7

Data Analytics in Smart Buildings

M. Victoria Moreno, Fernando Terroso-Sáenz, Aurora González-Vidal

and Antonio F. Skarmeta

Department of Information and Communications Engineering,

University of Murcia, 30100 Spain

7.1 Introduction
Cities are becoming more and more of a focal point for our economies
and societies at large, particularly because of on-going urbanisation, and the
trend towards increasingly knowledge-intensive economies as well as their
growing share of resource consumption and emissions. To meet public policy
objectives under these circumstances, cities need to change and develop, but
in times of tight budgets this change needs to be achieved in a smart way:
our cities need to become “smart cities”. In order to follow the policy of the
decarbonisation of Europe’s economy in line with the EU 20/20/20 energy and
climate goals, today’s ICT, energy (use), transport systems and infrastructures
have to drastically change. The EU needs to shift to sustainable production and
use of energy, to sustainable mobility, and sustainable ICT infrastructures and
services. Cities and urban communities play a crucial role in this process. Three
quarters of our citizens live in urban areas, consuming 70%1 of the EU’s overall
energy consumption and emitting roughly the same share of Green House
Gas (GHG). Of that, buildings and transport represent the lion’s share. Within
the worldwide perspective of energy efficiency, it is important to highlight
that buildings are responsible for 40% of total EU energy consumption and
generate 36% of GHG [1]. This indicates the need to achieve energy-efficient
buildings to reduce their CO2 emissions and their energy consumption.
Moreover, the building environment affects the quality of life and work of
all citizens. Thus, buildings must be capable of not only providing mechanisms

1
Source: European Commission 2013.

167
168 Data Analytics in Smart Buildings

to minimize their energy consumption (even integrating their own energy

sources to ensure their energy sustainability), but also of improving occupant
experience and productivity. In this chapter, we analyse the important role
that buildings represent in terms of their energy performance at city level
and, even, at world level, where they represent an important factor for
the energy sustainability of the planet. Analysis of the energy efficiency
of the built environment has received growing attention in the last decade
[2–4]. Various approaches have addressed energy efficiency of buildings
using predictive modelling of energy consumption based on usage profiles,
climate data and building characteristics. On the other hand, studies have
demonstrated the impact of displaying public information to occupants and
its effect in modifying individual behaviour in order to obtain energy savings
[5, 6]. Nevertheless, most of the approaches proposed to date only provide
partial solutions to the overall problem of energy efficiency in buildings,
where different factors are involved in a holistic way, but which, until now,
have been addressed separately or even neglected by previous proposals.
This division is frequently due to the uncertainty and lack of data and
inputs included in the management processes, so that analysis of how energy
in buildings is consumed is incomplete. In other words, a more integral
vision is required to provide accurate models of the energy consumed in
buildings [7].
The need for the robust characterization of energy use in buildings has
gained attention in light of the growing number of projects and developments
addressing this topic. Although much interest has been put into smart building
technologies, the research area of using real-time information has not been
fully exploited. In order to obtain an accurate simulation model, a detailed
representation of the building structure and its subsystems is required, although
it is the integration of all these pieces that requires the most significant
effort.
The integration and development of systems based on ICT and, more
specifically, the IoT [8], are important enablers of a broad range of applications,
both for industries and the general population, helping make smart buildings
a reality. IoT permits the interaction between smart things and the effective
integration of real world information and knowledge in the digital world.
Smart (mobile) things endowed with sensing and interaction capabilities
or identification technologies (such as RFID) provide the means to capture
information about the real world in much more detail than ever before.
Regarding this real-world data extraction, the great adoption of personal
handheld devices, like smartphones, has enabled the crowdsensing paradigm
 7.2 Addressing Energy Efficiency in Smart Buildings 169

as a prominent mechanism to capture a wide range of (mobile) data [9]. Unlike

other sensing approaches, in this case the collected data is directly generated
by the users’ personal contrivances, so it can be a useful solution for soliciting
feedback from a sheer number of people in an explicit or implicit manner.
From a smart building perspective, such feedback provides information about
its occupants’ preferences and habits that could be considered in order to come
up with customized energy-efficiency solutions.
Nevertheless, challenges related to: (1) the management of the huge
amount of data provided in real-time by a large number of IoT and crowd
based devices deployed, (2) the interoperability among different ICT, and
(3) the integration of many proprietary protocols and communication standards
that coexist in the ICT market applicable to buildings (such as heating, cooling
and air conditioning machines), need to be faced before flexible and scalable
solutions based on the IoT paradigm can be offered.
The structure of the present chapter is as follows: Section 7.2 describes
the key issues involved in energy efficiency in buildings. Among these issues,
relevant parameters affecting energy consumed in buildings are described and
proposed to be included as input data of building management for energy
efficiency. Then, Section 7.3 reviews the main related works which propose
partial solutions to the problem addressed in this chapter. Section 7.4 presents
a general architecture proposal for management systems of smart buildings,
which is modeled in three layers with different functionalities. Section 7.5
describes our proposal for an energy efficiency building management system.
This proposal tackles three different subproblems, each one of these is intro-
duced here. Section 7.6 summarizes the experiments carried out to evaluate
and validate the different proposals and mechanisms developed in this work.
Finally, Section 7.7 gives some conclusions and an outlook of future work.

7.2 Addressing Energy Efficiency in Smart Buildings

Optimizing energy efficiency in buildings is an integrated task that comprises
the whole lifecycle of the building. For buildings to have an impact at city
level in terms of energy efficiency, different challenges have been identified
in the building value chain (from design to end-of-life of buildings)2 , which
can be summarized as follows:
1. Design. The design of buildings should be integrated, holistic and multi
target.

2
http://www.ectp.org/
170 Data Analytics in Smart Buildings

2. Structure. The structure of buildings should provide features such as

safety, sustainability, adaptability and affordability.
3. Building envelope. This should ensure efficient energy and environmen-
tal performance. Prefabrication is a crucial step to guarantee energy
performance. Multifunctional and adaptive components, surfaces and
finishes to create added energy functionality, and durability should all be
built in.
4. Energy equipment and systems. Advanced heating/cooling and domestic
hot water solutions, including renewable energy sources, should focus
on sustainable generation as well as on heat recovery. Among these
systems, thermal storage (both heat and cold) is recognized as a major
breakthrough in building design. Distributed/decentralised energy gen-
eration should address the key requirement of finding smart solutions for
grid-system interactions on a large scale. ICT smart networks will form
a key component in such solutions. In [10], for instance, the authors
study the communication requirements for smart grids and describe the
most suitable communication protocols, wired and wireless, with special
attention to the latest proposals in this field.
5. Construction processes. These should consider ICT-aided construction,
improving the energy performance delivered, and automated construction
tools.
6. Performance monitoring and management. This should ensure inter-
operability among the different subsystems of the building, including
smart energy management systems that provide flexible actions to reduce
the gap between predicted and actual energy building performance,
occupancy modeling, the fast and reproducible assessment of designed
or actual performance, and continuous monitoring and control during
service life. Finally, knowledge sharing must be considered by means
of open data standards that allow collaboration among stakeholders and
interoperability among systems.
7. End of life. This should include decision-support concerning possible
renovation or the construction of a new building and associated systems.
During these phases it is necessary to continuously re-engineer the indexes
that measure energy efficiency to adapt the energy management system to
the building’s conditions. Hereinafter, we refer only to electrical energy
consumption since other kinds of energy such as fuel, gas or water are beyond
the scope of this work. Taking as reference the energy performance model
for buildings proposed by the CEN Standard EN15251 [11], it proposes
criteria for dimensioning the energy management of buildings, while indoor
 7.2 Addressing Energy Efficiency in Smart Buildings 171

environmental requirements are maintained. According to this standard, there

are static and dynamic conditions that affect the energy consumption of
buildings. Given that each building has a different static model according
to its design, we try to provide a solution for energy efficiency focusing on
analyzing how dynamic conditions affect the energy consumed in buildings.
Thus, we propose an initiative for the challenges involved in the living
stage buildings: Performance monitoring and management mentioned in the
above list. In this stage, we need to identify the main drivers of energy use
in buildings. After monitoring these parameters and analysing the associ-
ated energy consumed, we can model their impact on energy consumption,
and then, propose control strategies to save energy. The main idea of this
approach is to provide anticipated responses to ensure energy efficiency in
buildings.
Bearing in mind all these concerns, we enumerate below the stages [12]
that must be carried out to achieve efficiency building energy management:

