An Advanced IoT-based System for Intelligent Energy Management in Buildings

The energy sector is closely interconnected with the building sector and integrated Information and Communication Technologies (ICT) solutions for effective energy management supporting decision-making at building, district and city level are key fundamental elements for making a city Smart. The available systems are designed and intended exclusively for a predefined number of cases and systems without allowing for expansion and interoperability with other applications that is partially due to the lack of semantics. This paper presents an advanced Internet of Things (IoT) based system for intelligent energy management in buildings. A semantic framework is introduced aiming at the unified and standardised modelling of the entities that constitute the building environment. Suitable rules are formed, aiming at the intelligent energy management and the general modus operandi of Smart Building. In this context, an IoT-based system was implemented, which enhances the interactivity of the buildings’ energy management systems. The results from its pilot application are presented and discussed. The proposed system extends existing approaches and integrates cross-domain data, such as the building’s data (e.g., energy management systems), energy production, energy prices, weather data and end-users’ behaviour, in order to produce daily and weekly action plans for the energy end-users with actionable personalised information.


Introduction
One of the major challenges that the European Union (EU) faces within the scope of sustainable development is the increasing energy demand patterns of cities [1]. European cities should be places of advanced social progress and environmental regeneration, as well as places of attraction and engines of economic growth, based on a holistic integrated approach in which all aspects of sustainability are taken into account [2].
Cities are faced with a number of challenges associated with accommodation, atmosphere, transport and infrastructural development, making difficult for urban communities and cities to realise their objectives. In recent years, cities have been turning to advanced technologies to become Smart Cities. This term is used to describe Information and Communication Technological (ICT) solutions for cities and to highlight ICT importance and potential in helping the city to develop competitive advantages. More specifically, Smart Cities are comprised of cities that work in frugal and sound ways, by incorporating every one of its substructure and administrations into a unified whole and utilising insightful gadgets for observing and control, in order to guarantee maintainability and effectiveness [3,4].
Energy demand is one of the most crucial and multifaceted problems for Smart Cities [5,6]. As the quality of life is being improved, as well as the continuous increase of the population is given, it is obvious that the increase in energy demand is an irreversible situation. This continuous increase in of the current issues faced by the Smart Buildings market, as well as to improve knowledge reasoning and decision-making. Suitable rules are formed, aiming at the intelligent energy management and the general modus operandi of Smart Building. In this context, a web-based tool was implemented, which enhances the interactivity of buildings' energy management systems. The proposed tool collects, stores and represents in real-time the energy data of buildings. Based on real-time data (from heterogeneous and dynamic sources: building's data, energy production, energy prices, weather data and end-users' behaviour), as well as predicted data produced by prediction models (renewable energy production, energy consumption, indoor temperature and energy prices), the tool introduces a list of practical action plans for the buildings' occupants, structured upon a number of rules. The results from its pilot application are presented and discussed.
Apart from the introduction, the paper is structured along five sections. A review of the current state of the art, as well as the actual contribution of the proposed IoT-based system, is provided in Section 2. The internal architecture and the key features of the system (five data capturing modules, semantic framework and action engine) are presented in Section 3. Section 4 is devoted to the presentation of the proposed IoT-based system. Section 5 is devoted to the pilot application. Finally, the last section is summarizing the key issues that have arisen in this paper.

Literature Review
A number of available tools can support energy end-users in monitoring, managing and optimising their energy consumption. An innovative energy-aware IT ecosystem was presented by Fotopoulos et al. [21], providing personalized energy management and awareness services towards occupants' behavioural change. Moreover, an IoT Energy Platform has been developed for the management of IoT energy data [22]. Related products from leading companies in the market are the following: • Schneider Electric StruxureWare™ [23] is a platform of open, interoperable, and scalable software applications that provides energy managers with enterprise, operations or control level responsibility to optimise energy usage.

