Demand-controlled energy systems in commercial and institutional buildings: a review of methods and potentials

Abstract

Heating, ventilation and airconditioning (HVAC) are by far the most energy intensive systems in commercial and institutional buildings with office spaces. This makes HVAC systems attractive targets for energy efficiency improvement. New technological advancements can play significant role on improving energy efficiency. Such advancements have been also emerged in form of novel management and control strategies, which might lead to considerable energy savings with relatively minor investments. This paper evaluates demand control HVAC and lighting to assess the energy saving potential of upgrading the conventional building energy systems. This paper provides a summary of different methods and occupancy detection technologies. A range of technologies and methods are covered that vary in complexity, limitations and energy saving potential. Additional benefits such as demand response are evaluated and other emerging applications are discussed. Based on the review of methods and potentials, the paper assesses the state of the art in demand controlled energy systems and suggests areas for further research.

Full text

ECEEE SUMMER STUDY PROCEEDINGS 1443
Demand-controlled energy systems in
commercial and institutional buildings:
a review of methods and potentials
Shoaib Azizi
Department of Applied Physics and Electronics
Umeå University
90187 Umeå
Sweden
shoaib.azizi@umu.se
Gireesh Nair
Department of Applied Physics and Electronics
Umeå University
90187 Umeå
Sweden
gireesh.nair@umu.se
Thomas Olofsson
Department of Applied Physics and Electronics
Umeå University
90187 Umeå
Sweden
thomas.olofsson@umu.se
Keywords
occupancy detection, demand-side management
Abstract
Heating, ventilation and air-conditioning (HVAC) are by far
the most energy intensive systems in commercial and institu-
tional buildings with oce spaces. is makes HVAC systems
attractive targets for energy eciency improvement. New tech-
nological advancements can play signicant role on improving
energy eciency. Such advancements have been also emerged
in form of novel management and control strategies, which
might lead to considerable energy savings with relatively minor
investments. is paper evaluates demand control HVAC and
lighting to assess the energy saving potential of upgrading the
conventional building energy systems.
is paper provides a summary of dierent methods and oc-
cupancy detection technologies. A range of technologies and
methods are covered that vary in complexity, limitations and
energy saving potential. Additional benets such as demand
response are evaluated and other emerging applications are
discussed. Based on the review of methods and potentials, the
paper assesses the state of the art in demand controlled energy
systems and suggests areas for further research.
Introduction
Today, buildings are responsible for up to 40% of total nal
energy use in developed countries above industry or transport
sectors (U.N. 2009). Energy use in the building sector is grow-
ing due to population increase, rise in the time spent inside
buildings and improvement in building services and comfort
levels (Pérez-Lombard, Ortiz, and Pout 2008). HVAC systems
used for space conditioning to maintain thermal comfort and
ventilation requirements have become predominantly the most
important end user in the built environment in both residen-
tial and non-residential buildings (EIA2019). e increase in
energy use related to HVAC systems are predicted to continue
in non-residential buildings in the coming years (EIA2019).
Non-residential buildings accounted for the highest growing
rate in building sector in terms of energy use before the eco-
nomic crisis (Pérez-Lombard, Ortiz, and Pout 2008). During
the crisis, from 2007–12 the nal energy consumption in Eu-
rope decreased by 8% while on the contrary the nal energy
consumption of non-residential buildings remained quite sta-
ble (Saheb et al. 2015). Non-residential buildings account for
a quarter of total energy consumption in European building
stock and comprised of dierent typologies such as oces, ho-
tels and restaurant, educational, hospitals, wholesale and retail
(BPIE 2011). Dierent types of buildings have dierent distri-
bution of energy use although HVAC, followed by lighting are
the most energy intensive systems in European non-residential
buildings (Balaras et al. 2017). In Canada, they are responsible
for 66% of energy use in commercial and institutional build-
ings (NRCan 2015). Oce buildings are important typol-
ogy within non-residential building category and consume
between 2% and 3.2% of nal energy use (Pérez-Lombard,
Ortiz, and Pout2008). Since 2008 the electricity consumption
has increased by 74% mainly due to increase in required IT
devices, new telecommunication types, new appliances and
air-conditioning (D’Agostino, Zangheri, and Castellazzi 2017).
Oce buildings comprise of the second biggest category of
non-residential oor spaces. ey have similar cooling and

8-227-19 AZIZI ET AL
1444 ECEEE 2019 SUMMER STUDY
8. BUILDINGS: TECHNOLOGIES AND SYSTEMS …
heating conditions as residential buildings while they are used
in shorter periods (BPIE 2011).
Building energy systems are meant to provide a comfortable
environment for the occupants in the building. e level of oc-
cupancy has a great impact on building’s indoor environmental
qualities (IEQ) thus on energy use in the building (Elie and C.
2012). is is because occupants generate CO2, sensible and la-
tent heat and their behaviour such as opening windows and us-
ing equipment aect IEQ in the building. Demand-controlled
energy systems are meant to deliver the services only when
and where they are needed, in the amount that they are needed
(Shen, Newsham, and Gunay 2017). is strategy requires ac-
curate occupancy detection in real time. It is even obligatory in
some building codes and standards for certain building spaces
to have occupancy sensors for lighting control (IRC 2011).
