Energy saving in existing buildings by an intelligent use of interoperable ICTs

ORIGINAL ARTICLE

Energy saving in existing buildings by an intelligentuse of interoperable ICTs

Anna Osello & Andrea Acquaviva & Chiara Aghemo & Laura Blaso & Daniele Dalmasso &

David Erba & Giovanni Fracastoro & Damien Gondre & Marco Jahn & Enrico Macii &Edoardo Patti & Anna Pellegrino & Paolo Piumatti & Ferry Pramudianto &

Jerôme Savoyat & Maurizio Spirito & Riccardo Tomasi & Joseph Virgone

Received: 1 July 2012 /Accepted: 21 November 2012 /Published online: 5 May 2013# Springer Science+Business Media Dordrecht 2013

Abstract In this paper, we report a methodology, devel-oped in the context of Smart Energy EfficientMiddleware for Public Spaces European Project, aimedat exploiting ICT monitoring and control services toreduce energy usage and CO2 footprint in existing build-ings. The approach does not require significant construc-tion work as it is based on commercial-off-the-shelfdevices and, where present, it exploits and integratesexisting building management systems with new sensorsand actuator networks. To make this possible, the pro-posed approach leverages upon the following maincontributions: (a) to develop an integrated building auto-mation and control system, (b) to implement amiddleware for the energy-efficient buildings domain,

(c) to provide a multi-dimensional building informationmodelling-based visualisation, and (d) to raise people’sawareness about energy efficiency. The research ap-proach adopted in the project started with the selection,as case studies, of representative test and reference roomsin modern and historical buildings chosen for havingdifferent requirements and constraints in term of sensingand control technologies. Then, according to the featuresof the selected rooms, the strategies to reduce the energyconsumptions were defined, taking into account the po-tential savings related to lighting, heating, ventilation, andair conditioning (HVAC) systems and other device loads(PC, printers, etc.). The strategies include both the controlof building services and devices and the monitoring ofenvironmental conditions and energy consumption. In thepaper, the energy savings estimated through simulation,for both HVAC and lighting, are presented to highlightthe potential of the designed system. After the implemen-tation of the system in the demonstrator, results will becompared with the monitored data.

Keywords Interoperability . Middleware .WirelessSensor and Actuator Networks . Building managementsystem . Building information modelling . Energyawareness and efficiency

Introduction

The new era of smart Information and CommunicationTechnology (ICT), providing real-time access to

Energy Efficiency (2013) 6:707–723DOI 10.1007/s12053-013-9211-0

A. Osello (*) :A. Acquaviva : C. Aghemo : L. Blaso :D. Dalmasso :D. Erba :G. Fracastoro : E. Macii : E. Patti :A. Pellegrino : P. PiumattiPolitecnico di Torino,Torino, Italye-mail: [email protected]

D. Gondre : J. Savoyat : J. VirgoneUniversity Lyon1,Lyon, France

M. Jahn : F. PramudiantoFraunhofer Institute for Applied InformationTechnology FIT,Sankt Augustin, Germany

M. Spirito : R. TomasiIstituto Superiore Mario Boella (ISMB),Torino, Italy

monitoring information, is bringing a new concept ofenergy management. Pervasive devices also enable effec-tive control strategies allowing reduction in energy usageand CO2 footprint in existing public buildings withoutsignificant construction work. In particular, special atten-tionmust be paid to historical buildings where damage byextensive retrofitting must be avoided.

The aim of SEEMPubS research

The goals of the Smart Energy Efficient Middlewarefor Public Spaces (SEEMPubS 2010) project are sev-eral and fully integrated with each other as describedbelow:

& To develop an integrated electronic system tomonitor different buildings, building services,electronic devices, and operations in order to opti-mise and integrate all maintenance functions.

& To implement an interoperable Web-based soft-ware solution for real-time energy performancemonitoring and control of lighting and heating,ventilation, and air conditioning (HVAC) servicesthrough Wireless Sensor and Actuator Networks(WSAN). This is done through an integrated ap-proach based on the LinkSmart energy-awaremiddleware platform. In this way, the project isable to deploy the new solution in buildings byreusing already existing structures (e.g. proprietarysensor networks or management systems) and byintegrating them into an intelligent energy man-agement system, supported by a Graphical UserInterface.

& To provide multi-dimensional visualisation of pa-rameters of building operations and data sharingfrom technical systems. Real-time information iscollected by sensor networks detecting environ-mental and maintenance-oriented parameters ofperformance from lighting and HVAC services.The primary purpose of such a performance mon-itoring system is to provide facility managers andoperators with the means to easily assess the cur-rent and historical performance of the building as awhole, and of its significant energy consumingsystems and components. The performance moni-toring system includes not only the needed sen-sors, wiring and data acquisition devices, but alsomeans to calculate, display, and archive resultantparameters.

& To raise people’s awareness for energy efficiencyin public spaces: this is a key aspect and the firststep in the process of fostering sustainable andresponsible behaviour with respect to energy con-sumption. This includes the construction of a usercommunity portal to enable communication andinformation exchange among users. Such commu-nities, consisting of users with different levels ofknowledge of the energy domain, combined withambient awareness features, have the ability tocreate a new energy consciousness of citizens intopublic spaces.

All elements are validated through an iterativemethodology to ensure the gradual approximation toSEEMPubSwith simultaneous consideration of all stake-holder needs. The validation phase of the project isconducted utilising the most significant buildings in thePolitecnico di Torino Campus, selected as demonstrator.

The case studies

The construction period of the Politecnico di Torino’scampus ranges from the sixteenth century to thetwenty-first century, thus including buildings very dif-ferent in dimension, shell, construction characteristics,and performance. The short description below helps tounderstand the reason of its representativeness as anexample of a wide range of similar buildings inEurope.

The Valentino Castle on the river Po, seat ofthe Schools of Architecture is a historical site. Itsorigins date back to the beginning of the sixteenthcentury. A non-insulated and highly massive shellcharacterises the building, and the systems havebeen realised taking into account the building’sconservation requirements.

The Main Campus on Corso Duca degli Abruzzi,dedicated mainly to Engineering was inaugurated inNovember 1958. The building envelope was built withvery poor thermal insulation criteria and several retro-fits are being planned to increase the building’s energyperformance.