•
Honeywell Attune Advisory Services enable on-going monitoring and optimisation of building energy performance. Attune is powered by cloud-based and Software as a Service (SaaS) technologies and energy and automation experts, which help facilities to determine how to best save energy, time and money [24]. • Siemens Synco™ is a control system for small and medium-size multipurpose buildings, such as shops, offices and apartments. Ameresco's Intelligent Solutions (AIS) energy data platform is comprised of a suite of services with its core energy efficiency offerings [31].
Some of the above-mentioned tools are energy management systems at the building level and others are modelling tools with functions that help optimising the energy systems. The available solutions focus mainly on energy data visualizations and notifications. In fact, these tools can apply and process just some of the input data elaborated by the proposed IoT-based System, to provide, in some cases, only monitoring and controlling activities, in others, energy analyses to help users make decisions on reducing energy consumption at the building level.

Adopted Approach
In Table 1, the entire operating process is represented, underlying the inputs received by the models and the benefits for the overall environment and for the single users that the action plans could bring.  [29]. • Predictive Energy Optimization™ is Building IQ's software platform, designed to improve the energy efficiency of large, complex commercial, public, and academic buildings [30]. • Ameresco's Intelligent Solutions (AIS) energy data platform is comprised of a suite of services with its core energy efficiency offerings [31].
Some of the above-mentioned tools are energy management systems at the building level and others are modelling tools with functions that help optimising the energy systems. The available solutions focus mainly on energy data visualizations and notifications. In fact, these tools can apply and process just some of the input data elaborated by the proposed IoT-based System, to provide, in some cases, only monitoring and controlling activities, in others, energy analyses to help users make decisions on reducing energy consumption at the building level.

Adopted Approach
In Table 1, the entire operating process is represented, underlying the inputs received by the models and the benefits for the overall environment and for the single users that the action plans could bring. The proposed system extends existing approaches and integrates cross-domain data, such as building's data (e.g., energy management systems and other de-centralized sensor-based data), energy production, energy prices, weather data and end-users' behaviour, in order to produce daily and weekly action plans for the energy end-users with actionable personalised information. These action plans are based on the data captured and short-term predictions of the user's behaviour and energy usage. They include notifications for certain thresholds, analytical tailor-made recommendations and saving tips in the users' daily routines (e.g., load shifting, occupancy, set-point adjustment).

Benefits
CO 2 emissions reduction Energy consumption cut down Experience improvement Energy Cost The proposed system extends existing approaches and integrates cross-domain data, such as building's data (e.g., energy management systems and other de-centralized sensor-based data), energy production, energy prices, weather data and end-users' behaviour, in order to produce daily and weekly action plans for the energy end-users with actionable personalised information. These action plans are based on the data captured and short-term predictions of the user's behaviour and energy usage. They include notifications for certain thresholds, analytical tailor-made recommendations and saving tips in the users' daily routines (e.g., load shifting, occupancy, set-point adjustment). The added value of the IoT-based System consists in correlating various types of real-time data from different sources, hence integrating different systems, in order to achieve intelligent energy management of buildings, and, potentially, districts. Moreover, the degree of generalization of the system makes this advanced tool easily adaptable to buildings/cities with different features regarding, for example, types of buildings, energy infrastructures and energy demand and not just focused on specific sectors or building targets. It brings together traditional monitoring systems, low-scale energy management systems and IoT practices, in order to achieve smart energy management. Table 2 recaps the abovementioned systems, pointing out the types of input data applied.

Internal Architecture
The proposed IoT-based System combines a series of components, as follows ( Figure 1): • Five data capturing modules, which collect data from different source (building's data, energy production, energy prices, weather data and end-users' behaviour).

•
The semantic framework, which is a communication system that integrates data from multiple sources and domains using Semantic Web technologies.