Occupancy detection has long been practiced especially in
the eld of automated lighting systems (Guo et al. 2010) e
application of occupancy detection technologies in the HVAC
area is relatively recent and emerging eld (B. Dong and Lam
2011; Gunay, O&Amp, et al. 2016). Occupancy detection tech-
nologies provide information that can be used by building con-
trol systems to operate the energy systems proportional to the
number of occupants in the building. Accordingly, the energy
use would be optimized by integrated control of active and
passive heating, cooling, lighting and ventilation systems (Sun
et al. 2013). Webber et al. (2006) investigated the energy use
by oce equipment aer working hours and found the turn-
o rate of most types of equipment are under 50% signifying
considerable energy saving potential. A study in South Africa
showed higher energy use during non-working hours than
working hours in commercial buildings (Masoso and Grobler
2010). e occupant’s behaviour and poor building control are
presented as the reasons for such energy waste while occupan-
cy-based building automation systems can be a solution to im-
prove the energy eciency.
e objective of this paper is to review the conventional
and new innovative methods of detecting building occupancy.
Resolution and accuracy are dened as two important qualities
of occupancy detection. Conventional methods of occupancy
detection are identied and implicit sensors linked together in
network is introduced as important emerging approach. Sub-
sequently, energy eciency potential of demand control energy
systems are investigated and other possible applications are
presented. e nal section provides some concluding remarks
and presents areas for future research.
Qualities of occupancy detection
Dierent applications of occupancy detection require dier-
ent qualities of detection. Resolution and accuracy are distin-
guished the main qualities of occupancy detection in the litera-
ture (Mel et al. 2011).
RESOLUTION
Most common methods of occupancy detection are limited to
a binary output showing whether or not a space is occupied.
However, many applications require more detailed information
such as the number of people in the space. Mel et al. dene
resolution of a sensor as measure of the quality of information
that consists of 3dimensions (Figure1). As resolution of a sen-
sor increases the occupant becomes more dened, the space
becomes smaller and the information is available more quickly.
Four levels of occupant resolution include:
•
Level 1: Occupancy; whether there is at least one person
present in the zone
• Level 2: Count; how many people are present in the zone
• Level 3: Identity; who are the people present in the zone
• Level 4: Activity; what are the present people occupied with
ACCURACY
In order for the building management systems to operate ef-
ciently, the occupancy information has to be accurate and
reliable. e accuracy of occupancy detection can be dened
in two dierent ways based on whether considering the pres-
ence or absence of occupants (Gunay, Fuller, et al. 2016). e
accuracy of presence detection is the ratio of correct presence
detections and the accuracy of absence detection is the ratio of
correct absence detection. In order to calculate these ratios, the
acquired occupancy information is contrasted to the ground-
truth. e ground truth is obtained with a reliable method of
occupancy detection, e.g. in a study by Ghai et al. (2012) the
occupants manually tagged their location by using a desktop
application. Similarly, there are two types of occupancy detec-
tion errors (Shen, Newsham, and Gunay 2017). False negative
detection is when the zone is occupied but wrongly concluded
to be empty. On the other hand, false positive detection is when
a zone is empty but wrongly considered to be occupied.
False negative errors are more problematic type in the eld
of building energy management. For example, a false negative
error causes the automatic lighting systems to be switched o
when space is occupied leading to occupant’s dissatisfaction
and reduction in productivity. Occupants’ annoyance might
lead them to nd ways to resolve the problems by dismantling
the sensors or control systems leading to adverse energy per-
formance (O’Brien and Gunay 2014). Nagy et al. (2015) found,
even 90% accuracy of presence detection can cause considera-
ble dissatisfaction for the occupants in the case of lighting con-
trol. Nevertheless, the required accuracy might dier in dier-
ent applications. For example, occupancy detection for heating
control might require less accuracy because of the buildings’
thermal inertia. e concerns about false negatives usually lead
Figure 1. Resolution of occupancy detection (Melfi et al. 2011).


8. BUILDINGS: TECHNOLOGIES AND SYSTEMS …
ECEEE SUMMER STUDY PROCEEDINGS 1445
8-227-19 AZIZI ET AL
the designers to adjust long timeout period for occupancy sen-
sors (Nagy et al. 2015). e system must detect no occupancy
during the entire timeout period before it undertakes a control
action. is increases the condence that there is no occupancy
despite increase in false positive errors leading to lower energy
eciency potential. Increasing accuracy of occupancy detec-
tion usually associates with higher investment in deployment.
us, the decision on level of accuracy should be based on ac-
ceptable return on investment calculations that depends on the
application of occupancy detection system.
Technologies and Methods of sensing occupancy
PIR SENSORS
PIR sensors are one the most common types of sensors that
are oen used in building automation applications specically
for automatic lighting systems (Guo et al. 2010). ese devices
detect movements by sensing the changes in infrared radia-
tion thus, inferring human presence in the space (Dodier et al.
2006). Since movement is the key determinant for functionality
of these sensors, presence detection is not possible during the
periods of immobility. Moreover, these sensors require direct
line of sight, thus, they cannot detect movements happening
behind the objects in a room (e.g. a partition). In order to de-
crease false negative errors, usually an arbitrary delay period
is introduced. e more being conservative on occupant an-
noyance and discomfort, the longer delay period is selected.
Long delay period reduces energy eciency potential, which
is the primary intention of these systems; thus, the optimum
delay period is required to be determined. Nagy et al. (2015)
examined the accuracy of presence detection with dierent set
delay periods. ey inferred that the optimum delay period for
lighting control is when the accuracy level for presence detec-
tion reaches 95%. e delay period associated with this accu-
racy level lies between 4 and 20min depends on the space and
location of installed sensors. Figure 2 by Gunay et al. (2016)
shows similar approach that can be used to develop algorithms
to determine the optimum delay periods for sensors in their
specic location of installation.
