Energy and Buildings 40 (2008) 514–523

www.elsevier.com/locate/enbuild

Lighting energy savings in offices using different control systems

and their real consumption
B. Roisin a,*, M. Bodart b, A. Deneyer c, P. D’Herdt c
a
Université Catholique de Louvain (UCL), Unité d’Architecture, Place du Levant, 1, B-1348 Louvain-La-Neuve, Belgium
b
Fond National de la Recherche Scientifique (FNRS), Université Catholique de Louvain (UCL), Unité d’Architecture,
Place du Levant, 1, B-1348 Louvain-La-Neuve, Belgium
c
Belgian Building Research Institute (BBRI), Division of Energy and Climate, Avenue Pierre Holoffe, 21, 1342 Limelette, Belgium
Received 19 January 2007; received in revised form 22 March 2007; accepted 13 April 2007

Abstract
This paper compares the potential of lighting energy savings in office rooms by using different control systems, for three locations in Europe and
the four main orientations. The method is based on DAYSIM simulations to perform daylight calculations, on laboratory measurement to evaluate
precise system energy consumptions and on the implementation of a new algorithm to simulate a close-loop daylight dimming system. It appears
that the control of the electrical power in function of daylight leads to very high savings; they slightly depend on the room orientation and the
location. Savings vary from 45 to 61%. The performances of an occupancy sensor are also tested. Threshold values of occupancy rate for which
daylight dimming leads to higher gains than an occupancy control system vary between 27 and 44% depending on location and orientation. The
measurements of the energy consumption of the sensors and detectors also permit to conclude that systems with embedded DALI-compatible
ballast controllers should be abandoned in favour of a centralized DALI-compatible ballast controller or embedded analogue systems.
# 2007 Elsevier B.V. All rights reserved.

Keywords: Artificial lighting; Energy savings; Lighting control; Dimming; Daylight

1. Introduction with manual switching. Knight measured gains from 44 to 76%

using daylight control systems [2]. Galasiu et al. have evaluated
Recently, many different lighting control systems have been the impacts of window blinds control on two different daylight
developed. One of their first application field is the tertiary control systems [3]. The savings using an open loop dimming
sector and, in particular, office buildings. Their aims are to control system and different controlled shading systems vary
reduce the lighting energy consumption while maintaining a from 5 to 45%. In a Norwegian study, the saving potential due
high level of lighting comfort. While authors agree on the to daylight dimming systems is about 30–40% for south-facing
positive impact of these systems, there is a disagreement rooms and 20–30% for north-facing [4]. Li et al. evaluated the
in quantifying their saving potential. This quantification can performances of daylight dimming and on/off control system
be based either on monitoring in existing buildings or by [5]. They concluded that the daylight dimming system reduces
simulations. the energy consumption by 33%. The correlation curve between
Measured lighting energy savings show a very large range of electric lighting load and vertical daylight illuminance which
variation. Jennings et al. have tried to identify the impact of can be useful to determine energy savings is found in [5,6]. A
different control systems on office lighting consumption [1]. By survey in Turkey evaluates the gains obtained with a daylight
measuring the consumption of retrofitted installations, they dimming system according to the type of sky. For clear days the
concluded that an occupancy sensor can save up to 20% and a gains are 35%, 33% for mixed days and 16% for overcast days,
daylight dimming control system up to 26% by comparison with a mean of 31% for the whole year [7].
Simulations can be very useful, especially during the design
phase of a building. The simulation work can be divided into
* Corresponding author. Tel.: +32 10 47 91 52; fax: +32 10 47 21 50. two parts. First, the evaluation of dynamic daylight availability
E-mail address: [email protected] (B. Roisin). of the room, in order to be able to calculate the needed
0378-7788/$ – see front matter # 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.enbuild.2007.04.006
 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523 515