1. Monitoring. During the monitoring phase, information from heterogeneous

sources is collected and analysed before concrete actions are proposed to
minimize energy consumption, bearing in mind the specific context of a
given building. Since buildings with different functionalities have different
energy use profiles, it is necessary to carry out an initial characterization
of the main contributors to their energy use. For instance, in residential
buildings the energy consumed is mainly due to the indoor services provided
to their occupants (associated to comfort), whereas in industrial buildings
energy consumption is associated mostly with the operation of industrial
machinery and infrastructures dedicated to production processes. Considering
this, and taking into account the models for predicting the comfort response of
buildings occupants given by the ASHRAE [13], we describe below the
main parameters that must be monitored and analysed before implementing
optimum building energy management systems. In this way, from this set
of parameters affecting energy consumption in buildings, we can extract the
input data to be taken into account.
(a) Electrical devices always connected to the electrical network. In build-
ings, it is necessary to characterize the minimum value of energy consumption
due to electrical devices that are always connected to the electrical network,
since they represent a constant contribution to the total energy consumption
of the building. For this, it is necessary to monitor over a period of time the
energy consumed in the building when there is no other contributor to the total
energy being consumed. This value will be included as an input to the final
172 Data Analytics in Smart Buildings

system responsible for estimating the daily electrical consumption of the

building.
(b) Electrical devices occasionally connected. Depending on the kind of
building under analysis, different electrical devices may be used with different
purposes, such as increase of productivity and comfort. On the other hand,
the operation of such devices could be independent of the participation and
behaviour of the occupants; for example, in the context of a factory or an
office where there are timetables and rules. Whatever the case, recognition
of the operation pattern of devices must be included in the final system
responsible for estimating the daily electrical consumption of the building.
To obtain these patterns it is necessary to monitor previously the associated
energy consumption of every device or appliance. To monitor each component
separately in the total power consumption in a household or an industrial site
over time, cost effective and readily available solutions include Non-Intrusive
Load Monitoring (NILM) techniques [14].
(c) Occupants’ behaviour. Energy consumption of buildings due to the
behaviour of their occupants is one of the most critical points in every building
energy management system. This is mainly because occupant behaviour is
difficult to characterize and control due to its uncertain dynamic. First of
all, it is necessary to have solved the occupants’ localization before behaviour
models associated to them can be provided. Depending on the building context,
the impact of occupants behaviour on total energy consumption is different.
For example, in residential buildings the impact of the behaviour in the
energy consumed is one of the biggest, followed by environmental conditions.
However, in buildings with productive goals, the electricity consumed by the
appliances and devices working for such goals is usually the main contributor
to the total energy consumed in the building. Therefore, it is necessary to
monitor and analyse this issue to be able to provide behaviour patterns that
will be included in the final estimation of the daily energy consumption of the
building. To do so, different techniques, like crowd sensing, can be used to
extract a palette of underlying behavioural patterns. In that sense, occupants’
behaviour can be characterized for features such as:
• Occupants localization data.
• Activity level of occupants.
• Comfort preferences of occupants.
(d) Environmental conditions. Parameters like temperature, humidity, pres-
sure, natural lighting, etc. have a direct impact on the energy consumption of
buildings. Nevertheless, depending on the specific context of the building and
 7.2 Addressing Energy Efficiency in Smart Buildings 173

its requirements, this impact will differ and be greatest in the case of indoor
comfort services (like thermal and visual comfort). Therefore, forecasts of
the environmental condition should be also considered as input for the final
energy consumption estimation of the building.
(e) Information about the energy generated in the building. Sometimes,
alternative energy sources can be used to balance the energy consumption
of the building. Information about the amount of daily energy generated and
its associated contextual features can be used to estimate the total energy
generated in the future. This information allows us to design optimal energy
distribution or/and strategies of consumption to ensure the energy-efficient
performance of the building.
(f) Information about total energy consumption. Knowing the real value
of the energy consumed hourly or even daily permits the performance and
accuracy of the building energy management program to be evaluated, and
make it possible to identify and adjust the system in case of any deviation
between the consumption predicted and the real value. In addition, providing
occupants with this information is crucial to make them aware of the energy
that they are using at any time, and encourage them to make their behaviour
more responsible.
In this work we focus on residential buildings, where both comfort and
energy efficiency is required. As regards the comfort provided in buildings,
we focus on thermal and visual comfort.

2. Information Management. An intelligent management system must pro-

vide proper adaptation countermeasures for both automated devices and
users with the aim of providing the most important services in buildings
(comfort) and satisfying energy efficiency requirements. Therefore, energy
savings needs to be addressed by establishing a trade-off between the quality
of services provided in buildings and the energy resources required for the
same, as well as the associated cost.

3. Automation. Automation systems in buildings take inputs from the sensors

installed in corridors and rooms (light, temperature, humidity, etc.), and
use these data to control certain subsystems such as HVAC, lighting or
security. These and more extended services can be offered intelligently to
save energy, taking into account environmental parameters and the location
of occupants. Therefore, automation systems are essential to answer the needs
for monitoring and controlling energy efficiency requirements [15]. At this
respect, the 1888–2011 IEEE Standard for Ubiquitous Green Community
174 Data Analytics in Smart Buildings

Control Network Protocol [16] describes remote control architecture of digital

community, intelligent building groups, and digital metropolitan networks;
specifies interactive data format between devices and systems; and gives a
standardized generalization of equipment, data communication interface, and
interactive message in this digital community network.

4. Feedback and User Involvement. Feedback on consumption is neces-

sary for energy savings and should be used as a learning tool. Analysis
of smart metering, which provides real-time feedback on domestic energy
consumption, shows that energy monitoring technologies can help reduce
energy consumption by 5% to 15% [5]. As can be deduced, a set of subsystems
should be able to provide consumption information in an effective way. These
subsystems are:
• Electric lighting.
• Boilers.
• Heating/cooling systems.
• Electrical panels.
On the other hand, to date, information in real-time about building energy
consumption has been largely invisible to millions of users, who had to settle
with traditional energy bills. In this, there is a huge opportunity to improve the
offer of cost-effective, user-friendly, healthy and safe products for smart build-
ings, which increase the awareness of users (mainly concerning the energy
they consume), and permit them to bean input of the underlying processes
of the system. This would allow the collection of an unprecedented amount
of data related to users’ interactions and their associated contextual details
(e.g. identity, location and activity) by considering the active involvement of
users along with opportunistic sensing. Then, an appropriated processing of
that user-related data will enable the development of even more customized
services.
Taking into account all the aspects identified as relevant for their impact
in energy consumption of buildings, we review how related works from the
literature tackle them. In this way, we can extract the main limitations and
constraints of these works, and suggest proposals to address them.