Considering the PIR sensor’s line of sight, their accuracy is
sensitive to where they are installed. Furniture and room ge-
ometry can sometimes disrupt motion detection therefore; it
might be more ecient to use multiple sensors to cover a room
(Tiller et al. 2009). ere are also other limitations to some PIR
sensors on where they can be installed. Some PIR sensors are
incorporated with other devices such as switches and thermo-
stats in order to reduce the costs (Shen, Newsham, and Gunay
2017). Switches are designed to be installed 1m above the oor
while thermostats are supposed to be installed 1.5m above the
oor. e PIR sensors that are built-in parts of these devices
are likely to have suboptimal performance due to positioning
limitations (Shen, Newsham, and Gunay 2017).
ULTRASONIC SENSORS
Ultrasonic sensors are a type of occupancy sensors that similar
to PIR sensors detect motions. Unlike PIR sensors, ultrasonic
sensors are active which means they emit sound waves to their
surrounding environment. When they receive sound waves re-
ected from moving objects, they can perform detection due to
the change in the wavelength (Guo et al. 2010). Unlike PIR sen-
sors, ultrasonic sensors do not need a direct line of sight since ul-
trasonic waves can travel through the obstacles in a room. ese
sensors are more prone to false positive errors compared to PIR
sensors due to their sensitiveness to movement of inanimate ob-
jects such as blowing curtains (Maniccia and Luan 1994).
CO2 SENSORS
Both ultrasonic and PIR sensors can only have a binary output
to indicate the presence of occupant(s). e number of occu-
pants in the space is more oen desired in dierent applications
such as demand-controlled ventilation. An inactive person
would generate CO2 in the rate of 0.3L/min causing the CO2
concentration of a closed space to increase (ASHRAE 2005).
As a result, CO2 sensors’ output have the potential to estimate
the number of occupants in a space by sensing CO2 concentra-
tion (Wang, Burnett, and Chong 1999). e challenge with CO
2
sensors is that detection has some delay (Arora et al. 2015).
For example Arora et al. (2015) reported 30 min delay in their
experimental setup. e delay in observation might be due to
factors such as airtightness of the room, furniture layout and
distance between the occupants and sensors. Such disruptions
diminish the reliability of CO
2
sensors for occupancy detection.
Without considering the complex of dierent variables, it will
be misleading to apply models with assumption of perfectly
mixed indoor air. CO2 sensors can also be used to complement
PIR sensors especially when it is hard to provide a direct line
of sight for them (Lam et al. 2009).
ACOUSTIC SENSORS
Acoustic sensors are passive devices that do not emit energy
but await to detect changes in the energy they receive (Guo
et al. 2010). Unlike ultrasonic sensors that operate with ultra-
sonic waves, acoustic sensors receive energy in form of audible
sounds. ese sensors cannot distinguish between noises gen-
erated by human and other sources therefore have high rate of
false positive errors.
VISION-BASED OCCUPANCY DETECTORS
With the recent advancements in image processing techniques,
the video cameras can extract high-resolution occupancy in-
formation including presence, location, number and types of
activities in very high rates of accuracy (Benezeth et al. 2011).
Despite the usefulness of these systems in buiding service man-
Figure 2. Selection of PIR delay period based on the accuracy of
presence detection – adopted from Gunay et al. (2016).


8-227-19 AZIZI ET AL
1446 ECEEE 2019 SUMMER STUDY
8. BUILDINGS: TECHNOLOGIES AND SYSTEMS …
agement, their application is restricted mainly due to privacy
concerns. e concerns are aggravated when the videos and
information are sent to be processed and stored in a central
server. One solution to resolve the concerns could be to proc-
cess the images locally and send only limited occupancy infor-
mation to the building control system.
CHAIR SENSORS
Labeodan et al. (2016) experimentally evaluated three types
of chair sensors for occupancy detection. e results showed
that mechanical-switch sensors have better performance than
both strain and vibration sensors. Moreover, chair sensors
outperformed PIR sensors in terms of reliability and accuracy
although their durability and long-term reliability is uncertain
(Labeodan et al. 2016).
IMPLICIT OCCUPANCY SENSING
Mel et al. (2011) introduced implicit occupancy sensing as us-
ing existing building infrastructure to detect occupancy despite
they are not originally intended for this purpose. e data from
these systems is usually available for building control purposes.
Some examples of such systems include computer network traf-
c, connection of mobile devices to Wi-Fi, security access cards
and usage of elevators. Some other building systems can be used
to measure occupancy but require some modications. Some
examples of such systems are computer keyboard, mouse, web-
cam and microphone while the level of required modication
diers between them. Since such systems are already available
in the buildings, they usually can provide occupancy data with
no cost or little extra cost. e inferences made from each of
these data sources might be unreliable but the aggregation of
data from dierent sources can increase the accuracy and reli-
ability of occupancy detection (B. Dong and Lam 2011; Tiller et
al. 2009). Mel et al. (2011) categorized the implicit occupancy
sensing methods in three dierent tiers:
• Tier 1 does not require any modication to existing systems
and data is available to process and use.
• Tier 2 requires additional soware to make data accessible.
