Ei

DP

a

ARRA

KMcRSE

1

asetDtreEftCtab

f

0d

Energy and Buildings 43 (2011) 1392–1402

Contents lists available at ScienceDirect

Energy and Buildings

journa l homepage: www.e lsev ier .com/ locate /enbui ld

nabling machine understandable exchange of energy consumptionnformation in intelligent domotic environments

ario Bonino, Fulvio Corno ∗, Faisal Razzakolitecnico di Torino, Dipartimento di Automatica ed Informatica, Corso Duca degli Abruzzi 24, 10129 Torino, Italy

r t i c l e i n f o

rticle history:eceived 22 July 2010eceived in revised form 18 January 2011ccepted 25 January 2011

eywords:achine understandable power

onsumption information

a b s t r a c t

In the 21st century, all the major countries around the world are coming together to reduce the impactof energy generation and consumption on the global environment. Energy conservation and its efficientusage has become a top agenda on the desks of many governments. In the last decade, the drive to makehomes automated and to deliver a better assisted living picked pace and the research into home automa-tion systems accelerated, usually based on a centralized residential gateway. However most devisedsolutions fail to provide users with information about power consumption of different house appliances.The ability to collect power consumption information can lead us to have a more energy efficient society.

DFemantic energy information publishingnergy Profile ontology

The goal addressed in this paper is to enable residential gateways to provide the energy consumptioninformation, in a machine understandable format, to support third party applications and services. Toreach this goal, we propose a Semantic Energy Information Publishing Framework. The proposed frame-work publishes, for different appliances in the house, their power consumption information and otherproperties, in a machine understandable format. Appliance properties are exposed according to the exist-ing semantic modeling supported by residential gateways, while instantaneous power consumption is

odu

modeled through a new m

. Introduction

Energy conservation is a rising concern for many countriesround the world. The resources used to generate energy, theircarcity and the rising impact of those resources on the globalnvironment have made energy conservation a top agenda on theables of high government officials around the world. In USA, theepartment of Energy (DOE) launched a Weatherization Assis-

ance Program that enables low-income families to permanentlyeduce their energy bills by making their homes more energyfficient.1 The US Environmental Protection Agency defined annergy Conservation Action Plan which addresses opportunitiesor energy conservation in homes, schools, offices and indus-rial environments through the use of energy-saving innovation.2

hina introduced a medium and long term energy conserva-ion plan to push the whole society towards energy conservationnd energy intensity reduction, to remove energy bottlenecks, touild an energy saving society, and to promote sustainable social

∗ Corresponding author. Tel.: +39 011 564 7053; fax: +39 011 564 7099.E-mail addresses: dario.bonino@polito.it (D. Bonino),

ulvio.corno@polito.it (F. Corno), faisal.razzak@polito.it (F. Razzak).1 http://apps1.eere.energy.gov/weatherization/.2 http://www.epa.gov/.

378-7788/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.enbuild.2011.01.013

lar Energy Profile ontology.© 2011 Elsevier B.V. All rights reserved.

and economic development.3 The International Energy Associationpublished statistics of energy consumption by sector [1], accordingto which China uses 38.2% and 40.0% of its total energy on the resi-dential and industrial sectors, respectively. Europe uses 26.6% and32.2% of its total energy on the residential and industrial sectors.Currently, a trend can be seen that developing countries with grow-ing population use a major portion of their energy in the residentialand industrial sectors.

Ambient, ubiquitous and intelligent computing have providedstimulus to the research of a number of residential gateways [2–5]which provide control of appliances in a house and access to thegeneral appliance properties. Access to this information can be pro-vided locally through a software application or remotely over theweb. Due to the variety of approaches proposed or adopted in SmartHome research, we rely on a somewhat restricted definition, thatfocuses on the current applicability of Smart Home technologies.

In this paper we target Intelligent Domotic Environments (IDE),defined as “environments where commercial domotic systems4

are extended with a low cost device (embedded PC) allowing inte-gration and inter-operation with other appliances, and supporting

3 http://www.chinaenvironmentallaw.com/wp-content/uploads/2008/04/china-medium-and-long-term-energy-conservation-plan.doc.