7.3 Related Work

A complete review of previous solutions from the literature was carried
out during the development period of the present chapter. We tried to find
ways that would enable us to propose holistic solutions to building energy
 7.3 Related Work 175

management problems, which should address there relevant aspects men-

tioned previously, i.e. a complete monitoring phase, the efficient management
of information, using automation systems and involving occupants during the
system operation. Nevertheless, different proposals were found for different
goals, but none was integrated all the aspects. This was the first constraint
identified among previous solutions. Consequently, we decided to review the
main related work tackling each one of these aspects separately.
As regards the monitoring aspect, initial solutions to energy efficiency in
buildings were mainly focused on non-deterministic models based on simu-
lations. A number of simulation tools are available with varying capabilities.
In [17] and [18] a comprehensive comparison of existing simulation tools is
provided. Among these tools are ESP-r [19] and Energy Plus [20]. However,
this type of approach relies on very complex predictive models based on
static perceptions of the environment. For example, a multi-criteria decision
model to evaluate the whole lifecycle of a building is presented in [21]. The
authors tackle the problem from a multi-objective optimization viewpoint,
and conclude that finding an optimal solution is unrealistic, and that only an
approximation is feasible.
With the incessant progress made in the field of ICT and sensor networks,
new applications to improving energy efficiency are constantly emerging.
For instance, in office spaces, timers and motion sensors provide a useful
tool to detect and respond to occupants, while providing them with feedback
information to encourage behavioural changes. The solutions based on these
approaches are aimed at providing models based on real sensor data and
contextual information. Intelligent monitoring systems, such as automated
lighting systems, have limitations such as those identified in [22], in which
the time delay between the response of these automated systems and the actions
performed can reduce any energy saving, whilst an excessively fast response
can produce inefficient actions. These monitoring systems, while contributing
towards energy efficiency, require significant investment in an intelligent
infrastructure that combines sensors and actuators to control and modify the
overall energy consumption. The cost and difficulty involved in deploying
such networks often constrain their viability. Clearly, an infrastructure-less
system that uses existing technologies would provide a cheaper alternative
to building energy management systems. On the other hand, building energy
management must bare with the inaccuracy of sensors, the lack of adequate
models for many processes and the non-deterministic aspects of human
behaviour.
In this sense, there is an important research area that proposes techniques
of artificial intelligence as a way of providing intelligent building management
176 Data Analytics in Smart Buildings

systems. Rather than solving the above drawbacks. This approach involves
models based on a combination of real data and predictive patterns that
represent the evolution of the parameters affecting the energy consumption
of buildings. An example of such an approach is [23], in which the authors
propose an intelligent system able to manage the main comfort services
provided in the context of a smart building, i.e. HVAC and lighting, while
user preferences concerning comfort conditions are established according to
the occupants’ locations. Nevertheless, the authors only propose the inputs of
temperature and lighting in order to make decisions, while many more factors
are really involved in energy consumption and should be included to provide
an optimal and more complete solution to the problem of energy efficiency
in buildings. Furthermore, no automation platform is proposed as part of the
solution.
Regarding building automation systems, many works extend the domotics
field which was originally used only for residential buildings. A relevant
example is the proposal given in [24], where the authors describe an automa-
tion system for smart homes based on a sensor network. However, the system
proposed lacks automation flexibility, since each node of the network offers
limited I/O capabilities through digital lines, i.e. there is no friendly local
interface for users, and most importantly, integration with energy efficiency
capabilities is weak. The work presented in [25] is based on a sensor network
to cope with the building automation problem for control and monitoring
purposes. It provides the means for open standard manufacturer-independent
communication between different sensors and actuators, and appliances can
interact with each other with defined messages and functions. Nevertheless,
the authors do not propose a control application to improve energy efficiency,
security or living conditions in buildings.
The number of works concerning energy efficiency management in build-
ings using automation platforms is more limited. In [26], for instance, a
reference implementation of an energy consumption framework is provided,
but it only analyses the efficiency of ventilation system. In [27] the deployment
of a common client/server architecture focused on monitoring energy con-
sumption is described, but without performing any control action. A similar
proposal is given in [28], with the main difference that it is less focused
on efficiency indexes, and more on cheap practical devices to cope with a
broad pilot deployment to collect the feedback from users and address future
improvements for the system.
Regarding commercial solutions for the efficient management of building
infrastructures, there are proposals such as those given by the manufacturer
 7.3 Related Work 177

Johnson Controls3 , a company that provides products, services and solutions

that help increase energy efficiency and reduce the operation costs of its
clients’ buildings. Another well-known manufacturer is Siemens4 , who offer
a technical infrastructure for building automation and energy efficiency in
the form of market-specific solutions in buildings and public places. The
main differences between these commercial solutions and our proposal for
automation and energy efficiency management in smart buildings are those
related with the open and transparent character of our proposal, as well as its
capability to gather data from a large number of heterogeneous sources.
As regards user involvement, this can be done by means of their implicit
or explicit feedback. When implicit feedback is considered, an important
line of research focuses on the crowdsensing paradigm [9]. In brief, this
paradigm intends to uncover meaningful behavioural patterns by automati-
cally collecting the digital breadcrumbs of the different sensors that users’
personal devices are equipped with. At the same time, a novel course of
action has paid attention to social networks as a novel datasource to extend
the collection implicit user feedback [29]. Despite its inherent uncertainty,
several works are already able to extract meaningful behavioural patterns by
mainly using social-network feeds [30, 31]. As for explicit user’s feedback, the
crowdsourcing paradigm centers on providing tools to allow the management
of the information explicitly requested to sets of target users [32, 33]. In a
smart building context, crowdsensing or crowdsourcing paradigms have been
mainly used to flow management in indoor areas [34]. Last but not least, in
the building energy management field, some proposals have involved uses in
saving energy in buildings [5, 6]. However, few works have been addressed
this aspect. It is important to note that energy usage feedback in building
energy management systems needs to be provided to users frequently and
over a long time, offering an appliance-specific breakdown, while presented
in a clear and appealing way using computerized and interactive tools.
Concerning the fact that users have little awareness of the energy wastage
associated with their energy consumption behaviours is due partly to the
fact that most people do not know what the optimum comfort conditions are
according to environmental features and their needs. It is clear that, while each
person has his/her own comfort preferences and these preferences are strongly
conditioned by subjective concerns, there are a minimal and a maximum set of

3
http://www.johnsoncontrols.co.uk/content/gb/en/products/building efficiency.html
4
http://www.buildingtechnologies.siemens.com/bt/global/en/energy-
efficiency/Pages/Energy-efficiency.aspx
178 Data Analytics in Smart Buildings

comfort conditions recognized as common to everyone to ensure the quality of

life [35]. Therefore, the confidence and respect that users give to the intelligent
services that are offered to them in terms of comfort and energy efficiency
concerns in smart buildings, are crucial constraints in this type of system.
Nevertheless, thanks to pervasive computing practices, the integration and
development of systems based on IoT support and encourage the cooperation
between humans and devices in terms of:
• Facilitating communication between things and people, and between
things, by means of a collective network intelligence context.
• People’s ability to exploit the benefits of this communication through
their increasing familiarity with ICT.
• A vision where, in certain respects, people and things are homogeneous
agents endowed with fixed computational tools.
Smart buildings should prevent users from having to perform routine and
tedious tasks to achieve comfort, security, and effective energy management.
Sensors and actuators distributed in buildings can make user life more
comfortable; for example: i) room heating can be adapted to user preferences
and to the weather; ii) room lighting can change according to the daylight;
iii) domestic incidents can be avoided with appropriate monitoring and alarm
systems; and, iv) energy can be saved by automatically switching off electrical
equipment when not needed, or regulating their operating power according to
user needs, thus avoiding any energy overuse. In this sense, IoT is a key
enabler of smart services to satisfy the needs of individual users, who apart
from being users of the system, can also be seen as sensors in the same way as
temperature, thermal, humidity and presence sensors deployed in the building.
As can be noted, most of the approaches proposed to date only provide
partial solutions to the overall problem of energy efficiency in buildings,
where, although different factors are involved holistically, until now they
have been addressed separately or even neglected by previous proposals. This
division is frequently due to the uncertainty and lack of data and inputs in
the management processes, so that analysis of how energy in buildings is
consumed is incomplete. In other words, a more integral vision is required to
provide accurate models of the energy consumed in buildings [7]. In this sense,
no solutions have been proposed tackling the full integration of information
related with all relevant aspects directly involved in the energy consumption
of buildings (which are described in Section 7.2). For example, there are
not previous solutions that fully integrate information about the occupants of
buildings, despite of the fact that human behaviour has been recognized as
 7.4 A Proposal of General Architecture for Management Systems 179

one of the most important aspects affecting energy consumption in buildings.

