885377.
E-mail address: bruno.bueno@ise.fraunhofer.de (B. Bueno).
posed in order to represent the light scattering properties of CFS.
Kuhn et al. [1] presented a “Black-Box” model to predict solar gains
through CFS in building simulation programs. The model requires
angularly resolved solar heat gain coefficients, which can be ana-
lytically derived or obtained through calorimetric measurements
[2]. The model then introduces the radiant and convective effect of
solar heat gains into the energy balance of the building through a
two-layer approach.
Other modelling initiatives keep the separation between the
thermal and optical problems of the fenestration system. While
the thermal problem can be solved by a layer-by-layer heat
transfer equation, the optical problem is addressed by using
bi-directional scattering distribution function (BSDF) data. Klems
[3,4] developed a calculation method to generate BSDF data of
multi-layered fenestration systems from the angularly resolved
data of single layers, which could be independently measured
or calculated. Measurements of BSDF data can be carried out
with a goniophotometer [5], although just a few of these devices
currently exist that can fully characterize a CFS. Alternatively, one
can generate BSDF data through ray-tracing. The Radiance-based
program genBSDF computes these datasets from the geometry of
macroscopic systems and surface properties of the base materials

http://dx.doi.org/10.1016/j.enbuild.2015.02.038
0378-7788/© 2015 Elsevier B.V. All rights reserved.
 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20
[7]. The Klems’ method was implemented in the Window 6
program (http://windows.lbl.gov/software/window/6/).
BSDF data can be used within Radiance [8] in order to simu-
late the daylighting performance of CFS. One way to do this is by
applying the three-phase method [9]. This approach, based on the
daylight coefficient method, separates the light transport between
the outdoor and the indoor environments into three phases: exte-
rior transport, fenestration transmission through a BSDF dataset
and interior transport. The three-phase method has some advan-
tages over classic Radiance and Daysim [11] for the evaluation of
CFS, especially for systems composed of small structures and for
specular or refracting shading surfaces. This is due to the fact that
the calculation of the light transmission through a fenestration sys-
tem is not restricted to the backward raytracing method but can be
calculated by other methods such as photogoniometry and Photon-
mapping. The separation of the light transport in three phases, as
opposed to Daysim’s one-phase, brings computational efficiency
when the fenestration system can switch among multiple states
during a dynamic simulation (e.g. a Venetian blind system).
BSDF data is also used in building energy calculations. The build-
ing simulation program, EnergyPlus [12], allows the user to provide
BSDF data as the optical representation of the fenestration system,
while keeping the layer-by-layer heat transfer definition. Energy-
Plus includes a daylighting engine, DElight, based on the radiosity
method [6]. DElight is suitable for simple building geometries with
CFS but does not support dynamic shading control. In addition,
specularly and light-redirecting surfaces cannot be handled accu-
rately by the radiosity method.
ComFen (http://windows.lbl.gov/software/comfen/comfen.html),
a widely available simulation platform to evaluate fenestration
systems in office spaces, uses EnergyPlus’s daylighting engine
for annual daylighting simulations. The DIVA plug-in for Rhino
(http://diva4rhino.com/) uses Daysim as the calculation engine to
obtain climate-based daylighting metrics. Similarly, OpenStudio
(https://www.openstudio.net/) offers an integrated interface, in
which the user can carry out a daylighting analysis based on
the three-phase method and pass a resulting artificial lighting
schedule to EnergyPlus. This kind of interfaces, which handles
different calculation engines from a single definition of the
building properties, can also be found in Adeline [10]. ESP-r
(http://www.esru.strath.ac.uk/Programs/ESP-r.htm) offers time-
step coupling capabilities with Radiance and Daysim, but the run
time is inefficient and does not support BSDF data. However, none
of the reviewed methods that couple daylighting and thermal
simulation models are able to efficiently assess control strategies
for CFS that depend on thermal variables.
The present study proposes a new modelling approach, called
Fener. The main contribution is to be able to couple daylighting
and thermal simulations in a time-step basis, so that shading con-
trol strategies that depend on thermal variables, such as indoor air
temperature and energy load, can be simulated without iterating
between full-year simulations of a thermal model and a daylighting
model. Fener is a single-zone “shoe-box” energy model that uses the
three-phase method to predict the solar transmission and indoor
illuminance of office spaces with CFS. It allows for detailed anal-
yses of facade systems from the energetic and daylighting points
of view. Fener can also be used as a shading controller simulator,
whose output in terms of shading operation can then be used in
standard building simulation programs (e.g. EnergyPlus).
In this paper, a description of Fener is presented. Special atten-
tion is paid to the calculation of solar transmission and illuminance.
Then, the model is evaluated against EnergyPlus and Radiance for
different seasons and fenestration systems for a weather dataset for
Frankfurt/Main (Germany). A case study evaluating a new lamella
