Journal of Building Engineering 31 (2020) 101342

Contents lists available at ScienceDirect

Journal of Building Engineering

journal homepage: http://www.elsevier.com/locate/jobe

Lighting system control techniques in commercial buildings: Current trends

and future directions
Khairul Rijal Wagiman a, b, Mohd Noor Abdullah b, *, Mohammad Yusri Hassan c,
Nur Hanis Mohammad Radzi b, Ab Halim Abu Bakar d, Tan Chia Kwang d
a
Industrial Training Institute Selandar, Jalan Batang Melaka, 77500, Selandar, Melaka, Malaysia
b
Green and Sustainable Energy (GSEnergy) Focus Group, Faculty of Electrical and Electronic Engineering, Universiti Tun Hussein Onn Malaysia, 86400, Batu Pahat,
Johor, Malaysia
c
Centre of Electrical Energy Systems (CEES), School of Electrical Engineering, Faculty of Engineering, Universiti Teknologi Malaysia (UTM), 81310, Johor Bahru, Johor,
Malaysia
d
UM Power Energy Dedicated Advanced Centre (UMPEDAC), Level 4, Wisma R&D UM, University of Malaya, Jalan Pantai Baharu, 59990, Kuala Lumpur, Malaysia

A R T I C L E I N F O A B S T R A C T

Keywords: Artificial lighting is one of the major electricity consumption in commercial buildings and consumed about 17%
Daylighting of the total electrical energy. Therefore, there is a great potential to reduce energy consumption by implementing
Energy savings intelligent lighting control systems, such as integration of sensor technologies (occupancy and light sensors),
Lighting control system
advanced architectures (wireless- and network-based architectures), and intelligent control techniques (artificial
Lighting system control techniques
Optimization
intelligent and optimization). Moreover, an intelligent control system is capable of enhancing the visual comfort
of occupants, and reduce electricity consumption and greenhouse gas emission. The lighting control system can
be broadly categorized into three main techniques: controller-, optimization-based control, and hybrid. This
paper presents recent and significant state-of-the-art interior lighting system control techniques in commercial
buildings. The review focuses on several key research, including sensing technologies, objective functions and
constraints, techniques, tools, and energy performances. A survey trend analysis is presented graphically and the
findings are discussed extensively. Based on the comprehensive review of lighting control techniques, it is found
that the optimization-based control technique is widely used by 51% compared to other techniques as it has
superior performance to achieve higher energy savings while satisfying visual comfort of occupants, and at the
same time the technique solves multi-objective problems effectively. Moreover, future directions have been
drawn based on the highlighted gaps toward intelligent and sustainable buildings.

in the United States from the residential and commercial sectors was
around 39% of the total energy used [4]. Lighting is one of the building
1. Introduction energy systems (BESs) that accounts a large portion of energy con­
sumption of commercial buildings, for example, in the United States
Nowadays, energy demand has dramatically increased all over the with 17% of the total consumption [5]. For this reason, the European
world. In the United States, electricity energy consumption increases Council has implemented an energy efficiency action plan to reduce
approximately 1.6% every year [1]. This phenomenon is reflected in the energy consumption by 20% by 2020 in European Council countries [6].
number of population, economic growth, and infrastructure develop­ In order to attain lighting system energy efficiency in buildings,
ment. Consequently, the increase of greenhouse gas (GHG) emission has standard organizations such as the European Committee for Standardi­
serious impacts on the global environment (e.g., global warming and zation have produced the standards to guide and provide specifications
climate change). In the building sector, significant energy used globally and requirements about technical systems. The standards related to the
represents one-third of the total energy used with 40% and conse­ lighting system include the European Standard EN12464-1 (i.e., for
quently, carbon dioxide (CO2) emission with 30% [2]. For the European interior design) [7], the European Standard EN12464-2 (i.e., for exterior
Union (EU), building energy consumption contributes 40% of the total design) [8] and the European Standard EN15193 (i.e., for performance
energy and 36% of the total CO2 emission [3]. In 2017, the energy used

* Corresponding author.
E-mail address: [email protected] (M.N. Abdullah).

https://doi.org/10.1016/j.jobe.2020.101342
Received 4 September 2019; Received in revised form 10 February 2020; Accepted 7 March 2020
Available online 9 March 2020
2352-7102/© 2020 Elsevier Ltd. All rights reserved.
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

List of abbreviations IMC internal model control

IOES integer-order extremum seeking
ACO ant colony optimization IoT Internet of Things
ANFIS adaptive neuro-fuzzy inference systems LEED leadership in energy and environment design
AHP analytic hierarchy process LED light emitting diode
ANN artificial neural network LP linear programming
BB branch and bound MAS multi-agent system
BES building energy system MBC model based control
BMS building management system MD mobile device
BREEAM building research establishment environmental MIMO multiple input multiple output
assessment method MISOESN multi-input- single-output extremum seeking with
CC cloud computing Newton
CO convex optimization MOGA multi-objective genetic algorithm
CO2 carbon dioxide MOPSO multi-objective particle swarm optimization
COR competition over resources MSE mean squared error
CV-RMSE coefficient of the variation of the root mean square error NCO nonlinear constrained optimization
DA daylight autonomy NSGA II non dominated sorting genetic algorithm II
DC daylight coefficient PI proportional integral
DF daylight factor PID proportional integral differential
DGP daylight glare probability PIR passive infrared
DIC decentralized integral control PLC programmable logic controller
DMLR discrete-time Markov logistic regression PoE Power over Ethernet
DLP distributed linear programming PSO particle swarm optimization
ELECTRE ELimination Et Choix Traduisant la REalit�e PWM pulse width modulation
EP evolutionary programming R2 coefficient of determination
ES extremum seeking RBFNN radial basis function neural network
EU European Union RFID radio frequency identification
FFNN feedforward neural network RMSE root mean square error
FOES fractional-order extremum seeking SA simulated annealing
FoV field of view SBPSO standard binary particle swarm optimization
GA genetic algorithm SPEA strength Pareto evolutionary algorithm
GEO generalized extremal optimization EU European Union
GHG greenhouse gas VDC direct current voltage
HVAC heating, ventilation and air conditioning VLC visible light communication
IABPSO improved adaptive binary particle swarm optimization WinOSS WiFi-based non-intrusive occupancy sensing system
IBA illumination balancing algorithm

evaluation) [9]. presents the categories of control strategies and schemes of lighting
Lighting is one of the subsystems of BES and consumes a large systems that have been utilized in commercial buildings. Based on the
amount of electrical energy in buildings after heating, ventilation, and figure, yellow color represents control strategies and green color rep­
air conditioning (HVAC). Three main strategies toward energy efficient resents control schemes. A brief discussion of sensing-based lighting
and sustainable buildings [10] are retrofitting existing luminaires with control strategies and schemes are provided in the following sections.
energy efficient luminaires, such as light emitting diodes (LEDs)
[11–16]; reducing design illumination level [17–19]; and implementing
control systems (e.g., occupancy and daylight-linked controls). In terms
of lighting system design, a well-designed system can save energy and
maximize visual comfort to occupants in the buildings. Lamp technol­
ogies are also the contributors for energy efficiency in commercial
buildings, in which LED is a prominent energy efficient lamp with higher
photometric performance (e.g., luminous flux, color rendering index,
and luminous efficiency) and it is easy to control compared to other
lamps [20]. The control of a lighting system is the main contributor of
higher energy savings in buildings and has received high attention
among researchers in the last decade.

1.1. Classification of lighting system control strategies

Due to the rapid growth of sensing technologies, recent lighting

system control has adopted sensing-based strategies as crucial inputs to
improve the performance of lighting systems in terms of energy and
occupants’ visual comfort in a building. Sensing-based lighting system
control strategies can be categorized on the basis of their type of inputs:
(1) occupancy-based and (2) illuminance-based strategies. Fig. 1 Fig. 1. Categories of control strategies and schemes of lighting systems.

2
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

1.1.1. Occupancy-based control strategies controller, adaptive, Q-Learning, and model-based control (MBC).
In occupancy-based control strategies, passive infrared (PIR) sensors In order to improve the energy performance of artificial lighting, a
are commonly used for occupants’ detection due to its simplicity and mixed strategy [31] (i.e., a combination of occupancy- and
cost-effective [21]. Apart from PIR sensors, modern occupancy sensing illuminance-based strategies) has been considered in the literature, e.g.,
technologies have been utilized, such as radio-frequency identification in Refs. [28,32–35]. This strategy has been widely applied as it improves
(RFID) tags [22] and mobile devices (MDs) in WiFi environment [23]. energy efficiency and can capture occupancy patterns and illuminance
The output of a PIR sensor is a binary, i.e., either 0 (unoccupied) or 1 distribution together.
(occupied), and this value is sent to the controller to control artificial Based on the review conducted, the study of lighting system control
lighting based on its control schemes. There two types of control can be divided into three main topics: control functions, architectures,
schemes, namely switch on/off, denoted as OOF, and dimming control, and techniques, as illustrated in Fig. 2. In this article, the dimming
denoted as ODM. The OOF scheme is widely considered in the literature. control techniques of artificial lighting in commercial buildings have
This scheme turns on the lighting when it receives the signal ‘0’ and been discussed extensively, namely controller-, optimization-based, and
turns off the lighting when it receives the signal ‘1’ based on the spec­ hybrid techniques, as highlighted in Fig. 2. Moreover, the control
ified time delay (TD) setting. The most common TD settings are in the function has been highlighted for each reviewed paper for all three
range of 5–20 min [24,25]. However, so far, the shortest time of TD techniques controller-based, optimization-based, and hybrid sections.
considered in the literature is 1 min [26]. Meanwhile, for the ODM
scheme, artificial lighting is dimmed based on the occupants’ condition 1.2. Objective functions and constraints
(i.e., unoccupied and occupied). This scheme is in line with the
EN12464-1 standard, which is for the unoccupied condition, with an In a lighting system control, the main objective is to minimize energy
average illuminance level in a room/space contributed by controlling consumption and satisfy illuminance level preference of occupants in the
the lighting level is 300 lux. Meanwhile, for the occupied state, the light buildings. Energy consumption is linear to the dimming level of lumi­
level should be dimmed to achieve 500 lux of the average illuminance naires with pulse width modulation (PWM) signal-based drive [18,
level on the working plane. Based on our search, it is found that several 36–39]. Indeed, experimental results proved that the dimming level of
published papers reviewed occupancy-based control strategies [21,24, LED luminaires is linear in relation to its output power [40,41]. Thus,
25,27]. In Ref. [21], they reviewed the assessment criteria of the dimming levels of luminaires are considered as the objective func­
occupancy-based control strategies, lighting control design, and their tion for the control and optimization algorithms in most literature. The
performance. A review of sensing technologies, artificial lighting control following equation of the objective function and its constraint has been
strategies, their corresponding performance, and factors affecting per­ widely used:
formance for each control strategy is provided in Ref. [24]. Various
X
K
sensing technologies including PIR, ultrasonic, RFID, microwave, and Min f ðdÞ ¼ dk
others, and also the performance of occupancy-based control strategies k¼1
in terms of energy savings are reviewed in Ref. [25]. Meanwhile, s:t: (1)
occupancy-based lighting control strategies and their performance, as E � Em
well as sensor placement methods are reviewed in Ref. [27].
dkmin � dk � dkmax
1.1.2. Illuminance-based control strategies
Where fðdÞ is the objective function, which is the summation of dimming
In illuminance-based control strategies, light sensors (input devices)
are utilized to capture the illuminance values in a room/space and the levels of all luminaires, dk is the dimming level of kth luminaire/zone, E
values are sent to the controller. The values (input parameter) can be is the average illuminance levels measured on a working plane, dmink and
from either artificial lighting or daylight, or a combination of these (i.e., dmax
k are lower and upper limits of the dimming level of luminaire,
artificial lighting and daylight). The illuminance values from the com­ respectively, and Em is the maintained average illuminance level.
bination of artificial lighting and daylight are considered by the majority Apart from the dimming levels of luminaires as the objective func­
of the researchers as the inputs to the controller. The controller will tion, visual comfort (e.g., illuminance uniformity) of occupants is also
make a decision in controlling artificial lighting. These strategies can be been considered. Several studies have been taken into account, which
grouped into two schemes: switch on/off, denoted as IOF, and dimming combined the two objective functions including energy consumption of
control, denoted as IDM. In the first scheme (IOF), artificial lighting is artificial lighting and visual comfort of occupants [37,42]. Moreover, a
switched off when the illuminance value set-point is achieved by the combination of three objective functions, including energy use, light
light sensor (e.g., 250 [28] and 300 lux [29]). Meanwhile, the lighting quality, and spatial uniformity is also considered in order to simulta­
remains on when the captured illuminance values by the sensors do not neously minimize energy consumption and improve visual comfort [43].
achieve the illuminance value set-point. In the latter scheme (IDM), the The constraints of lighting control can be divided into two types:
dimming levels of luminaires are calculated by the controller based on luminaire-based and visual comfort-based constraints. In luminaire-
the captured illuminance values from the light sensors. The dimming based constraint, the dimming capability of luminaires is considered,
levels vary from 0%‒100% and these levels are determined when the as presented in Eq. (1). The values depend on the types of dimmers/
average illuminance level in a room/space satisfies the set-point value luminaires and the most commonly considered values are in the range of
max
(e.g., 500 lux follows the EN12464-1 standard). In order to compute the 0 (i.e., dmin
k ) to 1 (i.e., dk ). In visual comfort-based constraint, the
optimal dimming levels of luminaires, a well-performed control tech­ illuminance level is widely considered as a constraint in many litera­
nique is required for this purpose. The control techniques of tures. However, glare [44], daylight metrics (e.g., daylight factor and
illuminance-based dimming can be categorized into three main control daylight coefficient) [45], color quality (e.g., RGB color) [43,46,47],
strategies: controller-based, optimization-based, and hybrid control correlated color temperature (CCT) [48–50], color rendering index
strategies. Detailed discussions of all three control strategies are pre­ (CRI), and chromaticity [48,49] are considered to satisfy visual comfort
sented in Sections 2, 3, and 4. To the best of our knowledge, so far, only a and light quality. An extensive review on the measurement of lighting
single paper reviewed illuminance-based dimming control strategies in quality can be found in the work of Kruisselbrink et al. [51]. Apart from
offices, which is done by Ding et al. [30]. However, they only focused on that, others design parameters in lighting systems, such as aesthetic
controller-based strategies, which covered several control techniques, room perception [52] and circadian stimulation [53] are also
namely proportional (P), proportional integral (PI), proportional inte­ considered.
gral derivative (PID), artificial neural network (ANN), fuzzy logic (FL) In the literature, majority of the researchers have considered