Information about the identities of occupants, their locations and activities,
their comfort preferences, their levels of awareness with the problem of
the high energy consumption of buildings, their participation to get energy
saving, etc. must be included, jointly to other relevant information, in any
building energy management system. In this chapter, we present our own smart
system proposal, which is a holistic and flexible solution based on collecting
and analysing information of both the building context and its occupants,
and propose concrete actions which could be applied in the management of
any controllable infrastructure of buildings to ensure their energy efficient
performance. Our proposal of solution considers occupants as a key piece of
our management system, and we demonstrate the benefits of following this
approach in term of the energy saving achieved in various buildings used as
reference.

7.4 A Proposal of General Architecture for Management

Systems of Smart Buildings
The architecture of our proposal for smart building is modelled in layers which
are generic enough to cover the requirements of different smart environments
of cities, such as intelligent transport systems, security, health assistance or, as
is the case analysed in this chapter, smart buildings. This architecture promotes
high-level interoperability at the communication, information and services
layers. The layers of such architecture are depicted in Figure 7.1, and are
detailed below.

7.4.1 Data Collection Layer

Looking at the lower part of Figure 7.1, input data are acquired from a
plethora of sensor and network technologies such as the Web, local and
remote databases, wireless sensor networks, mobile devices, etc., all of them
forming an IoT ecosystem. In this sense, and considering the instance of this
architecture for the building management system proposed in this chapter, it
gathers information from sensors and actuators deployed in the building. As
for static sensors and actuators can be self-configured and controlled remotely
through the Internet, enabling a variety of monitoring and control applications.
Concerning mobile sensors, mechanisms to pro-actively or passively collect
their reported data is also included in this layer. Given the heterogeneity of
data sources and the necessity of seamless integration of devices and networks,
180 Data Analytics in Smart Buildings

Figure 7.1 Layers of the base architecture for smart buildings ecosystem.

a common language structure to represent data is needed to deal with this

issue. Therefore, the transformation of the collected data from the different
data sources into a common language representation is performed in this
stage.

7.4.2 Data Processing Layer

The data processing layer is responsible for processing the information
collected and making decisions according to the final application context. A
set of information processing techniques is applied to extract, contextualize,
 7.5 IoT-based Information Management System for Energy Efficiency 181

fuse and represent information for the transformation of massive input data
into useful knowledge, which can be distributed later towards the services
layer. Different algorithms can be applied for the intelligent data processing
and decision making processes, depending on the final desired operation of the
system (i.e. the services addressed). Considering the target application of smart
buildings, data processing techniques for covering, among others, security,
tele-assistance, energy efficiency, comfort and remote control services should
be implemented in this layer. And following a user-centric perspective for
services provided, intelligent decisions are made through behaviour-based
techniques to determine appropriate control actions, such as appliances and
lights, power energy management, air conditioning adjustment, etc.

7.4.3 Services Layer

Finally, the specific features for providing services, which are abstracted
from the final service implementation, can be found in the upper layer of
the proposed architecture (see Figure 7.1). Our approach offers a framework
with transparent access to the underlying functionalities to facilitate the
development of different types of final application. This generic proposal of
architecture for smart buildings has been instantiated in the system known
as City explorer. City explorer, which was developed at the University of
Murcia, integrates an automation platform which is divided into an indoor part,
and all the connections with external elements for remote access, technical
tele-assistance, security and energy efficiency/comfort providing services in
buildings. Figure 7.2 shows a schema of City explorer offering ubiquitous
services in the smart buildings field. The main components of City explorer
were presented in details in [36, 37]. The work developed in this chapter is
based on using City explorer as platform of experimentation and validation
of our proposal of building management to achieve energy efficiency. For
this, we have instantiated each generic layer of the architecture shown in
Figure 7.1, with the goal of offering a solution to energy efficiency in smart
buildings.

7.5 IoT-based Information Management System for Energy

Efficiency in Smart Buildings
As mentioned before, our proposal of IBMS uses the City explorer platform
applied to achieve energy efficiency in buildings. Our proposed system has
the capability, among others, to adapt the behaviour of automated devices
 182 Data Analytics in Smart Buildings

Figure 7.2 City explorer applied to smart buildings.

 7.5 IoT-based Information Management System for Energy Efficiency 183

deployed in the building in order to meet energy consumption restrictions,

while maintaining comfort conditions at the occupants’ desired levels.
More specifically, the goals of our intelligent management system are the
following:
• High comfort level: learn the comfort zone from users’ preferences,
guarantee a high comfort level (thermal, air quality and illumination)
and a good dynamic performance.
• Energy savings: combine the control of comfort conditions with an energy
saving strategy.
• Air quality control: provide CO2-based demand-controlled ventilation
systems.
Satisfying the above control requirements implies controlling the following
actuators:
• Shading systems to control incoming solar radiation and natural light as
well as to reduce glare.
• Window opening for natural ventilation or mechanical ventilation sys-
tems to regulate natural airflow and indoor air changes, thus affecting
thermal comfort and indoor air quality.
• Heating/cooling (HVAC) systems.
As a starting point, we focus only on the management of lights and HVAC
subsystems, since they represent the highest energy consumption at building
level. User interactions have a direct effect on the whole system perfor-
mance, because the occupants can take control of their own environment at
any time.
Thus, the combined control of the system requires optimal operation of
every subsystem (lighting, HVAC, etc.), on the assumption that each operates
normally in order to avoid conflicts arising between users’ preferences and
the simultaneous operations of such subsystems. Figure 7.3 shows a schema
of the different subsystems comprising the intelligent management system
integrated in City explorer, where the outputs of the system are forwarded to
the actuators deployed in the building.
As can be seen in Figure 7.3, the first task to solve is related with user
identification and localization, and the second problem is related with the
issues of comfort and energy efficiency in the management of the building. In
the following subsections we describe the different issues involved and which
were solved during this work, and represent our proposal of building energy
management system for energy efficiency.
 184 Data Analytics in Smart Buildings

Figure 7.3 Schema of the modules composing the management system in charge of the building comfort and energy efficiency.
 7.5 IoT-based Information Management System for Energy Efficiency 185

7.5.1 Indoor Localization Problem

In a smart building, embedded sensors measure and record user activities,
making it possible to predict their future behaviour, prepare everything one
step ahead according to the individual user’s preferences or needs, and provide
the most convenient energy efficient services. These services need to operate
by acquiring contextual information both from users and the environment.
Therefore, to make buildings smart and to be able to offer users customized
services, it is indispensable to previously solve the implicit indoor localization
problem. Furthermore, user identities need to be taken into account so that
the intelligent system can learn and manage devices according to their
behaviour and/or preferences. We obviously need to solve user identification in
smart buildings to provide customized comfort services committed to energy
efficiency, but while user privacy must also be respected because occupants
care about their private and social activities, and want full control of how their
personal location information and history are used. Hence, there is a need to
rely on non-intrusive, ubiquitous and cheap sensors to minimise infrastructure
deployment and prevent user dissatisfaction. Indeed, some sensors cannot be
installed in buildings; for instance, in Spain video cameras cannot be legally
used in offices. Problems like this make some localization systems unsuitable
for use in smart buildings.
In the scenario addressed in this work, the whole area of a smart building
is divided into locations (rooms, open areas, corridors, etc.) with different
comfort conditions in each one; for instance, optimum lighting conditions in a
corridor are different from those required in an office; or the optimum level of
air conditioning in an individual bedroom is different from that required in a
very crowded dining room. Furthermore, in each of these areas (an individual
bedroom, a dining room, an office, etc.), it is necessary to carry out a further
division depending on the service area of each comfort appliance deployed.
Therefore, our indoor localization system must be able to locate a user in terms
of regions, which correspond to the service areas of the appliances or devices
involved in her/his comfort condition. Recent years have seen great progress
in indoor localization systems, but there are still some weaknesses in terms of
the accuracy of location data, the time required for calibration processes, poor
robustness, or high installation and equipment costs [38]. Furthermore, when
user identification is needed, most of the systems proposed present difficulties
concerning complexity, computational load and inaccurate results. Since the
indoor localization problem does not have obvious solutions, we review
relevant solutions from the literature and identify the technological options
186 Data Analytics in Smart Buildings