The Cittadella Politecnica, recently built next tothe Main Campus in a former industrial area, is thePolitecnico’s research centre. The plan, started in 1994and at present under completion, includes recovery ofthe most interesting industrial buildings, and newbuildings. This campus was built adopting energy

708 Energy Efficiency (2013) 6:707–723

efficiency-oriented construction techniques, accordingto the Italian Law 10/91.

Based on the description above, it is clear that eachbuilding has specific energy needs and limitationsconcerning control systems which should be installed;a major challenge for the SEEMPubS project is to dealthis specific deployment's details, providing a genericframework, which is able to actually impact on energyefficiency.

The architecture of SEEMPubS building energymanagement system

As described previously, one of the main goals ofSEEMPubS is to develop a middleware for holistic build-ing energy management systems (BEMS). With regardsto system architecture design, two major requirementswere identified. First, such middleware needs to supportthe integration of heterogeneous technologies, such asWSAN, building management systems (BMS), or smartmeters. Second, the middleware must provide unifiedaccess to these technologies to support the implementa-tion of applications that use these technologies in aninteroperable way (e.g. implementation of building con-trol strategies). A further requirement from themiddleware perspective is that the developed componentsshould be extensible and reusable in similar settings.

In order to cope with these requirements, theSEEMPubS consortium has decided to limit customdevelopments and instead employ and extend an al-ready existing and proven technology, namely theLinkSmart Middleware.

The LinkSmart Middleware is a generic middlewarefor developing Ambient Intelligence (AmI) applications(Eisenhauer et al. 2009, 2011). It is the result of the FP6European Project HYDRA (HYDRAProject 2007), and itis currently employed in a large number of other EU andinternational research initiatives, e.g. (SEAM4US Project2013; Adapt4EE Project 2011; ebbits Project 2010). AmIis characterised by making use of various different tech-nologies and devices, which together perform ambient andintelligent computing tasks. Typical examples of AmI aresmart homes or ambient assisted living projects.

LinkSmart provides a framework and softwaredevelopment tool for integrating heterogeneous networkeddevices into AmI applications. Furthermore, LinkSmartcomes with software components that provide functional-ity typically for AmI applications (e.g. message encryp-tion, event management, or device discovery). The

LinkSmartMiddleware is applied and extended to developa dedicated middleware for the energy efficient buildingsdomain – under consideration of the aforementioned re-quirements. In the following we describe the SEEMPubSarchitecture and explain to which extent LinkSmart tech-nology is reused and extended with new functionality.

Figure 1 shows the layered architecture of theSEEMPubS system including its main components.On the LinkSmart Proxy Layer different kinds of tech-nologies (WSAN, BMS or smart meters) are integratedinto an interoperable system. We employ the Proxyapproach defined by the LinkSmart Middleware. Thisapproach allows to use the different technologies in aunified way. A LinkSmart Proxy is a component thatacts as bridge from a specific technology to theLinkSmart middleware by providing a defined serviceinterface. In this way, proxies allow unified access to thedifferent technologies through LinkSmart.

On the SEEMPubS Middleware Layer reside com-ponents that process data provided from the IntegrationLayer and provide further functionality to theApplication Layer (e.g. components to execute buildingcontrol strategies reside on this layer). The MiddlewareLayer is conceptually separated into two parts: theLinkSmart infrastructure components and the contextframework. The infrastructure components, namely net-work manager and event manager provide the corecommunication functionality for all LinkSmart compo-nents. The network manager enables network commu-nication among devices inside a LinkSmart network. Itcreates an overlay P2P network that implements SOAPTunneling as transport mechanism for Web service calls(Milagro et al. 2008), allowing direct communicationamong all devices inside a LinkSmart network, no mat-ter if they appear behind a firewall or Network AddressTranslator. The event manager allows developers tobuild event-based systems, which in the case of systemsdealing with sensors, is a core requirement. InSEEMPubS, the event manager is used to publish eventsat the Proxy Layer and to listen to events whereverneeded, for instance at context framework, for furtherprocessing. The context framework is an extension toLinkSmart which is specifically designed for energy-efficiency applications aiming at achieving semanticinteroperability. While syntactic interoperability isachieved by the Proxy Layer (by homogenising thedifferent standards and protocols of different technolo-gies), semantic interoperability in SEEMPubS means toabstract from the Proxy Layer to a layer that deals with

Energy Efficiency (2013) 6:707–723 709

domain objects like buildings, rooms, windows, appli-ances, sensors, etc. Such domain information can thenbe used by the system for (1) end-user applications, (e.g.for visualisation) and (2) for implementing buildingmanagement control strategies that work seamlesslyacross heterogeneous technologies. The context frame-work is described in detail following.

The Application Layer comprises end-user applica-tions for different kinds of users (e.g. an integratedmonitoring and control application for the building en-ergy manager would be part of the Application Layer aswell as applications for building occupants). With thehelp of the middleware layer, information should beeasy accessible by any kind of application.

Related work

Several research projects are dealing with increasingenergy efficiency in buildings. The FIEMSER (2007)

project aims at developing an energy managementsystem (BEMS) for existing and new residential build-ings. FIEMSER tries to achieve that goal by optimisingthe balance between energy generated locally, energyneeded from external providers and energy demandedby the building. Technically, FIEMSER employs acombination of sensors and actuators from traditionalPLC-based BEMS systems, connected via an OSGi+SOA REST architecture. While the goals of FIEMSERand SEEMPubS are to some extent comparable, theyboth aim for different domains.

Another similar project but with a different domain is3e-Housewhich deals with the integration of establishedICT technologies in social housing in order to providean innovative service for energy efficiency. Other pro-jects such as SaveEnergy (2011) focus on changingpeople’s behaviour through the use ICT—real-time in-formation and serious games in the case of SaveEnergy.A similar approach is implemented by the BeAware

Fig. 1 Layered system architecture


Project (2011), which also aims at increasing people’senergy awareness in private households.

