Energy and Buildings 127 (2016) 969979

Contents lists available at ScienceDirect

Energy and Buildings

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

Analysis of energy savings of three daylight control systems in a

school building by means of monitoring
R. Delvaeye a,b , W. Ryckaert a , L. Stroobant a,c , P. Hanselaer a , R. Klein a , H. Breesch a,
a
b
c

KU Leuven Technology Campus Ghent, Gebroeders De Smetstraat 1, 9000 Ghent, Belgium

Belgian Building Research Institute (BBRI), Division of Energy and Climate, Avenue Pierre Holoffe 21, 1342 Limelette, Belgium
Ghent University, Department of Information TechnologyIBCN, Gaston Crommenlaan 8 (Bus 201), 9050 Ghent, Belgium

a r t i c l e

i n f o

Article history:
Received 26 August 2015
Received in revised form 3 June 2016
Accepted 9 June 2016
Available online 15 June 2016
Keywords:
Energy savings
Daylight control system
Lighting
Monitoring
School buildings

a b s t r a c t
Daylight control systems, which automatically adjust the articial light levels depending on the daylight
penetration, can result in substantial energy savings. However, their energy saving potential cannot be
estimated accurately because it depends on several building and system parameters, climate conditions,
occupant behaviour and type and commissioning of the daylight control system. The objective of this
paper is to compare the energy saving potential and operation of different daylight control systems in
school buildings. One year monitoring has been carried out simultaneously in 3 neighbouring classrooms,
equipped with a different type of control system. The active power and the electric energy consumption
of the articial lighting were measured continuously on a minute-by-minute basis, as well as the occupancy of the classrooms and the global irradiance outside the building under an unobstructed horizon.
Momentary visual comfort assessments were carried out in the classrooms.
Although all classrooms have comparable occupancy and identical building characteristics, differences
between the annual energy savings of the different daylight control systems are found to be signicant:
the total annual energy savings varied from 18% to 46%. Under the given conditions, the open loop system
with the outward facing daylight sensor was noticed to yield the largest while the closed loop system
with centrally positioned sensor produced the smallest savings. However, it has to be made sure that the
energy savings are not at the expense of the visual comfort. The performance of the systems regarding both
energy savings and visual comfort is related to the operation and the initial commissioning. An in-depth
analysis of the monitoring campaign is discussed to explain the differences in energy savings and visual
comfort. The energy savings due to the implementation of a daylight control system are divided into on
the one hand dimming due to daylight penetration and on the other hand initial dimming to compensate
for the over dimensioning of the lighting system and to take into account constant illuminance control.
2016 Elsevier B.V. All rights reserved.

1. Introduction
Lighting is estimated to represent on average 40% of the total
electrical consumption in ofce buildings [1] and can amount to
70% of the total electrical consumption in school buildings without
mechanical ventilation [2]. Besides the replacement of the luminaires, lighting control systems can be installed to decrease the
energy consumption while preserving the visual comfort. Although
many different lighting control systems are available, it is rather difcult to quantify their energy saving potential. This is particularly
the case for photosensor-controlled electric lighting systems (hereafter called daylight control systems), which automatically adjust

Corresponding author.
E-mail address: Hilde.Breesch@kuleuven.be (H. Breesch).
http://dx.doi.org/10.1016/j.enbuild.2016.06.033
0378-7788/ 2016 Elsevier B.V. All rights reserved.

(switch on-off or dim) the luminous ux out of the luminaires as

a function of the available amount of daylight. The reason for this
is that the energy saving potential is dependent on a wide range of
parameters. Bodart and De Herde [3] examined the impact of the
window position, the fenestration area, the glazing transmission
factor and the internal wall reection coefcients on the lighting
consumption when using daylight control systems. Yang and Nam
[4] studied the impact of glazing ratios on lighting consumption and
on reduction of energy cost using daylight linked lighting control
system. Also the location and orientation of the building [5] and
possible external obstructions and the usage of blinds [6] affect
the energy saving potential of daylight control systems. The articial lighting and the lighting control systems themselves, including
for instance the number of rows of dimmed luminaires parallel to
the window side [7] and the commissioning of the daylight control
system [8], have to be considered as well.