3
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

Fig. 2. Diagram of lighting system control.

illuminance level as a constraint function of control and optimization of luminaires (correspond to the number of sensors to be employed) and
algorithms. Apart from illuminance level, illuminance uniformity has the appropriate position of the light sensors.
also been widely used as the parameter that reflects the quality of light
distribution in a room. A higher value means better quality of lighting 1.3. Lighting system control architecture
distribution (i.e., better visual comfort of occupants) and the highest
value is 1. Both parameters are specified in the EN12464-1. For example, Generally, the lighting system control architecture in buildings can
in the office room design, the minimum illuminance level and unifor­ be divided into two categories [58,59]: centralized [17,34,35,38,56,
mity are 500 lux and 0.6, respectively. 59–61] and distributed [34,35,59]. The lighting system technology
Majority of the researchers controlled the dimming level with respect grows rapidly by changing from conventional configurations (i.e., sen­
to a single luminaire (k in Eq. (1) refers to luminaire) [17,18,34,34,37, sors and luminaires are wired connected and hardware-based) to
54–56], which means that each luminaire is provided with a unique network-based systems for controllability, integration, and
level of dimming by the controller. This strategy considers a light sensor cost-effective purposes. Recently, two types of network-based system
co-located at the luminaire and is also known as the multi-sensor architectures have been used [62]: wireless distributed system and
approach. The significant advantage of this strategy is that it does not power-over-ethernet (PoE) centralized system. The selection of the
require any additional power for the sensor as it is integrated with the lighting system control architecture is crucial to determine the lighting
luminaire and at the same time higher energy savings are achieved than system control strategies, especially in terms of controller types and
the mono-sensor approach [26]. The main drawback of this strategy is it control algorithm development.
requires a higher number of light sensors (the number of sensors is equal
to the number of luminaires) and consequently, increases the complexity
of the controller and the cost of installation. In another strategy, the 1.4. Illuminance-based dimming level control techniques
luminaires are grouped into several control zones, where each control
zone has a single light sensor [57]. In this case, k in Eq. (1) refers to the As highlighted in Section 1.1, the control schemes of OOF, ODM, and
control zone. In this strategy, the dimming level produced by the IOF only require a simple control technique, which is turn on/off or dim
controller is shared with several luminaires within a control zone (a to either x or y state (i.e., the state refers to the illuminance level in lux).
group of luminaires). The main merit of this strategy is there is less On the other hand, the IDM scheme requires well-performed control
number of light sensors to be installed, which subsequently reduces the techniques (e.g., artificial intelligence methods) to determine the
cost of installation and the complexity of the controller. However, the optimal dimming levels of luminaires while satisfying occupants’ visual
drawback of this strategy lies in the determination of the group number comfort. Therefore, in this paper, the control techniques particularly
focused on the illuminance-based dimming control scheme (IDM). Based

4
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

on Fig. 2, the control techniques can be categorized into three, namely system control techniques for commercial buildings. The selection of the
controller-based, optimization-based, and hybrid techniques. These papers is based on the parameters considered, control objectives,
categories are based on the concept of techniques in designing the ar­ schemes and techniques, and their performance. The novelty of this
chitecture and computing the dimming levels of luminaires. In the paper is the systematic and extensive review of existing research works
controller-based technique, specific knowledge is required to design the that focus on lighting system control techniques with respect to input
architecture, e.g., proportional integral differential (PID), fuzzy logic parameters, methods, objective function and its constraints of
(FL) controller, and artificial neural network (ANN). Apart from that, an controller/optimization approaches, as well as their performance and
expert knowledge of the whole system operation of lighting is also shaping future directions toward sustainable commercial buildings. This
required to set the particular parameters of the techniques, e.g., PID and paper is structured as follows. Sections 2‒4 thoroughly review the inputs
FL, and a complete system data, e.g., ANN. FL requires the setting of and design parameters, control objectives and techniques, and the per­
fuzzification (i.e., based on inputs), defuzzification (i.e., based on out­ formance of controller-based, optimization-based, and hybrid tech­
puts), and rules (i.e., to make the relationship between the inputs and niques, respectively. Section 5 presents survey trend analysis of the
outputs of the system). ANN, which is the most popular predictive-based literature graphically. Section 6 discusses key findings of the literature
control technique, requires a set of data for training and testing pur­ and draws the future directions toward more sustainable and
poses. The data are essential to make the relationship between the inputs environmentally-friendly buildings. Finally, the conclusion of the study
and outputs in the neural network. It should be noted that in a lighting is presented in Section 7.
system, the inputs are represented by the illuminance values from the
light sensors and the outputs are represented by the dimming levels of 2. Controller-based techniques
luminaires from the controller. In these techniques, the optimal solution
(i.e., optimal dimming level of luminaires) fully depends on the quality In lighting system control and energy management in buildings, the
of the parameter settings (FL) and data (ANN). Moreover, the quality of primary goal is to minimize energy consumption and maximize visual
the design architecture and its parameters are other factors for deter­ comfort metrics of occupants. Generally, there are conventional and
mining the quality of solution (dimming levels), which are applicable for intelligent control systems. Specifically, in artificial lighting control,
ANN. The main advantage of this technique is the fast response to proportional integral (PI) and proportional integral differential (PID) are
determine the dimming levels of luminaires based on the measured the most commonly used methods in conventional controllers. Mean­
illuminance values by the light sensors. However, the significant while, artificial neural network (ANN) and fuzzy logic (FL) have been
drawbacks of this technique are the need of sufficient data for training widely used for intelligent controllers.
and testing purposes, many parameter settings, and the quality of tuning
of the controller parameters, which is time-consuming. Moreover, the 2.1. Conventional controllers
optimal solution (i.e., dimming levels of luminaires) of this strategy fully
depends on the quality of tuning of the controller parameters, where the In general, conventional controllers are closed-loop controls with
tuning is unique than other studies, which depends on the lit environ­ several drawbacks, such as possessing constant parameters, large time
ment and room parameters (e.g., geometry and surface reflection fac­ delay when used alone, and poor control performance [2]. In lighting
tors). A comprehensive review of this technique based on the literature control system applications, the controller has a simple and accurate
is presented in Section 2. tuning of artificial light output; however, it does not contribute to visual
In the optimization-based control, the determination of the dimming performance optimally [59].
levels of luminaires is fully optimized by the technique within the range Meugheuvel et al. [63] developed a PI controller for a lighting con­
of lower and upper limits of the dimming levels of luminaires (i.e., the trol system by considering daylight and occupancy. Standalone and
constraint in Eq. (1)). The significant merit of this technique is it does neighborhood exchangeable information scenarios were considered in
not require any specific knowledge of the whole system operation and their case study. From the results, the PI controller could reach up to
data for training and testing processes compared to the controller-based 10% energy savings compared to linear programming method. Peruffo
strategies. In this technique, the optimal solution (i.e., dimming levels of et al. [64] developed a PI controller by considering wireless light and
luminaires) fully relies on its parameter settings of the technique. For occupancy sensors. The control objective was to provide an adequate
instance, in swarm-based optimization method, the parameters that illuminance level at the workspace and to satisfy the pre-set mean illu­
need to be set are the number of problem dimensions, population size, minance level set by occupants. In addition, the performance of wireless
and the number of generations and specific control parameters of the communication in terms of packet losses was considered and compared
technique (e.g., for PSO: acceleration coefficients, inertia weight, and to wired communication. Based on the results, the performance of the
cognitive and social components). Based on the parameters, in order to wireless communication of PI controller was close to wired communi­
compute the optimal dimming levels of luminaires, the computational cation in controlling the dimming levels of luminaires for both types of
time becomes longer than controller-based techniques. However, the sensors. Caicedo et al. [54] proposed a PI-based controller to control the
setting of the parameters is discussed extensively in the literature. The dimming levels of a wireless lighting system by considering occupancy
examples of the techniques used in the literature are linear program­ and light sensor feedback information. In the study, information for both
ming (LP), convex optimization (CO), and particle swarm optimization light sensors (i.e., on the ceiling and workspace plane) was collected as
(PSO). A thorough review of this technique based on previous studies is the inputs to the central controller. The performance of the proposed
provided in Section 3. method was compared to the information of a single light sensor-driven
Hybrid techniques are a fusion of controller-based and optimization- lighting system on the ceiling plane, known as SIN–CON. From the re­
based control, for example, ANN and non-linear constrained optimiza­ sults, the proposed method provided higher energy savings compared to
tion (NCO). The purpose of this technique is to enhance the performance SIN–CON.
of a single technique. An extensive review of this technique based on Another PID controller was developed to control the dimming levels
published papers is presented in Section 4. of luminaires by considering the electric blind and cooling system for
A comprehensive review of previous works from year 2010 to 2019 energy efficiency and visual and thermal comforts by Soori and Vishwas
in lighting system control techniques is presented in this study. Three [65]. In the study, the PID controller consisted of three control loops: (1)
main techniques have been categorized based on various approaches the position of the blind, (2) luminaires, and (3) the cooling system. The
proposed by previous researchers. From all three techniques and spec­ results showed that the energy consumption decreased and the occu­
ified period of years (i.e., nine years), 37 published papers are consid­ pants’ comfort increased in terms of visual and thermal comfort with the
ered to understand the current research status and the trends in lighting integrated control system consisted of lighting, blind, and cooling

5
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

systems. results, the Ewp-based control strategy attained the least energy con­
sumption of the lighting system compared to other strategies. Similarly,
2.2. Intelligent controllers the annual simulation results showed that the Ewp-based control strategy
achieved the highest energy savings among other strategies but pro­
Recently, numerous researchers are focusing on using intelligent duced a higher glare level in the indoor environment.
controllers for advanced energy and occupants’ comfort due to their Chew et al. [71] developed a microcontroller-based control based on
performance (i.e., promising visual comfort accuracy). Two types of sensor data (i.e., occupancy and light). Based on the analysis, they
intelligent controllers from the literature survey are learning-based proved that the occupancy pattern is a main factor that contributes to
methods (e.g., fuzzy system, neural network, and regression method) higher energy savings of lighting in an indoor space: continuous usage
and model-based methods, such as model-based control (MBC). pattern (45.73%) and discrete usage pattern (62.86%). Moreover, they
Boscarino and Moallem [66] developed a quasi-real time adaptive revealed that the developed lighting control system using
multiple-input-multiple-output (MIMO) controller and a scene simulator microcontroller-based control achieved lower cost than other commer­
by considering daylighting control strategy. In the study, the proposed cial control systems (32%) and at the same time, capable of providing
controller was interacted with a Radiance lighting tool through a scene higher energy savings. The only research that integrated a photovoltaic
simulator in order to control the dimming levels of luminaires in the (PV) system with occupancy and daylight-linked controls was developed
software environment. The developed platform is useful and flexible for by Ozçelik
€ [72]. A self-adaptive algorithm was used and embedded in
any parameter settings and also practical in a simulation-based study. microcontroller-based for controlling lighting system. They revealed
Moreover, it does not require any development of mathematical models that the developed system has significant advantages, which are higher
for a lighting system (i.e., illuminance-based models from artificial energy savings, up to 94% than the baseline (i.e., no controller), and
lighting and daylighting). It should be noted that in the existing method, lower development cost than commercial lighting control systems.
illuminance-based models are required, where the models are derived Xu et al. [29] developed a programmable logic controller
from lighting simulation results using lighting software (e.g., DIALux). (PLC)-based control by considering general and combined (i.e., general
The models are than developed in programming platforms (e.g., MAT­ and task) control strategies. Objective and subjective performance
LAB) and subsequently, integrated to the control system model to obtain evaluations of the proposed method were carried out. In objective
the solutions. evaluation, the illuminance level, uniformity, and luminance were
In another method, Koroglu and Passino [33] proposed an illumi­ considered by conducting the measurement on the working plane.
nation balancing algorithm (IBA) to control the lamp output while Meanwhile, in subjective evaluation, the surveys were conducted on 220
maintaining illuminance uniformity. The experimental comparative persons based on the indicators of pleasure, clarity, proper luminance
results showed that the proposed method achieved higher energy sav­ contrast, color rendering, glare, and general satisfaction level. The
ings and quality of light distribution compared to the decentralized in­ developed a weight factor for each indicator was used the analytic hi­
tegral control (DIC) system. erarchy process (AHP) and expert group synthetic judgment matrix.
In a recent work by Wagiman and Abdullah [57], they developed a From the results of the subjective assessment, the satisfaction ratio of
radial basis function neural network (RBFNN) predictive controller to occupants had a linear relation with the illuminance level, and when the
control the dimming levels of artificial lighting that considered daylight. illuminance level was close to 300 lux (i.e., based on the recommended
The field of view (FoV) of the light sensor was considered in the problem value), the satisfaction ratio was more than 80%. It means that the
formulation. A short period (i.e., three days) with only clear sky in illuminance level of 300 lux is sufficient for occupants’ comfortability in
tropical condition was considered in the training process of RBFNN; an indoor space. Meanwhile, in objective assessment, the proposed
however, the developed method still achieved 34% energy savings and method satisfied the visual comfort metrics and at the same time ach­
the pre-set illuminance level was satisfied. This control strategy ieved up to 60% energy savings. Delvaeye et al. [26] developed a
employed only six light sensors (i.e., for 35 LED luminaires) based on the PLC-based method to control artificial lighting with different control
mathematical function [67,68], instead of the multi-sensor approach strategies. In the case study, the luminaires were divided into three
with 35 light sensors. In another study, Gunay et al. [28] developed an groups based on the row parallel to the window. Three strategies were
approximate discrete-time Markov logistic regression (DMLR) model to considered, for the first strategy (i.e., closed-loop control), each lumi­
control lighting system and blinds. The data were gathered from an naire had its own light sensor and individual dimming levels of lumi­
existing building management system (BMS) (i.e., illuminance levels, naires (i.e., multi-sensor approach). For the second strategy (i.e.,
light switch, and occupancy) and a local weather station (i.e., solar closed-loop control), the sensor was placed at the center of the class­
irradiance). The observations considered in the learning process of room (i.e., mono-sensor approach) and there were two dimming levels
DMLR were switched-on light and occupants’ behavior. The results based on the grouped luminaires. The third strategy (i.e., open loop
indicated that the proposed model achieved energy savings of 25% control) considered three dimming levels of luminaires with
compared to the existing lighting control system. In recent research, Dun outward-facing light sensors. The occupancy sensor was also considered
[69] proposed a fuzzy model reference adaptive system (FMRAS) algo­ for all three classrooms with the specified time delay (i.e., 1 min). The
rithm based on Takagi-Sugeno model to control the LED lighting system results showed that the third strategy achieved the highest energy sav­
by considering daylight. The comparative analysis showed that the ings of 46% compared to other strategies, followed by the second
proposed method performed better than a standard fuzzy controller in strategy (34%) and the first strategy (18%). Control strategy, light
terms of steady-state condition, overshoot, and speed response. sensor placement method, and group of luminaires are main factors that
Only one study used MBC to control integrated lighting (i.e., contribute to better energy performance of daylight-linked control.
dimming levels of luminaires) and shading (i.e., shade position) systems Kumar [73] developed the mapping models between the ceiling and
by considering visual comfort was proposed by Xiong and Tzempelikos a working plane in daylight-adaptive lighting control using a new data
[70]. In the study, the input information of the proposed method was fusion algorithm. Four models are involved (1) artificial light and
collected from exterior (i.e., direct/diffuse solar radiation) and interior daylight (i.e., the working plane and ceiling): separately (2) artificial
(i.e., vertical and horizontal illuminance values) sensors. Apart from the light and daylight (i.e., the vertical plane), separately (3) artificial light
sensor data, the annual daylight/glare data were generated using the (i.e., daylight illuminance at the working plane and ceiling was
simulation-based method for simulation purpose. The three control considered), and daylight (i.e., the vertical plane) separately, and (4)
strategies considered were daylight glare probability (DGP), vertical similar to (3) but with the additional feature of outlier rejection. Ac­
illuminance (Ev), and working plane (Ewp) implemented in offices dur­ cording to the results, the algorithm method (1) outperformed the
ing winter and spring seasons. Based on the real-time control strategies desired illuminance level of other methods. The proposed model can be