Many other research initiatives and projects aiming atincreasing energy efficient in different kinds of build-ings exist, each focusing on different specific topics.SEEMPubS tries to produce a result which can begeneralised—namely the middleware—that can be ap-plied in other projects dealing with energy efficiency inbuildings and the aforementioned issues of interopera-bility. While many different projects develop custommiddleware solutions, we will provide an open softwarethat can be applied for energy efficiency applications inbuildings in a general way. Using the existingLinkSmart infrastructure is supposed to support thisgoal.

Methodology

SEEMPubS is committed to an iterative approachfollowing the principles of the common standard ISO13407. The goal of this procedure is focused not onlyon the evolution of the applications and services butalso on the refinement of scenarios, requirements, andspecifications of the SEEMPubS platform and its im-plementation. This iterative process ensures the grad-ual approximation to all multi-disciplinary activitiesinvolved in SEEMPubS with simultaneous consider-ation of all stakeholders needs as described following.

Integration of devices

The strategy for smart building management and con-trol leverages upon an ICT infrastructure made ofheterogeneous monitoring and actuation devices, suchas WSAN. Moreover, in order to improve backwardscompatibility, the infrastructure supports wired de-vices work with different protocols, such as BACnet,LogWork, etc. An innovative Web service-orientedsoftware infrastructure has been developed to manageheterogeneous and commercial devices. As introducedbefore, the LinkSmart middleware (Eisenhauer et al.2009, 2011; HYDRA Project 2007) has been adoptedto provide interoperability between heterogeneous de-vices and networks, both existing and to be deployed.Moreover, the proposed infrastructure allows easy ex-tension to other networks, thus representing a contri-bution to the opening of a market for ICT-basedcustomised solutions integrating numerous products

from different vendors. The system manages energyefficiently and WSANs are preferred to simplify theintegration of new sensors into the system and also toavoid overloading of cables in historical buildings.The software infrastructure provides the followingmain functionalities:

& It enables the interfacing to the application layerby means of Web services, through which thesensor data are read and can be used for visualisa-tion or to feed energy management policies;

& It collects environmental data coming from thesensor nodes into the local database (DB), andthese data can be accessed in an asynchronousway and preserved from network failures;

& It allows the remote reconfiguration of sensor nodeparameters such as the sampling rates of physicalquantities which are to be monitored;

& It allows the remote control of actuator devices;& It enables interoperability among heterogeneous

networks, characterised by different communica-tion protocols, microcontrollers, and sensors.

As shown in Fig. 2, the software runs in a PC-Gateway (GW) and communicates directly with theheterogeneous networks. The dedicated Interface repre-sents the lowest layer of our proposed stack, and re-ceives information coming from various devices,regardless of the adopted communication protocols,hardware or the network topology. Hence, each networkneeds a specific software interface, which interprets theenvironmental information (e.g. temperature, humidity,etc.) and stores them in an integrated DB, in order tomake the whole infrastructure flexible and reliable withrespect to backbone network problems since data arelocally stored. The Web service layer, implementedusing LinkSmart, interfaces the device networks to theWeb, making the remote management and control eas-ier. Moreover it exports to the application client layer,the last in our stack, all the environmental data that werestored in the DB. At this layer, the information is avail-able to the end-user and ready to be post-processed or tobe shown via computers, tablets or smart phones.Particular emphasis was given to the possibility toreconfigure each node, changing, for instance, someparameters about power management. In this scenario,the end-user sends the new configuration via Web ser-vices to the GW and stores it in the DB. Then, the newsettings will be automatically sent to the receiver mote,when it wakes up from the sleeping period, through the


specific network software Interface. The configurableparameters change depending on the hardware and theOperating System running on the end node. However,using this software infrastructure, the user can chooseonly the right settings ignoring the real physical hard-ware related to the virtual device. About power man-agement, it is worth noting that there is no standard thatindicates how controls and configuration settings are tobe sent to the mote via the protocol packet payload.Hence, the proposed network software Interface is hard-ware dependent only for the way in which these param-eters are formatted. However, from the communicationpoint of view, it is protocol standards compliant. Theproposal and development of a standard for power man-agement would be desirable but is not the subject of thisproject. The goal of this research, from ICT side, is toenable the communication between heterogeneous andcommercial devices; this implies a certain dependenceon hardware.

Particularly in this case study, we have developeddifferent interfaces to manage respectively five com-mercial WSAN based on:

& Crossbow Telos rev B Datasheet (2003) open sourceend node to monitor air temperature, relative

humidity and illuminance, which exploits IEEE802.15.4 communication protocol;

& Our customised end node built on Texas Instruments’CC2530 Datasheet (2011) system on chip to monitorair temperature and illuminance leveraging the ZigBeeprotocol;

& ST Microelectronics Smart Plug (2013) commer-cial end node to monitor power energy consump-tion and to switch on/off the lighting plants,exploiting the ZigBee protocol;

& Plugwise (2006) Smart Plug commercial end nodeto monitor power energy consumption and toswitch on/off the appliances connected to themains exploiting the ZigBee protocol;

& EnOcean protocol stack (EnOcean Alliance 2013)commercial end nodes to monitor air temperature,relative humidity, illuminance, occupancy and toactuate heating and lighting plant, respectively.

Moreover, an interface has been developed to enablethe remote communication with the Siemens DesigoBMS that exploits the BACnet Protocol (1998) in orderto monitor and actuate Desigo-wired devices.

In a nutshell, the proposed Web-based infrastructureis a software that makes transparent to the end-user the

Fig. 2 Software infrastructure scheme to handle heterogeneous devices


underlying devices, abstracting all the information abouthardware, communication protocol stack and embeddedoperating systems. Furthermore, the use ofWeb servicesmakes the interoperability with third-party softwareeasy.

BIM and interoperability

To optimise the data exchange among architecture,mechanical electrical and plumbing (MEP), energysimulations and facility management (FM), at the be-ginning of the project a building information model-ling (BIM) methodology was used. For this reason, theinteroperability testing described in this paragraph isan important issue of SEEMPubS, both for processingand for data visualisation (Osello 2012).

The BIM approach started with the creation of 3Dmodels of each case study through Autodesk applica-tions as shown in Fig. 3. In particular, Revit Architecturewas used for architectural modelling and Revit MEP forthe modelling of heating/cooling and lighting systems(Autodesk Revit (2013)).