•
The action engine is an integrated solution for predicting the energy behaviour of buildings and to suggest actions to improve their energy efficiency. It can be integrated with existing middleware solutions to enhance them. The added value of the IoT-based System consists in correlating various types of real-time data from different sources, hence integrating different systems, in order to achieve intelligent energy management of buildings, and, potentially, districts. Moreover, the degree of generalization of the system makes this advanced tool easily adaptable to buildings/cities with different features regarding, for example, types of buildings, energy infrastructures and energy demand and not just focused on specific sectors or building targets. It brings together traditional monitoring systems, low-scale energy management systems and IoT practices, in order to achieve smart energy management. Table 2 recaps the abovementioned systems, pointing out the types of input data applied.

Internal Architecture
The proposed IoT-based System combines a series of components, as follows ( Figure 1): • Five data capturing modules, which collect data from different source (building's data, energy production, energy prices, weather data and end-users' behaviour).

•
The semantic framework, which is a communication system that integrates data from multiple sources and domains using Semantic Web technologies.

•
The action engine is an integrated solution for predicting the energy behaviour of buildings and to suggest actions to improve their energy efficiency. It can be integrated with existing middleware solutions to enhance them.

Data Capturing Modules
Specific data capturing modules has been developed for acquisition of site-specific data. Java or Python applications have been used for the development of the data capturing modules.

•
Decentralized sensors indicate the real-time conditions on the spot by providing measurements of specific parameters such as the energy consumption, indoor temperature and humidity, etc.

•
The module for Renewable Energy Sources (RES) production informs on the current level of self-production of energy of the connected renewable energy systems. • The weather forecast module is able to provide a comparison of the forecast and the actual field conditions, for the creation of real-time energy balances. • The energy prices module gives indication on the actual costs applicable for those who can adjust their energy contract to the current tariffs.

•
The occupants' feedback module is intended to gather the feedback about the comfort conditions of the occupants and other energy-related issues.

Semantic Framework
The second part of the procedure involved a communication system that integrates data from multiple sources (monitoring systems, Web Services, CSV files, etc.) and domains, with the purpose of contextualizing them in specific contexts, using Semantic Web technologies. It is based on the publish-and-subscribe communication pattern. More specifically, it has been implemented with the Ztreamy system, a semantic service implemented as a Python application. This service processes and contextualizes the data acquired from multiple sources. The Semantic Framework uses the Virtuoso triple-store as a data repository.
In this context, a relevant ontology was created (entitled OPTIMUS ontology) for all entities that are either included or related to the Smart Building environment and constitute the main vocabulary upon which the rules were based [41]. Figure 2 shows an excerpt of the OPTIMUS ontology referring to dynamic data, in particular to energy production sensors.

Data Capturing Modules
Specific data capturing modules has been developed for acquisition of site-specific data. Java or Python applications have been used for the development of the data capturing modules.

•
Decentralized sensors indicate the real-time conditions on the spot by providing measurements of specific parameters such as the energy consumption, indoor temperature and humidity, etc.

•
The module for Renewable Energy Sources (RES) production informs on the current level of selfproduction of energy of the connected renewable energy systems.

•
The weather forecast module is able to provide a comparison of the forecast and the actual field conditions, for the creation of real-time energy balances. • The energy prices module gives indication on the actual costs applicable for those who can adjust their energy contract to the current tariffs.

•
The occupants' feedback module is intended to gather the feedback about the comfort conditions of the occupants and other energy-related issues.