most suitable in light of our problem. Accuracy is usually the most important
requirement for positioning systems. In the location problem involved in
energy efficiency of buildings, we conclude that the accuracy required for
our localization system depends on the service areas of the appliances and
devices involved in the comfort and energy balance of the building.
In Figure 7.4 a rough outline of some positioning systems is presented,
with their accuracy ranges achieved until now according to the literature.
Since each localization technology has its particular advantages and disad-
vantages, we suggest that by combining several complementary technologies
and applying data fusion techniques, it is possible to improve the overall
system performance and provide a more reliable indoor localization system,
since more specific inferences can be achieved than when using a single kind of
data sensor. Therefore, after analysing Figure 7.4, we choose a hybrid solution
based on RF and non-RF technologies. Our technological solution to cover the
localization needs (i.e. those required by smart buildings to provide occupants
with customized comfort services) is based on a single active RFID system
and several Infra-Red (IR) transmitters. In Figure 7.5 we can observe the data
exchange carried out among the different technological devices that compose
our localization system.

Figure 7.4 Outline of some positioning technologies [38].

 7.5 IoT-based Information Management System for Energy Efficiency 187

Figure 7.5 Localization scenario.

The final mechanism implemented for indoor localization is shown in

Figure 7.6. In this figure, we can see that the first phase of our localization
mechanism is the space division through the installation of IR devices in the
walls of the building area where localization wants to be solved. Therefore,
for each space division, there is an IR identifier value (IDir) associated to this
region. For each one of these region, we implement a regression method based
on Radial Basis Functions (RBF) networks. The RBF estimates user positions
given different RFID tags situated in the roof. This RFID-based information
coming from the different building’s occupants conforms a data stream that
could be also processed by means of a crowdsensing approach so as to track
the flow of people within a building. In that sense, several proposals already
exist that intends to reconstruct the behaviour of people by using the type of
discrete locations [39].
In our localization mechanism, after the position estimation using the RBF
network, a Particle Filter (PF) is applied as a monitoring technique, which takes
into account previous user position data for estimating future states according
to the current system model. In the PF, we modify particle weights according to
the distances to the measurements during the correction stage, as the following
equation shows:

y t |−
p(−
→ →
x t ) · p(−
→
x t |−
→
x t−1 )
w(−
→
x t ) = w(−
→
x t−1 ) · →
− →
− →
− (7.1)
q( x t | x t−1 , y t )
 188 Data Analytics in Smart Buildings

Figure 7.6 Data processing for location estimation.

 7.5 IoT-based Information Management System for Energy Efficiency 189

where w(− →x t ) weights of the set of particles at instant t; p(− →y t |−

→
x t ) and
→
− →
−
p( x t | x t−1 ) gives the probabilistic behaviour of the output and the state
model of the system respectively, and q(− →
x t |−
→
x t−1 , −
→
y t ) is the approximation
of the expectedly function.
Algorithm 7.1 provides a summarized version of the general definition
of PF. The PF used in this work is slightly different from its generic
definition. The main difference of our filtering algorithm is in the cor-
rection stage, which applies the resample using the Sequential Importance
Sampling (SIS) algorithm [40] (step 13 of Algorithm 7.1). During this
step, information about the specific IR region at a given instant of time is
also used to benefit those particles which fall inside this area. Therefore,
before applying Equation (7.1), we filter according to the condition given by
Equation (7.2):

{If : yt ∈ Ωj ⇒ w(xit ) = 0 ∀ xit ∈

/ Ωj }, (7.2)

where Ωj represents the coverage area of the IR transmitter with identifier j,

and yt and w(xit ) denote, respectively, the measured parameter and the weight
of the set of particles i at the instant of time t. The main advantage of this
constraint is the faster convergence of the filter, because extra information is
available to carry out the correction stage.

Algorithm 7.1 Generic PF

Require: {xit−1 , wt−1
i
}N
i=1 , yt
s

i i Ns
Ensure: {xt , wt }i=1
1: Given a particle number Ns
2: Given a threshold NT value for resampling
3: for i = 1 to Ns do
4: Draw xit ∼ q(xt |xit−1 , yt )
5: Assign the particle a weight wti
6: end for
7: Calculate total weight: t = SUM[{wti }N
i=1 ]
s

8: for i = 1 to Ns do
9: Normalize: wti = t−1 · wti
10: end for
11: calculate N  cf f = Ns
1
i 2
i=1 (wt )
12: if Ncf f ≤ NT then
13: Correction stage.
14: end if
190 Data Analytics in Smart Buildings

7.5.2 Building Energy Consumption Prediction

The energy performance model of our BMS is based on the CEN Standard
N15251 [41]. This standard proposes the criteria of design for any build-
ing energy management system. It establishes and defines the main input
parameters for estimating building energy requirements and evaluating the
indoor environment conditions. The inputs considered to solve our problem
are the data coming from the RFID cards of users, the user interaction with the
system through the control panels or the web access, environmental parameters
coming from temperature, humidity and lighting sensors installed in outdoor
and indoor spaces, the consumption energy sensed by the energy meters
installed in the building, and the generated energy sensed by the energy meters
installed in the solar panels deployed in our testbed. After collecting the data, it
is mandatory to continue with their cleaning, preprocessing, visualization and
correlation study in order to find determining features, which can be used to
generate optimal energy consumption models of buildings (management layer
of the architecture presented in Section 7.4). Over the input set, we perform
the standardization and reduction of data dimensionality using Principal
Components Analysis (PCA) [42], identifying the directions in which the
observations of each parameter mostly vary.
Regarding theArtificial Intelligence (AI) techniques that have been already
applied successfully to generate energy consumption models of buildings
in different scenarios (as such we mentioned in the management layer
of the architecture presented in Section 7.4), we propose to evaluate the
performance of Multilayer Perceptron (MLP), Bayesian Regularized Neural
Network (BRNN) [43], SVM [44] and Gaussian Processes with RBF Ker-
nel [45]. They were selected because of the good performance that all of
them have already provided when they are applied to building modelling.
All these regression techniques are implemented following a model-free
approach, which is based on selecting – for a specific building – the optimal
input set and technique, i.e. such input set and technique that provides the
most accurate predictive results in a test data set. In order to implement
this free-model approach, we use the R [46] package named CARET [47] to
train the energy consumption predictive algorithms, looking for the optimal
configuration of their hyper-parameters.
The selected metric to evaluate the models generated for each technique
using test sets is the well-known RMSE (Root-Mean-Square Error), which
formulation appears in Equation (7.3).
 7.5 IoT-based Information Management System for Energy Efficiency 191

 n
1 
RM SE =  (yi − ŷi )2 (7.3)
n
i=1

This metric shows the error by means of the quantity of KWh that we deviate
when predicting, but in order to get a better understanding of the uncertainty
of the model, we also show its coefficient of variation (CVRMSE). This
coefficient is the RMSE divided by the mean of the output variable (energy
consumption) for the test set (Equation (7.4)), giving us a percentage of error
adjusted to the data, not just a number in general terms.
RM SE
CV RM SE = (7.4)
y

7.5.3 Optimization Problem

Once the building energy consumption is modeled we focus on the optimiza-
tion of its use trying to keep comfort conditions. As starting point, we establish
the comfort extremes considering location type, user activity and date [48].
Understanding the building thermal and energetic profiles allows us to quantify
the effects of particular heating-cooling set point decisions. To derive a heating
or cooling schedule, it is necessary to formulate the target outcome. In our
buildings, it is possible to:
1. Optimize the indoor temperature during occupation, i.e. minimize the
building temperature deviations from a target temperature.
2. Minimize daily energy consumption, or
3. Optimize a weighted mixture of the criteria, a so-called multi-objective
optimization.
The definition of building temperature deviation influences the results
strongly: taking the minimum building temperature will result in higher set
point choices and higher energy use than using e.g. the average of building
temperatures. Constraints on maximum acceptable deviation from target
comfort levels or an energy budget can be taken into account to ensure required
performance. For our optimization problem, we apply a genetic optimization
implemented in R (using the “genalg” package [49]) to our predictive building
models to derive schedules for heating/cooling setpoints.