To test the interoperability between architecturalmodelling and lighting simulations, it was necessaryto realise the 3D models of the real rooms and theirexternal environment. The 3D models were used torun energy simulations, which were validated through

the monitored data and then used to estimate thebuilding energy demand. In particular, Radiance(Larson and Shakespeare 1998) was used to validatethe models, comparing the simulated illuminance dis-tribution to the illuminance values measured in thecorresponding rooms, while Daysim (Reinhart 2001)was used to estimate the lighting energy demand andthe savings obtained with the specific control strate-gies proposed for each case study, as it allows runningannual simulation for a site, accounting for the specificdynamic climate conditions.

In order to run the lighting simulations, it was notpossible to import the parametric model from Revit intoRadiance/Daysim directly. The software EcotectAnalysis (Autodesk Ecotect Analysis 2013) was henceused as interface to launch Radiance/Daysim. As aresult, Revit was imported into Ecotect using differentprocedures, with the aim of finding the mostappropriate.

The first trial adopted the traditional approach basedon the exporting in Industry Foundation Classes format,but it did not succeed because some elements, such asthe window frames, were not exported or weredisplaced.

The second trial was based on the exporting fromRevit by a gbXML file, but some geometrical discrep-ancies in the surfaces generated from the solid

Fig. 3 BIM and interoperability as method to optimise the process


elements made the model unsuitable for the light-ing simulation. In fact, this type of exporting in-volved some model simplifications that turned outto be incorrect for lighting simulations, even if themodel was correctly generated and all the elementshad maintained their reciprocal positions.

The third procedure was the most appropriate, bothconcerning modelling in Revit Architecture environ-ment and for the analyses with Radiance and Daysim(thanks to Ecotect as a software interface), through theuse of 3DStudioMax as an intermediate software (toconvert the .fbx file exported from Revit in a .3ds fileimported into Ecotect) which gave good results in termsof geometrical consistency of the exported model.Nevertheless, also with this procedure it was necessaryto check the 3D models imported in Ecotect beforestarting the lighting simulations in Radiance andDaysim, to prevent some trivial mistakes connected toan oversight or elimination by the operator (for instancethe disappearance of some parametric elements).

Once the 3D models were successfully importedinto Ecotect, it was possible to proceed with the light-ing simulations in Radiance and Daysim. The numer-ical data obtained from the simulations weregraphically processed by using the Ecotect Analysisapplication, through the import of .dat format (comingfrom Radiance) or .da (coming from Daysim) andtherefore shown in 3D visualisation.

In a similar way it was possible to work betweenarchitectural (Revit Architecture) and thermic (Trnsys17) models using SketchUp as additional software.Compared to the classic method, the use of the 3Dmodel in Trnsys 17 required attention in some criticalpoints (north orientation, window frame and wallthickness). Therefore, before starting the simulationphase, we fixed the critical points. In detail:

& The model was optimised after the importation inSketchUp, before drawing the Thermal Zones;

& The window area was adjusted. In fact, since inRevit Architecture objects are parametric, it ispossible to calculate the exact area of the frameand of the glass surface for each window; thisvalue was put in Trnsys to improve the accuracyof the simulation results;

& The problem of the wall thickness between adjacentZones was fixed by moving the wall surface of oneof the two rooms. This had as a consequence avolume increase of one or both rooms; but this does

not have significant consequences on the results dueto the little impact of air volume in the heat balance.

Context awareness

The context framework is part of the energy efficientmiddleware. From a software architecture perspective itresides between the proxy layer and the applicationlayer. The aim of this context framework is to interpretdata provided by sensors and devices into semanticallyinteroperable information that can be used by the systemfor (1) end-user applications (e.g. for visualisation) and(2) implementing building management control strate-gies that work seamlessly across heterogeneous technol-ogies. Context-aware optimisation is a generic and“complex” task, since its main goal is to adapt thebuilding behaviour to detect and act upon a very broadand extensible set of situations. For this reason, theSEEMPubS context framework must provide flexiblemeans to describe generic “situations” which are signif-icant for energy consumption monitoring, optimisationand awareness.

As an example for such a situation, we define theenergy-wasting situation.

An energy wasting state is defined as a situationwhere energy is being spent without benefit for theusers. An example of energy wasting state might berepresented, for instance, by a room where the windowis open and the HVAC is running or a PC left idlewithout any use. Such information could be used toprovide feedback to end-users and motivate energy-saving behaviour. Another use-case showing the advan-tages of semantic interoperability is the technology-independent definition of control strategies in an inte-grated system.

In the following we describe how the SEEMPubScontext framework supports these concepts.

From sensor data to building information

The term semantic interoperability in our systemcharacterises the ability to access different kinds ofdevices, sensors, actuators, or systems through a uni-fied vocabulary (cf. Ontology manager) regardless ofvendor- or technology-specific implementations, dataformats, APIs, etc. Figure 4 shows how data aretransformed from the actual sensor to the contextframework. The Proxy layer is responsible for


processing raw sensor data and adding some basicmeta-information and forwarding it to an event man-ager. The context framework is responsible for asso-ciating the incoming data with a meaning.

The context framework

In this paragraph we present the implementation of thecontext framework that uses an Ontology-based knowl-edge repository and rule-based context recognition sys-tem that allows defining actions that should beperformed upon recognition. The Ontology models theabstraction from concrete devices, sensor etc. to domainspecific entities such as buildings, rooms, HVAC appli-ances, etc.

Ontology The Ontology provides a taxonomy for theenergy efficient buildings domain, describing a modelthat is used to support the development of energyefficiency applications. This model describes the(semi-)static environment of the building independent-ly from the type of implementation or platform used indevelopment of such applications. It models devices,sensors, actuators, rooms, and observable propertiessuch as temperature, humidity, and so on. To map

from the domain to the implementation, it also modelssoftware proxies, services, and events.

Context manager—rule engine and entity manager Thecontext manager consists of two major components,(1) an entity manager to manage the current and pastcontext of all entities and (2) a rule engine to store andevaluate context reasoning rules.