complementary electric light, for the whole year, according to energy consumption and that a daylight control system can save
the chosen time step. Secondly, the implementation of lighting up to 60%.
control systems.
2. Objectives and methodology
1.1. Evaluation of the daylight availability
The main objective of this study was to predict, by
It can be very time consuming to calculate the daylight simulation, the energy consumptions of lighting in offices
availability in a room, for a whole year. A solution to this according to different control systems. Therefore, we used
problem is to use the daylight coefficient method (DC) the simulation program DAYSIM to calculate the daylight
developed by Trengenza and Waters [8], like proposed by availability in an office room during the whole year [10].
Mardaljevic [9]. He used the software RADIANCE and found it Although DAYSIM is able to model several lighting control
accurate and faster than traditional methods. His conclusion systems, we did not use it for that part of the calculation. We
was confirmed by Reinhart and Herkel who implemented the wanted to test an individual daylight dimming control system
DC method, using RADIANCE, in a software called DAYSIM that could not be modelled in DAYSIM at that moment. We
and compared it to five other dynamic methods based on developed thus an algorithm for simulating individual daylight
RADIANCE daylight simulations [10,11]. The software dimming control systems according to a close-loop system. We
DAYSIM was validated by Reinhart and Walkenhorst by calculated then the necessary complementary artificial lighting
comparing results with measurements. The conclusion is that by implementing real consumption values of ballasts and
indoor illuminances can be modelled with comparable accuracy control systems we measured in laboratory.
for various blind settings under arbitrary sky conditions [12]. The second objective of this study was to measure the digital
controller energy consumption and to compare it with simple
1.2. Implementation of lighting control systems analog system. Indeed, nowadays, the digital addressable
lighting interface (DALI) is spreading. While its main
Some researcher have focused on the implementation of advantages are the flexibility, the independence between the
lighting control systems but made no link with the daylight luminaire control system and the electric wiring and the
modelling. Littlefair presents algorithms to quantify the savings possibility of pre-programming lighting scenes, some manu-
for various types of photoelectric and manual controls facturers have the tendency to provide one controller per
[13]. Ehrlich et al. focused on the accurate simulation of luminaire, in order to have the possibility of a standalone use
photosensor-based lighting controls in order to improve the and to facilitate the luminaire installation [23]. However, this is
comparison between such systems, their selection, placement not essential and it could induce (useless) extra energy
and commissioning [14]. Choi et al. developed a detailed consumption. With the measurements of the consumption of the
computer analysis model in order to investigate the perfor- auxiliaries, we can determine the minimum number of
mance of daylight responsive dimming systems [15]. luminaires that should be linked with one controller for the
same energy consumption as an analog system.
1.3. Link between daylight availability and lighting control This paper compares three different lighting control systems
systems installed in a single office, taking into account the real energy
consumption of auxiliary systems (electronic ballast and
Some authors worked on the link between daylight management system), the daylight availability over the year,
simulations and algorithms modelling lighting control systems. the orientation of the room and its location, in Europe. The
Li and Tsang did simulations based on a DC method using lighting controls are an individual daylight dimming system
RADIANCE and implemented a unique closed loop sensor. and an occupancy control system which can either switch off or
They compared results with measurements in a corridor and dim the light. The impact of each of these systems was
concluded that results of their method are in good agreement calculated by comparison with a simple scheduled automatic
with measurements, but they did not give any quantitative switch off system. The combination of these systems was also
energy lighting savings [16]. Clarke and Janak developed a evaluated.
method based on the conflation of the ESP-r and RADIANCE The energy consumption of the lighting systems (including
systems [17]. They concluded that optimized daylighting ballast, tube and control device) was accurately measured in
control system could save between 40 and 70% of the energy laboratory. The equipment, the methodology and the results are
consumption of artificial lighting [18]. Beside the simulations described in the first part of this paper.
of gains using daylight control systems, Mahdavi et al. focused The second part of this paper describes how accurate
on the implementation of control strategies to achieve some dynamic daylight simulations, obtained from real climatic
objective functions taking into account visual comfort and data’s and taking into account the position of the blinds, were
energetic considerations [19,20,21]. Reinhart proposed the used to predict the daylight penetration and the complementary
modelling of several lighting control systems (program electric lighting necessary to reach the set point illuminance
lightswitch) and implemented those algorithms in the DAYSIM level.
software [22]. By simulations, he concluded that a switch-off Additionally, the influence of the relative room occupancy
occupancy control system can save up to 20% of lighting (in function of the working hours) on the savings was analysed.
516 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523