970

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

As a result of the difference of these parameters from case to

case, widely divergent results for the energy saving potential of
daylight control systems have been reported. Atif and Galasiu [9]
determined by extrapolation of eld measurements in winter and
summer that the annual savings in electrical light consumption
would be 46% in the investigated case. Jennings et al. [10] speak
about 21% of lighting energy savings over a seven-month monitoring period with automatic daylight dimming controls. Yun et al. [11]
found from a monitoring campaign of 5 months in four ofces in
Korea that the application of automatic dimming control for lighting with a design illuminance of 500 lx can reduce lighting energy
consumption by up to 43%. Aghemo et al. [12] measured potential energy savings from 17% to 32% in ofces, taking into account
both the monitored annual electric energy consumption (for operation) and the parasitic energy consumption due to the installed
devices (luminaire ballasts, sensors and controllers). In 2 different
research projects of Li about daylight control in ofce buildings,
energy savings of 50% and 33% respectively are obtained [13,14].
Li et al. [15] also found that the monthly electric lighting energy
saving for the atrium corridors in a case study ranged from 14% to
65% using the present high frequency dimming controls. Haq et al.
[16] summarizes a savings potential of 2031% in ofce buildings
worldwide. Hackel and Schuetter [8] monitored an average energy
savings potential of 63% in 20 ofce buildings in the USA after commissioning. Williams et al. [17] concluded from a meta-analysis an
average savings range of daylight control of 39% in 73 ofce and
school buildings as a result from both measurements and simulations. When only measurements are considered, an average savings
potential of 28% is noticed (32 cases). Simulation results of Bodart
and De Herde [3] showed savings from 50 to 80%, depending on the
window conguration and the orientation.
However, each research studies a certain type of daylight control system. The impact of the type of daylight control system itself,
including the commissioning of the system, on the energy saving potential of daylighting may be important as well. Daylight
control systems are generally divided into open loop and closed
loop systems [18]. In an open loop control system, the photosensor is positioned to only detect daylight and thus is insensitive to
the articial light that it controls (no feedback). A lighting control
system is considered to be a closed loop system when the photosensor is able to detect both the available daylight and the articial
light that the system controls. Three main types of daylight sensors can be distinguished: outward facing daylight sensors (open
loop), centrally positioned daylight sensors (closed loop) and builtin daylight sensors in luminaires (closed loop), controlling one or
several luminaires. There are also control systems integrating both
open loop and closed loop into a dual loop system, combining the
advantages of both systems. However, these are rarely installed in
Europe and therefore not studied in this paper.
The objective of this study is to compare the energy saving
potential and operation of different daylight control systems in
school buildings. Therefore, a full year monitoring has been carried out simultaneously in 3 neighbouring classrooms in the same
building of a secondary school, all equipped with a different type
of daylight control system. Only continuous dimming systems are
studied. All rooms are equal in the matter of parameters such as
location, orientation and geometry of the room and windows. The
monitoring takes place in real circumstances, including real use of
the classrooms but also possible sub-optimal commissioning and
design of the lighting and control system.
Throughout this paper, at rst, the cases (geometry, building,
lighting and light control characteristics) which are monitored in
this study are presented. Secondly, the monitoring system (energy
consumption, active power, presence detection, global irradiance
and illuminance) and the data processing is discussed. Afterwards,
the results of the monitored energy savings of the different types of

Fig. 1. Monitored classroom of Don Bosco Haacht.

Table 1
Classroom characteristics.

Walls
Ceiling
Floor
Window

(%)

A (m2 )

LT

37
79
12

85.26
71.64
71.64
11.18

0.78

daylight control systems are discussed and compared, after which

the effect of initial dimming and dimming due to daylight penetration is studied. Finally, these energy savings are related to the
operation of the control systems and the obtained illuminance levels in the classrooms.
2. Description of the monitored cases
2.1. Building and environment
The three classrooms are located at the rst oor of a building of
a secondary school in Haacht (Belgium). The prevailing climate in
Belgium can be described as a temperate maritime climate. The
building was delivered in 2008 and the windows of the side-lit
classrooms are oriented north/north-west. The daylight penetration in the classrooms is not hampered by external obstructions,
since there are no trees or bushes in the neighbourhood and the
opposite building is a single-storey building at about 25 m distance.
This leads to a very low obstruction angle and thus an obstruction
factor of 1 according to Robinson and Selkwitz [19].
All 3 classrooms have the same geometry and interior nish.
Fig. 1 shows one of the classrooms. A general oor plan of the
classrooms with indication of the position of the luminaire rows
is shown in Fig. 2. The window conguration in the exterior wall is
shown in Fig. 3. Room and window characteristics are summarized
in Table 1. The classrooms are used for regular teaching activities
and are equipped with a computer, a whiteboard, a beamer and
a smart board. The oor area is fully occupied by chairs and tables
with a mean reection coefcient of 40%. The reection coefcients
of the walls and oor are rather low compared to the recommended
values of the European standard EN 12464-1 [20]. Obscuration of
the classrooms as well as sun protection is carried out using curtains. The curtains have a beige colour and are opaque.
The windows are made up of double glazing with a light transmission factor of 0.78. The Window-To-Floor-Ratio (WFR) in the
classroom is 16% corresponding to the guidelines of Reiter and De
Herde [21]. The ratio of the classrooms total depth to the windowhead-height measures 3 and is rather high compared to the rule of
thumb of Reinhart [22].

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

971

Fig. 3. Window conguration.

Fig. 4. Distribution of the daylight factors at working plane height.

Fig. 2. Floor plan classroom.

A simulation of the distribution of the daylight factors over the

oor area at working plane height (0,8m) is carried out with DIALux
4.11 [23] and shown in Fig. 4.The average daylight factor is 2.9%, a
medium daylight penetration according to the European standard
EN 15193 [24]. High daylight factors are noticed close to the windows. From a distance of approximately 2.5 m from the window,
i.e. the zone underneath row B (middle) and C (corridor side), the
daylight factors falls below 2%.

2.2. Articial lighting and light control systems

For the research project, the existing articial lighting in the
classrooms is replaced by new, dimmable, uorescent lighting.
Design, installation and initial commissioning of articial lighting
and light control system are done by the industry partners in this
project. All classrooms are also equipped with absence detection
and a different type of daylight control system is implemented in
each classroom.