system combined with a thermally dependent control strategy is
presented at the end.
11
2. Model description
2.1. Mission
The mission of Fener is to provide a simulation platform for
the evaluation of fenestration technologies applied to office spaces
independently of the complexity of the fenestration system. Fener
keeps a simple geometric definition, a rotatable rectangular shoe-
box space, while allowing a flexible definition of fenestration
systems and their control through the use of BSDF data. Although
building typologies could be extended in the future, the shoe-
box approximation is often used by the building performance
simulation community to showcase the impact of facade design
parameters [13]. Fener allows the user to define and control any
number and size of windows in each of the four facades and the
ceiling (including windows in more than one surface).
The computational cost of Fener is equivalent to the one
of Daysim, which is suitable for dynamic simulation, allow-
ing interactive comparisons of fenestration technologies. All the
time-consuming raytracing computation is carried out before the
dynamic simulation and can be saved (in matrix form) for other
simulations of similar scenarios.
Note that the goal of Fener is not to replace existing building
energy models. Therefore, the model is limited to one thermal zone,
no humidity calculation is included and infiltration/ventilation is
represented by a user-defined schedule of volume air changes in
the zone.
2.2. Three-phase method
The distinctive feature of Fener, as compared to other build-
ing simulation programs, is the integration of the Radiance-based
three-phase method for daylighting and solar transmission calcu-
lations.
The three-phase method separates the solar radiation from
the outdoor to the indoor environments into three phases: exte-
rior transport, transmission through the fenestration system and
interior transport. Each phase of energy transport is simulated inde-
pendently and stored in matrix form. The resultant irradiance (or
radiance, illuminance or luminance) at interior sensor points (I) is
obtained by the matrix multiplication I = VTDs, where V is a view
matrix that relates the interior sensor points with the inner BSDF
patches, T is a transmission matrix that relates the inner with the
outer BSDF patches, and D is a daylight matrix that relates the outer
BSDF patches with the sky patches. The final part of the equation
is a vector (s) that contains the average radiance of the sky patches
for a given time and sky condition. The combined VTD matrix can
therefore be understood as a daylight coefficient matrix that relates
outdoor radiance with indoor irradiance.
Fener uses the three-phase method to calculate the indoor hori-
zontal illuminance at a user-specified grid of sensors (at workplane
height), which is then used to analyse the daylight autonomy and
daylight distribution of the space. The resulting illuminance at a
control sensor is passed to a lighting control algorithm in order to
obtain the artificial lighting requirements. In addition, vertical illu-
minance can be calculated for the main viewing directions at the
workplaces in order to evaluate also the annual daylight glare [14].
The three-phase method is also used to calculate the transmit-
ted solar irradiance that is absorbed by indoor surfaces. A pair of
sensor grids (one facing towards the surface and another one facing
in the opposite direction) is placed 1 mm away from the surfaces.
The solar irradiance absorbed by a surface is the average of the irra-
diance difference between sensors looking at opposite directions.
The three-phase method has some advantages over classic Radi-
ance and Daysim for the evaluation of CFS. Systems such as small
venetian blinds or highly reflective systems require very time
12 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20
consuming Radiance parameters to be resolved or cannot be
resolved at all. The backward ray-tracing engine of Radiance also
fails for specular or refracting shading surfaces, as long as indi-
rect effects (e.g. reflection from the sun) are considered for the
illuminance calculation. In the three-phase method, all the com-
plexity of a fenestration layer is kept in its BSDF dataset, which can
be obtained experimentally. Ray tracing is used to calculate the
daylight and view matrices separately and independently from the
fenestration system. Hence, the computational cost is distributed
to the separate steps. A change in the shading system needs a
new BSDF for the CFS, but the exterior and interior ray-tracing
do not have to be repeated. Daysim needs one full ray-tracing
simulation for every shading position in order to obtain daylight-
ing coefficients. This constitutes an advantage of the three-phase
method for dynamic simulations with switchable shading systems.
The three-phase method also overcomes important limitations
encountered in EneryPlus’s embedded daylighting engine, DElight.
For example, the definition of surface properties is more flexible
and detailed through Radiance’s material inputs, and there are no
limits for the number and direction of sensor points allowing for
glare evaluations as explained above. Besides, EnergyPlus DElight
does not support dynamic shading control, which is one of the
motivations for the development of Fener.
2.3. Optical transmission
Fener uses the BSDF dataset of a fenestration system as the
transmission matrix in the three-phase method. It also uses the
angular-dependent absorption properties of each layer for the
thermal transmission calculation. The model requires BSDF spec-
ifications (transmittance and reflectance) for each layer of the