4 The word “domotics” is a contraction of the latin word domus, for house, withinformatics, and represents the residential extension of “building automation”.

Build

macc

taosssud

pmeintSglegcai

SpsuWaehn

ioDeSit

smatorpo(em

sms

t

D. Bonino et al. / Energy and

ore sophisticated automation scenarios” [6,7], as they currentlychieve advanced intelligence at a relatively low cost, enabling thereation of new building automation scenarios, with much moreomplex behavior and functionality.

If residential gateways provide energy consumption informa-ion then energy providers or 3rd party players could providepplications to increase energy awareness among consumers. Mostf these residential gateways [2–5,8] do not provide the energy con-umption information about different appliances in the house. Toupport different types of applications and services the energy con-umption information should be exposed in an open and machinenderstandable format, so that different applications can use theata according to their own diverse goals.

The goal of this paper is to enable residential gateways to exposeower consumed by different appliances installed in a house, in aachine understandable format, to support the development of

xternal applications. In the home environment we are interestedn active power only, since reactive power is much smaller and isot billed by most energy providers to residential users; thereforehroughout all of this paper, we always refer to active power, only.5

uch external applications, starting from data published by theateway, can provide visualization of energy information (eitherocally or on the web), can provide statistics and analysis of thenergy data, and in the near future may aim at achieving intelli-ent negotiation and consumption coordination. Exposing energyonsumption in a neutral and machine understandable format willllow multiple services to use the energy and power consumptionnformation according to their own application goals.

To achieve the aforementioned goal, this paper proposes aemantic Energy Information Publishing Framework (SEIPF) whichublishes, for different appliances in the house, their power con-umption information and other appliance properties, in a machinenderstandable format. Data is published according to the Semanticeb standards and best practices, to ensure application neutrality

nd intelligent machine processing. While appliance properties arexposed according to the existing semantic modeling supported byome gateways, power consumption is modeled by introducing aew modular Energy Profile (E.P.) ontology.

The proposed framework is consistent with publication ofnformation at different granularity levels (e.g., by aggregatingver device groups) and respecting different Authorization levels.epending on the type of application, different complexity lev-ls require different complexity in data representation, so that theEIPF is able to expose the data both as simple RDF triples (accord-ng to Linked Data requirements) and as full ontology instances, forhe benefits of applications needing intelligent processing.

A different complexity dimension is time, since energy con-umption is a real-time process showing variable speed, andeasuring and publishing power information must take into

ccount the time instants and the time intervals corresponding tohe published figures. The SEIPF proposed in this paper, being basedn Semantic Web and Linked Data standards, is easily extensible byeferring to standard ontologies, and can describe arbitrary com-lexity levels in information structure. Although this paper focusesn measuring and publishing instantaneous power consumptioni.e., with the measure taken at the time of querying), it will bextended in future to support time-related queries and aggregateeasures.

Today the ability to collect and share instantaneous power con-

umed by different devices in a house can enable the creation ofany applications that can lead towards a more energy efficient

ociety. These applications will be making consumers aware of

5 Incidentally, low-cost power meters compatible with affordable domotic sys-ems are usually not capable of measuring reactive power.

ings 43 (2011) 1392–1402 1393

their active power consumption through power meters, sharingtheir data over the web to create mash up applications, and/or con-sulting services to provide feedback on different appliances powerrequirements. In the future, intelligent negotiation and consump-tion coordination will allow third-party service providers to buildintelligent and automated services that use the energy consump-tion information to build dynamic services, such as automatic loadtransfer over intelligent energy grids. A basic requirement to fulfillthe scope of aforementioned applications is a standard, open andsemantic representation of a device’s power consumption infor-mation, so that applications may use this information according totheir own goals.

The remainder of this paper is divided into eight sections.Section 2 provides the target scenarios from which needs andrequirements driving the approach are derived. In Section 3, themain issues to achieve target scenarios are outlined. Section 4explains the basic principles on which the envisioned solution isfounded. In Section 5 a possible solution is presented, which isfurther detailed in Section 6. Section 7 provides implementationdetails of the proposed framework along with the experimentalresults by showing samples applications built to prove the feasibil-ity of the solution. Section 8 presents related works and Section 9concludes the paper and discusses possible extensions.