3. Equipment individual daylight dimming system with a movement

detection switching and with a movement detection dimming.
The measured equipment consists of an ETAP luminaire In these cases, the lights are dimmed according to the daylight
R2600/158 Isolum (608) equipped with tubes Philips Master availability when there is someone in the room and are either
TLD super 840 58W [24]. Two different electronic dimmable switched off or dimmed to 3% when there is nobody in the
ballasts were chosen: the first having an analogue (1–10 VDC) room.
command (Philips HF-R 158 TLD) and the other having a
digital (DALI) command (Philips HF-R DALI 158 TLD). In the 4. Methodology
later case (DALI ballast), a controller is required to transfer the
information from the detectors to the ballasts. The methodology developed in this work combines
The photometric curve of the luminaire is shown in Fig. 1. electrical and photometric measurements and computer
The different control systems considered here are: simulations. The first step was to analyse the system
‘‘ballast-tube-detector’’ in order to evaluate the real energy
 IDDS: Individual Daylight Dimming System. The lamp light consumption of the different combinations.
flux is controlled according to the daylight availability. The
sensors (one per luminaire working in close-loop) are fixed on 4.1. Electric and photometric measurements
the luminaires and measure the reflected illuminance of the
plane located under them (this product is known as ELS- 4.1.1. Total power measurement on an electronic ballast
ETAP Lighting System). This measurement was simply done by connecting the
 MDS: Movement Detection Switching. This system, based wattmeter to the ballast. By varying (using a voltage source)
on an infrared occupancy sensor, switches the light on and the control signal (1–10 VDC) of the analog ballast we obtained
off, according to movement detection. The length of the delay the power of the whole system for different output fluxes. For
can be chosen in order to limit the number of switch on and the digital ballast, we used the controller to control the output
switch off cycles. flux of the ballast. By modifying a set of dip switches on the
 MDD: Movement Detection Dimming. This system, as the controller the flux of the lamp could be controlled. But with this
MDS, is based on an infrared occupancy sensor, but dims the method, only five different output fluxes can be measured.
light to a chosen flux in case of absence. This flux can be Having measurements for more output fluxes would have
chosen by a set of dip switches located on the sensor. In our required sending a digital signal (DALI protocol) to the ballast.
case, we choose the minimum output flux (3% of the nominal
flux) to maximize the gains. 4.1.2. Detector power
The power of the detectors was also measured. The IDDS
As already mentioned in Section 2, these control systems sensors power is negligible whereas the presence detectors have
can be combined and we tested the combination of an a constant power of 0.5 W. For the DALI-compatible systems,
the power of the controller has to be added. That leads to a
power of 2–2.5 W depending on the state of the luminaire (On
or Standby position).

4.1.3. Relative photometric measurement

The luminous flux of the lamp was measured in order to
compare it with the manufacturer data’s. A photosensor (TAOS
TSL250R) was clipped at a random place of the tube and
pointed to it. We obtained the luminous flux in dimming mode
compared with full flux. A calibration in an Ulbricht sphere
allowed us to calculate the absolute nominal luminous flux.
Fig. 2 shows the relationship between the source electrical
power and the relative lamp luminous flux, for the chosen
ballasts.
Fig. 2 clearly shows a linear relationship between the source
electrical power and the lamp relative flux. The linear
regression line, which is quite the same for the two types of
ballasts, has the following expression:

9:02 ðWÞ Analog
P ¼ 0:46Frelat:% þ (1)
9:34 ðWÞ DALI-compatible

The constant term of this expression shows that the ballast

Fig. 1. ETAP R2600/158 Isolum (608) photometric curve. consumption is not nil even with the tube in turn-off state. Note
 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523 517

west and east) and placed in three different locations (Brussels—

50.98N; 4.538E, Athens—37.98N; 23.738E and Stockholm—
59.658N; 17.958E) in order to test the influence of the latitude on
the light energy savings possibilities. The location of Brussels
was of first importance for us and the two other locations were
chosen in order to test two extreme locations in Europe.

4.2.1. Daylight
Precise dynamic daylight simulations were made using
the software DAYSIM in order to calculate the daylight
illuminance for each daylight sensor location in the room, every
5 min, over the whole year. DAYSIM, which uses the
RADIANCE algorithm, uses hourly climatic data files in order
to calculate the illuminance according to a precise sky
modelling taking into account the sun position and the real
sky distribution [12]. Moreover, this software includes the
possibility to consider the use of shadings. Two modes to
simulate shading devices such as blinds are proposed by the
software: the simple dynamic device model or the advanced
dynamic device model. We used the simple model who models
a generic blind system that transmits 25% of diffuse daylight
and no direct sunlight compared to the case when the blinds are
retracted. This is a basic blind model that is sufficient for initial
design consideration when the type of shading control device is
still unspecified. According to this model, the blinds are
lowered when the workplane irradiance is over 50 W/m2 [22].
This value was chosen following the results of a monitoring
conducted by Reinhart and Voss [25].
Fig. 2. Electrical power as a function of relative luminous flux: (a) analog
ballast; (b) DALI-compatible ballast. 4.2.2. Artificial light
The number of luminaires and their position was calculated
that we can get the overall electrical power of the system by with DIALux. The best solution was to place four luminaires in
adding the detector power to these values (0.5 W for an analog two rows as shown in Fig. 4.
ballast, 2–2.5 W for a digital (DALI-compatible) ballast). With this configuration, the average artificial illuminance is
equal to 615 lx and the uniformity on the working plane is equal
4.2. Simulation method to 0.73. The lighting system is not too oversized, which is
important in order to compare the savings with realistic values.
The objective of this work was to evaluate the lighting
energy consumption of a typical office in different situations. 4.2.3. IDDS modelling
We create thus a theoretical office (width 3.05 m, length 6.55 m The IDDS system regulates the lamp flux in function of the
and height 3.05 m (see Fig. 3)). A window of 3.05 m  1.01 m daylight availability. At each time step t, as it works in close-
is located in one of the surfaces of the room and its sill is placed loop, the simulation of its comportment requires an iterative
at 1.01 m above the ground. This window is fitted with a double process. The goal of this modelling is to dim all the lamps
glazing of 77% visible transmittance. The room was arbitrarily adequately to obtain the set-point illuminance (500 lx) under
oriented according to the four main orientations (north, south, each sensor.
Under each sensor i, we can say that the sum of the artificial
light and the daylight must be equal to the set-point:
X
n
Edl;i ðtÞ þ t j ðtÞE j;i ¼ SP (2)
j¼1