Semantic Framework
The second part of the procedure involved a communication system that integrates data from multiple sources (monitoring systems, Web Services, CSV files, etc.) and domains, with the purpose of contextualizing them in specific contexts, using Semantic Web technologies. It is based on the publish-and-subscribe communication pattern. More specifically, it has been implemented with the Ztreamy system, a semantic service implemented as a Python application. This service processes and contextualizes the data acquired from multiple sources. The Semantic Framework uses the Virtuoso triple-store as a data repository.
In this context, a relevant ontology was created (entitled OPTIMUS ontology) for all entities that are either included or related to the Smart Building environment and constitute the main vocabulary upon which the rules were based [41]. Figure 2 shows an excerpt of the OPTIMUS ontology referring to dynamic data, in particular to energy production sensors. The ontology stands for a model of the static (e.g., building and technical systems features) and the dynamic (e.g., metering) characteristics of a building and their context (e.g., climate conditions The ontology stands for a model of the static (e.g., building and technical systems features) and the dynamic (e.g., metering) characteristics of a building and their context (e.g., climate conditions and energy costs). In the field of ontology engineering, it is considered to be a good practice to reuse existing ontologies or vocabularies to avoid reinventing the wheel and to increase the interoperability of the ontology. The developed ontology is based on already existing ontologies, such as Urban Energy ontology [42] and Semantic Sensor Network ontology [43].
With respect to the static data, the Urban Energy ontology has been extended to model the building and technical system features, such as building geometry, building thermal envelope, Domestic Hot Water (DHW) systems, space cooling/heating systems, and energy generator. The concepts and properties that are not included in this ontology have been created in a new ontology called OPTIMUS.
The Urban Energy ontology has been chosen because it conceptualizes the same domain as the OPTIMUS ontology, and because it is based on existing energy information standards, including: ISO/IEC CD 13273 Energy efficiency regulation and renewable energy sources; ISO/DTR 16344 Common terms, the definitions and symbols for the overall energy performance rating and certification of buildings; ISO/CD 16346 Assessment of overall energy performance of buildings; ISO/DIS 12655 Presentation of real energy use of buildings; ISO/CD 16343 Methods for expressing energy performance and for energy certification of buildings; and ISO 50001:2011 Energy management systems-requirements with guidance for use [44].
Concerning the dynamic data, the Semantic Sensor Network (SSN) ontology has been extended to include different metering systems. The SSN ontology can describe sensors and observations. It is based on the Stimulus-Sensor-Observation ontology design pattern. In particular, this ontology includes capabilities, measurement processes, observations and deployments in which sensors are used [45]. The ontology is aligned with an upper ontology (i.e., Dolce Ultra Light ontology) and it is compatible with SensorML and O&M (Observations and Measurements) standards of the Open Geospatial Consortium. The SSN ontology describes sensors (i.e., ssn:Sensor) as physical objects that observe and transform incoming stimuli into another representation, where stimuli (i.e., ssn:Stimulus) are changes or states in an environment that a sensor uses to measure a property and where observations (i.e., ssn:Observation) are contexts for interpreting incoming stimuli and fixing parameters such as time and location. Since the SSN ontology provides only core concepts, it needs to be extended with domain specific terms. The domain terms that already exist in the Urban Energy model have been used while those that are not included in it have been created as concepts of the OPTIMUS ontology.
The Semantic Framework can be found on the central open source platform, Github [46]. The main contextual data added are the type of sensor (e.g. building's data, energy production, energy prices, weather data and end-users' behaviour) and properties observed (e.g. PVSystem_Peak_Power and Solar_Irradiation). The contextual triples are generated according to the OPTIMUS ontology. For each stream, the following parameters have to be configured:

Action Engine
The action engine integrates prediction models, rules and a MariaDB database to store the results. The prediction models are data-driven models to forecast the energy behaviour of a building according to some specific indicators (e.g., renewable energy production, energy consumption, indoor temperature and energy prices). The prediction models are automatically estimated and customized per building given the measure to be forecasted and the data available (e.g., external variables and length of historical data). The estimated model can then be directly used to predict in a reliable and accurate way the measure across the upcoming week. Different types of models (times-series, Multiple Linear Regression-MLR, etc.) are considered and the best-fitted one is selected and parameterized per case to achieve the best performance. The prediction models have been implemented as R scripts and RapidAnalytics processes.
The rules (implemented as a Symfony PHP web application) are expert knowledge-based algorithms aimed at giving suggestions for intelligent energy management. They consist of simple logic-based rules (most of them based on logical sentences) that can be implemented and used for better managing a building that is already equipped with a network of sensors. The rules are divided into four individual sections, depending on three fields of application: • Building (management of occupancy, heating and cooling technical systems, indoor thermal comfort, air cooling through air-side economizer strategies); • Building and RES production (management of the generation and on-site RES production and exploitation); • Building, RES production and storage (management the operation of different energy flows towards energy cost reduction).
Each rule or a combination of them generates an Action Plan that is the suggestion for better managing the building with the purpose of decreasing its energy consumption.