6
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

used together with the findings in Ref. [74]. Moreover, the developed diagnose faults of LED luminaires. The FL was chosen to achieve illu­
model is useful when a lighting control system fully depends on the mination comfort by considering occupancy, artificial lighting, and
illuminance levels of the working plane. natural lighting information feedback and to control the health of lu­
Pandharipande et al. [75] proposed an integral multivariable minaires. Meanwhile, the PID was used to control the output levels of
controller to control the dimming levels of luminaires based on occu­ LEDs. Two input (i.e., ambient light and occupants’ motion) and output
pancy model. The model involved occupants’ information (i.e., illumi­ (i.e., LED current reference and fan power) variables were considered in
nance values, occupants’ position, and energy consumption pattern) and the FL. The proposed method demonstrated 50% energy savings and the
stored in database using the wireless network of the lighting control fault diagnosis of LED luminaires was achieved effectively.
system based on occupants’ MDs. In the case study, each luminaire was A recent study by Kandasamy et al. [32] proposed a hybrid method of
co-located with light and PIR motion sensors and anchor radio for ANN and internal model control (IMC) to control illuminance level on
device-centric positioning. It can be deduced that the developed model the working plane and satisfy occupants’ comfort. The outputs of the
has significant advantages in providing great performance of the control sensors and artificial lights were used as the training data for ANN. The
system in terms of the accuracy of the occupancy level (49.5%) and proposed ANN-IMC method was analyzed based on the variations of the
higher energy savings (50%). desired illuminance levels (i.e., 250 to 500 lux) and considered daylight
and no daylight scenarios. The results showed that the energy savings
with full daylight adaptive for the desired illuminance levels of 300 and
2.3. Hybrid controllers 500 lux were 54% and 40%, respectively. Moreover, the method per­
formed better in terms of robustness when dealing with the variations of
In order to improve the performance of a single method, hybrid the desired illuminance levels.
methods have been used in recent studies. Liu et al. [76] developed a The summary of related works that focused on controller-based
hybrid of FL and PI controllers to control the dimming levels of LED techniques is shown in Table 1.
luminaires subject to illuminance settings and LED luminaires dimming
limits by considering daylight and occupants’ information. The PI
controller was used to control the output levels of luminaires, whereas 2.4. Energy performance of controller-based techniques
the FL was implemented to achieve the pre-set illuminance level pref­
erence. The FL consisted of two inputs (i.e., daylight illuminance and Energy saving is the main indicator to evaluate energy performance
artificial lighting ranks) and an output (i.e., visual comfort rank). In the in most literature. As discussed in Section 1.1, the sensing-based lighting
results, both simulation-based and experimental approaches were re­ control is divided into two control strategies and each strategy is divided
ported, in which the proposed method in simulation-based approach into another two schemes. Thus, the total of four control schemes are
achieved energy savings of 30%, whereas in the experimental approach, considered to evaluate energy performance as follows: (i) occupancy-
the energy savings was 57%. In another hybrid method, the FL-PID was based on/off control scheme (OOF), (ii) occupancy-based dimming
proposed by Cimini et al. [77], to control the dimming levels and control scheme (ODM), (iii) illuminance-based on/off control scheme

Table 1
Summary of related works that focused on controller-based techniques.
Reference and Objective Input Design parameter Method Simulation tool/hardware Case studyh
year function parameter

ECa VCb Dc Od Eave Uof GIg

[32] 2018 ✓ – ✓ ✓ ✓ – – ANN-IMC DIALux and MATLAB LaboratorySE

[57] 2018 ✓ – ✓ – ✓ – – ANN DIALux and MATLAB OfficeS
[69] 2018 ✓ – ✓ – ✓ – – FMRAS MATLAB and ZigBee controller ClassroomR
[72] 2017 ✓ – ✓ – ✓ – – Self-adaptive algorithm Microcontroller ClassroomE
[29] 2017 ✓ √ (U) ✓ ✓ ✓ ✓ – AHP and expert group synthetic Ecotect, Daysim, MATLAB, and PLC OfficeSR
judgment matrix
[28] 2016 ✓ √ (G) ✓ ✓ ✓ – ✓ Approximate discrete-time Markov Daysim and MATLAB OfficeR
logistic regression (ADMLS)
[26] 2016 ✓ – ✓ ✓ ✓ – – Not mentioned DALI control system þ PLC ClassroomR
[70] 2016 ✓ √ (E & ✓ – ✓ – ✓ MBC MATLAB and LabVIEW OfficeSR
G)
[71] 2016 ✓ – ✓ ✓ ✓ – – Not mentioned Micro-controller Laboratory and
classroomER
[76] 2016 ✓ – ✓ ✓ ✓ – – FL-PI DIALux and Microcontroller OfficeR
[66] 2016 ✓ – ✓ – ✓ – – Adaptive MIMO Self-tuning multivariable controller, OfficeS
Radiance, API and REST
[54] 2016 ✓ – ✓ ✓ ✓ – – PI Not mentioned OfficeR
[77] 2015 ✓ – ✓ ✓ ✓ – – FL-PID Microcontroller IndustrialE
[64] 2015 ✓ – ✓ ✓ ✓ – – PI DIALux OfficeS
[75] 2014 ✓ – ✓ ✓ ✓ – – Integral multivariable controller DIALux OfficeS
[63] 2014 ✓ – ✓ ✓ ✓ – – PI DIALux OfficeS
[33] 2014 ✓ – ✓ – ✓ – – IBA Microcontroller OfficeE
[65] 2013 ✓ – ✓ ✓ ✓ – – PID DIALux and FX builder OfficeR

Note.
a
Energy consumption.
b
Visual comfort. E is illuminance level, U is illuminance uniformity, and G is glare.
c
Daylight.
d
Occupancy.
e
Average illuminance level.
f
Illuminance uniformity.
g
Glare index.
h
S is simulation, E is experimental, and R is real application.

7
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

(IOF), and (iv) illuminance-based dimming control scheme (IDM).

Minimize fi ðxÞ ði ¼ 1; 2; :::; MÞ
The summary of energy savings reported from previous studies is
s:t:
presented in Table 2. As can be seen from the table, for the single-control (2)
hj ðxÞ � hmin ðj ¼ 1; 2; :::; JÞ
scheme OOF, only a single study reported the energy savings of 30%, gmin � gk ðxÞ � gmax ðk ¼ 1; 2; :::; KÞ
whereas no control scheme was adopted for ODM. For schemes IOF and
IDM, the highest energy savings reported was 25% and 66%, respec­ where fi ðxÞ is the objective functions to be minimized, x ¼
tively. Meanwhile, for the combined control schemes, only one scheme ½x1 ; x2 ; :::; xn �T is the control variables vector, hj ðxÞis the inequality
was considered (i.e., combination of OOF and IDM), and the highest constraint functions, andgk ðxÞare the bounds. If i ¼ 1, the optimization
energy savings of 72% was recorded. problem becomes a single objective, whereas if i > 1, the optimization
problem is known as a multi-objective problem. In lighting system
2.5. Summary optimization, the objective function fi ðxÞ can be the energy consumption
of lighting systems, the visual comfort of occupants, and light quality.
Controller-based techniques can be divided into three groups of For example, in case of lighting design, Mandal et al. [78] considered
methods: conventional, intelligent, and hybrid. PI and PID methods (i.e., energy consumption, visual comfort (i.e., illuminance level, uniformity,
conventional controllers) are widely utilized for controlling artificial and glare), and cost of luminaires as the objective functions. Meanwhile,
lighting due to their simple design. In the last decade, intelligent con­ in the case of lighting control, a summation of dimming levels of lumi­
trollers, such as ANN, FL, and MBC are preferred by researchers due to naires has been considered as a prominent objective function in most
their robustness in solving indoor lighting systems. The combinations of literature as presented in Eq. (1) as it reflects the energy consumption of
conventional and intelligent controllers (i.e., called hybrid controllers) artificial lighting (as discussed in Section 1.1), and the application of this
are also applied in interior lighting systems for improving energy per­ objective function is discussed in the following sub-sections. Meanwhile,
formance and visual comfort of occupants, such as the methods of FL-PI, hj ðxÞcan be the illuminance level and uniformity, and gk ðxÞcan be the
FL-PID, and ANN-IMC. Based on energy performance from literature dimming level limits of luminaires.
review, the highest energy savings of 72% were achieved by considering In some situations, more than one objective functions are involved
the mixed scheme (OOF and IDM) by using self-adaptive algorithm and need to be optimized simultaneously. To solve this situation, the
(control method) and integrated with a microcontroller (controller) and multi-objective optimization problem can be used. In multi-objective
a PV system. optimization, there is no single optimal solution for all objectives, but
there is a set of solutions. In order to find the optimal solution of the
3. Optimization-based control techniques problem objectives, the trade-off between two or more conflicting ob­
jectives needs to be evaluated and it is called Pareto optimal solution.
Optimization is the action to obtain the best or most effective situ­
ation or performance under a given constraint. The main objective of 3.1. Single-objective applications
optimization is either to minimize or maximize results or decisions.
Optimization techniques can be divided into conventional and modern As mentioned in Section 1, for single-objective optimization, most
(meta-heuristic) techniques. Examples of conventional techniques are literature considered minimizing energy consumption of artificial
linear programming, dynamic programming, and non-linear program­ lighting as the objective function.
ming. Recently, modern methods are widely used in the engineering
field due to their performances, which are powerful to solve complex 3.1.1. Conventional methods
and non-linear problems. Meta-heuristic methods are inspired by be­ Interior lighting design in buildings is crucial to ensure the lighting
haviors, reactions, and communication mechanisms in nature. The system energy consumption is always at a minimum level. According to
methods include genetic algorithm (GA), ant colony optimization the EN12464-1, the main parameters related to visual comfort for
(ACO), and particle swarm optimization (PSO). An optimization prob­ consideration are illuminance level and uniformity for different work
lem can be expressed as follows: areas.

Table 2
Summary of energy savings for controller-based techniques.
Reference Room type and design illuminance level (lux) Type of study Method/controller/software Control schemed Energy savings (%)

Sa Eb Rc

[28] Office ‒ 250 ✓ Approximate discrete-time Markov logistic regression IOF 25

[29] Office ‒ 300 ✓ PLC OOF 30
IOF 25
OOF þ IDM 60
[72] Classroom ‒ 300 ✓ Self-adaptive algorithm and microcontroller þ PV OOF þ IDM 72
[76] Office ‒ 300 ✓ FL-PI and microcontroller OOF þ IDM 30
[66] Office ‒ 300 ✓ Self-tuning multivariable controller and Radiance IDM 66
[71] Laboratory and classroom ‒ 1200 ✓ Microcontroller IDM 55
[26] Classroom ‒ 500 ✓ DALI control system þ PLC OOF þ IDM 18–46
[73] Office ‒ 500 ✓ Data fusion algorithm IDM 60
[75] Office ‒ 500 ✓ Integral multivariable controller OOF þ IDM 50
[33] Room ‒ 300 ✓ IBA and microcontroller IDM 50
[57] Office ‒ 500 ✓ ANN IDM 34
[77] Industrial ✓ FL-PID and microcontroller OOF þ IDM 50
[32] Laboratory ‒ 300 Laboratory ‒ 500 ✓ ✓ ANN-IMC IDM 54 40

Note.
a
Simulation.
b
Experimental.
c
Real.
d
OOF is occupancy-based on/off control scheme, and IDM is illuminance -based dimming control scheme.