2. Scenarios and motivation

2.1. Scenario 1: home energy management system

Consider an energy provider, which provides its residentialconsumers a home energy management system. The system isconnected through a home network to a smart utility meter andelectrical appliances. To reach goals of energy awareness and effi-ciency, the system provides different applications to track thepower consumption of different appliances inside the house. It pro-vides tools to monitor current energy needs, delivering an analysison the power consumed over time and suggestions on better energymanagement plans. The system is a plug and play management sys-tem in which third-party applications can be installed to provideconsulting services. These services can give suggestions, such asthe vendors that provide more energy efficient devices, or plans tosave money by saving energy. Shifting the use of major appliancessuch as dishwashers and clothes dryers to hours with a lower over-all electricity usage can help utilities meet the energy demandsand help consumers save the energy under demand-based pricingplans. Providing real-time information linked to such dynamic pric-ing may be a winning combination for consumers who want to cutenergy costs.

To support the system, issues of gathering information relatedthe power consumed by appliances, publishing this information,providing an open and standard format for information have to beaddressed.

2.2. Scenario 2: 2020 intelligent energy grids

Consider the energy delivery and consumption landscape in2020 (or even before). In 2020 the world energy demand has grownby 76% with respect to 2007, requiring 4800 GW of capacity addi-tions, almost five times the 2009 capacity of the US [9]. In thisscenario, energy production and delivery dramaticaly relies onsmart grid solutions to effectively distribute the available energy

(mostly electrical) and to coordinate with consumption demandsto avoid peaks and abnormalities that today require oversizing ofdistribution and production systems.

One of the main contributions to the future ability to copewith such a high energy demand is the improved and automated

1 Build

chcmbcaaicbldfcesotetl

wdshvr

tttisdntW2ir

tttpwmfpseeadiea

3

go

394 D. Bonino et al. / Energy and

ooperation between consumption centers, be they residentialouses (around 30% of the current consumption) or industrial andommercial facilities. The Internet growth and the take off of Infor-ation and Communication technology and Artificial Intelligence

ased techniques has driven the energy distribution scenario to theurrent state where consumers and producers continuously andutonomously negotiate the best trade-off between energy needsnd availability. Take 2020 homes as an example, they are automat-cally communicating and coordinating their energy needs. In everyity district, single homes interact and communicate with neigh-ors to shape the global district consumption, to activate home

evel energy transformation and to coordinate local energy pro-uction, thus reducing the cumulative amount of energy requiredrom the main electric distribution network and almost eliminatingonsumption peaks. This amazing capability of coordinating differ-nt homes together can be observed every day: at each day hourome houses are producing energy thanks to solar cells installedn their roofs or to thermal co-generation of their heating sys-ems. The homes that are not generating, or that require additionalnergy, negotiate with neighbors energy transfers, minimizinghe need of ‘external supplies’ through the main power deliveryine.

Even countries are coordinating and collaborating in the sameay, while one hemisphere of the world is sleeping, energy pro-uction is mainly routed on the illuminated side of the earth,upporting the higher day-light consumption request. Everythingappens seamlessly, and if observed from a distant energy point ofiew the whole globe is traversed by a steady wave of energy, thategularly feeds human activities, 24 × 7.

This futuristic, but still realizable, scenario involves many sub-le issues, that need to be unveiled in order to guide research onhe technology infrastructure needed to support it. At the basis ofhe depicted scenario, the Internet acts as a connective tissue, flow-ng energy related information between different involved entitiesuch as homes, industries, offices, power delivery and power pro-uction plants. On top of this connection network, data exchangeeeds common, machine understandable formats to enable allhe above intelligent negotiations and consumption coordinations.

e do not go further in the analysis of the issues raised by the020 scenario, instead we focus on this information exchange

nfrastructure on which every advanced consumption policy isooted.