IoT-Based System
A web-based system was implemented, integrating the above-mentioned architecture. An important function of the tool is the immediate and complete virtual distribution on the Internet of the energy consumption in buildings. Thus, the user can be constantly updated on the energy consumption and other indicators (energy cost, CO 2 emissions, etc.) wherever located, always with the ease of use of the website.
On the first level, the proposed system collects, analyses and presents data amongst four major groups of indicators, facilitating the energy management (Table 3). Table 3. Indicators.

Index Indicator
Title Unit I GBT-11 Electricity per floor area KWh/m 2 I GBT- 12 Electricity per use per area kWh/m 2 for lighting, cooling, other uses I  Fuel used for heating per floor area lt/m 2 (either Heating oil or Natural Gas) I  Electrical Energy per floor area and user kWh/m 2 /user or kWh/m 2 /manhour I  Fuel used for heating per floor area and user lt/m 2 /user I  Electrical Power kW (constant metering) I  Electrical Power Factor cosφ I GBT-31 CO 2 emissions for Electricity per floor area tn/m 2 I GBT-32 CO 2 emissions for Heating per floor area Lt/m 2 I GBT- 33 Produced electricity by RES (PVs) kWh I  Cost of Electricity per floor area, €/m 2 I GBT- 42 Cost • The first group consists of five indicators and focuses on the building's energy consumption, either electricity or fossil fuels, which is directly compared with the building's surface. It includes both data from realized consumptions and projections for future ones (I GBT-11 , I GBT-12 , I GBT-13 ). The other two indicators that constitute the consumption group are a little bit more detailed (I GBT-14 , I  ). This group of indicators provides valuable information to the users groups, both for monitoring and taking action plans.

•
The second group of indicators is more technical and focuses on the power efficiency, in order to address any malfunctions (I GBT-21 , I GBT-22 ).

•
The third group emphasizes on the energy management's environmental impact, through calculating the damage that is done or is avoided, depending on the way that the consumed energy is produced. The group of indicators varies with the location of the building, as well as its features (I GBT-31 , I GBT-32 , I GBT-33 ). The values of indicators are based on virtual sensors, namely synthetic sensors whose date is obtained from existing sensors. For example, a virtual sensor can be created to obtain the total energy consumption of a building per square meter that can have several physical sensors for monitoring the data consumptions of the different sections.
At a second level, simple rules can be applied, giving suggestions for the improvement of energy management related to the management of the occupancy, heating and cooling technical systems, indoor thermal comfort, air cooling through air-side economizer strategies, generation and on-site RES production and exploitation, etc.
Once the user has logged in, the main screen provides some general information about the buildings and their energy performance (Figure 3). For the energy consumption monitoring of a specific category of assets, we are allowed, by choosing particular filters, to export relevant diagrams for the energy consumption. We can choose which asset(s) we wants to review, and as to what energy or environmental indicator.
Moreover, the user can chose among different options, such as action plans, historical data, weekly report and user activity (Figure 4). Another useful feature of the system is the fact that it provides the user with information related to the parameters affecting the action plans and the buildings. That way, the user can conclude whether there is any problem with the sensors and validate the suggestions provided by the algorithms of the system. • The first group consists of five indicators and focuses on the building's energy consumption, either electricity or fossil fuels, which is directly compared with the building's surface. It includes both data from realized consumptions and projections for future ones (IGBT-11, IGBT-12, IGBT-13). The other two indicators that constitute the consumption group are a little bit more detailed (IGBT-14, IGBT-15). This group of indicators provides valuable information to the users groups, both for monitoring and taking action plans.