With t is the time step under consideration, n the number of

luminaires and thus of sensors (four in our case), Edl,i(t) the
daylight illuminance at time t under the sensor associated to the
luminaire i (calculated with DAYSIM), tj(t) the dimming rate of
luminaire j at time t (the dimming rate of a luminaire is the flux
Fig. 3. Representation of the office. given by the luminaire at time t divided by the nominal flux of
518 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523

Fig. 4. Position of the luminaire and light distribution on the workplan.

this luminaire), Ej,i the illuminance due to the luminaire j at full the difference (1  (ti/ti1)) was less than 106. Usually, this
flux under the sensor associated to the luminaire i (this value required 10–15 iterations.
was calculated with DIALux), and SP is the set-point illumi-
nance. 4.2.4. Occupancy modelling (MDS/MDD)
The artificial light can be separated in two: the contribution To model occupancy, a 5 min time step occupancy profile
of the luminaire associated to the sensor and the contributions was generated automatically.
of the others luminaires. We can thus rewrite Eq. (2), assuming The dimming rate of each luminaire, taking into account the
di,j a term to make this separation: occupancy, is calculated by Eq. (6):
X
n
Edl;i ðtÞ þ di; j t j ðtÞE j;i þ ti ðtÞEi;i ¼ SP with di; j t i ðtÞ ¼ PðtÞton;i ðtÞ þ ð1  PðtÞÞtoff;i ðtÞ (6)
j¼1
With P(t) equal to 1 in case of presence and to 0 in case of
¼ 1 if i 6¼ j; di; j ¼ 0 if i ¼ j (3) absence, ton,i(t) the dimming rate of the luminaire i when there
We can now isolate the dimming rate of the luminaire i: is somebody in the room (1 for MDD/MDS, the value calcu-
P lated by Eq. (5) for IDDS + MDD or IDDS + MDS) and toff,i(t),
SP  Edl;i ðtÞ  nj¼1 di; j t j ðtÞE j;i the dimming rate of the luminaire i when there is nobody in the
t i ðtÞ ¼ (4)
Ei;i room (0 for MDS, 0.03 for MDD).
The delay between the departure of the room occupant and
We see in Eq. (4) that the dimming rate of the luminaire i the light switch off or dimming was set to 10 min.
depends on the dimming rate of the others luminaires. Thus to Beside the three tested control systems, we considered that
calculate it, we need the following iterative process. If we note k the lights are managed by a scheduled automatic shut off
the number of iterations, for each time step t, we have ti,k the system that switches off all lamps and sensors from 6 p.m. to 8
rate ti of the kth iteration: a.m. The positive impact of this scheduled shut off system is
P
SP  Edl;i ðtÞ  nj¼1 di; j t j;k1 ðtÞE j;i easily understandable for IDDS and MDD. It has also a positive
t i;k ðtÞ ¼ (5) impact on energy saving with the MDS control system as
Ei;i
it switches off the light sensors preventing their parasitic
At each time step t, the iteration requires an initial value of consumption during the night.
tj,0(t) chosen arbitrary between 0 and 1. In order to speed up the The results presented at Section 5.1 are based on presence in
process, the calculation takes the dimming rate of the precedent the room from 8 a.m. to 12 a.m. and from 1 p.m. to 6 p.m. In
time step as initial value. This leads to a higher convergence Section 5.2 we analyse the influence of the occupancy profile on
speed than if the initial value was chosen randomly. energy savings with MDD and MDS systems.
The algorithm will converge to the set-point for each sensor,
only if the full fluxes of the luminaires allow it. The dimming 4.2.5. Computing the energy consumption
rate values higher than 1 are indeed forbidden. It is also The year overall energy consumption can be estimated with
impossible to have values lower than the full dimming rate the results of previous parts. The relative flux (Frelat.%) in Eq. (1).
(typically 3%). The possibility of convergence and the speed can be replaced by the dimming rate of luminaires calculate with
depend thus on the size of the installation and on the number of the IDDS and/or presence modelling. Multiplying this equation
sensors. For our research, we decided to stop the iteration when by the time step of the simulation and making the sum over the
 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523 519