•
Tier 3 requires additional soware and hardware produce
and provide data to detect occupancy.
ere are various innovative technologies mentioned in the
literature used for implicit occupancy sensing. RFID (radio
frequency identication) is a technology based on electromag-
netic signal detection that is oen used in security access tags
and cards and is capable to determine the occupants’ location
in a cost eective manner (Zhen et al. 2008). Li and Becer-
ik-Gerber (2011) implies despite the potential of RFID-based
indoor location sensing solutions, widespread implementation
is not possible, as the adaptability to each indoor environment
needs to be further justied. Han et al. (2007) used only humid-
ity sensors to examine occupancy and calibrated the data with
size of windows. e results showed that humidity measure-
ments could potentially disclose occupancy information.
Newsham et al. (2017) examined occupancy in single-occu-
pant oces by using the data that can be easily collected from
occupant’s PC in low-cost such as mouse, keyboard and web-
cam. ese data sources are usually t into tier 2 as their data is
ready to use with small modications. e analysis showed such
data sources could be much more accurate than usual ceiling-
based PIR sensors with over 90% accuracy of occupancy detec-
tion. In this case, high accuracy associate with very low rate of
false negative errors enabled to reduce timeout period. Reduced
timeout period can signicantly increase energy savings in
lighting and HVAC systems. Ghai et al. (2012) used only context
sources that are commonly available in commercial buildings
to infer occupancy. ey used so sensor data (implicit sensor
data) such as Wi-Fi access points, area access badges, calendar
and instant messaging clients. e use of these readily available
data lead to 90% accuracy of presence detection.
NETWORK OF SENSORS
Shen, Newsham, and Gunay (2017) count a number of draw-
backs for the conventional single-sensor occupancy detection
approaches:
•
e conventional sensors are expensive and require high
investment costs. ey mentioned recent advances in wire-
less systems have decreased the installation cost although
such systems are not as reliable as wired sensors in terms
of data communication and their need to change batteries
is cumbersome.
•
e conventional sensors oen have low occupant resolu-
tion and do not provide any information on count, identity
and activity of occupants.
•
e conventional sensors are prone to false detection by a
shadow or ash such as headlight from a passing car.
•
Building systems that operate by a single sensor can easily
malfunction if the sensor fails indicating the low reliability
of the system.
e outcome from occupancy sensors usually associate with
uncertainty. Instead of using expensive high-end sensors, a
network of dierent sensors can be an answer to many of the
drawbacks assigned to conventional occupancy detection sys-
tems and enable to achieve more reliable and robust determina-
tion of occupancy. Moreover, with propagation of “Internet of
ings” (IoT), the implicit sensors data are more widely avail-
able that more oen have even more uncertainty than conven-
tional sensors. e challenge would be to develop an analysis
method to infer occupancy information from combination of
data from dierent sensors (Dodier et al. 2006).
ere are several methods in the literature for data fusion
and control strategies in real time. Dodier et al. (2006) applied
analysis models based on Bayesian probability theory to de-
termine occupancy from a network of PIR sensors. ey con-
clude this approach oers signicant benet as compared to
single-point sensing. Hailemariam et al. (2011) in their experi-
mental setup in an oce tested a heterogeneous sensor array.
ey used decision trees to classify the sensors and explored the
relationships between them. e results showed improvements
in accuracy when they used data from multiple PIR sensors.
On the contrary, combining dierent types of sensors worsened
the accuracy of detection when they analysed the data with de-
cision trees. Another study in an oce-type environment in
university premises used ambient sensing data such as lighting,
acoustics, motion and CO2 incorporated into an event-based
pattern detection algorithm (Gaussian Mixture Model). e

8. BUILDINGS: TECHNOLOGIES AND SYSTEMS …
ECEEE SUMMER STUDY PROCEEDINGS 1447
8-227-19 AZIZI ET AL
results showed the experimented occupancy detection system
could count the number of occupants with 83% accuracy al-
though the maximum number of occupants was 4 and accu-
racy might drop with higher occupant trac. Ghai et al. (2012)
used data from context sources such as area access badges and
Wi-Fi access points to measure occupancy. is study applied
machine-learning techniques including regression and classi-
cation to analyse opportunistic data (implicit sensors data) to
infer occupancy and could achieve 90% accuracy. Ekwevugbe,
Brown, and Fan (2012) used indoor climatic variables, indoor
events and energy data to infer occupancy patterns by using a
novel method from articial intelligence (AI) named Adaptive
Neuro-Fuzzy Inference System. ey conclude more reliabil-
ity is possible with this approach as compared to single-sensor
approach. Markov Chain model which is an approach based
on machine-learning was used in several studies (V L Erick-
son, Carreira-Perpiñán, and Cerpa 2011; B. Dong et al. 2010;
Varick L. Erickson and Cerpa 2010). Dong et al. (2010) tested
three dierent machine-learning approaches in an open-plan
oce building to estimate the number of occupants and could
achieve, in average, 73% accuracy by Hidden Markov models.
Energy eciency potential
Energy saving is found to be the primary reason for applying
occupancy sensing systems in most related literature. e en-
ergy eciency potential of occupancy based energy systems
are also dependent on intensity of occupation in the building.
Table 1. Comparison of research papers investigated energy saving potential of demand-controlled energy systems.