In particular, we must acknowledge that large-scale coordina-ion may not rely solely on the efforts of the utility providers,hat often lack the details about how energy is consumed byheir customers, and that cannot take into account increasing self-roduction of power by end users. In this context, users must beilling to share part of their consumption (and production) infor-ation, in real time, for the benefit of advanced monitoring and

orecasting applications: this implies that end-users should directlyublish their information and rust the utility providers and otherervice providers to use it an provide added-value services. How-ver, the intelligence of coordination application is expected toxpand over the next years, and to be able to encompass morend more sources of information. For this reason the publishedata should be application-agnostic, and available in an open and

nteroperable way. As Tim Berners-Lee literally shouted at the audi-nce [10], we need “Raw Data Now” to enable future intelligentpplications.

. Design issues

The core requirement of the defined scenarios is the ability toather the current power consumption information of an appliancer a group of appliances in the house and exposing that information

ings 43 (2011) 1392–1402

over the Internet. To accomplish scenarios mentioned in Section 2three main issues need to be addressed: gathering the power con-sumption information, publishing this information, and making theinformation usable by machines. Gathering power consumptioninformation means measuring the active power consumption ofhome devices at a given time instant. Such a measure is greatlyfacilitated if a smart or home automation plant is available in thehome. Publishing information means deciding, and most impor-tantly, enabling the householders to decide which informationto expose and at what granularity. Finally, distributing machineunderstandable information implies the adoption of an open andeffective data format that enables machines to interpret them;the Semantic Web and Linked Data research communities havealready paved the way in this direction. By going a little furtheron these three main issues, we can identify the following relatedneeds.

3.1. Energy consumption information

The power consumption information of an appliance or group ofappliances is exchanged between smart homes, energy providers,and/or any third-party applications. Homes will be equipped withappliances ranging from a lamp to a fully automated heating con-trol system. An appliance during its operation can have differentoperating states like on, off, stand-by, up, and down. In differentstates the appliance can have different power consumption lev-els: we need a mechanism to encode this state-dependent powerconsumption information. The residential gateway should providethe power consumption of appliances according to their currentoperating states. The availability of a house automation system,coupled with the knowledge of device characteristics, allows us toestimate and couple the power consumption information at a muchfiner level, with less expensive means compared to installing powermeters.

3.2. Energy information publishing in a machine understandableformat

Power consumption information should be distributed in anopen, machine understandable, semantic enabled and effectivedata format, so that the residential gateway can act as a singlepower consumption information point. This information point canprovide inputs for multiple and diverse third-party applications,services and potentially automated agents as well.

3.3. Information Publishing Control

Residential gateways are designed to provide automated controlto the house and thus they have information about every appli-ance in the house. The information may comprise the applianceproperties, its power consumption and procedures to automati-cally control the appliance. This information, if utilized by thirdparty applications, services or automated agents, has the potentialto provide the consumer better services. However, this informationsharing should be governed by a sound access control mecha-nism so that the basic consumer rights and privacy issues areaddressed. Privacy is a subjective issue, and different consumersor even people living in a single home might perceive it differ-ently from others. Several case studies in relation to the issue ofprivacy have been carried out [11,12] and the research is still ongoing.

4. Basic principles

The web today is witnessing a paradigm shift from a displayeddata era to a new era of well understood meaningful information

D. Bonino et al. / Energy and Buildings 43 (2011) 1392–1402 1395

evice

oao[aatab

4o

mddbbet

Sieaitit

Fig. 1. A sample d

ver the web. This is the vision of the future web, better knowns Semantic Web. W3C6 describes the Semantic Web as the webf data: it creates a universal medium for the exchange of data13]. The Semantic Web vision will enable automated negotiationnd retrieval of machine understandable information among webpplications, services and agents. The basic components that formhe basis of Semantic Web and the necessary ingredients needed forsolution of the three issues mentioned in Section 3 are explainedelow.

.1. Resource Description Framework (RDF), Linked Data andntology

RDF [14] is the framework presented by W3C for representingachine understandable information on the web. It is a simple

ata model to describe the resources over the web. RDF has beenesigned for situations in which information needs to be processedy applications, rather than being only displayed to people. Theasic structure of any expression in RDF is a collection of triples,ach consisting of a subject, a predicate and an object. A set of suchriples is called an RDF graph.