fenestration system, which are then combined into the BSDF data
(transmittance and absorptance) of the system by applying the
Klems method [3,4]. BSDF data of a transparent glazing layer can
be automatically generated from its optical properties at normal
incidence by applying the Roos model [15] (it requires adjusting
one parameter). BSDF data of complex fenestration layers, such as
venetian blinds, can be automatically generated from the Radiance-
based program genBSDF [7]. The model can also use BSDF data
generated by other means, such direct goniophotometer measure-
ments and calculations with the program Window.
The BSDF format imposes a number of assumptions that require
special attention. For example, it treats spatial inhomogeneities
of macroscopically structured samples as homogeneous. In this
approximation, the directly transmitted radiation from a venetian
blind would be a uniformly lit patch. An upgrade of the three-phase
method, the five-phase method, was developed to overcome this
problem for the direct component (redirected light is still homo-
geneous) at the cost of adding significant complexity to the matrix
multiplication. The BSDF format also collapses the fenestration sys-
tem into a infinitely thin layer, which introduces an error from the
effect of the window reveal. In order to reduce this error, thick sys-
tems (e.g. a window with venetian blinds) can be modelled in two
different positions: one position to calculate the daylight matrix,
corresponding to the outer layer of the real system, and another
position to calculate the view matrix, corresponding to the inner
layer.
The Klems method is also restricted by its assumptions [3].
For example, the method considers integrated spectral properties.
The condition for this assumption to be valid is that most layers
have spectrally flat optical properties, with at most one strongly
selective layer. If one of the layers is selective and the other layers
have an average transmission in the transmission region of the
selective layer that is different from its average transmission over
the full spectral region, the model fails. One solution for solar
transmission is to use the visible properties of the layers after a
spectrally selective solar control layer.
2.4. Thermal model
Fener uses a heat balance method to calculate indoor ther-
mal conditions and building energy demand. An energy balance
is applied to each indoor and outdoor surface, accounting for
conduction, convection, longwave and shortwave radiation heat
components.
Convection and longwave radiation heat fluxes (q) are calcu-
lated from a standard heat transfer coefficient formulation, q = hT,
where h is a heat transfer coefficient and T is a temperature dif-
ference. Indoor convective heat transfer coefficients are taken from
EnergyPlus documentation [16], depending on the relative position
between the surface and the indoor air. Outdoor convective heat
transfer coefficients are calculated as a function of the wind speed.
Radiative heat transfer coefficients are obtained from linearization
of the Stefan–Boltzmann equation, assuming only one bounce of
radiative heat fluxes between surfaces. Correlations for view fac-
tors between indoor surfaces are obtained from Mills [17] assuming
a “shoe-box” geometry. The outdoor radiative heat flux is calculated
from a fictive sky temperature obtained from the weather data and
a sky view factor (user defined).
The model solves the transient heat conduction through the
room enclosure by applying a finite differences method. Opaque
and transparent surfaces are numerically solved by treating lay-
ers as a simple RC network in matrix form. Temperature values
for the previous time step are used as initial values for the cur-
rent time step. The model calculates net surface heat fluxes at both
sides of the enclosure, which are then passed as Neumann boundary
conditions for the finite difference method. Similarly, the thermal
inertia of adiabatic internal surfaces is accounted for by solving the
same finite difference method but applying a zero-flux boundary
condition.
Angular-dependent absorption properties of the individual lay-
ers of the fenestration system are obtained by the Klems method
and used to calculate the solar radiation absorbed by these layers
for each time-step. Absorbed solar radiation is used as an energy
source in the RC network.
To calculate the dynamic evolution of indoor air temperature
between a cooling and a heating thermal setpoint, Fener solves a
sensible heat balance at the indoor air. The sensible heat balance
is composed of the convective heat fluxes from indoor surfaces,
the convective fraction of internal heat gains and the infiltra-
tion/ventilation sensible heat flux. Indoor and outdoor surface
energy balances are solved in Fener to provide boundary conditions
for the transient heat conduction equation.
The sensible building energy demand is calculated by applying
the same heat balances at the indoor air, but assuming that this is
at thermostat-setpoint conditions.
2.5. Shading control
One of the strengths of Fener is its flexibility for implementing
shading control algorithms. These can depend on daylighting vari-
ables, such as illuminance and glare, on thermal variables, such
as indoor air temperature and energy load, on weather variables,
such wind and solar radiation, on schedules and on combinations
of these and other variables. Cut-off control of flat venetian blinds
has also been introduced (see Section 4, Eq. (4)).
3. Model evaluation
In this section, Fener is evaluated against EnergyPlus and clas-
sic Radiance. The goal of this simulation-based evaluation is to
 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20 13