8
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

Convex optimization (CO) is where the objective and all constraint sensor at the ceiling and prior knowledge of sensor calibration. In this
functions are convex. Interior point method is one of the methods to research, daylight and no daylight cases were considered. Based on the
solve CO [79]. Tan et al. [61] proposed CO in searching for the optimal results, the proposed method performed better in achieving the
dimming levels of artificial lighting by considering occupants’ personal preferred illuminance level and higher energy savings when the light
preference using MDs. In the research, a centralized-based control was sensor was positioned on the working plane compared to the sensor
utilized by considering visible light communication (VLC) as a positioned on the ceiling.
communication protocol. Three MDs information required by the central Caicedo et al. [18] also proposed LP-based optimization to minimize
controller were neighboring luminaires, image, and illuminance pref­ the dimming levels of LEDs subject to pre-set illumination uniformity by
erence to compute the dimming levels of luminaires. Based on the re­ considering occupied and unoccupied conditions. Two values of beam
sults, the proposed method achieved higher detection accuracy (>80%) width of the LED lamp were considered: 10.5� and 60� . The results
for different dimming levels and frequencies, and satisfied the illumi­ showed that the proposed method with the beam width of 60� achieved
nance level set point. However, the proposed control fully depends on higher energy savings. Pandharipande and Caicedo [59] proposed
the occupants’ feedback to control lighting and this is an additional task LP-based optimization to optimize the dimming levels of artificial
for occupants in controlling the system, which is another challenge to be lighting. The proposed method was compared to the PI controller to
implemented in the future. Rossi et al. [38] proposed a new optimization evaluate the energy savings and average illuminance level for occupied
formulation that combined light sensor illuminance error and squared and unoccupied scenarios. Furthermore, two sensing-based strategies (i.
energy consumption, and solved using CO. Centralize-based control was e., occupancy and daylight) were also considered and both sensors were
considered together with occupancy and light sensors-driven. Based on co-located at the luminaires. The LP-based optimization approach ach­
the comparative results, the proposed method showed superior perfor­ ieved closer average illuminance level set point compared to the PI
mance than the standalone controller [63] in terms of no overshoot, controller. Meanwhile, the PI controller achieved higher energy savings
lower settling time, and higher energy savings. Moreover, both pre-set compared to the LP-based optimization approach.
illuminance values on the ceiling and working plane were achieved. Caicedo and Pandharipande [55] proposed a distributed linear pro­
However, the illuminance uniformity should be considered as a design gramming (DLP) algorithm to minimize energy consumption of a wire­
parameter for better visual comfort of occupants in the room. less network lighting system by considering distributed-based control.
Linear programming (LP) is an exact solution-based optimization The objective function was to minimize the dimming levels of luminaires
that deals with linear problems and its constraints. However, LP suffers by considering neighborhood distributed control information and to
from higher computational time when dealing with more variables and satisfy the constraints of neighborhood communication limit and pre-set
larger systems. LP can be solved using well-known techniques, such as illuminance level. The performance of the proposed method was
simplex and interior point methods [79]. Borile et al. [36] proposed an compared to the LP centralized-based control, in which the former
LP approach to find the optimal dimming levels of artificial lighting achieved energy savings within the limit of the LP centralized-based
based on the developed light sensor illuminance values estimation of a control. Din and Kim [44] proposed the LP algorithm to optimize en­
daylight model (i.e., on the ceiling). The developed model considered ergy consumption and satisfy occupants’ comforts (i.e., illuminance
the light sensors data (i.e., measured on the ceiling and the working preference and glare) for integrated artificial lighting and blind systems.
plane based on two different illuminance sources, i.e., artificial lighting The control strategies involved were artificial lighting (i.e., dimming
and daylighting). The data were then used as the training data for three level control) and blind (i.e., tilt angle control) systems. Based on the
different types of least squares estimators. The aim of the estimator was simulation results, the proposed method achieved up to 35% energy
to estimate the illuminance distribution from daylight at both ceiling savings compared to the system without optimized control.
and working plane in the room. In the study, the control strategies Another optimization-based method utilized is iterative method.
considered based on the data driven were the ceiling and working plane. Caicedo and Pandharipande [17,34] proposed an iterative method with
According to the performance analysis results, the ceiling-based control the aim of minimizing energy consumption by optimizing the dimming
excelled in achieving the desired illuminance level and higher energy levels of artificial lighting. The constraints considered were the pre-set
savings compared to the working plane-based control. However, the illuminance level, dimming level limit of luminaires, and spatial illu­
proposed method is time-consuming for training purpose as it requires a minance uniformity. The method proposed by Refs. [34] was compared
large amount of data. Wen and Agogino [80] proposed LP to optimize to another method that was not considered as spatial uniformity, known
the dimming levels of wireless network artificial lighting based on the as REF-SOL. The proposed method achieved better spatial uniformity of
information of daylight harvesting and occupancy. The proposed luminaires than REF-SOL but contributed to lower energy savings of 1%
approach was compared to a typical lighting system as the baseline, than REF-SOL. In the work by Caicedo and Pandharipande [17], the
which was a wired lighting system with on/off switches. According to proposed method was compared to the LP approach (daylight illumi­
the results, the proposed approach achieved significant energy savings nance information was required), and it showed that the proposed
of 45%–65% for various scenarios compared to the baseline. In another method was more robust in achieving the illuminance level preference
work, Wen and Agogino [60] developed a prototype of a wireless and smooth to change the dimming levels of luminaires compared to LP.
network control for an artificial lighting system. The types of develop­ Moreover, the proposed method did not require the adaptation of illu­
ment considered were prototypes of a wireless dimmer ballast actuator, minance from daylight. Nevertheless, the proposed method achieved
a wireless photosensor, and an LP-based control. The aim of LP was to lower energy savings than LP by 5%. In recent research, Zou et al. [23]
minimize the dimming levels of luminaires and satisfy the illuminance developed an occupancy adaptive mobile application, known as Win­
level desired. The proposed method achieved significant energy savings Light, to control artificial light by taking advantage of the existing WiFi
of 60.8%. Pandharipande and Caicedo [35] proposed the LP algorithm infrastructure. Moreover, the developed application also allowed MD
to find the optimum dimming levels of LED luminaires by considering users to control the brightness level of light of the nearby luminaires
the pre-set illuminance level (i.e., 500 lux for occupied and 300 lux for based on their personal preferences and zones. Such control used branch
unoccupied states) and uniformity (i.e., 0.3) as the constraints. The and bound (BB) algorithm in order to find the optimum level of
proposed algorithm was developed based on centralized and distributed brightness of the light. The occupancy data were driven using
architectures. From the results, the distributed-based control out­ WiFi-based non-intrusive occupancy sensing system (WinOSS). The test
performed the centralized-based control in terms of computational bed in the study was also equipped with PIR sensors. WinLight installed
effort. However, the optimum solution for both control methods was on an occupant’s MD could communicate mutually with the system
similar. Caicedo et al. [56] proposed LP optimization to determine the server. Thus, the brightness level of the nearby luminaires could be
optimum dimming levels of artificial lighting by considering a light controlled in the real-time application based on their illuminance level

9
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

preference. They compared the performance of the developed system performed better in terms of the optimal feasible trade-offs and simul­
with PIR sensors, in which WinLight achieved an average of 98.85% taneously satisfied the visual comfort preferences. So far, only this work
detection accuracy and energy savings of 93% compared to PIR sensors. considered psychovisual as an objective function in optimization prob­
lem formulation to represent occupants’ visual comfort in lighting
3.1.2. Meta-heuristic methods systems.
In BES optimization, meta-heuristic methods are extensively utilized, Another method, a generalized extremal optimization (GEO) algo­
including PSO [81,82] and GA [83,84], as well as other methods, which rithm used to find the optimal location and output power of the artificial
are firefly algorithm [85], evolutionary programming (EP), and ACO lighting subject to satisfy illuminance preference on the working plane
[86]. Indeed, PSO [87] and GA [88] methods were also successfully was proposed by Cassol et al. [95]. In the study, the two steps considered
performed for outdoor lighting systems. These techniques are popular to calculate the objective functions were (1) the least-square of the de­
optimization methods among researchers due to their abilities to find viations between the specified illuminance and measured illuminance
good solutions with less computational effort. on the working plane and (2) the total power of luminaires to provide
Si et al. [89] proposed PSO-based optimization to maximize the en­ luminous flux in (1). Three scenarios were considered based on different
ergy savings of wireless network artificial lighting by introducing light constraints added: (i) luminaire position was fixed (i.e., symmetry
penalty function to improve illuminance levels. The proposed method arrangement), (ii) 18 luminaire positions with all the luminaires were at
achieved up to 45% energy savings than the baseline (i.e., no control the same level of output power, and (iii) 18 luminaire positions with
system) and lower dimming levels of luminaires than the existing con­ nine different levels of output power. The experiment involved nine
trol system. However, the exact value of energy savings of the proposed luminaires and the results showed that the optimum output power of
method compared to the existing control system is missing. Another luminaires was recorded for scenario (iii). Moreover, the proposed
method, a modified competition over resources (COR) algorithm for method illustrated superior performance in terms of robustness and
optimizing energy savings of artificial lighting based on the sensor in­ accuracy to find the optimal solution with multiple constraints. In the
formation feedback of the targeted areas was proposed by Mendes et al. work of Afshari et al. [43], the authors considered energy consumption,
[90]. The experimental setup consisted of three components: spatial uniformity, and light as the objective functions for controlling a
microcontroller-based, simulation platform (LabVIEW), and a lighting lighting system. The objective functions were converted into a single
system prototype. The results showed that the modified COR out­ objective and solved by using gradient algorithm. The proposed method
performed the original COR with energy savings of 6.67% and lower achieved significant energy savings and satisfied lighting quality.
computational cost. Table 3 shows the summary of related works that focused on
optimization-based control techniques in lighting systems.
3.2. Multi-objective applications
3.3. Energy performance of optimization-based control techniques
Multi-objective optimization has been widely applied in engineering
fields, especially in electrical and energy systems. As an example, Cui In optimized-based control strategy, the energy performance evalu­
et al. [91] revealed that multi-objective optimization techniques, ation is similar as in Section 2.4. All four control schemes are considered
including multi-objective particle swarm optimization (MOPSO), in these techniques. The summary of energy savings reported from this
multi-objective genetic algorithm (MOGA), and non-dominated sorting strategy is illustrated in Table 4. From the table, for the single control
genetic algorithm II (NSGA II) have been successfully applied in a wide scheme, scheme OOF alone demonstrated the highest energy savings of
range of energy related system applications. Various techniques of 93% and IDM alone demonstrated the highest energy savings of 45%.
multi-objective optimization are commonly used in BES, such as MOPSO Meanwhile, for the combined control schemes, the combination of
[92] and NSGA II [93], as well as other methods, such as MOGA and schemes OOF and IDM recorded the highest energy savings of 66%. For
strength Pareto evolutionary algorithm (SPEA) [86]. the combination of schemes ODM and IDM, the highest energy savings
In the lighting system optimization context, according to the recorded was 54%. Lastly, for schemes IOF and IDM, the highest energy
EN12464-1, illuminance level and uniformity are the main parameters savings was 46%.
required to be considered in interior lighting system design. In order to
comply with the standard, multi-objective optimization is the solution 3.4. Summary
for solving two parameters simultaneously. Based on recent literature,
NSGA II is prominently utilized in the lighting system control when In optimization-based control strategies, the categories of this tech­
dealing with multi-objective problems for exterior [94] and interior [37, nique are based on the objectives, which are single- and multi-
42] of buildings. Madias et al. [37] utilized NSGA II to find the optimal objectives. The single-objective strategy can be divided into two types
energy consumption of artificial lighting by considering the objectives of of methods: conventional and meta-heuristic. Conventional methods are
minimizing the dimming levels of LED luminaires and maximizing widely used (specifically LP and CO) as the optimization engine in
illuminance uniformity by satisfying the pre-set illuminance level and interior lighting control. Meanwhile, several meta-heuristic methods
dimming limit level. In the study, the second objective function was have been applied, i.e., PSO and modified COR. In order to solve multi-
formulated based on the LED array theory. Based on the theory, the objective lighting system control problems, multi-objective of meta-
minimum value of the coefficient of the variation of root mean square heuristic methods is utilized. The most common objective functions in
error (CV-RMSE) will be reflected to obtain the maximization of illu­ lighting optimization-based control are the dimming levels of luminaires
minance uniformity on the working plane. The simulation results and occupants’ visual comfort. These functions are solved based on two
showed that the proposed method contributed to significant energy approaches, i.e., weighted sum method (i.e., both objectives are opti­
savings of 22% and lower computational time. However, the analysis mized individually) and Pareto-based solution (i.e., both objectives are
could have been improved by including indoor illuminance distribution optimized simultaneously). NSGA II is a popular approach among other
from daylight to increase energy savings. Villa and Labayrade [42] techniques as it can produce Pareto-based solution and also due to its
proposed NSGA II to optimize artificial lighting (i.e., ceiling and task robustness. In terms of energy performance, 93% energy savings was
lightings) in an individual office room by considering the energy de­ achieved by considering the OOF control scheme and by utilizing BB
mand and users’ psychovisual factor as the objective functions. The technique and WinLight controller platform.
users’ psychovisual data were obtained from experiment conducted
using integration of a virtual reality stereoscopic display device, a room
modeler and a lighting simulation tools. The proposed method

10
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

Table 3
Summary of related works that focused on optimization-based control techniques.
Reference and Objective function Input Design parameter Method Simulation tool/hardware Case studyi
year parameter

ECa VCb Otherc Dd Oe Eavf Uog GIh

[23] 2018 ✓ – – – ✓ ✓ – – BB CPLEX and DALI controller Offices, laboratories and living
roomR
[90] 2017 ✓ – – ✓ – ✓ – – Modified COR LabVIEW and OfficeE
microcontroller
[61] 2017 ✓ – – ✓ ✓ ✓ – – CO Not mentioned OfficeSR
[36] 2017 ✓ – – ✓ – ✓ – – LP DIALux and MATLAB OfficeS
[37] 2016 ✓ √ (U) – – – ✓ ✓ – NSGA II Relux and MATLAB OfficeS
[34] 2015 ✓ – – ✓ ✓ ✓ – – Iterative method DIALux OfficeS
[17] 2015 ✓ – – ✓ ✓ ✓ – – Iterative method DIALux OfficeS
[38] 2015 ✓ – – ✓ ✓ ✓ – – CO DIALux OfficeS
[59] 2015 ✓ – – ✓ ✓ ✓ – – LP DIALux OfficeS
[56] 2014 ✓ – – ✓ ✓ ✓ – – LP DIALux OfficeS
[43] 2014 ✓ √ (U & – ✓ – ✓ ✓ Gradient Robot Raconteur and RoomR
LQ) algorithm MATLAB
[55] 2013 ✓ – – ✓ ✓ ✓ – – DLP DIALux OfficeS
[35] 2013 ✓ – – ✓ ✓ ✓ ✓ – LP DIALux OfficeS
[42] 2013 ✓ √ (PsV) – – – ✓ ✓ – NSGA II 3DS MAX and MATLAB OfficeS
[44] 2013 ✓ – – ✓ – ✓ – ✓ LP Not mentioned RoomS
[80] 2011 ✓ – – ✓ – ✓ – – LP Radiance OfficeSR
[60] 2011 ✓ – – ✓ ✓ ✓ – – LP Radiance OfficeE
[18] 2011 ✓ – – ✓ – ✓ – – LP Not mentioned OfficeS
[89] 2011 ✓ – – ✓ – ✓ – – PSO Not mentioned OfficeS
[95] 2011 ✓ – √ – – ✓ – – GEO Radiance OfficeS
(Lum)
a
Energy consumption.
b
Visual comfort. U is illuminance uniformity, LQ is light quality, and PsV is psychovisual.
c
Lum is luminaire positions.
d
Occupancy.
e
Daylight.
f
Average illuminance level.
g
Illuminance uniformity.
h
Glare index.
i
S is simulation, E is experimental, and R is real application.