whole year would lead to the yearly energy consumption. Note

that if the dimming rate of luminaires is null (in case of absence
with an MDS), the consumption is null and not equal to the value
given by Eq. (1).

5. Results and analysis

5.1. Analysis of the different systems

The calculations were made for each orientation and

location.
For the reference situation, we considered that the lamps and
sensors are only managed by a scheduled automatic shut off
system.
An inefficient situation without night-time shut off would
lead to an over-consumption of 235% by comparison with the
reference case. Figs. 5–7 present the annual gains obtained for Fig. 6. Annual lighting consumption and gains for Brussels.
each system and combinations, for the four tested orientations
and for the three locations. The values above each bar are
the relative difference between the considered case and the
reference case calculated by the following equation:
C considered
difference ¼ 1  (7)
Creference
With Cconsidered the annual consumption of the considered case
and Creference the annual consumption of the reference case.
We observe that the only system whose gains are influenced
by the orientation and the location is the IDDS. That is easy to
understand as the savings obtained with the MDS and the MDD
control systems depend only on occupancy, which is not
affected by the location and the orientation. The advantage of
the south orientation is observed in each location and leads to a
gain of 7–12% (depending on the location) compared to a north
orientation. West and east lead to similar savings of 4.5–10%.
We can also observe that, even in the worst situation, the IDDS
control is efficient and leads to a minimum gain of 45% by Fig. 7. Annual lighting consumption and gains for Stockholm.
comparison with the reference case. Figs. 5–7 show that, when
there is an IDDS, the annual lighting energy consumption is
linked to the annual daylight availability (higher for Athens

Fig. 5. Annual lighting consumption and gains for Athens. Fig. 8. Monthly consumption with and without IDDS in Athens.
520 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523

Fig. 10. Monthly consumption with and without IDDS in Stockholm.

Fig. 9. Monthly consumption with and without IDDS in Brussels.

than for Stockholm); the savings are higher for a low latitude We observe that, even in a north orientated office in winter,
location (Athens) but remain very high (around 50%) for a high the impact of an IDDS system is quite important (about 30% of
latitude location, regardless of the window orientation. gains). The difference in consumption between a north and
Comparing the different control systems, the IDDS system is a south-orientated room is quite small compared to the
the most interesting and leads to the highest gains in case of a consumption without IDDS. This results from the fact that in a
single office with 90% occupancy. We can possibly consider south-orientated office, the blinds are closed more often than in
installing presence detector in combination with the IDDS a north orientated office (about 20% of the working hours for
system (IDDS + MDD or IDDS + MDS) but the additional the south and 0% for the north). During some months (autumn
gains are not really significant in this case. The influence of the and winter), east and west orientations could be better than
occupancy rate is discussed later in the paper. south but north is always the worst. Over the whole year, the
If we analyse the monthly energy consumption for the three south orientation leads to highest energy gains. Note that the
locations (Figs. 8–10), we can observe the difference between difference between the monthly energy consumptions without
summer and winter. IDDS system is due to the variation in the number of working

Fig. 11. Values of presence and probability to have presence for typical workdays with a presence rate of 100, 80, 60 and 20%.
 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523 521

Table 1
Occupancy rate threshold from which a daylight dimming sensor leads to more
savings than a MDS or a MDD sensor
MDS MDD