7.5.4 User Involvement in the System Operation

Following this approach to provide human-centric services in the context
of smart buildings, users can be seen as both the final deciders of actions,
192 Data Analytics in Smart Buildings

and system co-designers in terms of feedback that conditions future rules

and contributions to the software issuing these rules. In this sense, in our
energy building management system we consider the data provided directly
by users through their interactions when they change the comfort conditions
provided automatically by the system and, consequently, the system learns
and autoadjusts according to such changes and to the control comfort/energy
strategies defined by users using the graphic editor of City explorer. Further-
more, with the aim of offering users information about any unsuitable design
or setting of the system, as well as to help them easily understand the link
between their everyday actions and environmental impact, City explorer is
able to notify them about such matters (i.e. acting as a learning tool). On
the other hand, when the system detects disconnections and/or failures in the
system, it sends alerts by email/messages to notify users to check these issues.
All these features, which are included in our management system, contribute
to user behaviour changes and increase their awareness as time passes, or
detect unnecessary stand-by consumption of the controllable subsystems of the
building.
Finally, to understand the background of energy behaviour of users
involved in our experiments and to be able to form an initial context pattern
for the usability of the system under different constraints, we carried out
a follow-up study based on the feedback that users provide to City explorer
through the SCADA-web and the control panels installed in the smart building.
Another reason to carry out this study was the identified lack of research in
the building energy management area, where large-scale deployment needs
to be accompanied by a body of study on user behaviour, motivation and
preferences. The same was pointed out by [6]. In Figure 7.7 is shown the
schema of our final building energy management solution.

7.6 Evaluation and Results

7.6.1 Scenario of Experimentation
The reference building where our BMS for energy efficiency is deployed is the
Technology Transfer Centre (TTC) of the UMU5 . Every room of this building
is automated through a Home Automation Module (HAM) unit of the City
explorer platform. It permits us to consider a granularity at room level to
carry out the experiments.

5
www.um.es/otri/?opc=cttfuentealamo
 7.6 Evaluation and Results

Figure 7.7 Schema of the definitive module of our building energy management system.
193
194 Data Analytics in Smart Buildings

Figure 7.8 Tracking processes with a reference tag distribution of 1 m × 1 m.

7.6.2 Evaluation and Indoor Localization Mechanism

Different tracking processes are carried out in the environments considered
in our tests (the TTC building) applying for this the implementation of the
PF described in Algorithm 7.1. In Figure 7.8 an example of some tracking
processes are carried out considering transition between different spaces
of the TTC. For these paths, our system was configured to acquire data
every T = 10 s. (whereas for the rest of the tests a value of T = 5 min.
was considered). Taking into account the target location areas involved
(represented in different colors), and the real and estimated location data
provided by our mechanism, it can be safely said that it was able to monitor
the user locations with a high degree of accuracy and precision.
With an 1m × 1m distribution of reference RFID tags placed on the roof of
the test room, a 65% success percentage in localization is obtained having an
error lower than 1 m. 98% of cases have as much 2.5 m. of error. Therefore, it
 7.6 Evaluation and Results 195

can be safely said that our localization system is able to track users with
a sufficient level of accuracy and precision for the location requirements
associated with the comfort and energy management problem in buildings.
More details about this indoor localization system can be found in [50].

7.6.3 Evaluation. Energy Consumption Prediction

and Optimization
In Figure 7.9(a) it is shown the correlation heatmap between the electrical
consumption of the TTC building and the outdoor environmental conditions.
It is observed that energy consumption correlates significantly (α = 0.95)
and positively with temperature, radiation, wind speed variables, vapour pres-
sure deficit and dew point, and negatively with wind direction and humidity
variables. This means that we can use safely these variables as inputs of the
energy consumption model of our reference building, because they have clear
impact in the energy consumption. Otherwise, precipitations are so unusual
that they don’t have an association with the output.
Also, a logic differentiation between situations has been considered in
order to label behaviour. Situation 1: holidays and weekends, situation 2:
regular mornings, and situation 3: regular afternoons. The non-parametric
Kruskall Wallis test shows that energy consumption differs significantly
between situations (H(2) = 547.7, p < 0.01). Also, the post hoc pairwise
comparisons corrected with Holm’s method retrieve a p-value smaller than
0.01, supporting the decision of creating 3 different models [51].
Thus, for each of the three situations identified for the TTC building, we
have evaluated not only the punctual value of RMSE, but also we have vali-
dated whether one learning algorithm out-performs statistically significantly
the others using the non-parametric Friedman test [52] with the corresponding
post-hoc tests for comparison. Let xj i be the i-th performance RMSE of the jth
algorithm, for this building, we have used 5-times10-fold cross validation, so
i ∈ {1, 2, . . ., 50} and four techniques, so j ∈ {1, 2, 3, 4}. For every situation,
we find significant differences (α = 0.99) between every pair of algorithms,
except for SVM and Gauss RBF (p > 0.01), as it is shown in Figure 7.9(b)
for the particular case of situation 2.
The three models have in common that BRNN yields a better result than
the other tested techniques, based on the RMSE metric. Thus, BRNN is able
to generate a model with a very low mean error of 25.17 KWh – which only
represents the 7.55% of the sample (this is the most accurate result) in terms
of the CVRMSE. And for the worst case, BRNN provides a mean error of
196 Data Analytics in Smart Buildings

Figure 7.9 Modeling results.

 7.6 Evaluation and Results 197

43.76 KWh – which represents the 10.29% of the sample in the reference
TTC building – that is acceptable enough considering that the final aim is to
save energy.
To evaluate our GA-based optimization strategy, controlled experiments
were carried out in the TTC building with different occupant’s behaviours.
The results showed that we can accomplish energy savings between 15% and
31%. Trying to validate the application of our proposal we have applied it
in a different scenario with limited monitoring and automation technologies,
achieving energy saving of about 23%.

7.6.4 Evaluation. User Involvement

For the experiments described here, fifteen people took part in the focus
group studies which help us extract user-preferences and pinpoint design
concerns. Understanding user contexts, such as motivation for saving energy
and the constraints for implementing energy saving behavior, enables better
understanding of user preferences and how the energy monitoring system can
work with users to achieve the best possible behavioral changes.
During the data collection process performed in the experiment, the
subjects were asked to walk freely along the different scenarios consid-
ered, and to work or relax in the different areas designed specifically for
such goals. This experiment was repeated during 3 hours per day consid-
ering different conditions of user movements and activities, environmental
conditions, preferences, etc. At the time of writing, the system has just
completed the first 62 days of measurement, so this time is the base-
line period used to assess the impact of including users in the loop of
our system. During the first 31 days of the experiment, users lacked any
feedback about their energy consumption as well as any control capa-
bility over the setting of comfort and energy levels, but during the last
31 days of the experiment, users were empowered and were included as a
holistic component of the system. During this second phase of the system
operation, the system displayed real time energy usage in kW, cost of energy
usage, energy saving tips, energy usage history (hourly, daily, monthly),
etc. through both SCADA-web and the control panel installed in the target
scenario. Also, during this last phase, users could define their own strategies to
control any appliance or monitor any specific parameters sensed by the system.
Despite the relatively short time of evaluation (one month), a nearly
analysis shows that the system has already had a positive impact on user
behaviors, which can be translated into energy saving terms. Figure 7.10
 198 Data Analytics in Smart Buildings

Figure 7.10 Percentage of energy consumption savings in comfort services considering a user-centric building management efficient.
 7.7 Conclusions and Future Work 199

shows the energy savings achieved during the second month of operation of
our energy management system in contrast to the first experimental month. It
can be seen how we achieved a saving of up to 12% of the energy involved, and
the medium value of 9% for the experimental month. Furthermore, the results
reflect how the increased savings become more stable with time, specifically
from the 17th day of the system operation. The reason of this saving increasing
is because our system is able to learn and adjust itself to any feedback indicated
by users regarding their comfort associated profile, and to recognize patterns
of user behavior.