Author(s) Sensing
technology
Detection
accuracy
Occupancy
modelling
HVAC
energy
savings
Lighting
energy
savings
Climate Type of
building
Energy
saving
calculation
method
V L Erickson,
Carreira-
Perpiñán, and
Cerpa (2011)
Network
of low
resolution
cameras
80 % Markov chain
approach
42 % _Average of
3 climates
University
building
Computer
simulation-
EnergyPlus
Dong and Lam
(2011)
Network of
environment
sensors
83 % Gaussian
Mixture Model
based Hidden
Markov
Models
18.5 % _Humid
continental
climate-
Pittsburgh
University
building
Computer
simulation-
EnergyPlus
Varick L.
Erickson and
Cerpa (2010)
Sensor
network of
cameras
80 % Moving
Window
Markov Chain
20 % _Local
steppe
climate-
Fresno
NS* Computer
simulation-
EnergyPlus
Newsham et
al. (2017)
Ofce IT
equipment
>90 % Machine
learning-
genetic
programming
16.6–64 % 20–68 % semi-
continental
climate-
Ottawa
mock-
up ofce
environment
proof-of-
concept
demonstration
Floyd, Parker,
and Sherwin
(2002)
PIR sensors NS* _ _ 10–19 % humid
subtropical
climate-
Florida
Ofce
building
Field study
Agarwal et al.
(2010)
PIR sensors NS* _10–15 % _Tropical and
Subtropical
Steppe
Climate-
San Diego
University
ofces
Computer
simulation-
EnergyPlus
Goyal, Ingley,
and Barooah
(2013)
PIR and
ultrasound NS*
_50 % _Humid
Subtropical
Climate-
Gainesville
Ofce
building
Field study
Peng et al.
(2017)
Motion
and indoor
climate
sensors
93 % Machine
learning-
(KNN)
20.3 % _Tropical
Rainforest
Climate-
Singapore
Ofces in
commercial
building
Field study
Dong et al.
(2018)
PIR sensors 70 % Expectation
maximization,
nite state
automata,
uncertain
basis functions
20 % _NS* Ofce Field study
* Not specified

8-227-19 AZIZI ET AL
1448 ECEEE 2019 SUMMER STUDY
8. BUILDINGS: TECHNOLOGIES AND SYSTEMS …
Comparing the irregularly and regularly occupied spaces shows
6–40% dierence in their lighting energy saving potential (Guo
et al. 2010). Climate condition has great inuence on the en-
ergy saving potential associated with applying demand control
HVAC systems (V L Erickson, Carreira-Perpiñán, and Cerpa
2011). Buildings in colder climates are in general more energy
intensive, thus, have more potential to reduce energy use. is
was shown in a study related to university buildings in the
United States based on sensor network occupancy model pre-
dictions. eir investigations in three dierent climates showed
it is possible to achieve, in average, 42% annual energy savings
by using their model for demand controlled HVAC instead of
conventional control (V L Erickson, Carreira-Perpiñán, and
Cerpa 2011). It is worthy of note that baseline control strat-
egy of energy systems has signicant eect on energy saving
potential. Some studies might inate the accuracy and saving
potential by considering overnight periods while vacancy in
commercial buildings is self-evident and attributing energy
saving is irrelevant (Shen, Newsham, and Gunay 2017).
Dong and Lam (2011) deployed a complex environmental
sensor network consist of lighting, acoustics, motion and CO2
sensors in two university buildings. e occupancy output was
fed into a computer model and the comparative simulation
analysis showed 18.5% energy saving compared to conventional
temperature set-point schedule. Another study with similar ap-
proach also found 20% annual energy savings is possible by us-
ing acquired occupancy information in computer simulation of
building (Varick L. Erickson and Cerpa 2010). Newsham et al.
(2017) highlighted the importance of “Internet of ings” and
new emerging data sources that can be used with low cost for
building control and management. A combination of keyboard,
mouse and pixel changes in webcam image showed promising
results for occupancy detection. e enhanced accuracy of the
system compared to conventional PIR sensors could reduce
timeout period from 20 to 5 minutes leading to 25–45% higher
energy saving potential for lighting. Further analysis showed the
possibility to reach up to 64% reduction of HVAC energy use.
Floyd, Parker, and Sherwin (2002) investigated the energy sav-
ings by application of occupancy sensors to lighting control and
found 10–19% energy savings in the commercial buildings. e
summary of abovementioned studies and several other research
projects focused on demand control energy systems and their
entailed energy savings are presented in Table1.
Applications of occupancy sensing
Occupancy detection enables demand-controlled energy sys-
tems that lead to direct energy savings through HVAC and
lighting systems. However, occupancy information can be used
for other applications that may or may not be energy related.
Occupancy information can be used for converting or reusing
spaces by building service designers, for example emergency
evacuation plans (Suter, Petrushevski, and Šipetić 2014). Guo
et al. (2010) mentioned a real- case example of required oc-
cupancy information for reghters to evacuate the elevators
in a building. ey also mentioned the application of occu-
pancy detection in building security and highlighted the use
of network of sensors to enhance the reliability of occupancy
detection to avoid the high cost of false positive errors leading
to false alarm calls to police department.
Occupancy patterns are basic information required for
building design and are important information for building
simulation tools (Crawley et al. 2001). ese simulation tools
are oen used for building energy analysis and design of energy
systems. In order to achieve an optimal design, it is important
to have accurate occupancy pattern for each specic type of
building. Yu et al. (2010) used occupancy information to devel-
op a building energy demand predictive model. Such a model
can lead to accurate prediction of building energy consumption
to improve the energy performance of a building.