The rule engine has two main roles: (1) rules aredefined to recognise a situation based on the state ofentities in the system and (2) rules can also be used todefine an action that should be performed upon rec-ognition of a situation. We use the Drools (JBossCommunity 2013) rule engine—a well-known ruleengine for business applications—and modified therule language allowing it to define triggers that willbe used by the context manager to subscribe to spe-cific sensor event which monitor the context of theentities. When a sensor event comes, the context man-ager will update the context of all entities that areassociated with that event. The correlation betweenthe sensor event and the entities is done with the helpof Ontology reasoning. The drools engine will thenevaluate any rules that are associated with that entityand execute the actions defined in the rules. Suchaction could be the invocation of an actuator.

The entity manager is responsible for keeping trackof entities and their context attributes (e.g. temperaturein Room1 and humidity in Room1). An entity and itscontext attributes are monitored automatically oncethey are defined in a rule and their current state issaved. This approach enables the entity manager toonly store the current state of the system as needed.

By moving the storage of ever-changing context datafrom the Ontology to a dedicated entity manager, weavoid performance problems in the semantic store.

In the following, we provide an example that ex-plains how the context framework can be used torecognise situations and perform actions accordingly.

As stated before, the context framework allows usto deal with domain objects. This means, we candefine rules like the following: Every time the stateof Window1 or Radiator1 changes check if Window1 isin state open and Radiator1 is in state on. If they are,publish to any interested component that there is anEnergy Waste situation in Room1, involving Window1and Radiator1.The domain objects we are talkingabout here are Window1, Radiator1, and Room1.

Fig. 4 Abstraction of data


The advantage of this approach is that for the def-inition of such rules, no knowledge about underlyingtechnologies is needed. It does not matter if the radi-ator is controlled by a BACNet or EnOcean device orwhat kind of contact sensor is attached to Window1

This information is stored in the Ontology and willbe reasoned by the Ontology manager. Figure 5 shows(in a simplified way, leaving out sub-classes) how therelevant information is modelled in the Ontology.

Besides notification of situation changes the rule-based approach also allows for implementing controlstrategies, such as: Every time the presence property ofRoom1 changes and it is false, switch off the light inthat room.

Beyond this simple example it is worth observingthat presence detection, or more in general occupancydetection (EnOcean Alliance 2013) is a complex pro-cess and can thus give more complex types of outputsrather than “true/false” presence information.Therefore, occupancy values might have been evalu-ated beforehand by other rules or by dedicated com-ponents, for instance an occupancy proxy that takescare of optimising occupancy recognition by usingmultiple input sources and adding probabilities tooccupancy states.

As for recognition the context framework also han-dles the invocation of an actuator for switching the light.Therefore, the context manager consults the Ontology tofind the respective actuator. A standardised proxy inter-face allows executing the correct service via the proxythat communicates with this actuator.

Control rules and control strategies (middleware)

Since the project aims at controlling HVAC and lightingsystems using temperature, luminance and occupancysensors, new control rules must be designed and testedin order to properly balance energy savings and comfortissues (Fracastoro et al. 2012). The thermal inertia ofbuildings leads to a low responsiveness of the HVACsystem, which requires an anticipation of possible futureuse of a room in order to meet comfort requirementwithin a reasonable time when occupancy is detected.The test cases have been modelled in Daysim (lightingsimulations) and Trnsys (thermal simulations) and sev-eral control rules have been implemented and tested.According to the information on the rooms’ features,achieved during the first phase of the project, somelighting control strategies were initially proposed. Inparticular, for spaces with high daylight availabilityand medium/high users absence probability, a combina-tion of daylight harvesting and occupancy detection wasproposed as lighting control strategies, while, for spaceswith low daylight penetration, only occupancy detectionwas supposed to be used. In any case users are allowedto override the system with manual control. Thermalcontrol strategies are guided by the maintenance orimprovement of user comfort combined to the bestenergy savings. As the uncertainty in occupancy sched-ules is detrimental to savings and/or comfort, a differ-ence has been made between personal offices wherepresence detection is the only available information,and common rooms (meeting rooms or classrooms for

Fig. 5 Ontology instantiation


instance) where users have to book the room, whichallows strategies based on anticipation.

Another difficulty is to evaluate the time necessary toreach the set point temperature when preheating isplanned. It depends on the inertia of the building butalso on the weather conditions and the indoor tempera-ture. A precise calculation of this “recovery time”wouldlead to the best compromise in terms of energy savingsand user comfort. Figure 6 below displays the maincontrol rules considered for both lighting and HVACsystems.

Results

SEEMPubS is still under way but first simulation resultsare already available. Simulations are part of a prelimi-nary work that aims at optimising energy conservation

strategies that are implemented in the demonstrator. It isalready planned to compare numerical results with ex-perimental results as soon as results are available. Thefollowing paragraphs presents some simulation resultsof the ongoing work on energy saving in the HVAC andlighting systems.

Estimated energy savings, heating and cooling

This paragraphs aims at presenting detailed results forone of the six case studies. The test office is part of theValentino Castle, the historical building made of heav-y walls without thermal insulation. Its energy con-sumption is high and its thermal responsiveness isvery low.

As savings strategies are based on an optimisation ofvacancy periods, results heavily depend on occupancyhypothesis used in the simulations. Two different kind

Fig. 6 Control rules considered for lighting and HVAC systems (Tsp set point temperature, trec recovery time)


of occupation profiles are used in the simulation. The“theoretical occupancy” starts at 8 AM and ends at 6 PM

(10 h a day, 5 days a week, which means 50 h a week)and corresponds to the time range when occupancyprobability is high. In the presence detection algorithm,the theoretical occupation profile does not provide a lotof details since it is hard to forecast the actual occupan-cy, while in the presence anticipation algorithm, thetheoretical occupancy profile is detailed as much aspossible since it is used for classrooms and meetingrooms where it is easy to know in advance what thefuture occupation should be. The “real occupancy” isdirectly linked to the occupancy detection sensor but inthe simulation it is set as fixed ranges which include twofull occupancy days (onMondays and Fridays) with a 1-h lunch break and two half occupancy days (onTuesdays and Wednesdays) with a 5-h occupancy divid-ed into three periods (8 AM–10 AM, 1 PM–2 PM, and 4 PM–6 PM). No occupancy at all is considered on Thursdays.For instance, it may represent a day off, a seminar day ora day out of office. With regards to this simulationhypothesis, the real occupancy profile used in the simu-lation represents 56 % of the theoretical occupancy.