Table 4
Summary of energy savings for optimization-based control techniques.
Reference Room type and design illuminance level (lux) Type of study Method/controller/software Control schemef Energy savings (%)

Sa Eb Rc

[60] Office - 350 ✓ ✓ LP and DALI controller, and Radiance OOF þ IDM 60
[18] Office – 500/300 ✓ LP ODM þ IDM 33
[80] Office 300 ✓ LP IOF þ IDM 35-46 45
IDM
[17] Office – 500 & 300 ✓ LP ODM þ IDM 48–54
[34] Office ‒ 500 ✓ Iterative method ODM þ IDM 25
[23] Offices, laboratories and living room ✓ ✓ BB þ DALI controller, and WinLight OOF 80–93
[89] Office ‒ 800 ✓ PSO IDM 30–45
[37] Office ‒ 500 ✓ NSGA II IDM 22

Note.
a
Simulation.
b
Experimental.
c
Real.
f
OOF is occupancy-based on/off control scheme, ODM is occupancy-based dimming control scheme, IOF is illuminance-based on/off control scheme, and IDM is
illuminance-based dimming control scheme.

4. Hybrid techniques 4.1. Control techniques

Hybrid techniques consider the hybridization of controller-based (e. In a recent study, Yin et al. [96] developed a combination of PID and
g., PID and ANN) and optimization-based control techniques (e.g., fractional-order extremum seeking (FOES) to control the dimming levels
extremum seeking, PSO, and non-linear constrained optimization) to of luminaires and minimize energy consumption. The PID was chosen to
enhance the performance of a single technique. A comprehensive review control the dimming of a group of artificial lighting and also to satisfy
of hybrid techniques is presented in the following section. the illuminance level set-point. The experiment was conducted using
Arduino microcontroller, light sensors, groups of luminaires, and
MATLAB-Simulink. The proposed method was compared to
integer-order extremum seeking (IOES). The PID-FOES performed better

11
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

than the PID-IOES in terms of energy savings, convergence rate, and 4.3. Summary
accuracy. An optimal control using a PID controller and multi-input
single-output extremum seeking control with Newton algorithm Based on literature survey, only four hybrid techniques have been
(MISOESN) was proposed by Yin et al. [97]. The PID was utilized to applied in research works: PID-FOES, PID-MISOESN, ANN-NCO, and
control the brightness of artificial lighting and fulfill the illuminance ANN-PSO. Both PID and ANN techniques can be easily hybrid with other
level preference. Newton-based extremum seeking (ES) was used to find techniques, e.g., PSO. According to the energy performance of hybrid
the optimum energy consumption based on the PID output. The pro­ techniques, it shows that ANN-NCO recorded the highest energy savings,
posed method was developed using Arduino microcontroller and it was up to 66% compared to ANN-PSO with the same control scheme.
integrated with MATLAB-Simulink. The performance of the
PID-MISOESN was compared to the gradient- and Newton-based ES. The 5. Literature survey trend analysis
results illustrated that the accuracy of PID-MISOESN outperformed the
other two methods and achieved optimum lighting energy consumption. Based on the literature survey throughout this article (i.e., Sections
The hybridization of ANN and non-linear constrained optimization 2‒4), the summary of graphical trend analysis is provided, including
(NCO) algorithm was proposed by Tran and Tan [98] for minimizing the publications, room/building types, types of research, objective func­
dimming levels of LEDs based on adaptive sensorless illumination and tions, control strategies and techniques, and simulation tools, in the
occupancy control. However, the study did not consider daylight effect following sections.
across the room. A feed-forward neural network (FFNN) was utilized to
generate the dimming levels of LEDs by considering illuminance distri­
5.1. Publications
bution from artificial lighting on the working plane. Meanwhile, the
optimizer was to find the optimum dimming levels of luminaires by
Fig. 3 shows the articles published in related journals or conferences,
considering the information of generated illuminance levels from the
with Energy and Buildings [100] contributed nearly 33% of the total
FFNN model and occupancy status. The results showed that the FFNN
journals published. It is the most prominent journal related to lighting
model achieved 90% accuracy, whereas the overall proposed method
systems in buildings. The journal is well-known for the energy used in
provided energy savings of 66%. Si. et al. [99].proposed a combination
buildings (e.g., lighting and HVAC systems). The second are IEEE jour­
of PSO and RBFNN to optimize the dimming levels of artificial lighting.
nals [101], with nearly 14% of the total journals published. A major
In the study, RBFNN was chosen to predict the illuminance distribution
portion of IEEE journals is contributed by IEEE Sensors Journal, which is
from artificial lighting. The position of the light sensors (xy coordinates)
related to sensor-driven strategy in lighting control strategies in build­
and the order of ith lamp were considered as the input data, and the
ings. Renewable and Sustainable Energy Reviews Journal [102] is the
illuminance level was measured by the light sensors as the output data
main platform that provides high quality review articles of BES toward
for the RBFNN. The output of RBFNN was then processed by PSO to find
low carbon buildings.
the optimal dimming levels of luminaires based on the occupancy sensor
information that satisfied the targeted illuminance level. The simulation
results showed that the proposed method contributed to energy savings 5.2. Room/building types
of 44%.
The summary of related works that focused on hybrid techniques is Fig. 4 presents the types of rooms/buildings considered as test-beds
shown in Table 5. to evaluate the performance of the proposed control techniques. It can
be seen that offices recorded the highest percentage among other rooms/
4.2. Energy performance of hybrid techniques building types with 68%. This is followed by the classrooms (16%),
laboratories (8%), industries (4%), and others (4%). Others represent
Similar control schemes (in Sections 2.4 and 3.2) are considered in the conference rooms and rooms that are not specified in the literature,
this control strategy to evaluate energy performance. However, only as examples. Most of the papers considered office rooms located in
combined schemes of occupancy-based on/off control (OOF) and educational buildings, e.g., in Refs. [29,57,70,76]. However, several
illuminance-based dimming control (IDM) are considered in this strat­ works did not mention the specific location of the office rooms (i.e.,
egy. The summary of energy savings reported from this strategy is either in educational or other buildings), e.g., in Ref. [17,66]. The office
illustrated in Table 6. It can be seen that the combined scheme of OOF is widely considered as the test-bed due to its higher density and time
and IDM achieved up to 66% energy savings, which is the highest sav­ duration of occupants. Moreover, data availability and accessibility to
ings achieved. the room/building are other reasons for this popularity. Mostly, the
works considered the laboratory for testing the developed prototypes (i.
e., small scale), e.g., in Ref. [32,71].

Table 5
Summary of related works that focused on hybrid control techniques.
Reference and year Objective function Input parameter Design parameter Method Simulation tool/hardware Case studyh
a b c d e f g
EC VC D O Eav Uo GI

[96] 2018 ✓ – ✓ – ✓ – – PID-FOES MATLAB- Simulink and microcontroller OfficeE

[97] 2017 ✓ – ✓ – ✓ – – PID-MISOESN MATLAB-Simulink and microcontroller OfficeE
[98] 2014 ✓ – ✓ ✓ ✓ – – ANN-NCO DIALux and MATLAB OfficeS
[99] 2013 ✓ – ✓ ✓ ✓ – – ANN-PSO DIALux OfficeS
a
Energy consumption.
b
Visual comfort.
c
Daylight.
d
Occupancy.
e
Average illuminance level.
f
Illuminance uniformity.
g
Glare index.
h
S is simulation, E is experimental, and R is real application.

12
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

Table 6
Summary of energy savings for hybrid control techniques.
Reference Room type and design illuminance level (lux) Type of study Method/controller/software Control schemed Energy savings (%)
a b c
C E R

[98] Office - 350 ✓ ANN-NCO OOF þ IDM 28–66

[99] Office - 600 ✓ ANN-PSO OOF þ IDM 44

Note.
a
Simulation.
b
Experimental.
c
Real.
d
OOF is occupancy-based on/off control scheme, and IDM is illuminance-based dimming control scheme.

Fig. 3. Publications.

Fig. 5. Types of research.

Fig. 4. Room/building types.

(24%) studies. In simulation-based study, the results can be obtained

Based on the review, the industrial building is rarely considered as faster, and hardware and equipment set-up are not required. However,
the test-bed in the lighting control field. The only paper that considered the accuracy of the results is the main issue due to the assumed pa­
the industrial building is [77]. This is due to geometry constraints (e.g., rameters related to the case study, for example, interior reflection fac­
higher ceiling/roof and wider floor area) and the complicated layout (e. tors and weather conditions (i.e., do not illustrate the actual condition,
g., machines/equipment) [103]. e.g., daylight contribution from clear/overcast skies are considered for
the whole day of the simulation). Indeed, Williams et al. [31] revealed
5.3. Types of research that the simulation results in terms of average energy savings were 20%
higher than the real-application study. Moreover, in terms of daylight
In the lighting control system field, research studies can be classified factor (DF) contribution, the simulation results were 20% higher than
into three: simulation, experimental, and real-application studies, as the field measurement [45]. In a real study, the widely used hardware
shown in Fig. 5. It shows that simulation-based study is the most popular are microcontrollers and PLCs. In comparison, microcontrollers are the
study with 56% compared to experimental (20%), and real-application most popular option among researchers to conduct experiments in

13
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

lighting control/optimization systems than PLCs. This is because glare-related metrics (e.g., GDP), and color quality (e.g., CCT). Apart
microcontrollers have a smaller size, lower power, and costs. Nonethe­ from objective assessment, subjective assessment has been considered in
less, microcontrollers have a specific purpose and are more suitable for the literature as an objective function, i.e., in Ref. [42]. In the research, a
small-scale applications due to their lower processor capacity, and input psychovisual function was considered as the objective function, where
and output numbers. Meanwhile, PLCs are well-known controllers in the function was modeled based on the collected psychovisual data
automation and industrial applications. For this reason, PLCs are suit­ through laboratory experiment among 36 participants. The psychovi­
able for larger-scale applications of lighting systems due to shorter sual function was optimized simultaneously with the energy consump­
processing time and the availability of many inputs and outputs that can tion of artificial lighting (i.e., dimming levels).
be extended. However, the cost of PLCs is far higher than Apart from visual comfort, the location of luminaires was also
microcontrollers. considered as the objective function in Ref. [95]. This function was
considered together with the power consumption of lighting, and both
5.4. Objective functions functions were converted into a single objective function using a
weighted sum method to find the optimum solution.
Fig. 6 illustrates objective functions that have been considered for
control and optimization methods, in which energy consumption is 5.5. Control strategies
prominent among researchers with 86%, followed by visual comfort
(12%), and other objectives (2%). Energy consumption is the objective Three control strategies have been considered in lighting systems,
function represented by the dimming levels of luminaires. By mini­ namely occupancy, daylight, and mixed strategies, as shown in Fig. 7.
mizing energy consumption of a lighting system, electrical energy, The mixed-control scheme is the primary control scheme with 63%. This
maintenance cost, and pollutant emissions can be reduced. Visual strategy has fully utilized occupancy (i.e., occupancy-based strategy)
comfort as the objective functions includes illuminance-based (e.g., and light sensors (i.e., illuminance-based strategy), and higher energy
illuminance uniformity), glare, and color quality. Luminaire position is savings have been recorded, up to 72% (i.e., in this paper based on
represented by other objectives, as the objective function in lighting Tables 2, 4, and 6). The single strategy, which consists of occupancy and
system control strategy. daylight strategies, made up the remaining 34% and 3%, respectively.
As mentioned in Section 1.2, the summation of dimming levels of
luminaires (i.e., Eq. (1)) is widely used as the objective function for 5.6. Control techniques
representing energy consumption of artificial lighting of a control sys­
tem. Meanwhile, for visual comfort, an interesting objective function to Various control techniques have been utilized in lighting system
represent illuminance uniformity called CVRMSE was developed in control of buildings as presented in Fig. 8, on the basis of controller-
Ref. [37]. The developed CVRMSE model was treated simultaneously based (orange), optimization-based (blue), and hybrid (green) tech­
with the summation of the dimming levels of luminaires (i.e., Eq. (1)) to niques. It can be seen that the optimization-based control technique has
find the equal solution of both competing objective functions (i.e., 50% the largest portion among control techniques, around 51%, followed by
for each objective). The CVRMSE model does not only correspond to the the controller-based technique with 38%, and finally the hybrid tech­
illuminance uniformity but also to the average illuminance level as well. nique with 11%. Majority of the researchers utilized a single technique
However, the study did not consider the daylighting in the indoor of both controller-based and optimization-based control.
environment. In Ref. [43], two visual comfort metrics (i.e., color quality The highest percentage is contributed by LP with 22%. The second
and spatial uniformity) and an energy consumption model were devel­ highest is PI with 8%, which is categorized under the controller-based
oped as the objective functions. All three objectives were combined strategy. This is followed by CO, iterative method, and NSGA II is
linearly and provided a single optimum solution (using close to 5% (i.e., the optimization-based control technique) and the
optimization-based technique). Correlated color temperature (CCT) is remaining techniques are 3%. The popularity of LP as the optimization
one of the important metrics that represented the color quality of light technique in lighting control systems is due to fewer parameter settings
[104]. and ease of implementation, and most proposed problem formulation in
In terms of glare, daylight glare probability (GDP) was considered as a dimming control scheme has a linear function [35].
one of the objective functions in Ref. [70]. Apart from GDP, in the study, In terms of hybridization, ANN and PID are the most popular tech­
illuminance-based on two planes (i.e., working and vertical) were niques compared to other hybrid techniques, e.g., PSO, NCO, FOES, and
considered and all three objectives were controlled together to find the MISOESN. ANN has demonstrated superior performance as a prediction
optimum position of shading and the dimming levels of luminaires. engine due to its accuracy, ease of implementation, and integration with
Majority of the works considered objective assessment of visual other control techniques. ANN has been used for different prediction
comfort in the indoor environment, which is to determine the visual functions in lighting systems: the dimming levels of luminaires [57,98]
comfort level based on the specific mathematical equations/models and illuminance distribution [99]. Two types of ANNs are utilized in this
mentioned in the standards or literature, such as illuminance-based field based on the literature, namely FFNN [98] and RBFNN [99].
metrics (i.e., average illuminance level and illuminance uniformity),

Fig. 6. Objective functions. Fig. 7. Control strategies.