•
The second group of indicators is more technical and focuses on the power efficiency, in order to address any malfunctions (IGBT-21, IGBT-22).

•
The third group emphasizes on the energy management's environmental impact, through calculating the damage that is done or is avoided, depending on the way that the consumed energy is produced. The group of indicators varies with the location of the building, as well as its features (IGBT-31, IGBT-32, IGBT-33). The values of indicators are based on virtual sensors, namely synthetic sensors whose date is obtained from existing sensors. For example, a virtual sensor can be created to obtain the total energy consumption of a building per square meter that can have several physical sensors for monitoring the data consumptions of the different sections.
At a second level, simple rules can be applied, giving suggestions for the improvement of energy management related to the management of the occupancy, heating and cooling technical systems, indoor thermal comfort, air cooling through air-side economizer strategies, generation and on-site RES production and exploitation, etc.
Once the user has logged in, the main screen provides some general information about the buildings and their energy performance (Figure 3). For the energy consumption monitoring of a specific category of assets, we are allowed, by choosing particular filters, to export relevant diagrams for the energy consumption. We can choose which asset(s) we wants to review, and as to what energy or environmental indicator.
Moreover, the user can chose among different options, such as action plans, historical data, weekly report and user activity (Figure 4). Another useful feature of the system is the fact that it provides the user with information related to the parameters affecting the action plans and the buildings. That way, the user can conclude whether there is any problem with the sensors and validate the suggestions provided by the algorithms of the system.
The system can be appropriately customized to the users' requirements and building characteristics. All aspects of the system are open sourced. The code for data capturing modules, semantic framework, prediction models and rules are showcased in a distinct repository in order to be easily reusable [46].

Impact Assessment Methodology
The general framework for the assessment of the impact is based on the comparison of the energy consumption before and after the real implementation of the action plans (pre-action vs. postaction). The general framework includes the following four phases [47]: • Pre-action phase: The energy consumption can be assessed by means of inverse models or forward models. The inverse models are built through the real-time data collection related both to climate and users (input data) and to historical energy consumption (output data). The forward models are fed by data related to climate, users, equipment, lighting (input data) and by building features (fixed parameters). The historical energy consumption data are then used to calibrate the model.

•
Pre action tailoring with post-action input data: In order to make the comparison between preaction and post-action energy consumption consistent, the models (inverse or forward) developed in the previous phase need to be tailored. This means that the calibrated models are tailored considering the boundary conditions (climate and user) occurring in the post action phase.

•
Post-action phase: In the post-action phase, the energy consumption can be assessed in two different ways: modification of the forward model through the application of the inference rules (when the implementation of the action plans is simulated); energy monitoring (when the action plans are actually implemented). • Impact assessment: The impact of the action plans application can be assessed in four different ways: inverse model vs. inverse model; monitoring vs. forward model; monitoring vs. inverse model; forward model vs. forward model.

Building's Characteristics
The presented IoT-based system was applied to the building premises of the Decision Support System Laboratory of the School of Electrical and Computer Engineering, National Technical University of Greece (DSS Lab). The pilot operation aimed at the combination and interconnection of an advanced IoT-based system, with smart automation systems and smart technologies and equipment (smart meters, sensors, etc.).
The DSS Lab building is located in Athens, Greece. It is an old building, which was built in 1979. The building is covering an area of 244 m 2 and consists of two floors, where the offices of the employees are situated, and also one meeting room. Normal working hours for the offices are 10:00 a.m. to 8:00 p.m. The building is occupied by 40 employees on an average day. The system can be appropriately customized to the users' requirements and building characteristics. All aspects of the system are open sourced. The code for data capturing modules, semantic framework, prediction models and rules are showcased in a distinct repository in order to be easily reusable [46].