According to Tim Berners-Lee’s web architecture note [10], theemantic Web is not just about exposing machine understandablenformation over the Internet but making links between differ-nt exposed information sets, so that a machine, an application,service or a person can find related information. The availabil-

ty of data from different sources in a universal format and linkedogether is known as ‘Linked Data’. Linked Data7 (LD) assumes thatnformation is available in RDF format. Formally, Linked Data is aerm used to describe recommended best practices for exposing,

6 http://www.w3.org/.7 http://linkeddata.org/.

model in DogOnt.

sharing, and connecting pieces of data, information, and knowl-edge on the Semantic Web using URIs and RDF. To capture domainknowledge in a generic way, and provide a commonly agreedunderstanding of a domain, which may be reused and sharedacross applications and groups, the concept of Ontology is used.An ontology is a formal specification of a shared conceptualiza-tion [15]. W3C provides the Ontology Web Language (OWL) todescribe ontologies about a particular domain [16]. OWL has threeincreasingly expressive sub languages: OWL Lite, OWL DL, and OWLFull.

4.2. DogOnt

DogOnt [6] is an ontology able to model the domotic systemof a house with focus on supporting the intelligent operations ina domotic environment. DogOnt has been designed along 7 mainhierarchical trees, corresponding to the Building Thing, BuildingEnvironment, Functionality, Command, DomoticNetworkComponent,State, StateValue concepts. Classes descending from Building Envi-ronment and Building Thing are used to describe the environmentof the house. DogOnt provides the ability to model differentdevices that exist in a house through the Building Thing con-cept and DogOnt has a clear separation between the Controllableand Not Controllable devices. The Functionality class is relatedto the different functionalities of the device. The Command andNotification concepts are linked to specific functionality instances.Network modeling is achieved through DomoticNetworkCompo-nent. State and StateValue are used to model the different states

of a device. A very basic device model is shown in Fig. 1. Itillustrates a device sample lamp which has two inherited func-tionalities named OnOffFunctionality and QueryFunctionality (notshown). The sample lamp has a state OnOffState and is locatedin sample living room. The interested readers are referred to [6]

1396 D. Bonino et al. / Energy and Build

oD

4

ttisootfs

4

bTatnpiddmeaTtAscic

Fig. 2. WoD layer inside Dog.

r to the DogOnt website8 for more details and examples aboutogOnt.

.3. Dog

Dog [7] is an ontology-powered Domotic OSGi Gateway (Dog)hat is able to expose different domotic networks as a single,echnology neutral, home automation system. It is versatile ast is built on top of the OSGi framework and the adoption ofemantic modeling techniques allows Dog to support intelligentperations inside the home environment. Dog uses the DogOntntology to model a house environment. Dog provides the abilityo control different devices installed in a house and to query dif-erent device properties ranging from location to current operatingtate.

.4. Web of Domotics (WoD)

WoD [17] is a collaboration architecture derived by com-ining components of three main concepts, namely: Internet ofhings, Ubiquitous Computing and Domotics. It is an Internetrchitecture enabling mobile users to access device informa-ion and operate them in a ubiquitous manner, independent ofetwork-specific location dependence. It addresses issues likeroximity-based device identification (e.g., visual tags), network

ndependent detection of service access points (through DNS-basedevice de-referencing), user identification through OpenID, openata exchange, service/device description through Linked Data for-ats and device operation through REST-based interaction. WoD

xtends Dog to include a new WoD layer (Fig. 2), that comprisesHttpAccess unit and an Authentication and Authorization unit.

he HttpAccess unit provides the ability to get information ando control devices in a domotic system over the Internet. Theuthentication and Authorization unit provides an authenticationtub that forwards the incoming request to third-party authenti-

ation services like OpenID or OAuth.9 The Authorization portions used to authorize the incoming requests according to local poli-ies.