2.2.1. Articial lighting

All classrooms are equipped with ceiling-mounted recessed
luminaires. The type of luminaires and the daylight control systems
were proposed by the manufacturers in accordance with common
practice in this type of buildings. The lighting is split up into (1)
general lighting to illuminate the horizontal working plane of the
classroom and (2) board lighting to illuminate the vertical working
plane of the classroom (i.e. the whiteboard). Characteristics of the
general lighting are shown in Table 2. Each classroom has the same
luminaire conguration with 3 rows of 3 luminaires. Each luminaire
is equipped with a dimmable electronic ballast. The lighting systems of the classrooms have been designed to obtain a maintained
mean illuminance value of 500 lx. The installed board lighting differs from classroom to classroom, but includes an asymmetrical
reector in any case. The luminaire conguration is shown in Fig. 2.
2.2.2. Lighting control systems
At rst, all classrooms are equipped with an absence detector.
An absence detector automatically turns off the lighting system if
the last person present does not do this manually when he leaves.
The switch off delay of all detectors is set to 10 min.
In all 3 classrooms, a different type of daylight control system
has been implemented. It was opted for only examining continuous dimming systems. All 3 rows of luminaires in the classrooms
are controlled by the daylight control system. By approximation,

972

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

Table 2
Characteristics of general lighting.

Description lamps (type; power per lamp)

Colour Rendering Index of lamps
Correlated Colour Temperature of lamps
Light Output Ratio of luminaires
Number of luminaires
Number of lamps per luminaire
Total measured power load (lamps and electronic ballasts)a
Maintained illuminance at the horizontal working plane
Normalized power density [25]
a

Classroom 1

Classroom 2

Classroom 3

TL5 HE 25 W
80% +
4000 K
94%
9
2
486 W
572 lx
1.19 W/(m2 .100 lx)

TL5 HE 25 W
80% +
4000 K
94%
9
2
531 W
572 lx
1.30 W/(m2 . 100 lx)

TL5 HO 54 W
80% +
4000 K
95%
9
1
540 W
478 lx
1.58 W/(m2 . 100 lx)

Total measured power load is the maximum measured power load when the articial lighting is not dimmed.

there is a linear relationship between the dimming level and the

system power (i.e. lamp power and ballast power losses) in the
dimming range between 20% (30%45% of the total power load)
and 100% (100% of the total power load). For lower dimming levels,
the ballast power losses predominates the total power (+10% of
the total power load). This is valid for DALI and 110 V dimming
control strategies.
Within the scope of this project, all daylight control systems only
work automatically in order to focus on the energy saving potential
of the daylight control system itself. This means that the users of
the classrooms cannot adjust the dimming level manually.
Table 3 gives an overview of the installed daylight control systems: 2 different types of closed loop daylight control systems are
implemented in classrooms 1 and 2, an open loop system with an
outward facing sensor is installed in classroom 3. In classroom 1,
each luminaire has its own daylight sensor while in classroom 2
a centrally positioned daylight sensor is installed. This means that
in classroom 1, all luminaires can be dimmed separately. In order
to prevent large luminance contrast variations, the manufacturer
programmed the daylight system in a way that it reduces the light
output with only 50% of the incoming daylight instead of 100%. This
means that the luminaire is adjusted to its minimum dimming level
when the daylight penetration is more than twice the maximum
light level that can be reached with the articial lighting system.
Hence, for daylight penetrations of more than 1000 lx1500 lx, the
light output of the luminaire(s) is minimal (+1% light output corresponding with +13% of the maximum system power).
In classroom 2, the row of luminaires at the window side is
dimmed at one dimming level and the rows in the middle and at
the corridor side are dimmed together at a second level. The luminaires in each group are adjusted to their minimum dimming level
when the amount of incoming solar radiation exceeds a threshold
value.
In classroom 3, a separate dimming level for each row of luminaires parallel to the window side is programmed. If there is a
large amount of daylight in the room, each luminaire group can
be switched off if the light value remains under 12% for longer than
5 min.
The commissioning of the daylight control systems was carried
out by a representative of the manufacturer of the daylight control
systems. The manufacturers representative was asked to commission the system in such a way that the illuminance level would be
close to 500 lx in all circumstances.

2.3. Use of the classrooms

In a normal school week, lessons take place from Monday to Friday between 8h35 and 16h35 (see Table 4), except on Wednesday,
when there are no lessons in the afternoon. The duration of a lesson
is 50 min. A maximal occupancy of 30 h (36 lessons of 50 min) per
week is expected.