Athens
North 31.5 19.5
South 27 14.5
West 28.5 16
East 28.5 16

Brussels
North 38.5 27
South 35 23
West 36 25
Fig. 12. Annual gains using a MDD/MDS system in function of the presence East 36 25
rate.
Stockholm
North 43.5 32.5
South 38 26.5
days in each month (in our study the offices are empty on West 40 29
Saturday and Sunday). For instance there are 23 working East 40.5 29.5
days in January (resulting in a consumption of 50.67 kWh)
and 20 days in February (resulting in a consumption of
44.63 kWh).
The energy gains by using an occupancy sensor are not as 3% (of full flux) and then presents a power of 11 W (10.4 W
high as those reported in literature [1,26,27]. First, this is due to from remaining flux (see Fig. 2(a) or Eq. (1)) and 0.5 W from
the fact that we consider a timer that shuts off the lights during the detector) when there is nobody in the room. With the MDS
nights and weekends, in our reference case. Secondly, in this system, when nobody is present, the only consumption is the
study we consider a room occupied the whole day except during consumption of the detector, which has a power of 0.5 W.
the lunch time. Many offices have a more variable presence However, the MDD system is interesting for landscape
resulting in higher gains using presence detection. offices as it prevents people to be placed in a bright spot,
compared to the average room illuminance, when working
5.2. Impact of the presence on the MDD and MDS system alone in the office room. It keeps a low general room
illuminance, to prevent too high contrasts between the
In this section, we analyse the annual energy consumption occupant’s working plane and the rest of the room. In single
using a MDD or MDS system with a variable presence offices the MDS control system is preferred.
schedule. The presence schedule, during working hours, was The curves allow us to evaluate the occupancy rates
automatically and randomly generated by steps of 10% from 0 threshold for which an occupancy sensor is more interesting
to 100%. For each time step, the value 0 or 1 of the presence than a daylight dimming sensor, for each location and
(P(t)—see Eq. (6)) is generated following a probabilistic orientation. These results are presented in Table 1. For example
reasoning. If there is presence at a time step t, the probability to in a south oriented building in Brussels, we observe that the
have presence at a time step t + 1 is high and decrease with MDS system is more interesting than the IDDS system when
the time; if there is absence a time t, the probability to have occupation drops below 35%. With these values, we can say
presence at time t + 1 is low and increase with the time. The that when occupancy rate in a single office is less than 27%,
values of high and low probabilities as well as the speed of the MDS is always preferable to IDDS and when occupancy rate is
increasing/decreasing depend on the presence rate. Fig. 11 greater than 44% the best solution is to use IDDS. For
shows some examples of presence schedules with their occupancy rates between these two extremes, the best system
respective probability for 20, 60, 80 and 100% of presence rate. depends on the location and the orientation.
Fig. 12 presents the energy gains, for MDD and MDS
systems as a function of the presence rate, regardless of the 5.3. Using DALI-compatible control systems
location. Firstly, we observe that the curves are not linear. This
is due to the fact that the time delay of the control system before Results presented in Sections 5.1 and 5.2 are valid for the
extinction has a greater influence for shorter presence time. analogue ballasts. Using luminaires with a digital (DALI-
Secondly, we see that the gains are not equal to 100% when compatible) ballast and an embedded controller would decrease
the presence is null. The consumption of the detectors (for the the performance of the systems. For a 2310 working hours year,
MDS) and the remaining flux (for the MDD) are responsible for the energy consumption of the controllers of the four luminaires
this fact. The advantage of using MDS compared to MDD is (18.5 kWh) should be added to the lighting consumption. It
clear; with the MDD system, for each luminaire, the tube is lit at represents only 3% of the consumption obtained with the
522 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523