Table 1
Input parameters of the reference case study used to evaluate Fener against Ener-
gyPlus and Radiance.

Location Frankfurt/Rome
Latitude 50.10 ◦ /41.80 ◦
Longitude 8.68 ◦ /12.58 ◦
Orientation south/west
Zone length (interior) 3.6 m
Zone width (interior) 8.2 m
Zone height (interior) 2.8 m
Sky view factor 0.5
Ground albedo 0.2
Wall reflectance 0.5
Ceiling reflectance 0.7
Floor reflectance 0.2
Worktime schedule 6-18 LT
Infiltration/ventilation 0.5 ACH
Equipment heat gain 8.0 W m−2 (worktime)
Occupation 1 person (worktime)
People radiant fraction 0.5
Sensible activity level 63.72 W person−1
Lights installed power 10.1 W −2
Lights radiant fraction 0.72 Fig. 1. Reference office space with one south-facing window.
Illuminance setpoint 300 lux
Illuminance sensor height 0.8 m
Heating thermostat setpoint 20 ◦ C (worktime)
15 ◦ C (rest) the present study, infiltration and mechanical ventilation are com-
Cooling thermostat setpoint 26 ◦ C (worktime) bined in the same parameter, 0.5 ACH. Heat gains from people
30 ◦ C (rest) are calculated according to their sensible activity level (assuming
Opaque surface construction Exterior finish (1 cm) 63.72 W person−1 ). Internal gains due to equipment are 8.0 W m−2
 = 0.25 W m−1 K−1
Insulation (5 cm)
during worktime. The lighting is controlled by an on/off algorithm.
 = 0.03 W m−1 K−1 When the illuminance calculated at a sensor point located in the
Massive material (8 cm) middle of the room at work height is less than 300 lux due to day-
 = 2.3 W m−1 K−1 light, the artificial lights in the space are assumed to be on. The
c = 2.3e6 J m−3 K−1
installed lighting power is 10.1 W m−2 .
Interior finish (1 cm)
 = 0.25 W m−1 K−1 The office is located in Frankfurt/Main (Germany) (50.10◦ N,
Window dimenssions 3.0 m × 1.5 m 8.68◦ E). Weather conditions are obtained from Meteonorm
Window U-value 1.39 W m−2 K−1 database (http://meteonorm.com/).
Window SHGC 0.576
Window visible transmittance 0.744
Shading device Fabric  = 0.21 3.2. Comparison with EnergyPlus and Radiance
External grey venetian blinds (80 mm)
Simulations are carried out with Fener, Radiance and EnergyPlus
for three different systems: one composed of just the double-pane
test the code implementation by using models that have already glazing, the same fenestration system protected by an external diff-
been extensively used and evaluated within their respective com- usive fabric of transmittance 0.21 and reflectance 0.62 and the same
munities, building energy and daylighting. EnergyPlus is used to fenestration system protected by 80-mm grey external venetian
assess the cooling and heating energy demand and the transmit- blinds.
ted solar irradiance calculated by Fener, while Radiance is used In this analysis, EnergyPlus reads the BSDF dataset of the fen-
to test the calculation of transmitted solar irradiance and indoor estration system calculated by Fener. Radiance, on the other hand,
illuminance. does not use the BSDF dataset of the system but solves a geometri-
cal model of it by backwards ray tracing. For the glazing description
in Radiance, a BTDF-function is used in combination with the Roos
3.1. Model setup
model [15] in order to get similar angular transmission and reflec-
tion properties than the Window 6 model.
Input parameters for this case study are extracted from [13]
Fig. 2 compares hourly values of building energy demand calcu-
and summarized in Table 1. It consists of a single-zone office
lated by Fener with those calculated by EnergyPlus for one sunny
space with one south-facing external facade (3.6 m × 8.2 m × 2.8 m,
deeper in the north-south direction), which represents an office
located in an intermediate floor of a multi-storey building (Fig. 1). Table 2
Other surfaces in the space are assumed adiabatic. The window Thermal and optical properties of glazing panes.
is 3.0 m × 1.5 m composed of a double-pane insulating glazing Property Outer pane Inner pane
(5.7 mm × 12.7 mm × 6 mm), with low e-coating and argon filling,
Thickness (m) 0.0057 0.006
and with a U-value of 1.39 W m−2 K−1 , a visible transmittance of Solar transmittance 0.771 0.530
0.744 and a solar heat gain coefficient (SHGC) of 0.576 (Table 2). In Solar front (outer) reflectance 0.070 0.260
order to simplify the analysis, no window frame is considered. Solar back reflectance 0.070 0.179
The window shading device consists of grey external venetian Visible transmittance 0.884 0.839
Visible front reflectance 0.080 0.042
blinds. The lamellas are 80 mm width separated 72 mm among Visible back reflectance 0.080 0.057
each other (pure diffuse surface reflectance of 0.61). The office IR transmittance 0.000 0.000
is single-occupied during worktime. Thermostat setpoints for the Front emissivity 0.840 0.090
heating period are 20 ◦ C during worktime and 15 ◦ C otherwise. Back emissivity 0.840 0.837
Conductivity (W m−1 K−1 ) 1.000 1.000
They are 26 ◦ C and 30 ◦ C for the cooling period (DIN 18599). In
14 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20

Fig. 2. Comparison of hourly cooling and heating energy demand of the office space with glazing, glazing + shade and glazing + blinds calculated by EnergyPlus (EP) and by
Fener (FN) during a sunny summer day (23/07) and an overcast winter day (15/02) in Frankfurt. Cooling is indicated as negative energy demand.

Table 3 Fener and EnergyPlus may explain this difference. In summer,

Weather data information for 23/07 and 15/02, according to the weather file of
Frankfurt am Main (Germany) obtained from Meteonorm database.
VARIABLE 23/07
Min. air temperature (◦ C) 13.4
Max. air temperature (◦ C) 30.8
Average air temperature (◦ C) 22.4
Max. direct normal irradiance (W m−2 ) 938
Max. diffuse horizontal irradiance (W m−2 ) 100
day in summer (23/07) and one overcast day in winter (15/02).
Weather information for these two days is summarized in Table 3.
Lighting energy demand is calculated by Fener and passed as a
schedule to EnergyPlus. Cooling energy demand is represented as
negative energy demand.
It can be seen that Fener reproduces the diurnal cycle of heating
and cooling energy demand calculated by EnergyPlus. Statistical
results of the comparison for the full hourly series of February and
July are presented in Table 4. The results for two additional cases are
also presented: the same south-oriented office in Rome (weather
data from Meteonorm), and the same office in Frankfurt rotated
90 ◦ in order to have a west-oriented window.
The agreement between Fener and EnergyPlus is better in sum-
mer than in winter. The root-mean-square error (RMSE) is around
2 W m−2 in winter and around 1 W m−2 in summer. The differ-
ent heat transfer models (convection, conduction, etc.) used in
transmitted solar radiation dominates the cooling energy demand
making the relative difference between the two models smaller.
15/02 The RMSE is similar for the different fenestration systems,
−7.1 which is expected as both models are using the same BSDF
−1.3 datasets. The error is also similar for the three cases (Frank-
−3.7 furt/south, Rome/south, Frankfurt/west). Note that the average
0
energy demand (reference value) in February in Rome is negative
100
indicating a predominance of cooling over heating energy demand.
The mean-bias error (MBE) is small and has positive and negative
values both in winter and in summer, so no systematic difference
between the two models can be inferred.
The comparison of transmitted solar irradiance for the six
cases (sunny and overcast day, glazing, glazing+shade and glaz-
ing+blinds) is illustrated in Fig. 3. EnergyPlus, Fener and Radiance
are compared. In order to appreciate the differences for both
the cases with high solar transmission and with low solar trans-
mission, three different scales are used. The agreement of the
models for all cases is good. Table 5 shows that the RMSE
between Fener and Radiance is below 0.2 W m−2 . The error is
higher (1 W m−2 ) for the case with the highest transmitted solar
values, which is probably related with the resolution of the
Klems patches. The comparison of transmitted solar irradiance
demonstrates that the three-phase method is able to accurately
capture transmitted energy given accurate BSDF datasets are pro-
vided.