The major holidays at Belgian secondary schools are listed in

Table 5. The total duration of the holidays per year is approximately
14 weeks, which results in a theoretical maximal occupancy of the
classrooms of 1140 h per year.
3. Description of the monitoring system and the data
processing
3.1. Monitoring of energy consumption, occupancy and global
horizontal irradiance
The long-term monitoring in the 3 classrooms is carried out
by means of a central PLC system, a Beckhoff CX9020 [26]. Several types of terminals are connected via EtherCAT to the PLC. The
embedded PC runs 3 monitoring applications, 1 for gathering data
from each classroom. In all classrooms the active power and electric
energy consumption of the lighting and the occupancy of the classrooms are continuously monitored. Beside these parameters, also
the global irradiance under an unobstructed horizon is monitored.
This monitoring was carried out from 01/12/2013 until 30/11/2014.
3.1.1. Energy consumption and energy savings
Measurements of the power and the energy consumption are
carried out by means of electronic single-phase energy meters (type
Eltako WSZ12DE-32A) [27]. Pulses are generated every 0.5 Wh and
are registered digitally by the PLC system together with the timestamp (current local time in Central European Time Zone). These
measurements are carried out for each individual row of luminaires
parallel to the window side (see Fig. 2).
Eq. (1) denes the obtained energy savings W% (%) due to the
implementation of a daylight control system as a function of the
lighting energy consumption (kWh) with (WL,withDCS ) and without
(WL,withoutDCS ) a daylight control system during light on time Ton
W% =

WL,without DCS WL,with DCS

100
WL,without DCS

(1)

Light on time Ton is dened as the time that the users of the
classrooms need articial lighting during their presence (i.e. when
the lighting is switched on manually). WL,withDCS in Eq. (1) is the
energy consumption monitored by the single-phase energy meters.
WL,withoutDCS in Eq. (1) is the theoretical energy consumption of the
articial lighting during light on time if no daylight control system
would have been installed. It is calculated as the product of the
total measured power load (of the full lighting system or of a single
luminaire row) and the light on time Ton .
The energy savings due to the implementation of a daylight
control system can be divided into 2 parts: energy savings due to
(1) initial dimming without daylight W%initial and due to (2) daylight penetration W%daylight . Initial dimming is used to compensate
for the over dimensioning of the lighting system and to take into
account constant illuminance control [24]. Eq. (1) is consequently
rewritten as Eq. (2) where Pboundary is dened as the average power

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

973

Table 3
Characteristics of daylight control systems.

Type
Open loop/closed loop
Control system (communication)
Number of different dimming levels
Power loss

Classroom 1

Classroom 2

Classroom 3

Individual daylight sensor per luminaire

Closed loop
110 V
9
7 mW

Centrally positioned daylight sensor

Closed loop
DALI
2
< 2W

Outward facing daylight sensor

Open loop
DALI
3
< 4W

level of an individual row of luminaires during night when the

curtains are opened.
W% = W% inital + W%daylight =

WL,without DCS Pboundary Ton

Pboundary Ton WL,with DCS

WL,without DCS

WL,without DCS

100

(2)

3.1.2. Presence detection and occupancy

The monitoring of the occupancy is carried out by means of
a presence detector (type ESYLUX PD-C360i/24 DC24Vplus) [28],
which detects the presence of people by recording movement of
infrared radiation. It has been installed in the middle of the classroom, at ceiling height (2.8 m). At this height, the circular detection
area, accurate enough to detect sedentary work, has a diameter
of 10.3 m. As the detector is only used for monitoring purposes
(another absence detector is used to switch off automatically the
lighting system when no movement is detected), the switch off
delay is set to its minimum of 1 min. Every change of the occupancy
state is monitored, i.e. event based monitoring. To discard inaccuracies of the detector, a lter is applied to the raw occupancy data. On
the one hand, the lter lls up gaps of absence of less than 2 min
between short periods of occupancy. On the other hand, the last
minute of each period is deleted by the lter as this occupancy is
the result of the delay time and is thus only apparent.
3.1.3. Global horizontal irradiance
To interpret the operation and energy savings of the daylight
control systems, it is important to know the outdoor environment.
Global horizontal irradiance is chosen as parameter because the
measurements are part of a larger monitoring campaign where also
indoor and outdoor temperatures are monitored and analysed. In
addition, global horizontal irradiance data gives a good indication
of and can be converted to daylight illuminance data for detailed
daylight simulations [29].
The monitoring of the global horizontal irradiance is carried out
using a silicon pyranometer (type Kipp and Zonen SP Lite2) [30],
positioned outdoors under an unobstructed horizon. The eld of
view of the pyranometer measures 180 and the pyranometer has
a spectral range from 400 nm to 1100 nm. The response time of the
sensor is lower than 500 ns at 95% and has a directional error which
is smaller than 10 W/m2 (up to 80 with 1000 W/m2 beam).
3.2. Monitoring of illuminance
Although the main objective of the implementation of daylight
control systems is to save energy, it has to be made sure that the
energy savings are not at the expense of the visual comfort. To evaluate the obtained visual comfort in every classroom, temporary
measurements of the interior horizontal illuminance were carried
out on 25/09/2014 between 18h30 and 19 h on 24 locations per
classroom and during the weekend of 11-12/10/2014 on two locations (see Fig. 2). The measurements are performed at a working
plane height of 0.8 m above nished oor by means of photomet-

Table 4
Schedule of classrooms.
Time period

Activity

8h35
8h359h25
9h2510h15
10h1510h25
10h2511h15
11h1512h05
12h05
12h0513h05
13h0513h55
13h5514h45
14h4514h55
14h5515h45
15h45
15h4516h35
16h35