Energy use in commercial buildings is strongly related to
their occupancy. Demand response is referred to the eort to
shi the energy peak load and has become more important in
building management due to uctuation of energy prices in dif-
ferent times of a day. Chaney, Hugh Owens, and Peacock (2016)
investigated the inuence of occupancy pattern on applying
demand response and they inferred occupancy information is
important to enable demand response programs. Timm and
Deal (2016) investigated the role of energy information dash-
boards such as occupancy information in changing the energy
behaviours of occupants in a behaviour change campaign. e
occupants behavioural change can lead to signicant energy
savings in commercial buildings without any technical inter-
vention while occupancy information can lead to exploit this
potential (Meier 2006).
With multiple applications of occupancy detection systems,
it will be easier to justify cost of their deployment although dif-
ferent systems must be integrated with each other. Currently in
the existing buildings, even energy systems are not integrated
so that the occupancy data that is used for lighting control can-
not be easily used by HVAC systems without modications.
Implicit data sources are oen described to be easily accessi-
ble and inexpensive to be used to detect occupancy although
there might be hidden costs that are oen not considered by
the researchers. Such data sources are seldom integrated with
the building management systems and required modications
that incur extra expenses. Ghai et al. (2012) mentioned access
control information might be considered sensitive information
and is rarely used for lighting and HVAC control despite being
a promising source of occupancy information. Integration of
dierent data sources with control systems is a major barrier
to use implicit sensing approaches. In future, with increase in
market penetration of wireless sensors, the data would be more
easily accessible through the IoT environment.
Conclusions and areas for further research
Occupancy detection is important approach to tailor the build-
ing services and to improve their eciency. is paper inves-
tigated dierent aspects of occupancy detection in building
management specically with respect to the application in
demand-controlled energy systems. e subjects that are cov-
ered include conventional and emerging occupancy sensors
and methods, implicit occupancy sensing, network of sensors,
multi-sensor data fusion, energy eciency potential of demand-
controlled energy systems and other possible applications for
occupancy sensing systems.
Using a network of dierent sensors is an advantageous ap-
proach to improve the reliability of occupancy detection by us-
ing inexpensive accessible data from implicit sources. ere are

8. BUILDINGS: TECHNOLOGIES AND SYSTEMS …
ECEEE SUMMER STUDY PROCEEDINGS 1449
8-227-19 AZIZI ET AL
many dierent multi-sensor data fusion methods to analyse such
data although there is not enough research to compare these
methods for dierent kinds of sensors and data sources. Further
research is required to investigate the optimal use of each fusion
method in respect to dierent combination of data sources and
contextual factors such as probable occupant trac.
e recent advancements in detection sensors to track the
occupants and their activities would open up new applications
and possibilities. Nevertheless, this level of detailed informa-
tion has already caused concerns about privacy issues. Some
examples of the sensors that caused concerns on privacy in-
clude cameras and webcams with image processing features
and security access cards. Future research should consider such
concerns and try to improve the functionality of such technolo-
gies while mitigating the privacy concerns.
e focus on sensing technologies has currently caused a
discrepancy between development of new sensing technologies
and the requirements of applications in practice (Li and Becerik-
Gerber 2011). Some of emerging hot topics such as building-to-
grid integration and using buildings for demand-response may
not require highly detailed and accurate occupancy detection.
It is important that future research involves more deciplines
related to building industry into this area to identify the value
and functionality of sensing technologies by developing their
applications in the built environment.
References
Agarwal, Yuvraj, Bharathan Balaji, Rajesh Gupta, Jacob Lyles,
Michael Wei, and omas Weng. 2010. “Occupancy-Driv-
en Energy Management for Smart Building Automation.”
In Proceedings of the 2nd ACM Workshop on Embedded Sens-
ing Systems for Energy-Eciency in Building, 1–6. ACM.
Arora, Abhay, Manar Amayri, Venkataramana Badarla,
Stéphane Ploix, and Sanghamitra Bandyopadhyay. 2015.
“Occupancy Esitmation Using Non Intrusive Sensors in
Energy Ecient Buildings.” In Proceedings of the 14th
Conference of International Building Performance Simula-
tion Association, 1441–48. ASHRAE. 2005. “Handbook of
Fundamentals.” ASHRAE Atlanta, GA.
Balaras, Constantinos, Elena Dascalaki, Popi Droutsa, Meletia
Micha, S Kontoyiannidis, and Athanassios Argiriou. 2017.
“Energy Use Intensities for Non-Residential Buildings.”
In ZBORNIK RADOVA PROCEEDINGS. Belgrade, Sava
Center.
Benezeth, Y, H Laurent, B Emile, and C Rosenberger. 2011.
“Towards a Sensor for Detecting Human Presence and
Characterizing Activity.” Energy and Buildings 43 (2):
305–14.
BPIE. 2011. “Europe ’ s Buildings under the Microscope,
Buildings Performance Institute Europe.” Brussel.
Chaney, Joel, Edward Hugh Owens, and Andrew D Peacock.
2016. “An Evidence Based Approach to Determining
Residential Occupancy and Its Role in Demand Response
Management.” Energy & Buildings 125: 254–66.
Crawley, Drury B, Linda K Lawrie, Frederick C Winkelmann,
W F Buhl, Y.Joe Huang, Curtis O Pedersen, Richard K
Strand, et al. 2001. “EnergyPlus: Creating a New-Gener-
ation Building Energy Simulation Program.” Energy and
Buildings 33 (4): 319–31.
D’Agostino, Delia, Paolo Zangheri, and Luca Castellazzi. 2017.