14
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

Fig. 8. Control techniques.

Meanwhile, PID is a simple controller (i.e., requires only three param­ optimizing energy consumption and visual comfort of occupants (i.e.,
eter settings: kp , ki , and kd ) that performs better in controlling the objective functions) to produce Pareto-based solution. Meanwhile, both
dimming levels of artificial lighting systems when the parameters are GEO and gradient algorithm techniques are used for optimizing
tuned appropriately. multi-objective problem in artificial lighting by converting all objective
Based on literature review, only three techniques are utilized for functions to a single objective function using a weighted sum method (i.
solving multi-objective problems for lighting control: NSGA II [37,42], e., provided a single optimal solution). NSGA II has the highest fre­
GEO [95], and gradient algorithm [43]. NSGA II is utilized for solving quency compared to GEO and gradient algorithm because the approach
multi-objective problems for lighting control by simultaneously can optimize multiple objective functions simultaneously and generate a

Fig. 9. Simulation tools. (a) Lighting simulation, and (b) control and optimization.

15
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

set of solutions, known as Pareto-optimal front. The approach is also integrating such method with other sensing technologies (e.g., light
reliable and easy to implement (i.e., available in MATLAB global opti­ sensors) in order to improve the efficacy of the system for sustainable
mization toolbox) [37]. buildings. In terms of light sensor locations, most literature applied the
sensor co-located with the luminaire in the control strategy, e.g., in
5.7. Simulation tools Ref. [17,18,34,34,37,54–56]. However, it has many shortcomings, such
as higher number of sensors, and increased capacity and configuration of
Fig. 8 shows simulation tools related to lighting and they can be the controller.
categorized into two groups: lighting simulation, and control and opti­ The development of an illuminance mapping model (also known as
mization. Lighting simulation tools that have been used in the literature light sensor calibration) between horizontal (e.g., ceiling and working
are shown in Fig. 9(a). It clearly shows that DIALux software is the most plane) and vertical (e.g., walls) planes is one of the important bodies of
used simulation tool. This is because DIALux is freeware, in which a research in lighting control systems, e.g., in Refs. [36,38,54,56,63,70,
luminaire library is linked to the database of various luminaire manu­ 73,74]. Based on literature findings, the light sensors placed on the
facturers and offers various types of office furniture and sky conditions working plane showed superior performance among other planes in
(i.e., for daylight analysis) [105]. Moreover, the software applies terms of illuminance and energy savings. This is based on the mea­
radiosity and raytracing calculation methods. surement of illuminance levels on the working plane carried out for all
For control and optimization tools, MATLAB [106] is widely used types of research (i.e., simulation and real). It should be noted that the
compared to LabVIEW software, as illustrated in Fig. 9(b). MATLAB is a illuminance values on the working plane are proportional to the illu­
well-known computing programming language platform as reported by minance values on the ceiling (i.e., light sensor locations) [74].
Ref. [107]. Apart from that, MATLAB also provides control toolboxes (e. The dimming levels of luminaires as an objective function are useful
g., neural network and fuzzy logic) and optimization toolboxes (e.g., and well-established to represent energy consumption in lighting. In
linear, non-linear, and mixed-integer linear programming and terms of visual comfort, the developed CVRMSE model (i.e., represents
multi-objective), as well as Simulink (e.g., graphical, customized sys­ illuminance uniformity) in a recent work as the objective function is
tems, and solvers for modelling and simulation), which are the main interesting, and it is useful to optimize the CVRMSE model simulta­
factors for its popularity. neously with the dimming levels of luminaires. The purpose of this
approach is to achieve a balanced solution of both objectives. However,
6. Discussion the developed model did not include daylighting into the proposed
model formulation. Daylighting is a crucial parameter that should be
This section discusses the key findings from literature survey. Real­ considered to achieve higher energy savings. However, considering the
istic future directions have also been included. daylighting parameter into illuminance uniformity-based visual comfort
model formulation will increase the complexity of the model as the
6.1. Literature survey key findings illuminance values from the daylight fully depend on time (i.e., the
values vary every time), resulting in lower illuminance uniformity
Lighting is a crucial subsystem of BES that provides light in a room so across the room.
that the occupants can perform various tasks and activities. Lighting Lighting system control strategies can be divided into two: occu­
systems have significant potential improvement with two main objec­ pancy- and illuminance-based strategies. A combination of these two
tives, which are energy consumption and visual comfort of occupants. strategies improved the performance of energy efficiency up to 72% and
From the review conducted, visual comfort of occupants can be cate­ this strategy is a prominent control strategy in the last decade. From
gorized into objective and subjective assessments. For objective assess­ each strategy, it can be divided into another two schemes: switching on/
ment, comfort can be measured either by real measurement or off and dimming control. The dimming control scheme is based on the
simulation, in terms of the value of illuminance level and uniformity. illuminance-based strategy that is widely used in the literature as it can
Meanwhile, for subjective assessment, it is based on questionnaire sur­ contribute to higher energy savings and satisfy visual comfort of occu­
vey analysis conducted on experts and non-expert persons related to pants. However, this scheme requires specific control techniques. Thus,
lighting systems in buildings. To achieve better results from the ques­ various techniques have been developed and proposed for controlling
tionnaire survey conducted, the respondents should be knowledgeable the dimming levels of luminaires in order to reduce energy consumption
and expert in the lighting system field and the survey questions must be while maintaining visual comfort of occupants in a building. Based on
more specific and technical to lighting systems. Intelligent control sys­ the survey, these control techniques (i.e., illuminance-based dimming
tems present a strategy to improve energy performance and visual control) can be categorized into three: controller-based, optimization-
satisfaction of occupants in buildings. Utilization of new lighting tech­ based, and hybrid techniques (a combination of controller-based and
nologies, such as LEDs can reduce significantly energy use and enhance optimization-based control). From the survey trend analysis, the
the quality of light (e.g., CCT and CRI) in buildings. LEDs have been optimization-based control technique recorded the highest percentage
proven to have good control under PWM signal to provide a linear among other techniques, where LP method is widely used for this
relation between dimming level and output power of a luminaire. For technique. This is because the technique directly optimized the dimming
this reason, majority of the researchers used LED luminaires in their levels of luminaires (i.e., decision variables) and fully satisfied visual
studies. An advanced lighting control system involves centralized and comfort metrics (i.e., constraints in optimization formulation), resulting
distributed architectures. Centralized architecture is a prominent ar­ in lower energy consumption (i.e., objective function). Moreover, this
chitecture as it uses a single controller and it can be integrated with an technique can solve multi-objective problems simultaneously by
existing network infrastructure to minimize the initial cost of providing a set of solutions (Pareto-optimal front), e.g., using NSGA II.
installation. The integration of artificial lighting and blind systems in daylight-
Recently, mobile devices (MDs) have become a necessity for linked control strategy is an interesting research topic among re­
communication, work, and social purposes. Therefore, recent lighting searchers [28,44,65]. The main reason of this consideration is because
system control has been developed by taking the advantage of the usage the higher amount of illuminance levels across the room is mainly from
of MDs and existing wireless network infrastructure (i.e., WiFi) to detect, daylight penetrating through windows. Hence, the higher amount of
trace, and count the occupants in buildings accurately [23,61,75]. illuminance levels benefits energy performance (i.e., reduces energy
Moreover, the occupants could control the lighting system based on consumption of lighting). At the same time, it contributes to higher glare
their personal preference using MDs, such as illuminance level and level and consequently causes discomfort to occupants. Therefore,
on/off switch. The developed methodology can be expanded by controlling artificial lighting and blind systems simultaneously can

16
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

ensure the balance between energy consumption and occupants’ visual uniformity-based visual comfort model in future development should
comfort. consider other metrics related to daylight, such as DF, daylight co­
Based on literature survey, the highest energy savings are recorded efficient (DC), and daylight autonomy (DA).
by considering a single scheme, i.e., the OOF scheme of 93% by using the � As highlighted in Sections 2‒4, there are several techniques utilized
integration of BB technique and WinLight (i.e., developed controller in lighting control schemes, particularly in hybrid techniques. Thus,
platform). Meanwhile, by considering mixed schemes, i.e., OOF and various hybrid methods need to be researched in the future for better
IDM, the highest energy savings of 72% were achieved under self- results compared to single methods, such as ANN-fuzzy and adaptive
adaptive algorithm technique. However, the energy performance of neuro-fuzzy inference system (ANFIS). In a distributed control sys­
the techniques in the literature is affected by several factors, including tem, a multi-agent system (MAS) can be explored as the control
TD and occupancy patterns for the OOF scheme; illuminance distribu­ system because the architecture of MAS is similar to a distributed
tion of daylight (e.g., sky conditions, window characteristics and ge­ control system and the system has been successfully applied in HVAC
ometry, and building’s location and orientation) for IDM scheme; and systems [92,109]. Hybridization of other intelligent controller- and
robustness of the control techniques for both schemes. optimization-based control techniques (i.e., becomes hybrid tech­
nique) needs to be further explored, such as ANN-GA, FL-GA,
6.2. Challenges and future directions ANN-COR, and others.
� Simulation-based study accounted more than half (56%) of the total
This section highlights the challenges in lighting control strategies research works, as highlighted in Section 5. Moreover, in daylight
based on literature survey. From these challenges, future directions are analysis, majority of the studies considered a single type of sky
pointed out in the following points. conditions, i.e., clear sky (e.g., in Ref. [34,55,57]). Therefore, in
future works, experimental studies (e.g., developing prototype/real
� As highlighted in Section 1.2, there is a lack of lighting control control systems) need to be carried out to validate simulation results
strategies that consider a fewer number of light sensors as the input (e.g., illuminance values for different types of skies as this parameter
devices in the literature, where majority of the researchers consid­ influences the validity of the results, particularly in calculating the
ered sensors co-located with the luminaires (i.e., known as the multi- dimming levels of luminaires) and ensure the systems can be
sensor approach). This strategy (i.e., with fewer number of sensors) implemented in the real world. For future research, in terms of
needs more research in the future as it has many benefits, such as controllers, single-board computers (e.g., Raspberry Pi, Orange Pi,
ease of configuration, small capacity of controllers, and lower and Banana Pi) can be explored as the computers resemble personal
computational effort and cost of installation. The future directions of computers (PCs) with less processor capacity than PCs but higher
this strategy should consider the determination of the optimal than microcontrollers and at lower cost. Moreover, such controllers
number of light sensors to be employed and their positions in the support the Internet of Things (IoT) technology.
lighting system control, and also the number of luminaires in each � As discussed in Sections 2‒4, most of the studies developed lighting
control zone. Moreover, comparative analysis should also be carried systems implemented on the test-bed (e.g., office room) without any
out between this strategy and the multi-sensor strategy in terms of integration of cloud computing (a few studies applied IoT, e.g.,
the number of sensors to be employed, energy performance, and Ref. [23]). Therefore, the integration of the latest sensor technolo­
visual comfort metrics (e.g., illuminance level and uniformity) to gies, occupancy and environmental models, occupants’ preference
ensure the strategy performs well, is stable, and cost-effective. model, advanced control system, PV system, energy management
� As mentioned in Tables 1, 3, and 5, in terms of visual comfort, most system (EMS), IoT, and cloud computing (CC) [110] can be devel­
of the works focused on illuminance-based metrics (i.e., average oped as integrated lighting systems, which should be more acces­
illuminance level and uniformity) as the objective functions or con­ sible, controllable, and sustainable in the future. So far, only one
straints in the control systems, e.g., in Refs. [17,34,36,37]. Only two research integrated a PV system with an intelligent lighting control
papers considered glare in the problem formulation, i.e., daylight system, i.e., in Ref. [72]. Moreover, all the data gathered in the cloud
glare probability (GDP) [70] and glare (the type is not specified) can be utilized in the future for descriptive, diagnostic, predictive,
[28]. In future works, various glare-related metrics, such as unified and prescriptive for each sub-system and model of the system. For
glare rating (UGR), British glare index (BGI), visual comfort proba­ instance, in the predictive function in Ref. [98], an artificial lighting
bility (VCP), discomfort glare index (DGI), and many more, i.e., can illuminance-dimming levels of luminaires model was developed to
be found in Ref. [104,108], should be considered. Considering both establish sensorless (i.e., light) indoor lighting systems. However, the
illuminance-based and glare-related metrics, the lighting system data for the model are limited to the day and daylight is not
control solution fully satisfies the international lighting-related considered. In future sensorless indoor lighting control systems in
standards for indoor spaces, e.g., the EN12464-1 standard, and terms of illuminance-based strategy, the model of artificial lighting
provides great visual comfort of occupants in the indoor illuminance-daylighting illuminance-dimming levels of luminaires
environment. needs to be considered. Similarly, in occupancy-based strategy, the
� As discussed in Sections 2 and 3, a few previous works developed development of occupancy-light status/dimming model should be
new mathematical models related to visual comfort in the indoor considered to attain the same goal, i.e., sensorless lighting control
environment, specifically in terms of illuminance uniformity (e.g., in systems. In addition, both systems should be integrated into a com­
Ref. [37]), where this parameter is mentioned in many international plete system.
lighting standards (e.g., the EN12464-1 standard). The model (i.e.,
illuminance uniformity-based visual comfort model) is also crucial to 7. Conclusion
be considered as the objective function and is optimized simulta­
neously with well-known objective function, which is the dimming Lighting consumes substantial electricity energy in commercial
levels of luminaires (i.e., represent energy consumption of lighting) buildings and there is a higher potential to save energy by considering
and solved using multi-objective optimization methods, such as intelligent control systems. The main objective of this article is to
NSGA-II and GEO (i.e., methods have already been used in previous extensively review lighting system control techniques in commercial
works), as well as other methods, such as MOPSO, MOGA, and SPEA buildings, which covers controller-, optimization-based control, and
II in future applications. This consideration is useful to balance or hybrid techniques. The utilization of intelligent lighting control systems
find the best compromised solution between energy consumption has many benefits, including maximizing visual comfort of occupants,
and visual comfort of occupants. The new model of illuminance minimizing energy consumption in buildings, and reducing pollutant