Start of school day

Lesson 1
Lesson 2
Break
Lesson 3
Lesson 4
End of school day at Wednesday
Lunch break
Lesson 5
Lesson 6
Break
Lesson 7
End of school day in case of a 7-lessons day
Lesson 8
End of school day in case of a 8-lessons day

Table 5
Major holidays (based on calendar year 2014).
Holiday

Duration

Period

Christmas holidays
Spring half-term
Easter holidays
Summer holidays
Autumn half-term

2 weeks
1 week
2 weeks
2 months
1 week

End of DecemberBeginning of January

Beginning of March
Mid-April
July + August
End of October

ric sensors (Gigahertz-Optik; type detector head: VL-3701-2; type

measuring instrument: P9710-1) [31].
4. Results and discussion
4.1. Energy consumption and savings
Table 6 gives an overview of the monthly energy consumption
of the lighting systems, the light on time and the occupancy in the 3
classrooms for the monitored period (1/12/2013 until 30/11/2014).
Signicant differences in energy consumption between the
classrooms are noticed. The annual energy consumption in classroom 2 is respectively 31% and 52% higher than in classroom 1 and
3, while the light on time is in the same order of magnitude for the
3 classrooms. Moreover, monthly energy consumption differs a lot
depending on occupancy, user behaviour and daylight availability.
Although classroom 1 has almost the same occupancy in January
(winter) and May (spring), the light on time is in May 20% lower
than in January. In addition, the energy consumption in May is only
50% of the consumption in January.
More in-depth conclusions can be drawn from Fig. 5, which compares the energy savings of the individual luminaire rows as well as
the total articial lighting system in the 3 classrooms on an annual
basis. Although the 3 classrooms are equivalent in the matter of
geometry, orientation and interior nish, the reported energy savings tend to differ quite strongly. While the closed loop daylight
control system in classroom 1, using an individual daylight sensor
per luminaire, yielded 34% energy savings in total, the energy savings W% in classroom 2 came to only 18%. In classroom 3, in which

974

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

Table 6
Monthly energy consumption, light on time and occupancy in the 3 classrooms.
Energy consumption (kWh)

December 13
January 14
February 14
March 14
April 14
May 14
June 14
July 14
August 14
September 14
October 14
November 14
Total

Light on time Ton (h)

Occupancy (h)

Classroom 1

Classroom 2

Classroom 3

Classroom 1

Classroom 2

Classroom 3

Classroom 1

Classroom 2

Classroom 3

15.1
32.5
30.6
16.6
7.1
16.3
7.5
0.1
0.4
19.8
19.1
26.8
192.0

24.3
36.7
33.9
16.2
7.4
22.1
11.1
0.0
0.3
29.7
28.7
41.9
252.3

15.6
25.6
26.5
8.7
1.0
6.4
2.6
0.0
0.0
18.5
26.7
34.0
165.7

41.7
86.0
88.5
57.2
28.9
66.3
32.8
0.3
1.6
71.7
55.1
72.8
602.9

50.6
79.2
77.8
39.0
19.8
55.5
27.3
0.0
0.7
73.1
68.6
88.7
580.2

43.7
68.7
79.7
49.6
13.3
50.2
15.7
0.0
0.5
90.4
72.6
82.2
566.6

45.7
95.2
101.7
72.0
36.2
98.1
57.7
0.0
1.1
109.6
81.5
94.3
793.1

57.0
103.5
108.0
65.2
30.8
96.0
49.7
0.0
1.5
114.5
81.4
98.6
806.1

51.2
87.1
96.1
71.6
35.9
76.7
49.7
0.0
0.8
113.0
80.3
93.1
755.4

Fig. 5. Annual lighting energy savings (%).

an outward facing daylight sensor is used, 46% lighting energy was

saved by installing the daylight control system.
Fig. 5 also compares the lighting energy savings per luminaire
row. While the 44% energy savings of luminaire row A (next to the
window) in classroom 2 is comparable to the savings of row A in
classroom 1 (39%) and classroom 3 (50%), the energy savings of rows
B (middle) and C (next to the wall at the corridor side) in classroom
2 are found to be much lower than the savings in the other 2 classrooms: luminaire rows B and C in classroom 2 consumed only 6%
less energy compared to the situation in which no daylight control
system would have been installed.

4.2. Operation
The results of the energy savings can be explained by evaluating the operation of the different daylight control systems. Fig. 6
shows the cumulative frequency of occurrence of the power on
luminaire rows A and B in the 3 classrooms in May 2014 during Ton .
It is clear that the closed loop system with a centrally positioned
daylight sensor (classroom 2) has a peculiar behaviour compared
to the other daylight control systems. The active power (derived
from the measured energy consumption) of row B in classroom 2
has a narrow spread close to 100% of the total power load: 92% of
the power measurements have a value between 85% and 100% of
the total power load. This implies that the daylight control system
dims this luminaire row very limited in real terms. The same result
is expected for luminaire row C. Despite the fact that the energy
savings of row A in classroom 2 seemed to be normal compared
to the gained energy savings in classrooms 1 and 3, Fig. 6 shows