As the presence anticipation algorithm was origi-nally designed for common rooms such as classroomsor meeting rooms, results discussed later will mainlyfocus on a the presence detection algorithm.

In order to optimise the restarting time of theHVAC system in the early morning or after a periodof vacancy, a study was carried out in order to char-acterise the building inertia. The idea was to find thebest correlation between the time Δτ necessary toreach the set point temperature (referred to as recoverytime later in the text) and the ambient conditions suchas the outdoor temperature, the indoor temperature Tinand the set point temperature Tsp. Restarting curveshave been analysed for 154 mornings of a yearlysimulation. Monday mornings were analysed separate-ly since the thermal behaviour of the building is dif-ferent after two unheated days.

The best correlations linked the recovery time to thetemperature difference between set point and indoortemperature at the restarting time. As this very simplelinear equation (lines 1 on Fig. 7 and Table 1) is notsufficient to properly describe the building behaviour, itsdirect use leads to good energy savings but also leads todiscrepancy of user comfort (see Fig. 8), due to theunderestimation of the recovery time for about half ofthe time. In order to optimise the balance between

energy savings and user comfort, two updated correla-tions are used (lines 2 in Fig. 7 and Table 1). As it is notsufficient to improve comfort as much as expected, asecond set of updated correlations is used (lines 3 inFig. 7 and Table 1).

Energy savings values are given in percentage ofsavings compared to simulation results of the actualcontrol rules used in the case study (reference room).The main differences with the test room lie in the absenceof presence detection and in the use of only two set pointtemperatures (one at night, and another one during work-ing hours). As indoor temperatures may be higher in thereference room than in the test room (especially duringworking hours when occupancy is not “detected”), non-negligible heat transfers occur from the reference room tothe test room. This leads to an over-estimation of thereference room consumption and an under-estimation ofthe test room consumption. Results can then be flatteringand … false! In order to calculate the consumption thatwould actually be observed if the whole building wasequipped with the same control system, two simulationsare run separately: one with the actual conservation strat-egy (reference), and another one with the proposed con-servation strategy (test). Comfort analysis is based on theoperative temperature distribution in the first hour ofoccupancy (8 AM–9 AM) when occupancy is “detected”(Fig. 8). Results are very close to the present situation interms of comfort and this strategy allows a 14.6 % reduc-tion on the overall annual energy demand (Table 1).

These first simulation results show that the estimationof the recovery time can be improved in the future. Themore accurate the estimation of recovery time is, the lessenergy is needed in order to reach a given comfort level(and themore efficient are the control rules). Neverthelessthe recovery time heavily depends on the building char-acteristics, but also on the HVAC system specifications(water and air flow rates and temperatures of the fan coilsunit) that are difficult to know with precision. It thenrequires to fit simulation results with monitoring resultsin order to find the most likely HVAC specifications.These control rules will therefore require a lot of dataand studies to be extended to new buildings with differentcharacteristics. That is the reason why we are currentlystudying the feasibility of developing a self-learning sys-tem that would itself calculate an estimation of the recov-ery time, based on data collection of the past few days ormonths. It would avoid thermal inertia calculation (andassociated data collection) since the system would calcu-late it itself. An algorithm based on presence anticipation


also requires the recovery time calculation and also needsto be linked to a schedules DB. Simulation on the sameoffice showed slightly better results in terms of comfortbut higher energy consumption for presence anticipationstrategy. It is probably due to the high inertia of thebuilding, which leads to slow temperature variation whenthe HVAC system is turned off. For vacancy periodshorter than a couple of hours, the presence detectionstrategy does not call for heating or cooling while thepresence anticipation strategy restarts the HVAC systemsin advance, which leads to higher power consumption.Presence anticipation strategy is then probably more rel-evant for low occupancy rooms in low inertia buildings.Results showed that it is possible to save up to 22.6 %(32.5 % on heating and 3.9 % on cooling) in the class-rooms case study.

The development of HVAC strategies will continuein two perspectives: extending detailed analysis ofpredictable energy savings to similar spaces (offices)of other types of buildings (Main Campus andCittadella Politecnica), and analysing the resultsobtained from prototypes that will be installed in thesix test rooms

Estimated energy savings, lighting

To assess the effectiveness of the new lighting controlsolutions, the energy savings achievable with respect tothe actual reference situation (manual lighting control),were calculated from the estimated total annual electriclighting energy consumptions. The total annual energyused for lighting depends on different aspects, mainlyrelated to the indoor daylight availability, the buildingusage (occupancy profile and lighting requirements), theusers’ behaviour in terms of interaction with the lightingsystems, and the lighting plant characteristics. The sim-ulations carried out with Daysim are based on the annualweather data for Torino, on the actual offices occupancyprofile (monitored during le previous phase of analysis)and lighting requirement (target illuminance of 500 lux),on the plant power consumption including electricityconsumption associated to the control devices, and onstochastic user behaviour models implemented inDaysim to mimic how building occupants interact withmanual controls of lighting plants or shading devices. Inparticular simulations were initially carried out consid-ering a “mixed users’ behaviour” that means users are

Table 1 Correlations used in the three test cases for calculation of the HVAC restarting time and estimation of associated energysavings

Correlations used for estimation of recovery time Energy savings (%)

Overall Heating Cooling

Test case 1 t ¼ 0:30 Tsp " TinðiÞ! "

(Monday morning) 22.1 25.7 9.2t ¼ 0:12 Tsp " TinðiÞ

! "(rest of the week)


(Monday morning) 19.6 22.5 9.4t ¼ 0:20 Tsp " TinðiÞ

! "(rest of the week)


þ 0:33 (Monday morning) 14.6 16.0 9.6t ¼ 0:25 Tsp " TinðiÞ

! "þ 0:33 (rest of the week)

Fig. 7 Recovery time as afunction of the temperaturedifference Δt=Tsp−Tin(i)during the heating period


partly active and partly passive with respect to the use ofelectric lighting and blinds. Active user is a user whooperates the electric lighting in relation to ambient day-light condition, opens blinds in the morning, and partlycloses them to avoid visual discomfort, while a passiveuser is a user who keeps the electric lighting switched onthroughout the working day and keeps the blindslowered throughout the year.