7.7 Conclusions and Future Work

The proliferation of ICT solutions (IoT among them) represents new oppor-
tunities for the development of new intelligent services, contributing to more
efficient and sustainable cities. In this sense, with the increasing urbanization
seen in recent decades, there is an urgent need to achieve energy-efficient
environments to ensure the energy sustainability of cities. But to achieve
this goal, it is first necessary to solve energy efficiency concerns at building
level, since this constitutes the cornerstone of the overall problem. For greater
energy efficiency in buildings, smart solutions are required to monitor and
control the capabilities offered by wide sensor and actuator networks deployed
as part of the system. Furthermore, occupants play an important role in this
type of system, since they are the recipients of the indoor services provided
by electrical appliances installed in buildings, most of them responsible for
providing them with comfort conditions. In this sense, it is required to propose
building management systems able to tackle energy efficiency requirements
while user comfort conditions are also taken into account. To date, however, the
solutions proposed are mainly based on determinist models with few accurate
predictions, and are not able to consider real-time data in most cases. Indeed,
they do not even come close to reflecting reality.
In this chapter, we propose a building energy management system powered
by IoT capabilities and part of a novel context and location-aware system
that covers the issues of data collection, intelligent processing to save energy
according to user comfort preferences and features that modify the operation
of relevant indoor devices. An essential part of our energy efficiency system
are the key aspects of integrating user location and identity, so that customized
services can be provided to them while any useless energy consumption in the
building is avoided. Furthermore, another relevant feature is users involvement
200 Data Analytics in Smart Buildings

with the system, through their interactions and their participation to get energy
savings in the building.
The applicability of our system has been demonstrated through its installa-
tion in a reference building. Thus, using user location data, considering target
regions of occupancy for comfort and energy management in the building,
and finally including users in the loop of the system operation, we show that
energy consumption in buildings can be reduced by a mean of about 23%.
If we translate this mean value of energy saving to city level, assuming that
buildings represent 40% of the total energy consumption at European level,
a reduction of 9% at city level could be achieved by installing this energy
management system in buildings.
The ongoing work is focused on the inclusion of people behaviour during
the operational loop of this kind of systems for smart cities. Thus, for the case
of smart building applications, users will be encouraged to participate in an
active way through their engagement to save energy. On the other hand, in the
case of the public tram service, data coming from crowd-sensing initiatives
will be integrated to improve the estimation of the urban mobility patterns.

Acknowledgments
This work has been partially funded by MINECO TIN2014-52099-R Project
(grant BES-2015-071956) and ERDF funds, by the European Commis-
sion through the H2020-ENTROPY-649849 and the FP7-SMARTIE-609062
Projects, and the Spanish Seneca Foundation by means of the PD program
(grant 19782/PD/15).

References
[1] D. Petersen, J. Steele, and J. Wilkerson, “Wattbot: a residential electricity
monitoring and feedback system,” in Proceedings of the 27th interna-
tional conference extended abstracts on Human factors in computing
systems, pp. 2847–2852, ACM, 2009.
[2] Y. Agarwal, B. Balaji, R. Gupta, J. Lyles, M. Wei, and T. Weng,
“Occupancy-driven energy management for smart building automation,”
in Proceedings of the 2nd ACM Workshop on Embedded Sensing
Systems for Energy-Efficiency in Building, pp. 1–6, ACM, 2010.
[3] T. D. Pettersen, “Variation of energy consumption in dwellings due to
climate, building and inhabitants,” Energy and buildings, vol. 21, no. 3,
pp. 209–218, 1994.
 References 201

[4] R. Lindberg, A. Binamu, and M. Teikari, “Five-year data of measured

weather, energy consumption, and time-dependent temperature varia-
tions within different exterior wall structures,” Energy and Buildings,
vol. 36, no. 6, pp. 495–501, 2004.
[5] S. Darby, “The effectiveness of feedback on energy consumption,” A
Review for DEFRA of the Literature on Metering, Billing and direct
Displays, vol. 486, p. 2006, 2006.
[6] C. Fischer, “Feedback on household electricity consumption: a tool for
saving energy?,” Energy efficiency, vol. 1, no. 1, pp. 79–104, 2008.
[7] K. Voss, I. Sartori, A. Napolitano, S. Geier, H. Gonçalves, M. Hall, P.
Heiselberg, J. Widén, J. A. Candanedo, E. Musall, et al., “Load matching
and grid interaction of net zero energy buildings,” 2010.
[8] C. Perera, A. Zaslavsky, P. Christen, and D. Georgakopoulos, “Sensing
as a service model for smart cities supported by internet of things,”
Transactions on Emerging Telecommunications Technologies, vol. 25,
no. 1, pp. 81–93, 2014.
[9] B. Guo, Z. Yu, X. Zhou, and D. Zhang, “From participatory sensing
to mobile crowd sensing,” in Pervasive Computing and Communi-
cations Workshops (PERCOM Workshops), 2014 IEEE International
Conference on, pp. 593–598, March 2014.
[10] A. Llaria, J. Jiménez, and O. Curea, “Study on communication technolo-
gies for the optimal operation of smart grids,” Transactions on Emerging
Telecommunications Technologies, 2013.
[11] E. 15251:2006, “Indoor environmental input parameters for design and
assessment of energy performance of buildings – addressing indoor air
quality, thermal environment, lighting and acoustics,” 2006.
[12] M. Hazas, A. Friday, and J. Scott, “Look back before leaping forward:
Four decades of domestic energy inquiry,” IEEE pervasive Computing,
vol. 10, pp. 13–19, 2011.
[13] L. Berglund, “Mathematical modelsfor predicting thermal comfortre-
sponse of building occupants,” in Ashrae Journal- American Society
of Heating Refrigerating and Air Conditioning Engineers, vol. 19,
pp. 38–38, Amer Soc Heat Refrig Air-Conditioning Eng Inc 1791 Tullie
Circle Ne, Atlanta, GA 30329, 1977.
[14] A. Zoha, A. Gluhak, M. A. Imran, and S. Rajasegarar, “Non-intrusive
load monitoring approaches for disaggregated energy sensing: A survey,”
Sensors, vol. 12, no. 12, pp. 16838–16866, 2012.
[15] A. I. Dounis and C. Caraiscos, “Advanced control systems engineering
for energy and comfort management in a building environment—a
202 Data Analytics in Smart Buildings