that the daylight control system does not behave like a continuous
dimming system. 75% of the power measurements of row A in classroom 2 occurred between 19% and 22% of the total power load. The
daylight control system thus generally adjusts its dimming level
to the minimum level of the system, although it was expected to
work as a continuous dimming system. Due to the initial decision
to observe the operation in real circumstances, it is impossible to
explain if this problem is caused by possible malfunctioning of sensor or controller, or less efcient control-strategies or sub-optimal
parameter choice during initial commissioning.
The daylight control systems in classrooms 1 and 3 can be classied as continuous dimming systems. In row B of classroom 1 for
example, where each luminaire is equipped with a daylight sensor,
all dimming levels between 35% and 70% occur more or less in the
same frequency during the examined period. However, in classroom 3 with outward facing daylight sensor, the luminaire rows
A and B switch off at high daylight levels causing a large leap at
0% power load. In addition, a minimum power load for dimming
is clearly determined for all systems. In classroom 1, the minimal
power of row A amounts to 26% of the total power load. In classroom 3, the minimum level for rows A and B is 20% of the total
power load.
Henceforth, the analysis is concentrated on the obtained measurement results of the lighting system of classroom 1 and 3.
Seasonal differences in energy savings are explained in Fig. 7
by comparing the operation of the daylight control system with
individual daylight sensor per luminaire between January and May.
A large effect of the availability of daylight exists on the power
occurrence of all luminaire rows. The differences are the largest for

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

975

Table 7
Annual lighting energy consumption and energy savings, divided in initial savings and daylight savings in classroom 1 and 3.

Energy consumption (kWh)

Energy savings due to initial dimming (kWh)

Energy savings due to daylight penetration (kWh)

Total
Row A
Row B
Row C
Total
Row A
Row B
Row C
Total
Row A
Row B
Row C

Classroom 1

Classroom 3

192.0
59.6
55.3
77.1
56.4
15.1
29.5
11.8
44.6
23.0
12.8
8.8

165.8
50.3
50.7
64.8
41.3
1.5
24.7
15.1
97.4
49.7
26.1
21.6

Fig. 6. Cumulative frequency of occurrence of power of luminaire rows A and B in the 3 classrooms in May 2014 during light on time.

row A (window side): in January, the measured power exceeds 70%

of the total power load for more than 80% of the time while the
power is larger than this value for less than 5% of time during May.
The main reason for this is that the amount of daylight is the largest
in this zone and thus largest variations in daylight availability occur
here. The smallest differences are generally found on luminaire row
B. This can be attributed to the dual nature of the energy savings, i.e.
savings by initial dimming and savings by daylight dimming. Fig. 7
also shows that row C in classroom 1 is hardly dimmed in January.
4.3. Initial dimming vs. dimming due to daylight penetration
Fig. 6 shows that the total measured power load of the luminaire
rows A and B in classroom 1 and 3 is barely reached in May. The
maximum measured value of the power of row B amounts to 71%
and 75% in classroom 1 and 3 respectively. This can be explained
by initial dimming of row B, even when there is no daylight available. The initial dimming consists of two parts. Firstly, there is a
time dependent part that is related to the maintenance factor. The
daylight control system compensates for the over dimensioning
that is provided for setting off the ageing of the articial lighting
system. Secondly, when the articial lighting is turned on without
being dimmed, the illuminance level in the middle of the classroom
(below luminaire row B) is much higher than the illuminance level
at the window (below row A) and the corridor side (below row C).
The daylight control system is commissioned in such a way that
the illuminance level approaches 500 lx in all circumstances. This
means that row B can be dimmed initially (without daylight) much
more than the rows of luminaires at the sides. The difference in

initial dimming and dimming due to daylight penetration is illustrated for classroom 1 in Fig. 8. It shows the course of the power
of all luminaire rows and of the global irradiance during 1 day in
December 2013. The power of row B (in the middle of the room)
is lower than the power of the other 2 rows when the daylight
availability is limited at the start of the lessons due to more initial dimming. As the daylight level rises in the morning, the power
of row A (at the window side) responds the most to the increasing
irradiance. Once a certain global irradiance is exceeded (in this case
around 09:25 a.m.), the power of row A is lower than row B (and C)
due to more daylight penetration on the window side.
Table 7 denes the annual energy savings due initial dimming
and daylight dimming compared to the monitored lighting energy
consumption during light on time Ton for all luminaire rows in
classroom 1 and 3. Furthermore, the total energy consumption and
energy savings are mentioned. Large energy savings due to initial
dimming are noticed for row B in both classrooms. Contrary, row
A has larger total energy savings due to daylight dimming in both
classrooms. This results in classroom 1 in larger energy savings for
row B (i.e. 42.3 kWh) than row A (i.e. 38.1 kWh) and in classroom 3
in equal savings (about 51 kWh) for row A and B (see also Fig. 5).
The effect of daylight availability during the year on the division
in initial and daylight savings for each luminaire row in classroom
1 is shown in Fig. 9. Based on the denition in Eq. (2), the initial
savings for each row are nearly constant throughout the year. On
the contrary, the daylight savings depend on the season, with much
larger daylight savings in spring/autumn than during winter.

976

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

Fig. 7. Cumulative frequency of power of the 3 luminaire rows in classroom 1 in January and May during light on time.

Fig. 8. Initial dimming vs. dimming due to daylight penetration in classroom 1.