“Towards Nearly Zero Energy Buildings in Europe: A
Focus on Retrot in Non-Residential Buildings.” Energies
Dodier, Robert H, Gregor P Henze, Dale K Tiller, and Xin Guo.
2006. “Building Occupancy Detection through Sensor
Belief Networks.” Energy and Buildings 38 (9): 1033–43.
Dong, Bing, Burton Andrews, Khee Poh Lam, Michael
Höynck, Rui Zhang, Yun-Shang Chiou, and Diego
Benitez. 2010. “An Information Technology Enabled
Sustainability Test-Bed (ITEST) for Occupancy Detection
through an Environmental Sensing Network.” Energy and
Buildings 42 (7): 1038–46.
Dong, Bing, and Khee Poh Lam. 2011. “Building Energy and
Comfort Management through Occupant Behaviour
Pattern Detection Based on a Large-Scale Environmental
Sensor Network.” Journal of Building Performance Simula-
tion 4 (4). Taylor & Francis: 359–69.
Dong, Jin, Christopher Winstead, James Nutaro, and Teja
Kuruganti. 2018. “Occupancy-Based HVAC Control
with Short-Term Occupancy Prediction Algorithms for
Energy-Ecient Buildings.” Energies 11 (9):2427.
EIA. 2019. “Annual Energy Outlook 2019 with Projections to
2050.”
Ekwevugbe, T, N Brown, and D Fan. 2012. “A Design Model
for Building Occupancy Detection Using Sensor Fusion.”
In 2012 6th IEEE International Conference on Digital Eco-
systems and Technologies (DEST), 1–6.
Elie, Azar, and Menassa Carol C. 2012. “Agent-Based Modeling
of Occupants and eir Impact on Energy Use in Commer-
cial Buildings.” Journal of Computing in Civil Engineering 26
(4). American Society of Civil Engineers: 506–18.
Erickson, V L, M Á Carreira-Perpiñán, and A E Cerpa. 2011.
“OBSERVE: Occupancy-Based System for Ecient Re-
duction of HVAC Energy.” In Proceedings of the 10th ACM/
IEEE International Conference on Information Processing
in Sensor Networks, 258–69.
Erickson, Varick L., and Alberto E. Cerpa. 2010. “Occupancy
Based Demand Response HVAC Control Strategy.” In Pro-
ceedings of the 2nd ACM Workshop on Embedded Sensing
Systems for Energy-Eciency in Building – BuildSys ’10, 7
Floyd, D. B., D. S. Parker, and J. R. Sherwin. 2002. “Measured
Field Performance and Energy Savings of Occupancy
Sensors: ree Case Studies.” Florida Solar Energy Center
(Online Publication), no. Nlpip: 97–105.
Ghai, S K, L V anayankizil, D P Seetharam, and D
Chakraborty. 2012. “Occupancy Detection in Commercial
Buildings Using Opportunistic Context Sources.” In 2012
IEEE International Conference on Pervasive Computing
and Communications Workshops, 463–66.
Goyal, Siddharth, Herbert A Ingley, and Prabir Barooah.
2013. “Occupancy-Based Zone-Climate Control for
Energy-Ecient Buildings: Complexity vs. Performance.”
Applied Energy 106: 209–21.
Gunay, H.Burak, O&Amp, Apos, William Brien, Ian Beaus-
oleil-Morrison, Philippe Bisaillon, and Zixiao Shi. 2016.
“Development and Implementation of Control-Oriented
Models for Terminal Heating and Cooling Units.” Energy
& Buildings 121: 78–91.
Gunay, H Burak, Anthony Fuller Fuller, William O’Brien, and
Ian Beausoleil-Morrison. 2016. “Detecting Occupants’

8-227-19 AZIZI ET AL
1450 ECEEE 2019 SUMMER STUDY
8. BUILDINGS: TECHNOLOGIES AND SYSTEMS …
Presence in Oce Spaces: A Case Study.” ESim 2016, no.
October.
Guo, X, Dk Tiller, G P Henze, and C E Waters. 2010. “e
Performance of Occupancy-Based Lighting Control Sys-
tems: A Review.” Lighting Research & Technology. London,
England.
Hailemariam, Ebenezer, Rhys Goldstein, Ramtin Attar, and
Azam Khan. 2011. “Real-Time Occupancy Detection
Using Decision Trees with Multiple Sensor Types.” In Pro-
ceedings of the 2011 Symposium on Simulation for Architec-
ture and Urban Design, 141–48. SimAUD ’11. San Diego,
CA, USA: Society for Computer Simulation International.
Han, Jun, Abhishek Shah, Mark Luk, and Adrian Perrig. 2007.
“Don’t Sweat Your Privacy : Using Humidity to Detect
Human Presence.” In Proceedings of 5th International
Workshop on Privacy in UbiComp, 1–6.
IRC, N R C. 2011. “National Energy Code of Canada for
Buildings.” Ottawa, On: Government of Canada.
Labeodan, Timilehin, Kennedy Aduda, Wim Zeiler, and
Frank Hoving. 2016. “Experimental Evaluation of the
Performance of Chair Sensors in an Oce Space for
Occupancy Detection and Occupancy-Driven Control.”
Energy and Buildings 111: 195–206.
Lam, Khee Poh, Michael Höynck, Bing Dong, Burton An-
drews, Yun-shang Chiou, Diego Benitez, Joonho Choi,
and Robert Bosch Llc. 2009. “OCCUPANCY DETEC-
TION THROUGH AN EXTENSIVE ENVIRONMEN-
TAL SENSOR NETWORK IN AN OPEN-PLAN OFFICE
BUILDING.” In Eleventh International IBPSA Conference,
1452–59. Glasgow, Scotland.