reference system but reaches nearly 9% of the consumption of 7. Conclusion

the best system (IDDS + MDS in a south-orientated office
in Athens). So the current trend to install DALI-compatible This paper tried to analyse the performance of different
systems with one controller per luminaire can be practical for lighting control systems for a single office.
their standalone use (e.g. if the DALI line is not yet installed) The first part treats the power measurement of the different
but reduces the energetic gains. Furthermore, if embedded parts in a lighting installation, including detectors and control
controllers are installed in each luminaire without a night time systems. The relationship between the lamp power (including
cut off, the yearly parasitic consumptions of the controllers ballast) and its relative light output was measured. This
would reach nearly 70 kWh, a third of the consumption of relation is very useful to calculate the energy consumption
the best system! A DALI-compatible controller begins to be of a system when it is dimmed. We also measured the power
energetically profitable in an MDS/MDS installation where of different detectors and found that an IDDS sensor has
one controller (and its associated detector) can replace the a negligible power, a MDS/MDD detector has a power of
analogue detectors of at least four luminaires. At this moment, 0.5 W and a DALI-compatible controller and its detector
the consumption of one DALI-compatible controller and its have a power of 2 W. Evaluating these values, we can say
detector (2 W) equals the consumption of four analogue that a DALI-compatible system can be profitable only
detectors (0.5 W each). when one controller and detector manages at least four
luminaires.
6. Discussion The control systems are adjusting the electric light
according to available daylight or presence. Combinations
All along this study, we have considered a passive occupant of these two types of control systems were also considered. The
who does not care about switching on or off the lights in study shows that electric lighting savings are high when
function of available daylight. As presented by Reinhart, the regulating the light according to daylight availability and
results presented here could be better with a user who cares slightly depend on the orientation and location. The best
about daylight [22]. Moreover, in this study, we do not consider configuration we simulated is the south-orientated office in
wall switches. The light is either always turned on during the Athens. In this case, the saving potential is equal to 61% of
days when there is no presence detector or automatically the annual power consumption. The worst case is a north-
switched on when an occupant enters the room when there are orientated office in Stockholm for which the potential gain is
presence detectors. Some researches have shown a relation nearly 45%. Addition of a presence detector offers supple-
between workplane illuminance and the switch-on probability mentary gains of 1–4% when the office is occupied during the
[28]. The study with an absence detector (manual on/automatic whole day except the lunch time.
off) instead of an MDS/MDD detector would have surely When adjusting the light in function of occupation, energy
shown better gains because the occupant would not turn on the savings are higher if the lights are switched off than if they are
light every time he enters his office. Even in an office with simply dimmed in case of absence. A shut off system is
daylight control systems, sometimes the users do not turn on the preferable for single offices or meeting rooms whereas a
light even if the 500 lx are not reached. This would save even dimming system is interesting for large landscape offices, for
more energy but could have an impact on comfort, profitability visual comfort reasons.
and eyestrain. A second impact on the electric consumption of When an occupancy sensor is used, lighting energy
the use of an absence detector is due to the fact that the sensors savings increase when the occupation rate decreases,
are switched off when the lights are off. In this case, the but the relationship between the occupation rate and the
parasitic consumption of the sensors (0.5 W each) is null energy consumption is not linear as the time delay (time
whenever nobody is present in the room [29]. before switch off or dimming after the occupant leave
In our study, the chosen simulation time step was 5 min. the room) has a greater influence as the occupation rate
This is very short and is justified by results of research by decreases.
Walkenhorst et al. [30]. By comparing lighting electric This study permits also to determine when a daylight
consumption of simulations using irradiances from 1 h and dimming regulation is preferable to an occupancy sensor
1 min data sets, they found that the consumptions are switching. If the occupancy rate is higher than 44%, the IDDS is
underestimated by 6–18% when using 1 h irradiances. Because always better and if it is less than 27%, an occupancy sensor is
1 min data sets are not easily available, they have implemented preferable.
a modified Skartveit–Olseth method to make 1 min data sets
from 1 h data sets [31]. During our study, we did the same Acknowledgments
comparison between results based on 1 h irradiance time step
data files and 1 min irradiance time step data files, made This work presents the results of a master’s thesis; we would
following their modified Skartveit–Olseth method. Contrary to like to thank the promoter Christian Eugène for his help in
Walkenhorst et al., we found less than 1% of differences electrical and photometric measures. Thanks to the ‘‘Région
between 1 h and 1 min simulations for simulation for a whole Wallonne’’ which has funded a part of this research and to the
year. The difference between these two studies could be further Belgian lighting manufacturer ETAP for the loan of the tested
investigated. equipment.
 B. Roisin et al. / Energy and Buildings 40 (2008) 514–523 523