Impact Assessment Methodology
The general framework for the assessment of the impact is based on the comparison of the energy consumption before and after the real implementation of the action plans (pre-action vs. post-action). The general framework includes the following four phases [47]:

•
Pre-action phase: The energy consumption can be assessed by means of inverse models or forward models. The inverse models are built through the real-time data collection related both to climate and users (input data) and to historical energy consumption (output data). The forward models are fed by data related to climate, users, equipment, lighting (input data) and by building features (fixed parameters). The historical energy consumption data are then used to calibrate the model.

•
Pre action tailoring with post-action input data: In order to make the comparison between pre-action and post-action energy consumption consistent, the models (inverse or forward) developed in the previous phase need to be tailored. This means that the calibrated models are tailored considering the boundary conditions (climate and user) occurring in the post action phase.

Building's Characteristics
The presented IoT-based system was applied to the building premises of the Decision Support System Laboratory of the School of Electrical and Computer Engineering, National Technical University of Greece (DSS Lab). The pilot operation aimed at the combination and interconnection of an advanced IoT-based system, with smart automation systems and smart technologies and equipment (smart meters, sensors, etc.).
The DSS Lab building is located in Athens, Greece. It is an old building, which was built in 1979. The building is covering an area of 244 m 2 and consists of two floors, where the offices of the employees are situated, and also one meeting room. Normal working hours for the offices are 10:00 a.m. to 8:00 p.m. The building is occupied by 40 employees on an average day.
In this context, equipment was placed across the premises of the DSS Lab and a pilot connection to the proposed IoT-based system was implemented ( Figure 5). A number of sensors and energy meters were placed across the premises and the areas of the lab, in order to provide useful data to the system. More specifically, the installed equipment is used to provide measurements about indoor and outdoor temperature, humidity, lighting function and energy consumption for each area (47 data streams in total), as follows: In this context, equipment was placed across the premises of the DSS Lab and a pilot connection to the proposed IoT-based system was implemented ( Figure 5). A number of sensors and energy meters were placed across the premises and the areas of the lab, in order to provide useful data to the system. More specifically, the installed equipment is used to provide measurements about indoor and outdoor temperature, humidity, lighting function and energy consumption for each area (47 data streams in total), as follows:

Baseline Scenario
The building is supplied by medium voltage electricity, which is converted to low voltage. In the case of black out, the building makes use of a back-up generator that operates on diesel. Electricity is used to cover the needs for heating, cooling, lighting, etc.

Impact Analysis
The "monitoring vs. forward model" has been applied for the impact analysis. In this case, the tailored forward model developed in the phase "pre action tailoring with post-action input data" is capable of estimating the energy consumption according to different boundary conditions without considering the effect of the action plans. The resulting simulated energy consumption can be

Baseline Scenario
The building is supplied by medium voltage electricity, which is converted to low voltage. In the case of black out, the building makes use of a back-up generator that operates on diesel. Electricity is used to cover the needs for heating, cooling, lighting, etc.

Impact Analysis
The "monitoring vs. forward model" has been applied for the impact analysis. In this case, the tailored forward model developed in the phase "pre action tailoring with post-action input data" is capable of estimating the energy consumption according to different boundary conditions without considering the effect of the action plans. The resulting simulated energy consumption can be considered as a baseline. In order to assess the actual effect of the action plans (impact), a comparison between the actual energy consumption consequent to the application of the action plans and the output of the forward baseline model is carried out.
During the pilot implementation, the following action plans were applied. The building's data (indoor air temperature and energy consumption), weather conditions (outdoor air temperature), energy prices and end-users' behaviour data concerning thermal comfort are integrated. The rules are automatically generated during the configuration of the system:

•
Optimising the boost time of the heating/cooling system taking into account the forecasting of the indoor air temperature and the occupancy levels of the building.