Fig. 9. Lighting energy savings, divided in initial savings and daylight savings, in classroom 1 (per luminaire row per month).

4.4. Illuminance
The results of the energy savings have to be related to the
obtained visual comfort, in particular to the illuminance level on

the workplane in the classrooms. Table 8 shows momentary measurements of the illuminance at workplane height in all classrooms,
averaged per luminaire row. Two scenarios are considered: (1) only

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

977

Fig. 10. Illuminance and power in classroom 1 and 3, measured on 11-12/10/2014.

articial lighting (obscured room) and (2) articial lighting and

daylight.
In classroom 1, the illuminance levels on the workplane below
luminaire rows B and C approaches the target level of 500 lx with
and without daylight. Only row A is not dimmed to 500 lx in the
scenario with daylight. In classroom 2 with centrally positioned
daylight sensor, high illuminance levels are measured in the whole
room in both scenarios. None of the zones below the luminaires
rows is dimmed to less than 700 lx, which is a lot higher than the

desired 500 lx. These high illuminance values are in line with the
low energy savings on Fig. 5 and high power loads during operation
on Fig. 6. In classroom 3, the measured illuminance without daylight approaches the target value of 500 lx for all rows. However,
row B and C are dimmed to extreme low values in scenario with articial lighting and daylight. The average values are 302 and 340 lx
respectively, which conrms that the dimming curves of rows B
and C are too steep. This result puts the large energy savings of row

978

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

Table 8
Illuminance (lx) on workplane height, measured on 25/09/2014 between 18h30 and
19 h.

Classroom 1
Classroom 2
Classroom 3

Articial lighting
Articial lighting + daylighting
Articial lighting
Articial lighting + daylighting
Articial lighting
Articial lighting + daylighting

ErowA (lx)

ErowB (lx)

ErowC (lx)

469
826
711
726
537
635

513
589
858
945
536
302

528
553
775
913
507
340

B and C in this classroom with a control system with outward facing

daylight sensor in Fig. 5 into perspective.
These momentary results are conrmed by continuous measurements of the illuminance in Fig. 10 on two locations in
classroom 1 and 3 (see Fig. 2) during 2 successive days with continuous articial lighting (absence detection was overruled) and
without occupancy. After set up of the measurements at night, the
illuminance on both locations in classroom 1 equals approximately
the desired 500 lx. At location 1 (underneath row A, close to the
window, see Fig. 2), an illuminance value on the workplane significantly larger than 500 lx is noticed during the 2nd day as a result
of daylight penetration. At that moment, row A is dimmed to a
minimum power of 43 W or 26% of the maximum power. In classroom 3, the illuminance on both locations does not always reach
the required value of 500 lx, e.g. at the end of the 2nd day or during
night at location 1. This is caused by inappropriate operation of the
control system: lighting is switched off but is not turned on again
when irradiance outside is low: e.g. row A is dimmed to 0% on the
1st day at 10 h and stays off during the rest of the measurement
period.
5. Conclusions
The monitored energy savings and the operation of 3 different
types of daylight control systems under real-life conditions in
classrooms of a secondary school in Belgium are compared in this
paper.
Although the 3 classrooms have equivalent parameters such as
geometry, orientation and interior nish, the lighting energy savings, obtained by the implementation of a daylight control system,
tend to differ signicantly. The closed loop daylight control system with individual daylight sensor per luminaire in classroom 1
and the open loop system with outward facing daylight sensor in
classroom 3 yielded 34% and 46% energy savings respectively. The
energy savings in classroom 2, with a closed loop system using a
centrally positioned sensor came to only 18%.
It is remarkable that comparison of the energy savings per luminaire row shows that the middle luminaire row in classroom 1 and
3 has an equal or larger energy saving potential than the row next
to the window, although daylight levels decrease as a function of
the distance from the windows. A considerable part of the energy
savings in the middle row does not result from dimming by daylight harvesting, but is the result of the compensation of the initially
higher articial lighting levels.
These energy savings are related to the operation of the control systems and the obtained illuminance levels in the classrooms.
The closed loop system with a centrally positioned sensor does not
behave like a continuous dimming system but more as an on/off
system. In addition, the luminaire rows central (i.e. row B) in this
classroom and next to the corridor (i.e. row C) hardly dimmed
during the monitoring period, which results in low total energy
savings compared to the energy savings in the other classrooms.
The illuminance levels on the workplane underneath these rows in
classroom 2 are larger than the desired value of 500 lx. The measured illuminance in classroom 1 is very close to the required value.