Li, Nan, and Burcin Becerik-Gerber. 2011. “Performance-
Based Evaluation of RFID-Based Indoor Location Sensing
Solutions for the Built Environment.” Advanced Engineer-
ing Informatics 25 (3): 535–46.
Maniccia, D, and X Luan. 1994. “Methods for Assessing the
Maintained and Initial Detection Performance of Oc-
cupancy Sensors.” Journal of the Illuminating Engineering
Society 23 (2). Taylor & Francis: 108–15.
Masoso, O T, and Louis Johannes Grobler. 2010. “e Dark
Side of Occupants’ Behaviour on Building Energy Use.”
Energy and Buildings 42 (2). Elsevier: 173–77.
Meier, Alan. 2006. “Operating Buildings during Temporary
Electricity Shortages.” Energy and Buildings 38 (11):
1296–1301.
Mel, R, B Rosenblum, B Nordman, and K Christensen. 2011.
“Measuring Building Occupancy Using Existing Network
Infrastructure.” In 2011 International Green Computing
Conference and Workshops, 1–8. IEEE.
Nagy, Zoltán, Fah Yik Yong, Mario Frei, and Arno Schlueter.
2015. “Occupant Centered Lighting Control for Comfort
and Energy Ecient Building Operation.” Energy and
Buildings 94: 100–108.
Newsham, Guy R, Henry Xue, Chantal Arsenault, Julio
J Valdes, Greg J Burns, Elizabeth Scarlett, Steven G
Kruithof, and Weiming Shen. 2017. “Testing the Accuracy
of Low-Cost Data Streams for Determining Single-Person
Oce Occupancy and eir Use for Energy Reduction of
Building Services.” Energy and Buildings 135: 137–47.
NRCan. 2015. “Improving Energy Performance in Canada.”
O’Brien, William, and H Burak Gunay. 2014. “e Contextual
Factors Contributing to Occupants’ Adaptive Comfort
Behaviors in Oces – A Review and Proposed Modeling
Framework.” Building and Environment 77: 77–87.
Peng, Yuzhen, Adam Rysanek, Zoltán Nagy, and Arno
Schlüter. 2017. “Occupancy Learning-Based Demand-
Driven Cooling Control for Oce Spaces.” Building and
Environment 122: 145–60.
Pérez-Lombard, Luis, José Ortiz, and Christine Pout. 2008. “A
Review on Buildings Energy Consumption Information.”
Energy & Buildings 40 (3): 394–98.
Saheb, Yamina, Katalin Bódis, Sándor Szabo, Heinz Ossen-
brink, and Strahil Panev. 2015. “Energy Renovation: e
Trump Card for the New Start for Europe.” Luxembourg.
Shen, Weiming, Guy Newsham, and Burak Gunay. 2017.
“Leveraging Existing Occupancy-Related Data for Opti-
mal Control of Commercial Oce Buildings: A Review.”
Advanced Engineering Informatics 33 (C): 230–42.
Sun, B, P B Luh, Q.-S. Jia, Z Jiang, F Wang, and C Song. 2013.
“Building Energy Management: Integrated Control of Ac-
tive and Passive Heating, Cooling, Lighting, Shading, and
Ventilation Systems.” IEEE Transactions on Automation
Science and Engineering 10 (3): 588–602.
Suter, Georg, Filip Petrushevski, and Miloš Šipetić. 2014. “Op-
erations on Network-Based Space Layouts for Modeling
Multiple Space Views of Buildings.” Advanced Engineering
Informatics 28 (4): 395–411.
Tiller, D K, X Guo, G P Henze, and C E Waters. 2009. “e
Application of Sensor Networks to Lighting Control.”
LEUKOS – Journal of Illuminating Engineering Society of
North America 5 (4): 313–25.
Timm, Stephanie N, and Brian M Deal. 2016. “Eective or
Ephemeral? e Role of Energy Information Dashboards
in Changing Occupant Energy Behaviors.” Energy Re-
search & Social Science 19: 11–20.
U.N. 2009. “Buildings and Climate Change: Summary for
Decision-Makers.” United Nations Environmental Pro-
gramme, Sustainable Buildings and Climate Initiative.
Wang, Shengwei, John Burnett, and Hoishing Chong.
1999. “Experimental Validation of CO2-Based Occu-
pancy Detection for Demand-Controlled Ventilation.”
Indoor and Built Environment 8 (6). Basel, Switzerland:
377–91.
Webber, Carrie A, Judy A Roberson, Marla C McWhinney,
Richard E Brown, Margaret J Pinckard, and John F Busch.
2006. “Aer-Hours Power Status of Oce Equipment in
the USA.” Energy 31 (14): 2823–38.
Yu, Zhun, Fariborz Haghighat, Benjamin C M Fung, and Hi-
roshi Yoshino. 2010. “A Decision Tree Method for Build-
ing Energy Demand Modeling.” Energy and Buildings 42
(10): 1637–46.
Zhen, Z, Q Jia, C Song, and X Guan. 2008. “An Indoor Locali-
zation Algorithm for Lighting Control Using RFID.” In
2008 IEEE Energy 2030 Conference, 1–6.
Acknowledgement
e authors gratefully acknowledge the nancial support from
the European Union under RUGGEDISED project.