review,” Renewable and Sustainable Energy Reviews, vol. 13, no. 6,

pp. 1246–1261, 2009.
[16] C. Ninagawa, H. Yoshida, S. Kondo, and H. Otake, “Data transmission
of ieee1888 communication for wide-area real-time smart grid applica-
tions,” in Renewable and Sustainable Energy Conference (IRSEC), 2013
International, pp. 509–514, IEEE, 2013.
[17] M. S. Al-Homoud, “Computer aided building energy analysis techniques,
”Building and Environment, vol. 36, no. 4, pp. 421–433, 2001.
[18] D. B. Crawley, J. W. Hand, M. Kummert, and B. T. Griffith, “Contrasting
the capabilities of building energy performance simulation programs,”
Building and environment, vol. 43, no. 4, pp. 661–673, 2008.
[19] J. Clarke, J. Cockroft, S. Conner, J. Hand, N. Kelly, R. Moore, T.
O’Brien, and P. Strachan, “Simulation-assisted control in building energy
management systems,” Energy and buildings, vol. 34, no. 9, pp. 933–940,
2002.
[20] D. B. Crawley, L. K. Lawrie, F. C. Winkelmann, W. F. Buhl, Y. J. Huang,
C. O. Pedersen, R. K. Strand, R. J. Liesen, D. E. Fisher, M. J. Witte, et
al., “Energy plus: creating a new generation building energy simulation
program,” Energy and Buildings, vol. 33, no. 4, pp. 319–331, 2001.
[21] Z. Chen, D. Clements-Croome, J. Hong, H. Li, and Q. Xu, “Amulticriteria
lifespan energy efficiency approach to intelligent building assessment,”
Energy and Buildings, vol. 38, no. 5, pp. 393–409, 2006.
[22] V. Garg and N. Bansal, “Smart occupancy sensors to reduce energy
consumption,” Energy and Buildings, vol. 32, no. 1, pp. 81–87, 2000.
[23] H. Hagras, V. Callaghan, M. Colley, and G. Clarke,“A hierarchical fuzzy–
genetic multiagent architecture for intelligent buildings online learning,
adaptation and control,” Information Sciences, vol. 150, no. 1, pp. 33–57,
2003.
[24] D.-M. Han and J.-H. Lim, “Design and implementation of smart home
energy management systems based on zigbee,” Consumer Electronics,
IEEE Transactions on, vol. 56, no. 3, pp. 1417–1425, 2010.
[25] P. Oksa, M. Soini, L. Sydänheimo, and M. Kivikoski, “Kilavi platform
for wireless building automation,” Energy and Buildings, vol. 40, no. 9,
pp. 1721–1730, 2008.
[26] D. O’Sullivan, M. Keane, D. Kelliher, and R. Hitchcock, “Improving
building operation by tracking performance metrics throughout the build-
ing lifecycle (blc),” Energy and buildings, vol. 36, no. 11, pp. 1075–1090,
2004.
 References 203

[27] G. Escrivá-Escrivá, C. Álvarez-Bel, and E. Peñalvo-ópez, “New indices

to assess building energy efficiency at the use stage,” Energy and
Buildings, vol. 43, no. 2, pp. 476–484, 2011.
[28] V. Sundramoorthy, G. Cooper, N. Linge, and Q. Liu, “Domesticating
energy-monitoring systems: Challenges and design concerns,” IEEE
Pervasive Computing, vol. 10, no. 1, pp. 20–27, 2011.
[29] G. Bello-Orgaz, J. J. Jung, and D. Camacho, “Social big data:
Recent achievements and new challenges,” Information Fusion, vol. 28,
pp. 45–59, 2016.
[30] B. Pan, Y. Zheng, D. Wilkie, and C. Shahabi, “Crowd sensing of traffic
anomalies based on human mobility and social media,” in Proceedings
of the 21st ACM SIGSPATIAL International Conference on Advances
in Geographic Information Systems, SIGSPATIAL’13, (New York, NY,
USA), pp. 344–353, ACM, 2013.
[31] Massive Online GeoSocial Networking Platforms and Urban Human
Mobility Patterns: A Process Map for Data Collection, ch. 197,
pp. 1586–1593.
[32] D. E. Difallah, M. Catasta, G. Demartini, P. G. Ipeirotis, and P. Cudré-
Mauroux, “The dynamics of micro-task crowdsourcing: The case of ama-
zon mturk,” in Proceedings of the 24th International Conferenceon World
Wide Web, pp. 238–247, International World Wide Web Conferences
Steering Committee, 2015.
[33] C. Cardonha, D. Gallo, P. Avegliano, R. Herrmann, F. Koch, and S.
Borger, “A crowdsourcing platform for the construction of accessibil-
ity maps,” in Proceedings of the 10th International Cross-Disciplinary
Conference on Web Accessibility, W4A’13, (New York, NY, USA),
pp. 26:1–26:4, ACM, 2013.
[34] A. Piscitello, F. Paduano,A.A. Nacci, D. Noferi, M. D. Santambrogio, and
D. Sciuto, “Danger-system: Exploring new ways to manage occupants
safety in smart building,” in Internet of Things (WF-IoT), 2015 IEEE
2nd World Forum on, pp. 675–680, Dec 2015.
[35] A. Handbook, “Fundamentals,” American Society of Heating, Refriger-
ating and Air Conditioning Engineers, Atlanta, vol. 111, 2001.
[36] M. A. Zamora-Izquierdo, J. Santa, and A. F. Gómez-Skarmeta, “An
integral and networked home automation solution for indoor ambient
intelligence,” Pervasive Computing, IEEE, vol. 9, no. 4, pp. 66–77, 2010.
[37] J. Santa, M. A. Zamora-Izquierdo, M. V. Moreno, A. J. Jara, and A. F.
Skarmeta, “Energy-efficient indoor spaces through building automation,”
204 Data Analytics in Smart Buildings

in Inter-cooperative Collective Intelligence: Techniques and Applica-

tions, pp. 375–401, Springer, 2014.
[38] H. Liu, H. Darabi, P. Banerjee, and J. Liu, “Survey of wireless indoor
positioning techniques and systems, ”Systems, Man, and Cybernetics,
Part C: Applications and Reviews, IEEE Transactions on, vol. 37, no. 6,
pp. 1067–1080, 2007.
[39] H. Ji, L. Xie, C. Wang, Y. Yin, and S. Lu, “Crowdsensing: A
crowd-sourcing based indoor navigation using rfid-based delay tolerant
network,” Journal of Network and Computer Applications, vol. 52,
pp. 79–89, 2015.
[40] A. Haug, “A tutorial on Bayesian estimation and tracking techniques
applicable to nonlinear and non-Gaussian processes,” MITRE Corpora-
tion, McLean, 2005.
[41] E. Standard et al., “Indoor environmental input parameters for design
and assessment of energy performance of buildings addressing indoor
air quality, thermal environment, lighting and acoustics,” EN Standard,
vol. 15251, 2007.
[42] H. Abdi and L. J. Williams, “Principal component analysis,” Wiley
Interdisciplinary Reviews: Computational Statistics, vol. 2, no. 4, pp.
433–459, 2010.
[43] L. Hawarah, S. Ploix, and M. Jacomino, “User behavior prediction in
energy consumption in housing using bayesian networks,” in Artificial
Intelligence and Soft Computing, pp. 372–379, Springer, 2010.
[44] Y. Fu, Z. Li, H. Zhang, and P. Xu, “Using support vector machine to
predict next day electricity load of public buildings with sub-metering
devices,” Procedia Engineering, vol. 121, pp. 1016–1022, 2015.
[45] M. Alamaniotis, D. Bargiotas, and L. H. Tsoukalas, “Towards smart
energy systems: application of kernel machine regression for medium
term electricity load forecasting,” SpringerPlus, vol. 5, no. 1, pp. 1–15,
2016.
[46] R Core Team, R: A Language and Environment for Statistical Computing.
R Foundation for Statistical Computing, Vienna, Austria, 2015.
[47] M. Kuhn, “Building predictive models in R using the caret package,”
Journal of Statistical Software, vol. 28, no. 5, pp. 1–26, 2008.
[48] J. A. Orosa, “A new modelling methodology to control hvac systems,”
Expert Systems with Applications, vol. 38, no. 4, pp. 4505–4513, 2011.
[49] E. Willighagen, “Genalg: R based genetic algorithm,” R package version
0.1, vol. 1, 2005.
 References 205

[50] M. V. Moreno, M. Zamora-Izquierdo, J. Santa, and A. F. Skarmeta,

“An indoor localization system based on artificial neural networks
and particle filters applied to intelligent buildings,” Neurocomputing,
vol. 122, pp. 116–125, 2013.
[51] J. M. Andy Field and Z. F. Niblett, Discovering Statistics Using R. Sage
Publications Ltd, 1st ed., 2012.
[52] J. Demšar, “Statistical comparisons of classifiers over multiple datasets,”
The Journal of Machine Learning Research, vol. 7, pp. 1–30, 2006.