References [16] D.H.W. Li, E.K.W. Tsang, An analysis of measured and simulated day-
light illuminance and lighting savings in a daylight corridor, Building and
[1] J.D. Jennings, F.M. Rubinstein, D. DiBartolomeo, S.L. Blanc, Compar- Environment 40 (2005) 973–982.
ison of control options in private offices in an advanced lighting [17] J.A. Clarke, M. Janak, Simulating the thermal effects of daylight-con-
controls testbed, Journal of the Illuminating Engineering Society trolled lighting, Building Performance (BEPAC) (1) (1998).
[18] J.A. Clarke, J.W. Hand, M. Janak, Daylight performance: daylight quality
(2000).
[2] I.P. Knight, Measured energy savings due to photocell control of indivi- and control of energy consumption—Chapter in Daylight Performance of
dual luminaire, Lighting Research Technology 31 (1) (1999) 19–22. Buildings, James and James, 1999, ISBN: 1 873936 87 7.
[3] A.D. Galasiu, M.R. Atif, R.A. MacDonald, Impact of window blinds on [19] A. Mahdavi, S. Chang, V. Pal, Exploring model-based reasoning in
lighting systems control, Journal of the Illuminating Energy Society 29
daylight-linked dimming and automatic on/off lighting controls, Solar
Energy 76 (2003) 523–544. (1) (2000) 34–40.
[4] B. Brekke, E.H. Hansen, Energy saving in lighting installations by the [20] A. Mahdavi, S. Chang, A hybrid system for daylight-responsive lighting
control, in: Proceedings of the Seventh International IBPSA Conference,
utilization of daylight, in: Proceedings of the Right Light, vol. 3, 1995, pp.
Rio de Janeiro, Brazil, August 13–15, (2001), pp. 849–856.
875–886.
[5] D.H.W. Li, T.N.T. Lam, S.L. Wong, Lighting and energy performance for [21] A. Mahdavi, B. Spasojevic, K.A. Brunner, Element of a simulation-
an office using high frequency dimming control, Energy Conversion & assisted daylight-responsive illumination system control in buildings,
in: Proceedings of the Ninth International IBPSA Conference, Montréal,
Management 47 (2006) 1133–1145.
[6] D.H.W. Li, J.C. Lam, Evaluation of lighting performance in office Canada, August 15–18, (2005), pp. 693–700.
buildings with daylighting controls, Energy and Buildings 33 (2001) [22] C.F. Reinhart, Lightswitch-2002: a model for manual and automated
793–803. control of electric lighting and blinds, Solar Energy 77 (2004) 15–28.
[23] DALI AG of Zvei, Division Luminaires, DALI Manual. AG DALI,
[7] S. Onaygil, Ö. Güler, Determination of the energy saving by daylight
responsive lighting control systems with an example from Istanbul, Frankfurt am Main Richard Pflaum Verlag, 2001.
Building and Environment 38 (2003) 973–977. [24] www.etaplighting.com.
[8] P.R. Trengenza, I.M. Waters, Daylight coefficients, Lighting Research & [25] C.F. Reinhart, K. Voss, Monitoring manual control of electric lighting and
blinds, Lighting Research & Technology 35 (3) (2003) 243–260.
Technology 15 (2) (1983) 65–71.
[9] J. Mardaljevic, Simulation of annual daylighting profiles for internal [26] W. Morrow, B. Rutledge, D. Maniccia, M. Rea, High performance lighting
illuminance, Lighting Research & Technology 32 (3) (2000) 111–118. controls in private offices: a field study of user behavior and preference, in:
World Workplace ’98 Proceedings, International Facilities Management
[10] C.F. Reinhart, S. Herkel, The simulation of annual daylight illuminance
distributions—a state-of-the-art comparison of six RADIANCE-based Association, Chicago, IL, 1998.
methods, Energy and Buildings 32 (2000) 167–187. [27] B. VonNieda, D. Maniccia, A. Tweed, An analysis of the energy and cost
[11] C.F. Reinhart, Daylight availability and manual lighting control in office savings potential of occupancy sensors for commercial lighting systems,
in: Illuminating Engineering Society of North America 2000 Annual
buildings—simulation studies and analysis of measurement, PhD Thesis,
Fraunhofer Institut Solar Energiesysteme, 2001. Conference: Proceedings, IESNA, New York, 2000, pp. 433–459.
[12] C.F. Reinhart, O. Walkenhorst, Validation of dynamic RADIANCE-based [28] A.D. Galasiu, J.A. Veitch, Occupant preference and satisfaction with the
daylight simulations for a test office with external blinds, Energy and luminous environment and control system in daylit offices: a literature
review, Energy and Buildings 38 (2006) 728–735.
Buildings 33 (2001) 683–697.
[13] P.J. Littlefair, Predicting lighting energy use under daylight linked [29] C.F. Reinhart, M. Morrison, F. Dubrous, The lightswitch wizard—reliable
lighting controls, Building Research & Information 26 (4) (1998) daylight simulations for initial design investigation, in: Proceedings of the
Buildings Simulation, vol. III, Eindhoven, The Netherlands, August 11–
208–222.
[14] C. Ehrlich, K. Papamichael, J. Lai, K. Revzan, A method for simulating 14, (2003), pp. 1093–1100.
the performance of photosensor-based lighting controls, Energy and [30] O. Walkenhorst, J. Luther, C.F. Reinhart, J. Timmer, Dynamic annual
Buildings 34 (2002) 883–889. daylight simulation based on one-hour and one-minute means of irradi-
ance data, Solar Energy 72 (2) (2002) 385–395.
[15] A.-S. Choi, K.-D. Song, Y.-S. Kim, The characteristics of photosensors
and electronic dimming ballasts in daylight responsive dimming systems, [31] A. Skartveit, J.A. Olseth, The probability density and autocorrelation of
Building and Environment 40 (2005) 39–50. short-term global and beam irradiance, Solar Energy 49 (6) (1999) 477–487.