Table 4
Mean bias error (MBE) and root mean square error (RMSE) between Fener and EnergyPlus for non-zero hourly values of energy demand during one month. Cooling is indicated
as negative energy demand. The reference value (REF) is the average of the non-zero values calculated by Fener.

Location Orientation Case Month MBE RMSE REF

(W m−2 ) (W m−2 ) (W m−2 )

Frankfurt South Glazing February −0.4 2.2 2.2

Frankfurt South Glazing July −0.5 0.9 −20.5
Frankfurt South Glazing + shade February −0.6 2.3 6.1
Frankfurt South Glazing + shade July −0.5 0.9 −15.1
Frankfurt South Glazing + blinds February −0.1 2.2 5.8
Frankfurt South Glazing + blinds July 0.2 0.7 −16.8
Rome South Glazing February −0.6 1.0 −14.2
Rome South Glazing July −0.9 1.1 −29.4
Rome South Glazing + shade February 0.1 2.0 −1.9
Rome South Glazing + shade July −0.6 0.9 −25.7
Rome South Glazing + blinds February 0.9 1.9 −2.9
Rome South Glazing + blinds July 0.3 0.9 −27.2
Frankfurt West Glazing February 0.7 2.2 6.4
Frankfurt West Glazing July 1.1 1.4 −21.6
Frankfurt West Glazing + shade February −0.0 1.9 6.6
Frankfurt West Glazing + shade July 0.2 0.8 −15.6
Frankfurt West Glazing + blinds February 0.2 2.0 6.4
Frankfurt West Glazing + blinds July 0.4 1.2 −17.3
 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20 15

Fig. 3. Comparison of hourly transmitted solar irradiance of the office space with glazing, glazing + shade and glazing + blinds calculated by EnergyPlus (EP), by Fener (FN)
and by Radiance (RD) during a sunny summer day (23/07) and an overcast winter day (15/02) in Frankfurt. Three different scales for the same comparison are shown.

Fig. 4 compares average workplane illuminance calculated by In the daylighting literature, it is common to use the relative
Fener and Radiance for the six cases, plotted with three different RMSE, defined as:

scales. Scores of the comparison are shown in Table 6. The conclu-  2
sions are similar as those for transmitted solar irradiance, although 
1  xsim − xref
the agreement is slightly worse than before, especially for the case RMSE = 2
(1)
N xref
with blinds.
16 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20

Fig. 4. Comparison of hourly average workplane illuminance of the office space with glazing, glazing + shade and glazing + blinds calculated by Radiance (RD) and by Fener
(FN) during a sunny summer day (23/07) and an overcast winter day (15/02) in Frankfurt. Three different scales for the same comparison are shown.

where xsim and xref are the calculated and reference illuminance
values, respectively, and N is the total number of values. For this
case study, RMSErel ranges between 4 and 8% for the case with high
illuminance value (just glazing without shading device). The case
with shades in winter produces a RMSErel = 52%, but this is because
the absolute illuminance values are low. The case with blinds
produces RMSErel around 16%. This is lower than the expected
RMSErel reported for the program Daysim [18] (35%), considered the
state-of-art in dynamic daylighting simulations until the develop-
ment of the three-phase method.
The distribution of daylight in the office space is compared in
Fig. 5 at midday for the sunny day. Due to the resolution of the
Klems patches, the three-phase method scatters the light passing
through the fenestration system. This effect is more clearly appre-
ciated for the case without shading devices although some noise
is also observed for the other two cases. The graph reproduces
 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20 17

(a) (b) (c)

Fig. 5. Comparison of illuminance distribution in the office space with glazing (left), glazing + shade (middle) and glazing + blinds (right) calculated by Radiance (grey) and
by Fener (black) during a sunny summer day (23/07) at 12:00 in Frankfurt. Illuminance units: lux.