17
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

emissions. Intelligent controllers have attracted the most attention [15] C.K. Gan, A.F. Sapar, Y.C. Mun, K.E. Chong, Techno-economic analysis of LED
lighting: a case study in UTeM’s faculty building, Procedia Eng 53 (2013)
among researchers due to their performances in terms of the dimming
208–216, https://doi.org/10.1016/j.proeng.2013.02.028.
levels of luminaires and targeted illuminance levels. Based on the re­ [16] K.R. Wagiman, M.N. Abdullah, M.A.M. Ariff, M.Y. Hassan, Techno-economic
view, it is found that optimization-based technique shows more exten­ analysis of lighting system retrofit at the industrial training institute, in: IET Conf.
sive utilization of 51% compared to controller-based and hybrid Proc., Institution of Engineering and Technology, 2018, pp. 1–6, https://doi.org/
10.1049/cp.2018.1332.
techniques due to its optimal achievement in the dimming levels of [17] D. Caicedo, A. Pandharipande, Daylight and occupancy adaptive lighting control
artificial lighting and satisfaction of occupants’ visual comfort. More­ system: an iterative optimization approach, Light. Res. Technol. 48 (2015)
over, multi-objective optimization can be used to solve multi-objective 661–675, https://doi.org/10.1177/1477153515587148.
[18] D. Caicedo, A. Pandharipande, G. Leus, Occupancy-based illumination control of
problems when dealing with energy consumption and visual comfort LED lighting systems, Light. Res. Technol. 43 (2011) 217–234, https://doi.org/
of occupants simultaneously. In order to improve the performance of 10.1177/1477153510374703.
existing methods, future research can focus more on optimization-based [19] K.R. Wagiman, M.N. Abdullah, Lighting system design according to different
standards in office building: a technical and economic evaluations, J. Phys. Conf.
control and hybrid techniques. Ser. 1049 (2018), https://doi.org/10.1088/1742-6596/1049/1/012010.
[20] M.M. Aman, G.B. Jasmon, H. Mokhlis, A.H.A. Bakar, Analysis of the performance
Funding of domestic lighting lamps, Energy Pol. 52 (2013) 482–500, https://doi.org/
10.1016/j.enpol.2012.09.068.
[21] C. de Bakker, M. Aries, H. Kort, A. Rosemann, Occupancy-based lighting control
This research was supported by Ministry of Education of Malaysia in open-plan office spaces: a state-of-the-art review, Build. Environ. 112 (2017)
through Research Grant Scheme (FRGS/1/2018/TK07/UTHM/02/3) 308–321, https://doi.org/10.1016/j.buildenv.2016.11.042.
[22] F. Manzoor, D. Linton, M. Loughlin, Occupancy monitoring using passive RFID
and partially sponsored by Universiti Tun Hussein Onn Malaysia.
technology for efficient building lighting control, in: 2012 Fourth Int. EURASIP
Work. RFID Technol, 2012, pp. 83–88, https://doi.org/10.1109/RFID.2012.10.
Declaration of competing interest [23] H. Zou, Y. Zhou, H. Jiang, S.-C. Chien, L. Xie, C.J. Spanos, WinLight: a WiFi-based
occupancy-driven lighting control system for smart building, Energy Build. 158
(2018) 924–938, https://doi.org/10.1016/j.enbuild.2017.09.001.
The authors declare that they have no known competing financial [24] M.A. ul Haq, M.Y. Hassan, H. Abdullah, H.A. Rahman, M.P. Abdullah, F. Hussin,
interests or personal relationships that could have appeared to influence D.M. Said, A review on lighting control technologies in commercial buildings,
the work reported in this paper. their performance and affecting factors, Renew. Sustain. Energy Rev. 33 (2014)
268–279, https://doi.org/10.1016/j.rser.2014.01.090.
[25] X. Guo, D.K. Tiller, G.P. Henze, C.E. Waters, The performance of occupancy-based
Appendix A. Supplementary data lighting control systems: a review, Light. Res. Technol. 42 (2010) 415–431,
https://doi.org/10.1177/1477153510376225.
[26] R. Delvaeye, W. Ryckaert, L. Stroobant, P. Hanselaer, R. Klein, H. Breesch,
Supplementary data to this article can be found online at https://doi. Analysis of energy savings of three daylight control systems in a school building
org/10.1016/j.jobe.2020.101342. by means of monitoring, Energy Build. 127 (2016) 969–979, https://doi.org/
10.1016/j.enbuild.2016.06.033.
[27] K.R. Wagiman, M.N. Abdullah, M.Y. Hassan, N.H.M. Radzi, A review on sensing-
References based strategies of interior lighting control system and their performance in
commercial buildings, Indones. J. Electr. Eng. Comput. Sci. 16 (2019) 208–215,
[1] U.S. Energy Information Administration, Annual Energy Review, EIA., 2017. http https://doi.org/10.11591/ijeecs.v16.i1.pp208-215.
s://www.eia.gov/totalenergy/data/annual/index.php. (Accessed 5 August [28] H.B. Gunay, W. O’Brien, I. Beausoleil-Morrison, S. Gilani, Development and
2018). implementation of an adaptive lighting and blinds control algorithm, Build.
[2] P.H. Shaikh, N.B.M. Nor, P. Nallagownden, I. Elamvazuthi, T. Ibrahim, A review Environ. 113 (2017) 185–199, https://doi.org/10.1016/j.buildenv.2016.08.027.
on optimized control systems for building energy and comfort management of [29] L. Xu, Y. Pan, Y. Yao, D. Cai, Z. Huang, N. Linder, Lighting energy efficiency in
smart sustainable buildings, Renew. Sustain. Energy Rev. 34 (2014) 409–429, offices under different control strategies, Energy Build. 138 (2017) 127–139,
https://doi.org/10.1016/j.rser.2014.03.027. https://doi.org/10.1016/j.enbuild.2016.12.006.
[3] A.S. Ahmad, M.Y. Hassan, M.P. Abdullah, H.A. Rahman, F. Hussin, H. Abdullah, [30] X. Ding, J. Yu, Y. Si, Office light control moving toward automation and
R. Saidur, A review on applications of ANN and SVM for building electrical humanization: a literature review, Intell. Build. Int. (2018) 1–32, https://doi.org/
energy consumption forecasting, Renew. Sustain. Energy Rev. 33 (2014) 10.1080/17508975.2018.1555087.
102–109, https://doi.org/10.1016/j.rser.2014.01.069. [31] A. Williams, B. Atkinson, K. Garbesi, E. Page, F. Rubinstein, Lighting controls in
[4] EIA, How Much Energy Is Consumed in U.S. Residential and Commercial commercial buildings, Leukos 8 (2012) 161–180, https://doi.org/10.1582/
Buildings?, 2018. https://www.eia.gov/tools/faqs. (Accessed 17 September LEUKOS.2012.08.03.001.
2018). [32] N.K. Kandasamy, G. Karunagaran, C. Spanos, K.J. Tseng, B.-H. Soong, Smart
[5] EIA, Commercial Buildings Energy Consumption Survey, 2012. https://www.eia. lighting system using ANN-IMC for personalized lighting control and daylight
gov/consumption/commercial/reports/2012/lighting/. (Accessed 17 September harvesting, Build. Environ. 139 (2018) 170–180, https://doi.org/10.1016/j.
2018). buildenv.2018.05.005.
[6] The European Parliament and of the Council, Directive 2010/31/EU of the [33] M.T. Koroglu, K.M. Passino, Illumination balancing algorithm for smart lights,
European parliament and of the Council of 19 may 2010 on the energy IEEE Trans. Contr. Syst. Technol. 22 (2014) 557–567, https://doi.org/10.1109/
performance of buildings, Off. J. Eur. Union (2010) 1–23. TCST.2013.2258399.
[7] European Committee for Standardization, European Standard EN 12464-1: Light [34] D. Caicedo, A. Pandharipande, Sensor-driven lighting control with illumination
and Lighting - Lighting of Work Places - Part 1: Indoor Work Places, 2011. and dimming constraints, IEEE Sensor. J. 15 (2015) 5169–5176, https://doi.org/
[8] European Committee for Standardization, European Standard EN 12464-2: Light 10.1109/JSEN.2015.2436338.
and Lighting - Lighting of Work Places - Part 2: Outdoor Work Places, 2014. [35] A. Pandharipande, D. Caicedo, Adaptive illumination rendering in LED lighting
[9] European Committee for Standardization, European Standard EN 15193: Energy systems, IEEE Trans. Syst. Man, Cybern. Syst. 43 (2013) 1052–1062, https://doi.
Performance of Buildings - Energy Requirements for Lighting, 2017. org/10.1109/TSMCA.2012.2231859.
[10] M.-C. Dubois, F. Bisegna, N. Gentile, M. Knoop, W.O.B. Matusiak, E. Tetri, [36] S. Borile, A. Pandharipande, D. Caicedo, L. Schenato, A. Cenedese, A Data-Driven
Retrofitting the electric lighting and daylighting systems to reduce energy use in Daylight Estimation Approach to Lighting Control, IEEE Access, 2017, p. 1,
buildings: a literature review, Energy Res. J. 6 (2015) 25–41. https://doi.org/10.1109/ACCESS.2017.2679807.
[11] M.A. ul Haq, M.Y. Hassan, H. Abdullah, H.A. Rahman, M.P. Abdullah, F. Hussin, [37] E.-N.D. Madias, P.A. Kontaxis, F. V Topalis, Application of multi-objective genetic
M. Bandi, Energy saving in lighting from T5 lamp retrofits - a case study, in: IEEE algorithms to interior lighting optimization, Energy Build. 125 (2016) 66–74,
Student Conf. Res. Dev., 2013, 2013, pp. 187–190, https://doi.org/10.1109/ https://doi.org/10.1016/j.enbuild.2016.04.078.
SCOReD.2013.7002569. [38] M. Rossi, A. Pandharipande, D. Caicedo, L. Schenato, A. Cenedese, Personal
[12] G.S.B. Ganandran, T.M.I. Mahlia, H.C. Ong, B. Rismanchi, W.T. Chong, Cost- lighting control with occupancy and daylight adaptation, Energy Build. 105
benefit analysis and emission reduction of energy efficient lighting at the (2015) 263–272, https://doi.org/10.1016/j.enbuild.2015.07.059.
Universiti tenaga nasional, Sci. World J. (2014) 1–12, 2014. [39] Y. Gao, Y. Cheng, H. Zhang, N. Zou, Dynamic illuminance measurement and
[13] T.M.I. Mahlia, H.A. Razak, M.A. Nursahida, Life cycle cost analysis and payback control used for smart lighting with LED, Measurement 139 (2019) 380–386,
period of lighting retrofit at the University of Malaya, Renew. Sustain. Energy https://doi.org/10.1016/j.measurement.2019.03.003.
Rev. 15 (2011) 1125–1132, https://doi.org/10.1016/j.rser.2010.10.014. [40] S. Ozenc, M. Uzunoglu, O. Guler, Experimental evaluation of the impacts of
[14] F.P. Vahl, L.M.S. Campos, N. Casarotto Filho, Sustainability constraints in techno- considering inherent response characteristics for lighting technologies in building
economic analysis of general lighting retrofits, Energy Build. 67 (2013) 500–507, energy modeling, Energy Build. 77 (2014) 432–439, https://doi.org/10.1016/j.
https://doi.org/10.1016/j.enbuild.2013.08.039. enbuild.2014.03.062.
[41] L.T. Doulos, A. Tsangrassoulis, P.A. Kontaxis, A. Kontadakis, F.V. Topalis,
Harvesting daylight with LED or T5 fluorescent lamps? The role of dimming,