Simulations and comparisons between energy de-mand with the existing control system (manual control)and the new proposed lighting control rules were carriedout for all rooms selected as case studies for theSEEMPubS project. The calculated energy savingranges between 5 and 16 % for offices with manuallycontrolled blinds, while it rises up to 30 % when bothelectric lighting and blinds are automatically controlled.Energy savings are calculated as relative difference be-tween the simulated consumption of existing referencecontrol solution (manual control) and the new proposedlighting control rules.

Analysing the results achieved from the lighting sim-ulations, whose goal was to verify the effectiveness, interms of energy saving, of the proposed control strate-gies for each room, it was observed that, in most of thecases, obtained energy savings were below the expecta-tions. The initially proposed control strategies weremainly based on switching lights on and off as a conse-quence of users presence detected by the occupancysensors, and on the automatic dimming of lights as aconsequence of daylight availability. To optimise theenergy saving results, it was decided to simulate acontrol rule which provided for users to turn the lightson when the luminous environment is perceived toodark and for occupancy sensors to switch them off

automatically when users absence is detected. Whenswitched on, in rooms with high daylight availability,lights are dimmed in order to maintain the target illumi-nance on the working plane. Figure 9 shows the newproposed control rules for lighting.

The reduction of electric lighting consumption forthe different rooms now ranges between 27 and 48 %for rooms with high daylight availability and it reaches13 % for the open plan office where only switching offbased on occupancy detection was considered. Thelast hypothesis that was checked was based on adifferent user behaviour: as it is supposed that theSEEMPubS platform should increase people’s aware-ness by means of a Web-based tool to directly interactand communicate information to users, the last set ofsimulations were carried out considering an “activeuser”, that is a user who behaves consistently withthe actual indoor and outdoor environmental condi-tions (Reinhart 2004). The energy savings obtainedwith the described assumptions rise to a maximum ofapproximately 70 %. In this last case, both savings dueto optimised control of lighting and improvement ofuser behaviour are considered.

A summary of the calculated electric energy sav-ings for lighting is presented in Table 2.

The technical architecture used for lighting control doesnot differ from previous commercial solutions. The type ofsensors used in SEEMPubS project (photosensors andoccupancy sensors) are commonly used to control lightingand the sensors layout aswell does not differ from standardsolutions. Control strategies instead, have been analysed inorder to find the most effective solution by comparingpossible alternative control logics for each room. In mostexisting buildings, the same control logics are applied in

Fig. 8 Indoor operative temperature distribution during heating season between 8:00 and 9:00 in the office of Valentino Castle fordifferent control strategies


roomswith very different features, in this way reducing thesaving potentials of the automatic control system.

This is clearly highlighted by the savings resultsobtained by comparing, through simulation, the lightingcontrol solution typically proposed by many commer-cial products and adopted in many offices (auto on/offand dimming) with control logics, based on occupancyas well as daylighting availability, but with a differentdegree of manual and automatic control (e.g. savingsrising for a single room from 16 to 39 %).

Furthermore, the SEEMPubS middleware, unlikecommon commercial systems, allows the monitoringand recording of both lighting energy, environmentallighting conditions and plant status, thus allowing facil-ity managers to check the system operation and in caseof fault or ineffective operation to quickly intervene torecalibrate the system. It is in fact commonly known thata system behaviour not meeting users’ expectations isone of the main causes of automatic control systemsrejection.

Fig. 9 Optimised control rules for lighting

Table 2 Summary of the energy savings obtained for different control rules and user behaviour

Energy savings (Energyreference control rules−Energyproposed control rules)/Energyproposed control rules; %)

Auto on/off+dim (mixeduser behaviour)

Auto on/off(mixed userbehaviour)

Manual on+autooff+dim (mixeduser behaviour)

Manual on+autooff (mixeduser behaviour)

Manual on+autoof+dim (activeuser behaviour)

Manual on+autooff (active userbehaviour)

DAUIN single office 16 39 71

Administrative office 6 27 70

DITER office 30 48 64

DAUIN open plan office 5 13 14


Results obtained with the simulation carried out inthis phase of SEEMPubS project are consistent with incase of active user behaviour or greater than the sav-ings achieved with automatic lighting control in othersimilar research activates (Li et al. 2010; Lee andSelkowitz 2006; Yun et al. 2012).

Conclusions

In this study, we tested how ICT makes it possible tointegrate different processes, applications, systems andtechnologies in order to reduce energy consumption.First of all, to obtain a smart building management andcontrol leverages, we defined an ICT infrastructuremade of heterogeneous monitoring and actuation de-vices, and we developed the LinkSmart middleware tocollect environmental data coming from the sensornodes in a local DB. Then we proposed a Web-basedinfrastructure to make transparent to the end-user theunderlying devices, abstracting all the information abouthardware, communication protocol stack and embeddedoperating system. Moreover, we set a context frame-work that uses an Ontology-based knowledge repositoryand rule-based context recognition system to allow thedefinition of actions that should be performed uponrecognition. Finally, we used BIM and interoperabilityto process and visualise all data essential for energysimulations and for FM. Our results concern controlrules and test on HVAC and lighting systems usingtemperature, luminance and occupancy sensors, in orderto properly balance energy savings and comfort issues.Future work should optimise the increase in people’sawareness by a Web-based tool to directly interact andcommunicate information to users.

Acknowledgement The research is funded by EU, FP7 Col-laborative project—2010: SEEMPubS.

References

SEEMPubS (2010). http://seempubs.polito.it/. Accessed June2012.

Eisenhauer, M., Rosengren, P., Antolin, P. (2009) A Develop-ment Platform for Integrating Wireless Devices and Sen-sors into Ambient Intelligence Systems.Sensor, Mesh andAd Hoc Communications and Networks Workshops, 2009.SECON Workshops '09. 6th Annual IEEE Communica-tions Society Conference on, vol., no., pp.1–3, 22–26 June.