•
Scheduling the set-point temperature by taking into consideration thermal comfort of the occupants. The users were able to choose different schedules and set point temperatures for each office.
The total energy consumption for heating/cooling in 2014 (baseline), as well as the actual/predicted energy consumption for heating/cooling during the period 2015-2016 (pilot operation period), are presented in Figures 6 and 7. The predicted energy consumption is based on the energy consumption for heating/cooling in 2014 and the degree days. The reduction of energy consumption for heating/cooling is estimated at 8.1% in 2015 and 10.9% in 2016. As a result, a significant decrease of the building's operating cost was achieved, estimated at 11.3%. between the actual energy consumption consequent to the application of the action plans and the output of the forward baseline model is carried out. During the pilot implementation, the following action plans were applied. The building's data (indoor air temperature and energy consumption), weather conditions (outdoor air temperature), energy prices and end-users' behaviour data concerning thermal comfort are integrated. The rules are automatically generated during the configuration of the system:

•
Optimising the boost time of the heating/cooling system taking into account the forecasting of the indoor air temperature and the occupancy levels of the building.

•
Scheduling the set-point temperature by taking into consideration thermal comfort of the occupants. The users were able to choose different schedules and set point temperatures for each office.
The total energy consumption for heating/cooling in 2014 (baseline), as well as the actual/predicted energy consumption for heating/cooling during the period 2015-2016 (pilot operation period), are presented in Figures 6 and 7. The predicted energy consumption is based on the energy consumption for heating/cooling in 2014 and the degree days. The reduction of energy consumption for heating/cooling is estimated at 8.1% in 2015 and 10.9% in 2016. As a result, a significant decrease of the building's operating cost was achieved, estimated at 11.3%.   The results revealed the significant potential for energy savings through the installation and operation of the proposed IoT-based system. Taking into consideration the initial cost for the installation and operation of the system, the average annual energy savings as derived from the system pilot application and the annual maintenance cost, the payback period is estimated at two years approximately.  The results revealed the significant potential for energy savings through the installation and operation of the proposed IoT-based system. Taking into consideration the initial cost for the

Future Prospects
The idea of the Smart Buildings could be generalized, including additional application areas, such as the following: • "Pillars" applications, focusing on the street and road lighting control by analysing the lamps' failures and reports' crucial data for the local authorities. • "Electrical Vehicle" applications, processing data from electric vehicles charging stations, namely those parking spaces where electric vehicles supply equipment, is used to charge vehicles.
The above could provide local authorities with tools to manage their energy consumption in different categories of infrastructures encountered in Smart Cities.
Moreover, this extra information could be exploited in order to create more rules and visualizations that could further enhance the cognitive comprehension of each person interested in how the system operates and make it user friendly for further sustainability. The Smart Appliances REFerence (SAREF) ontology could also be used as an input for the ontology generation [48].
The deployment of innovative award incentive mechanisms for the energy end-users could also enable behavioural energy efficiency. The innovative aspect of such mechanisms is that the energy end-users will be able to automatically generate their own coins (through a virtual energy currency), by reducing their energy consumption.

Conclusions
ICT-based solutions that exploit Internet of Things (IoT) technologies can contribute significantly to energy saving, by motivating and supporting behavioural change of the buildings' occupants. In this context, the proposed IoT-based system facilitates energy end-users to know how much energy is consumed in total and what is the contribution of the specific end-user and other peers to that, as well as get personalized recommendations of actions for energy conservation and load shifting, along with an estimation of their impact on energy use and user comfort.
The main aim was to provide a flexible, easy to expand and easily customizable system from an administrator (all permissions) and from a user perspective (view customization) scalable, ICT platform. The system uses data sensors that are installed in the building and measure real-time data as regards consumption of systems and appliances, occupancy data, behavioural data, set points, system setting, etc. It simplifies the complexity of the information gathered by those systems, and put it in the hands of energy end-users (buildings' occupants), in context (e.g., end-users know how to improve the building behaviour when he/she is in the building, performing a specific action). Moreover, it could be used from the city authorities for the monitoring and management of the city's energy status in buildings.
Semantic Web technologies could actually play a key role in the rapid development of various aspects of the Smart City infrastructure where various other research areas could come into play. The intersection of Web Services, Semantic Web and energy management could help city authorities towards a smoother transition to the future of cities.