Table 5 the illuminance overprediction shown in Fig. 4 for the case with
Mean bias error (MBE) and root mean square error (RMSE) between Fener and Radi-
blinds.
ance for non-zero hourly values of transmitted solar irradiance during one day in
Frankfurt. The reference value (REF) is the average of the non-zero values calculated
by Fener for each variable.
4. Application: evaluation of a new lamella system and
Case Day MBE RMSE REF control strategy
(W m−2 ) (W m−2 ) (W m−2 )

Glazing February −0.2 0.2 1.8 In this section, Fener is applied to asses the performance of
Glazing July 0.2 1.0 12.6 a new blind system, called Winglamella, developed by Barten-
Glazing + shade February 0.1 0.2 0.4 bach GmbH in the context of the EU’s research project Inspire
Glazing + shade July −0.1 0.2 3.5
(http://www.inspirefp7.eu/).
Glazing + blinds February −0.0 0.0 0.4
Glazing + blinds July 0.0 0.1 2.6 The Winglamella system is made of highly reflective Aluminium
material (Fig. 6), which can be (i) perforated and covered with a
foil (0.16 transmittance) or (ii) non-perforated. These two variants
are put together to build a split-daylighting device composed of
two separately mechanically controlled partitions. The perforated

Table 6
Mean bias error (MBE), root mean square error (RMSE), relative MBE and relative
RMSE between Fener and Radiance for non-zero hourly values of average workplane
illuminance during one day in Frankfurt. The reference value (REF) is the average of
the non-zero values calculated by Fener for each variable.

Case Month MBE RMSE REF MBErel RMSErel

(lux) (lux) (lux) (%) (%)

Glazing February −12.4 13.3 279.3 −4 4

Glazing July 89.2 197.2 1945.4 −1 8
Glazing+shade February 17.9 19.4 52.3 52 52
Glazing+shade July 3.8 17.3 383.5 −3 6
Glazing+blinds February 6.0 6.5 42.7 16 16 Fig. 6. Image of a Winglamella composed of perforated aluminium covered with
Glazing+blinds July 31.2 38.0 260.3 15 17 Haverkamp foil.
Source: http://www.inspirefp7.eu/
18 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20

Table 7
Algorithms used to calculate the Winglamella cut-off and Retro(60 ◦ ) tilt angles of
the lamellas as a function of the profile angle.

Profile angle (◦ ) Tilt angle (◦ ) Tilt angle (◦ )

Retro(60 ◦ ) Cut-off

0 ≤ ε < 1.5 90 90
1.5 ≤ ε < 6.9 93 − 2 · ε 93 − 2 · ε
6.9 ≤ ε < 30 82.62 − 2ε 93 − 2 · ε
30 ≤ ε < 44 82.62 − 2ε 9.2 · 10−7 · ε6 − 0.0177832 · ε4
+1.69134 · ε3 − 67.5536 · ε2
+1270.62 · ε − 9219.94
ε
44 ≤ ε ≤ 90 82.62 − 2
0