18
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

Energy Build. 140 (2017) 336–347, https://doi.org/10.1016/j. [70] J. Xiong, A. Tzempelikos, Model-based shading and lighting controls considering
enbuild.2017.02.013. visual comfort and energy use, Sol. Energy 134 (2016) 416–428, https://doi.org/
[42] C. Villa, R. Labayrade, Multi-objective optimisation of lighting installations 10.1016/j.solener.2016.04.026.
taking into account user preferences – a pilot study, Light. Res. Technol. 45 [71] I. Chew, V. Kalavally, N.W. Oo, J. Parkkinen, Design of an energy-saving
(2013) 176–196. controller for an intelligent LED lighting system, Energy Build. 120 (2016) 1–9,
[43] S. Afshari, S. Mishra, A. Julius, F. Lizarralde, J.D. Wason, J.T. Wen, Modeling and https://doi.org/10.1016/j.enbuild.2016.03.041.
control of color tunable lighting systems, Energy Build. 68 (2014) 242–253, [72] M.A. Ozçelik,
€ The design and implementation of PV-based intelligent distributed
https://doi.org/10.1016/j.enbuild.2013.08.036. sensor LED lighting in daylight exposed room environment, Sustain. Comput.
[44] I. Din, H. Kim, Joint blind and light control for lighting energy reduction while Informatics Syst. 13 (2017) 61–69, https://doi.org/10.1016/j.
satisfying light level and anti-glare requirements, Light. Res. Technol. 46 (2013) suscom.2017.01.001.
281–292, https://doi.org/10.1177/1477153513497740. [73] R. Kumar, New algorithms for daylight harvesting in a private office, 2015 18th,
[45] X. Yu, Y. Su, X. Chen, Application of RELUX simulation to investigate energy Int. Conf. Inf. Fusion. (2015) 383–392.
saving potential from daylighting in a new educational building in UK, Energy [74] K.-W. Park, A.K. Athienitis, Workplane illuminance prediction method for
Build. 74 (2014) 191–202, https://doi.org/10.1016/j.enbuild.2014.01.024. daylighting control systems, Sol. Energy 75 (2003) 277–284, https://doi.org/
[46] F. Dong, A.C. Sanderson, A dynamic adaptive light field sampling approach for 10.1016/j.solener.2003.08.013.
smart lighting control, Light. Res. Technol. 46 (2013) 593–614, https://doi.org/ [75] A. Pandharipande, D. Caicedo, X. Wang, Sensor-driven wireless lighting control:
10.1177/1477153513502030. system solutions and services for intelligent buildings, IEEE Sensor. J. 14 (2014)
[47] S.K. Ng, K.H. Loo, Y.M. Lai, C.K. Tse, Color control system for RGB LED with 4207–4215, https://doi.org/10.1109/JSEN.2014.2351775.
application to light sources suffering from prolonged aging, IEEE Trans. Ind. [76] J. Liu, W. Zhang, X. Chu, Y. Liu, Fuzzy logic controller for energy savings in a
Electron. 61 (2014) 1788–1798, https://doi.org/10.1109/TIE.2013.2267696. smart LED lighting system considering lighting comfort and daylight, Energy
[48] K.A.G. Smet, W.R. Ryckaert, M.R. Pointer, G. Deconinck, P. Hanselaer, Build. 127 (2016) 95–104, https://doi.org/10.1016/j.enbuild.2016.05.066.
Optimization of colour quality of LED lighting with reference to memory colours, [77] G. Cimini, A. Freddi, G. Ippoliti, A. Monteriù, M. Pirro, A smart lighting system
Light. Res. Technol. 44 (2012) 7–15, https://doi.org/10.1177/ for visual comfort and energy savings in industrial and domestic use, electr,
1477153511432250. Power Components Syst 43 (2015) 1696–1706, https://doi.org/10.1080/
[49] W. Dun, Evaluation model of lighting environment for subway station space 15325008.2015.1057777.
based on back- propagation neural network, Light Eng. 26 (2018) 66–73. [78] P. Mandal, D. Dey, B. Roy, Optimization of Luminaire Layout to Achieve a
[50] H.-S. Lee, S.-Y. Kwon, J.-H. Lim, A development of a lighting control system based Visually Comfortable and Energy-Efficient Indoor General Lighting Scheme by
on context-awareness for the improvement of learning efficiency in classroom, Particle Swarm Optimization, Leukos (2019) 1–16, https://doi.org/10.1080/
Wireless Pers. Commun. 86 (2016) 165–181, https://doi.org/10.1007/s11277- 15502724.2018.1533853.
015-2811-6. [79] S.S. Rao, Engineering Optimization: Theory and Practice, fourth ed., John Wiley
[51] T. Kruisselbrink, R. Dangol, A. Rosemann, Photometric measurements of lighting & Sons, Inc., 2009.
quality: an overview, Build. Environ. 138 (2018) 42–52, https://doi.org/ [80] Y.-J. Wen, A.M. Agogino, Personalized dynamic design of networked lighting for
10.1016/j.buildenv.2018.04.028. energy-efficiency in open-plan offices, Energy Build. 43 (2011) 1919–1924,
[52] C. Moscoso, B. Matusiak, Aesthetic perception of a small office with different https://doi.org/10.1016/j.enbuild.2011.03.036.
daylighting systems, Indoor Built Environ. 27 (2017) 1187–1202, https://doi. [81] N. Delgarm, B. Sajadi, F. Kowsary, S. Delgarm, Multi-objective optimization of the
org/10.1177/1420326X17711490. building energy performance: a simulation-based approach by means of particle
[53] I. Acosta, R.P. Leslie, M.G. Figueiro, Analysis of circadian stimulus allowed by swarm optimization (PSO), Appl. Energy 170 (2016) 293–303, https://doi.org/
daylighting in hospital rooms, Light. Res. Technol. 49 (2015) 49–61, https://doi. 10.1016/j.apenergy.2016.02.141.
org/10.1177/1477153515592948. [82] K.-P. Lee, T.-A. Cheng, A simulation–optimization approach for energy efficiency
[54] D. Caicedo, S. Li, A. Pandharipande, Smart lighting control with workspace and of chilled water system, Energy Build. 54 (2012) 290–296, https://doi.org/
ceiling sensors, Light. Res. Technol. 49 (2016) 446–460, https://doi.org/ 10.1016/j.enbuild.2012.06.028.
10.1177/1477153516629531. [83] S. Papantoniou, D. Kolokotsa, K. Kalaitzakis, Building optimization and control
[55] D. Caicedo, A. Pandharipande, Distributed illumination control with local sensing algorithms implemented in existing BEMS using a web based energy management
and actuation in networked lighting systems, IEEE Sensor. J. 13 (2013) and control system, Energy Build. 98 (2015) 45–55, https://doi.org/10.1016/j.
1092–1104, https://doi.org/10.1109/JSEN.2012.2228850. enbuild.2014.10.083.
[56] D. Caicedo, A. Pandharipande, F.M.J. Willems, Daylight-adaptive lighting control [84] Z. Ma, S. Wang, Supervisory and optimal control of central chiller plants using
using light sensor calibration prior-information, Energy Build. 73 (2014) simplified adaptive models and genetic algorithm, Appl. Energy 88 (2011)
105–114, https://doi.org/10.1016/j.enbuild.2014.01.022. 198–211, https://doi.org/10.1016/j.apenergy.2010.07.036.
[57] K.R. Wagiman, M.N. Abdullah, Intelligent lighting control system for energy [85] F. Wahid, R. Ghazali, L.H. Ismail, Improved firefly algorithm based on genetic
savings in office building, Indones. J. Electr. Eng. Comput. Sci. 11 (2018) algorithm operators for energy efficiency in smart buildings, Arabian J. Sci. Eng.
195–202, https://doi.org/10.11591/ijeecs.v11.i1.pp195-202. (2019), https://doi.org/10.1007/s13369-019-03759-0.
[58] M.H.T. Imam, S. Afshari, S. Mishra, An experimental survey of feedback control [86] M.W. Ahmad, M. Mourshed, B. Yuce, Y. Rezgui, Computational intelligence
methodologies for advanced lighting systems, Energy Build. 130 (2016) 600–612, techniques for HVAC systems: a review, Build. Simul. 9 (2016) 359–398, https://
https://doi.org/10.1016/j.enbuild.2016.08.088. doi.org/10.1007/s12273-016-0285-4.
[59] A. Pandharipande, D. Caicedo, Smart indoor lighting systems with luminaire- [87] A. Castillo-Martinez, J. Ramon Almagro, A. Gutierrez-Escolar, A. del Corte, L.
based sensing: a review of lighting control approaches, Energy Build. 104 (2015) J. Castillo-Sequera, M.J. G�omez-Pulido, J.-M. Guti�errez-Martínez, Particle swarm
369–377, https://doi.org/10.1016/j.enbuild.2015.07.035. optimization for outdoor lighting design, Energies 10 (2017), https://doi.org/
[60] Y.-J. Wen, A.M. Agogino, Control of wireless-networked lighting in open-plan 10.3390/en10010141.
offices, Light. Res. Technol. 43 (2011) 235–248. http://journals.sagepub.com/doi [88] G.F.M. Lima, J. Tavares, I.S. Peretta, K. Yamanaka, A. Cardoso, E.L. Jr.,
/pdf/10.1177/1477153510382954. Optimization of lighting design using Genetic algorithms, in: 9th IEEE/IAS Int.
[61] F. Tan, D. Caicedo, A. Pandharipande, M. Zuniga, Sensor-driven, human-in-the- Conf. Ind. Appl., INDUSCON), 2010.
loop lighting control, Light, Respir. Technol. (2017) 1–21, https://doi.org/ [89] W. Si, H. Ogai, K. Hirai, H. Takahashi, M. Ogawa, An improved PSO method for
10.1177/1477153517693887, 0. energy saving system of office lighting, SICE Annu. Conf. (2011) 1533–1536,
[62] A. Pandharipande, G.R. Newsham, Lighting controls: evolution and revolution, 2011.
Light. Res. Technol. 50 (2018) 115–128, https://doi.org/10.1177/ [90] L.A. Mendes, R.Z. Freire, L. dos S. Coelho, A.S. Moraes, Minimizing computational
1477153517731909. cost and energy demand of building lighting systems: a real time experiment
[63] N. van de Meugheuvel, A. Pandharipande, D. Caicedo, P.P.J. van den Hof, using a modified competition over resources algorithm, Energy Build. 139 (2017)
Distributed lighting control with daylight and occupancy adaptation, Energy 108–123, https://doi.org/10.1016/j.enbuild.2016.12.072.
Build. 75 (2014) 321–329, https://doi.org/10.1016/j.enbuild.2014.02.016. [91] Y. Cui, Z. Geng, Q. Zhu, Y. Han, Review: multi-objective optimization methods
[64] A. Peruffo, A. Pandharipande, D. Caicedo, L. Schenato, Lighting control with and application in energy saving, Energy 125 (2017) 681–704, https://doi.org/
distributed wireless sensing and actuation for daylight and occupancy adaptation, 10.1016/j.energy.2017.02.174.
Energy Build. 97 (2015) 13–20, https://doi.org/10.1016/j.enbuild.2015.03.049. [92] R. Yang, L. Wang, Multi-objective optimization for decision-making of energy and
[65] P.K. Soori, M. Vishwas, Lighting control strategy for energy efficient office comfort management in building automation and control, Sustain. Cities Soc. 2
lighting system design, Energy Build. 66 (2013) 329–337, https://doi.org/ (2012) 1–7, https://doi.org/10.1016/j.scs.2011.09.001.
10.1016/j.enbuild.2013.07.039. [93] N. Nassif, S. Kajl, R. Sabourin, Optimization of HVAC control system strategy
[66] G. Boscarino, M. Moallem, Daylighting control and simulation for LED-based using two-objective genetic algorithm, HVAC R Res. 11 (2005) 459–486, https://
energy-efficient lighting systems, IEEE Trans. Ind. Informatics. 12 (2016) doi.org/10.1080/10789669.2005.10391148.
301–309, https://doi.org/10.1109/TII.2015.2509423. [94] H. Rocha, I.S. Peretta, G.F.M. Lima, L.G. Marques, K. Yamanaka, Exterior lighting
[67] ALERA, Installed daylight sensors guide. http://www.aleralighting.com, 2014. computer-automated design based on multi-criteria parallel evolutionary
(Accessed 4 September 2018). algorithm: optimized designs for illumination quality and energy efficiency,
[68] LUTRON, Daylight sensor design and application guide. http://www.lutron.com, Expert Syst. Appl. 45 (2016) 208–222, https://doi.org/10.1016/j.
2014. (Accessed 12 September 2018). eswa.2015.09.046.
[69] W. Dun, Optimization of intelligent illumination in university classroom based on [95] F. Cassol, P.S. Schneider, F.H.R. França, A.J. Silva Neto, Multi-objective
FMRAS control algorithm, Light Eng. 26 (2018) 52–59. optimization as a new approach to illumination design of interior spaces, Build.
Environ. 46 (2011) 331–338, https://doi.org/10.1016/j.buildenv.2010.07.028.

19
K.R. Wagiman et al. Journal of Building Engineering 31 (2020) 101342

[96] C. Yin, X. Huang, S. Dadras, Y. Cheng, J. Cao, H. Malek, J. Mei, Design of optimal [103] L. Wang, H. Li, X. Zou, X. Shen, Lighting system design based on a sensor network
lighting control strategy based on multi-variable fractional-order extremum for energy savings in large industrial buildings, Energy Build. 105 (2015)
seeking method, Inf. Sci. 465 (2018) 38–60, https://doi.org/10.1016/j. 226–235, https://doi.org/10.1016/j.enbuild.2015.07.053.
ins.2018.06.059. [104] S. Carlucci, F. Causone, F. De Rosa, L. Pagliano, A review of indices for assessing
[97] C. Yin, S. Dadras, X. Huang, J. Mei, H. Malek, Y. Cheng, Energy-saving control visual comfort with a view to their use in optimization processes to support
strategy for lighting system based on multivariate extremum seeking with building integrated design, Renew. Sustain. Energy Rev. 47 (2015), https://doi.
Newton algorithm, Energy Convers. Manag. 142 (2017) 504–522, https://doi. org/10.1016/j.rser.2015.03.062.
org/10.1016/j.enconman.2017.03.072. [105] DIAL GmbH, DIALux lighting design software. https://www.dial.de/en/dialux/,
[98] D. Tran, Y.K. Tan, Sensorless illumination control of a networked LED-lighting 2017. (Accessed 20 August 2018).
system using feedforward neural network, IEEE Trans. Ind. Electron. 61 (2014) [106] Mathwork Inc, Matlab (n.d.), https://www.mathworks.com/products/matlab.
2113–2121, https://doi.org/10.1109/TIE.2013.2266084. html. (Accessed 29 August 2018).
[99] W. Si, H. Ogai, T. Li, K. Hirai, A novel energy saving system for office lighting [107] A.A. Baloch, P.H. Shaikh, F. Shaikh, Z.H. Leghari, N.H. Mirjat, M.A. Uqaili,
control by using RBFNN and PSO, in: IEEE 2013 Tencon - Spring, 2013, Simulation tools application for artificial lighting in buildings, Renew. Sustain.
pp. 347–351, https://doi.org/10.1109/TENCONSpring.2013.6584469. Energy Rev. 82 (2018) 3007–3026, https://doi.org/10.1016/j.rser.2017.10.035.
[100] Elsevier, Energy and buildings journal, ScienceDirect (n.d.), https://www.journ [108] A. Galatioto, M. Beccali, Aspects and issues of daylighting assessment: a review
als.elsevier.com/energy-and-buildings/. (Accessed 2 February 2018). study, Renew. Sustain. Energy Rev. 66 (2016) 852–860, https://doi.org/10.1016/
[101] IEEE, IEEE journals, inst. Electr. Electron. Eng. (n.d.), https://www.ieee.org/pu j.rser.2016.08.018.
blications_standards/publications/periodicals/journals_magazines.html. [109] R. Yang, L. Wang, Development of multi-agent system for building energy and
(Accessed 28 January 2018). comfort management based on occupant behaviors, Energy Build. 56 (2013) 1–7,
[102] Elsevier, Renewable and sustainable energy reviews journal, ScienceDirect (n.d.), https://doi.org/10.1016/j.enbuild.2012.10.025.
https://www.journals.elsevier.com/renewable-and-sustainable-energy-reviews. [110] M. Molina-Solana, M. Ros, M.D. Ruiz, J. G� omez-Romero, M.J. Martín-Bautista,
(Accessed 7 January 2018). Data science for building energy management: a review, Renew. Sustain. Energy
Rev. 70 (2017) 598–609, https://doi.org/10.1016/j.rser.2016.11.132.

20