Eisenhauer, M., Jahn, M., Pramudianto, F., Sabol, T., Hreno, J.2011. Towards a generic Middleware for developing Am-bient Intelligence Applications. 2nd Workshop oneeBuildings Data Models at CIB W078–W102, 26–28October 2011, Sophia Antipolis.

HYDRA Project (2007). http://www.hydramiddleware.eu/. Eu-ropean Commission permalink: http://cordis.europa.eu/projects/rcn/79422_en.html. Accessed April 2012

SEAM4US Project (2013). Sustainable energy management forunderground stations. http://seam4us.eu/. European Com-mission permalink: http://cordis.europa.eu/projects/rcn/100734_en.html. Accessed June 2012

Adapt4EE Project (2011). http://www.adapt4ee.eu. EuropeanCommission permalink: http://cordis.europa.eu/projects/rcn/100798_en.html. Accessed June 2012

ebbits Project (2010). Business-Based Internet of Things andServices. http://www.ebbits-project.eu. European Commis-sion permalink: http://cordis.europa.eu/projects/rcn/96598_en.html. Accessed June 2012

CrossbowTelos revBDatasheet (2003). http://bullseye.xbow.com:81/Products/Product_pdf_files/Wireless_pdf/TelosB_Datasheet.pdf.Accessed June 2012.

Texas Instruments’ CC2530 Datasheet (2011). http://www.ti.com/lit/ug/swru191b/swru191b.pdf. Accessed June2012.

ST Microelectronics Smart Plug (2013). http://www.st.com/internet/evalboard/product/250409.jsp. Accessed June2012.

Plugwise (2006). Smart wireless solutions for energy saving,energy monitoring and switching. http://www.plugwise.com.Accessed June 2012.

EnOcean Alliance (2013). http://www.enocean-alliance.org/en/home/?rnd=1. Accessed June 2012.

BACnet Protocol (1998). http://www.bacnet.org/. AccessedJune 2012.

Osello, A. (2012). The future of drawing with BIM for engi-neers and architects. In A. Osello, A. Acquaviva, E. Macii,E. Patti, C. Aghemo, L. Blaso, G. Fracastoro, A.Pellegrino, D. Dalmasso, P. Piumatti, & J. Savoyat (Eds.),Data standardization in an interdisciplinary research (pp.189–253). Palermo: Flaccovio.

Autodesk Revit (2013). http://www.autodesk.co.uk/adsk/servlet/pc/index?siteID=452932&id=14645193. AccessedJune 2012.

Larson G. W., Shakespeare R. A. (1998). Rendering withRadiance. The Art and Science of Lighting Visualiza-tion. Morgan Kaufmann Publishers, ISBN 1-55860-499-5.

Reinhart C F (2001). Daylight Availability and Manual LightingControl in Office Buildings. Simulation Studies and Anal-ysis of Measurements. Ph.D. thesis, Technical University ofKarlsruhe, Faculty of Architecture, October 2001.

Autodesk Ecotect Analysis (2013). http://usa.autodesk.com/ecotect-analysis/. Accessed June 2012.

JBoss Community (2013). https://www.jboss.org/drools/. AccessedJune 2012.

G. V. Fracastoro, J. Virgone, C. Aghemo, A. Pellegrino, L.Blaso, J. Savoyat, K. Johannes (2012). Energy efficiencyin public buildings through ICT based control and moni-toring system. In: 5th International Building Physics Con-ference (IBPC): The Role of Building Physic in Resolving


http://seempubs.polito.it/

http://www.hydramiddleware.eu/

http://cordis.europa.eu/projects/rcn/79422_en.html


http://seam4us.eu/



http://www.adapt4ee.eu/



http://www.ebbits-project.eu/



http://bullseye.xbow.com:81/Products/Product_pdf_files/Wireless_pdf/TelosB_Datasheet.pdf

http://bullseye.xbow.com:81/Products/Product_pdf_files/Wireless_pdf/TelosB_Datasheet.pdf

http://www.ti.com/lit/ug/swru191b/swru191b.pdf

http://www.ti.com/lit/ug/swru191b/swru191b.pdf

http://www.st.com/internet/evalboard/product/250409.jsp

http://www.st.com/internet/evalboard/product/250409.jsp

http://www.plugwise.com/

http://www.enocean-alliance.org/en/home/?rnd=1

http://www.enocean-alliance.org/en/home/?rnd=1

http://www.bacnet.org/

http://www.autodesk.co.uk/adsk/servlet/pc/index?siteID=452932&id=14645193

http://www.autodesk.co.uk/adsk/servlet/pc/index?siteID=452932&id=14645193

http://usa.autodesk.com/ecotect-analysis/

http://usa.autodesk.com/ecotect-analysis/

https://www.jboss.org/drools/

Carbon Reduction Challenge and Promoting HumanHealth in Buildings, Kyoto, Japan, May 28–31, 2012 pp.825–832.

Reinhart, C. F. (2004). Lightswitch-2002: a model for manualand automated control of electric lighting and blinds. SolarEnergy, 77(2004), 15–28.

Li, D. H. W., Cheung, K. L., Wong, S. L., & Lam, T. N. T.(2010). An analysis of energy-efficient light fittings andcontrols. Applied Energy, 87(2010), 558–567.

Lee, E. S., & Selkowitz, S. E. (2006). The New York TimesHeadquarters daylighting mockup: monitored performanceof the daylighting control system. Energy and Buildings,38(2006), 914–929.

Yun, G. Y., Kim, H., & Kim, J. T. (2012). Effects of occupancyand lighting use patterns on lighting energy consumption.In Energy and Buildings, 46(2012), 152–158.

Milagro, F., Antolin, P., Kool, P., Rosengren, P., Ahlsén, M., Müller,J., et al. (2008). SOAP tunnel through a P2P network ofphysical devices. Sophia Antopolis: In Internet of ThingsWorkshop.

FIEMSER (2007). http://www.fiemser.eu/. Accessed September2012.

SaveEnergy (2011). http://www.ict4saveenergy.eu/. AccessedSeptember 2012.

BeAware Project (2011). http://www.energyawareness.eu/beaware/. Accessed September 2012.


Documents

Energy saving in existing buildings by an intelligent use of interoperable ICTs