lamellas are used in the lower part of the window (0.88-m height),
while the non-perforated is used in the upper part (0.8-m height).
When the lower part is closed, the perforation of the lamella still Fig. 7. Calculated annual lighting, heating and cooling consumption of the office
space located in Villafranca di Verona, Italy. The three pairs of bars represent three
permits visual contact to the exterior. different orientations of the fenestration system: south (left), west (middle) and east
The Winglamella system is compared with a conventional sys- (right). For each pair of bars, the one on the left corresponds to the Warema system
tem composed of white aluminium Warema venetian blinds cut-off cut-off controlled and the one on the right to the Winglamella system controlled
controlled (pure diffuse surface reflectance of 0.77, 35 mm width according to Table 6.
and 30 mm spaced). The geometry and operation of the test room is
the same as the office space defined in the previous section. In this to Table 8 and the Warema system controlled with the following
case study, the artificial lighting is dimming-controlled by two sen- cut-off algorithm:
sors, one located in the first half of the room next to the window and  
the other one in the other half. The fenestration system is a double- cos(˛p )
ˇcutOff = arcsin d − ˛p , (2)
skin window composed of the following layers (from outside to w
inside): a single-pane glass of 7.8-mm width, an air gap of 220 mm
where ˇcutOff is the cut-off angle, d is the distance between slats, w
that locates the shading system and a double-pane glass with low-e
is the slat width (assuming flat slats) and ˛p is the profile angle of
coating (7.8 × 16 × 8.5 mm). In this analysis, the in-between air gap
the sun on the window, defined as:
is assumed not to be ventilated. The total U-value of the fenestration  
system is 0.98 W m−2 K. The window-to-facade ratio is 56%. The cli- tan(˛sun )
mate information for this case study is extracted from Meteonorm ˛p = arctan , (3)
cos(sun − surf )
database for Villafranca di Verona, Italy. This is the location where
the new Winglamella system will be installed within the context where ˛sun is the solar altitude and  sun and  surf are the solar and
of the Inspire project. surface azimuth, respectively.
The control strategy designed by Bartenbach GmbH for the For this analysis, cooling consumption is calculated from the
Winglamella system takes into account the occupation schedule of cooling energy demand applying a constant coefficient of opera-
the office space, the daylight availability, the indoor air temperature tion COP = 2.5. Heating energy consumption is obtained from the
and the direct radiation on the facade. According to these variables, heating energy demand with an efficiency of 0.9. The simulation
the tilt angle of the blinds in the lower partition changes between is carried out for three different orientations of the fenestration
three states: open (0 ◦ ), intermediate (45 ◦ ) and closed (90 ◦ ). The system: south, east and west.
tilt angle of the upper partition’s blinds can vary between open and For all orientations, the two configurations present similar heat-
close positions in 5 ◦ -angle steps according two different shading ing consumption. The artificial lighting consumption is 9% lower
control strategies: Cut-off and Retro (60 ◦ ), described in Table 7. The for the south-oriented Winglamella case and 39% lower for the
possible combinations of variables are summarized in Table 8. west- and east-oriented Winglamella cases. The cooling energy
Fig. 7 compares the annual heating, cooling and lighting con- consumption of the Winglamella is significantly lower compared
sumption between the Winglamella system controlled according to the Warema system (37% for south and around 28% for east and
west).
In order to better understand the aggregated results, two
Table 8 monthly average days for July were generated for the south and the
Control state of the two partitions of the Winglamella system for a combination west orientations (Fig. 8). Solar transmission, cooling and artificial
of variables. Variables refer to presence/absence of occupants, daytime/nighttime,
lighting consumption are plotted together.
indoor air temperature below/above 23 ◦ , presence/absence of direct solar radiation
on the facade. During summer, the solar transmission of the Winglamella sys-
tem is lower and more evenly distributed during the day compared
Case Occupancy Daytime Air temp < Dir rad on Upper Lower
to the Warema system, which results in a lower cooling consump-
23 ◦ facade partition partition
tion. This can be explained by the following reasons:
1 0 0 0 0 Open Open
2 0 0 1 0 Closed Closed
3 0 1 0 0 Closed Closed
• At closed position, the Winglamella system is able to block the
4 0 1 0 1 Closed Closed solar radiation better than the conventional system, despite the
5 0 1 1 0 Open Open perforations. Conventional lamella systems cannot close more
6 0 1 1 1 Open Open than 55–70 ◦ . At maximum tilt angles of 65 ◦ (as in this case), the
7 1 0 0 0 Open Closed
8 1 0 1 0 Closed Closed
sun’s beam is inevitably hitting the backside of the neighbouring
9 1 1 0 0 Cut-off Closed above lying lamella, which results in increased transmitted
10 1 1 0 1 Retro(60 ◦ ) Closed energy. This is particularly relevant for east and west facade ori-
11 1 1 1 0 Open Intermed entations in which the blind system has to be closed in order to
12 1 1 1 1 Cut-off Closed
avoid direct sun penetration.
 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20
(a)
(b)
Fig. 8. Calculated monthly averaged diurnal cycle of cooling consumption, solar
transmission and lighting consumption in July for the Winglamella system con-
trolled according to Table 6 and the Warema system cut-off controlled. Two
orientations are shown: south (top) and west (bottom).
(a)
Fig. 9. Calculated spatial daylight autonomy of the office space for the Winglamella system controlled according to Table 6 (grey) and the Warema system cut-off controlled
(black). Three orientations are shown: south (left), west (middle) and east (right).
19
• When direct radiation is present, the Retro(60 ◦ ) control algo-
rithm is applied to upper part of the Winglamella system, while
the lower part is closed. Despite being tilted to a vertical position,
the perforated blinds allow a more even transmission of day-
light in the space. This fact is evident in the morning hours when
the Warema system presents a higher lighting consumption. The
Winglamella’s control algorithm also maintains the blinds closed
during night and non-occupied hours whereas the Warema sys-
tem maintains the cut-off control at all times when there is direct
radiation on the window and sets the blinds open otherwise inde-
pendently of the season.
During winter, the Winglamella’s control algorithms allows
higher and more evenly distributed solar heat gains throughout
the daylight hours, a fact that reflects in the lower lighting demand
especially during the morning.
Fig. 9 compares the spatial daylight autonomy of the space for