In classroom 3, on the other hand, the luminaires in the middle and

next to the corridor are dimmed to signicantly less than 500 lx.
This is caused by inappropriate operation: lighting is switched off
by the control system but is not turned on again when irradiance
outside is low. This stresses the importance of correct installation, functioning and commissioning of the daylight control system
regarding energy savings and visual comfort.
Acknowledgements
The authors appreciate the nancial support of the Institute for
the Promotion of Innovation by Science and Technology in Flanders (IWT-Vlaanderen)TETRA-project 120124 Impact of daylight
control systems on design and renovation of school buildings.
References
[1] M. Krarti, Energy Audit of Building SystemsAn Engineering Approach, CRC
Press, Boca Raton Florida, 2000.
[2] VITO, Energy use and energy savings potential in elementary and secondary
schools in Flanders (in Dutch), 2001.
[3] M. Bodart, A. De Herde, Global energy savings in ofces buildings by the use
of daylighting, Energy Build. 34 (2002) 421429.
[4] I.-H. Yang, E.-J. Nam, Economic analysis of the daylight-linked lighting control
system, Sol. Energy 84 (8) (2010) 15131525.
[5] B. Roisin, M. Bodart, A. Deneyer, P. DHerdt, Lighting energy savings in ofces
using different control systems and their real consumption, Energy Build. 40
(2008) 514523.
[6] C.F. Reinhart, Effects of interior design on the daylight availability in open
plan ofces, in: Proceedings of the ACE3 2002 Summer Study on Energy
Efciency in Buildings, Pacic Grove, USA, 2002.
[7] D.H. Li, K. Cheung, S. Wong, T.N. Lam, An analysis of energy-efcient light
ttings and lighting controls, Appl. Energy 87 (2010) 558567.
[8] S. Hackel, S. Schuetter, Commissioning for Optimal Savings from Daylight
Control, Energy Center of Wisconsin, Minnesota Department of Commerce,
Division of Energy Resources, 2013.
[9] M.R. Atif, A.D. Galasiu, Energy performance of daylight-linked automatic
lighting control systems in large atrium spaces: report on two
eld-monitored case studies, Energy Build. 35 (2003) 441461.
[10] J. Jennings, F. Rubinstein, D. DiBartolomeo, S. Blanc, Comparison of control
options in private ofces in an advanced lighting controls testbed, J. Illum.
Eng. Soc. 29 (2000) 3960.
[11] G.Y. Yun, H. Kim, J.T. Kim, Effects of occupancy and lighting use patterns on
lighting energy consumption, Energy Build. 46 (March) (2012) 152158.
[12] C. Aghemo, L. Blaso, A. Pellegrino, Building automation and control systems: a
case study to evaluate the energy and environmental performances of a
lighting control system in ofces, Autom. Constr. 43 (July) (2014) 1022.
[13] D.H. Li, J.C. Lam, Evaluation of lighting performance in ofce buildings with
daylighting controls, Energy Build. 33 (2001) 793803.
[14] D.H. Li, T.N. Lam, S. Wong, Lighting and energy performance for an ofce
using high frequency dimming controls, Energy Convers. Manage. 47 (2006)
11331145.
[15] D.H. Li, A.C. Cheung, S.K. Chow, E.W. Lee, Study of daylight data and lighting
energy savings for atrium corridors with lighting dimming controls, Energy
Build. 72 (April) (2014) 457464.
[16] M.A.U. Haq, M.Y. Hassan, H. Abdullah, H.A. Rahman, M.P. Abdullah, F. Hussin,
D.M. Said, A review on lighting control technologies in commercial buildings,
their performance and affecting factors, Renew. Sustain. Energy Rev. 33
(2014) 268279.
[17] A. Williams, B. Atkinson, K. Garbesi, E. Page, F. Rubinstein, Lighting controls in
commercial buildings, LEUKOS, LEUKOS J. Illum. Eng. Soc. North Am. 8 (3)
(2012) 161180.
[18] IEA, Daylight in Buildings: A Source Book on Daylighting Systems and
Components, Lawrence Berkeley National Laboratory, Berkeley, CA, USA,
2000.
[19] A. Robinson, S. Selkwitz, Tips for Daylighting with Windows (Updated
Version), Lawrence Berkeley National Laboratory, Berkeley, CA, USA, 2013.
[20] EN 12464-1, Light and lighting Lighting of work places Part 1: Indoor work
places, 2011.
[21] S. Reiter, A. De Herde, Natural Lighting of Buildings (in French), Presses
Universitaires de Louvain, 2004.
[22] C.F. Reinhart, A simulation-based review of the ubiquitous
window-head-height to daylit zone depth rule-of-thumb, in: IBPSA,
Conference, Montral, Canada, 2005.
[23] DIAL http://www.dial.de/DIAL/en/dialux.html, [Online] (accessed 4.11.14).
[24] EN 15193, Energy performance of buildingsenergy requirements for
lighting, 2007.
[25] W. Ryckaert, C. Lootens, J. Geldof, P. Hanselaer, Criteria for energy efcient
lighting in buildings, Energy Build. 42 (2010) 341347.
[26] Beckhoff Automation, [Online]. Available: http://www.beckhoff.com/
(accessed 28.10.14), 2014.

R. Delvaeye et al. / Energy and Buildings 127 (2016) 969979

[27] Eltako Electronics, [Online]. Available: http://www.eltako.com/ (accessed
28.10.14).
[28] ESYLUX, [Online]. Available: http://www.esylux.com/en/en/ (accessed
28.10.14).
[29] C.F. Reinhart, Tutorial on the use of Daysim simulations for sustainable
design, 2010.

979

[30] Kipp and Zonen, [Online]. Available: http://www.kippzonen.com/ (accessed

28.10.14).
[31] Gigahertz-Optik, [Online]. Available: http://www.gigahertz-optik.de/
(accessed 28.10.14).