the three orientations. The Winglamella system achieves a deeper
daylight penetration in the space for all three orientations, although
the differences are particularly important for the east and west ori-
entations. This result is coherent with Fig. 7, in which the artificial
lighting consumption of the east- and west-oriented Winglamella
system is significantly lower than the one of the Warema system.
Final remark: the current analysis does not include glare con-
siderations, which should be checked before finally recommending
the new system.
5. Conclusions
A Radiance-based modelling approach for the analysis and con-
trol of CFS in office spaces has been presented. The new model,
called Fener, couples detailed daylighting and thermal simulations
in a time-step basis, which makes it easy to implement innovative
shading control algorithms. The use of BSDF datasets to represent
(b) (c)
20 B. Bueno et al. / Energy and Buildings 94 (2015) 10–20
the light scattering properties of CFS permits the evaluation of a
broad variety of systems.
Fener has been satisfactorily evaluated against EnergyPlus and
classic Radiance. It has been proven that Fener is able to reproduce
the diurnal cycle of heating and cooling energy demand calculated
by EnergyPlus (for the considered case studies). The comparison
also demonstrates that the three-phase method used in Fener is
able to accurately capture transmitted energy given good BSDF
datasets are provided. The results show a slight overprediction of
indoor illuminance calculated by Fener as compared to classic Radi-
ance. This difference is more pronounced for a fenestration system
with blinds than for a system with a diffusive fabric or just glazing.
Nevertheless, the accuracy of the dynamic daylighting simulation
lays within what is considered the state-of-the-art, which is accept-
able given the scope of the model.
An example of application is presented to illustrate the fact
that the new model can be used to assess the energy and day-
lighting performance of innovative shading systems and control
algorithms. A new perforated aluminium lamella system, combined
with a sophisticated control algorithm that depends on thermal
variables, is compared with a conventional blind system. The new
system achieves a deeper daylight penetration as well as significant
reductions in cooling and lighting consumption as applied to an
office space located in Villafranca di Verona, Italy. The new system
is particularly suitable for east and west facade orientations, reduc-
ing the transmitted solar radiation more than a conventional blind
while allowing for view contact with the outdoor environment. This
is partially the result of a counterplay between the perforation of
the new system and the limited tilting angles of the conventional
system.
The possibility to easily test shading control algorithms that
depend on radiation and thermal variables is a unique feature of
the new model.
Acknowledgements
The present study has been partially funded by the Federal Min-
istry of Economic Affairs and Energy by a resolution of the German
parliament. This study has also been partially supported by the
iNSPiRe project. The iNSPiRe project has received funding from
the European Union’s Seventh Programme for research, technolog-
ical development and demonstration under GA no. 314461. The
European Union is not liable for any use that may be made of the
information contained in this document which is merely repre-
senting the authors view. The authors acknowledge the work of
Bartenbach GmbH, who designed the Winglamella system and its
control and provided Fraunhofer ISE with BSDF datasets of their
lamella system. The authors thank also Elena Guidolin for her con-
tribution to this work.
References
[1] T.E. Kuhn, S. Herkel, F. Frontini, P. Strachan, G. Kokogiannakis, Solar control:
a general method for modelling of solar gains through complex facades in
building simulation programs, Energy Build. 43 (1) (2011) 19–27.
[2] T.E. Kuhn, Calorimetric determination of the solar heat gain coefficient g with
steady-state laboratory measurements, Energy Build. 84 (0) (2014) 388–402.
[3] J. Klems, A new method for predicting the solar heat gain of complex fenes-
tration systems. I Overview and derivation of the matrix layer calculation, Am.
Soc. Heat. Refrig. Air Cond. Eng. Trans. 100 (1) (1994) 1065–1072.
[4] J. Klems, A new method for predicting the solar heat gain of complex fenestra-
tion systems. II Detailed description of the matrix layer calculation, Am. Soc.
Heat. Refrig. Air Cond. Eng. Trans. 100 (1) (1994) 1073–1086.
[5] M. Andersen, J. de Boer, Goniophotometry and assessment of bidirectional
photometric properties of complex fenestration systems, Energy Build. 38 (7)
(2006) 836–848.
[6] F.C. Winkelmann, S. Selkowitz, Daylighting simulation in the DOE-2 building
energy analysis program, Energy Build. 8 (1985) 271–286.
[7] A. McNeil, C. Jonsson, D. Appelfeld, G. Ward, E. Lee, A validation of a ray-
tracing tool used to generate bi-directional scattering distribution functions
for complex fenestration systems, Sol. Energy 98 Part C (0) (2013) 404–414.
[8] G. Ward, R. Shakespeare, Rendering with Radiance: The Art and Science of
Lighting Visualization, Morgan Kaufmann, San Francisco, 1998.
[9] G. Ward, R. Mistrick, E. Lee, A. McNeil, J. Jonsson, Simulating the daylight
performance of complex fenestration systems using bidirectional scattering
distribution functions within radiance, J. Illum. Eng. Soc. N. Am. 7 (4) (2011)
1–19.
[10] H. Erhorn, J. de Boer, M. Dirksmller, ADELINE – an integrated approach to light-
ing simulation, in: Proceedings of the Fifth International IBPSA Conference,
Prag, Czech Republic, 1997.
[11] C. Reinhart, O. Walkenhorst, Dynamic RADIANCE-based daylight simulations
for a full-scale test office with outer venetian blinds, Energy Build. 33 (7) (2001)
683–697.
[12] D.B. Crawley, L.K. Lawrie, F.C. Winkelmann, W. Buhl, Y. Huang, C.O. Pedersen,
R.K. Strand, R.J. Liesen, D.E. Fisher, M.J. Witte, J. Glazer, Energyplus: creating
a new-generation building energy simulation program, Energy Build. 33 (4)
(2001) 319–331.
[13] C. Reinhart, J. Jakubiec, D. Ibarra, Definition of a reference office for standardized
evaluations of dynamic faade and lighting technologies, in: Proceedings of the
13th Conference of International Building Performance Simulation Association,
August 26–28, Chambry, France, 2013.
[14] J. Wienold, Dynamic daylight glare evaluation, in: Proceedings of the Eleventh
International IBPSA Conference, Glasgow, Scotland, 2009, July, pp. 27–30.
[15] A. Roos, P. Polato, P.A. van Nijnatten, M.G. Hutchins, F. Olive, C. Anderson,
Angular-dependent optical properties of low-e and solar control windows-:
simulations versus measurements, Sol. Energy 69 (Suppl. 6(0)) (2001) 15–26.
[16] DOE, EnergyPlus Engineering Reference, EnergyPlus, 2010.
[17] A. Mills, Heat and Mass Transfer, Irwin Heat Transfer Series, Taylor & Francis,
1995.
[18] C.F. Reinhart, P.F. Breton, Experimental of autodesk 3ds max design 2009 and
daysim 3.0., Leukos 6 (1) (2009).