8 http://elite.polito.it/dogont-tools-80.9 http://oauth.net/.

ings 43 (2011) 1392–1402

5. Proposed solution

To provide residential gateways with the ability to expose thepower consumption information of an appliance or a set of appli-ances installed in the house and also do so by addressing theissues raised in Section 3, this paper proposes a Semantic EnergyInformation Publishing Framework (SEIPF). The SEIPF exposes thepower consumption information of appliances along with differ-ent appliance properties in RDF format over the web. The EnergyConsumption Information modeling issue (Section 3.1) is addressedby defining a new Energy Profile (E.P.) ontology (defined in Sec-tion 5.1). This ontology is based on the modularity pattern andmodels the energy consumption information about any appli-ance that modeled through the underlying domotic ontology, i.e.,DogOnt. The modularity pattern provides separation to model dif-ferent aspects of a system through separate ontologies and may beplugged on top of various ontologies. The Machine UnderstandableFormat issue (Section 3.2) is addressed by adopting RDF as the stan-dard format to expose information because it provides meaningfulrepresentation of information which can be semantically post pro-cessed, as explained in Section 4. The complete approach is definedin Section 5.3. The Information Publishing Control issue (Section 3.3)is currently addressed by using the Authentication and Authoriza-tion unit available inside the WoD architecture. However, in thefuture we intend to incorporate an ontology based access controlpolicy. It is further explained in Section 5.2.

5.1. Energy Profile ontology (E.P.)

To model the energy consumption information, an Energy Pro-file (E.P.) ontology has been developed. It models the energyconsumption information about different appliances in the house.The E.P. ontology is developed according to the modularity pattern,so that it can be attached to any ontology that can describe thedomotic environment of a building (DogOnt in our case). The basicconcepts of the E.P. Ontology are DeviceProfile and Consumption(as shown in Fig. 3).

1. DeviceProfile: This class describes energy profiles of all the majordevice categories in the house. The energy profile informationcan be related to different appliances, such as lamp, coffeemaker,and dishwasher. This class has two properties.• hasDevice: This property specifies the instance of the device

to which this DeviceProfile applies. The property maps ontothe DogOnt device instances (i.e., instances of the Controllableclass).

• hasConsumption: Every device may have different levels ofpower consumption, depending on the operating state of theappliance: each DeviceProfile collects various power Consump-tion object instances, one for each allowed device state.

2. Consumption: This class encodes the power consumed by theappliance in a given state. For each device (e.g., Lamp), differentstates (e.g., LampOn and LampOff) are described, each corre-sponding to a different power consumption level. Consumptioninstances have four properties:• associatedState: This property specifies the state of the appli-

ance whose power consumption we are describing with thisinstance. The property maps onto the DogOnt ontology, whereeach device is described in terms of its allowed states.

• nominalValue: This property shows the nominal power con-sumption of the appliance in the given state. It gives the

estimated power consumption of a device in a state.

• realValue: This property is the measured power consumptionof an appliance in a given state. This property is used if thedevice has a power characterization available, otherwise thenominal value is used.

D. Bonino et al. / Energy and Buildings 43 (2011) 1392–1402 1397

y Profi

tcctiptD

aLaa

5

htcpcc

Fig. 3. Energ

• hasUnit: This property defines the unit of power for the powerconsumed by the appliance, expressed as one pre-definedinstance of the MetricUnit class in the Measurement UnitsOntology.10

The E.P. ontology defines two extension points through whichhe ontology defining the domotic system (in our case DogOnt)an be attached. The first is the hasDevice property of DeviceProfilelass, which attaches a energy consumption information structureo a device or an appliance. For example, in DogOnt it relates tonstances of Controllable concept. The second is the associatedStateroperty of the Consumption class, which relates a given state ofhe appliance or device to its consumption level. For example, inogOnt it relates to the instances of the StateValue concept.

A fragment of E.P. instances is shown in Fig. 4, where we definesingle Device Profile instance named SimpleLamp EP. Simple-

amp EP is attached to two consumption instances SimpleLamp Onnd SimpleLamp Off, which are instances of LampOnConsumptionnd LampOffConsumption classes, respectively.

.2. Information access control

The residential gateway houses different chunks of data about aome ranging from appliance properties and operations to sensinghe presence of people and their choices. Some of this information

an be utilized to provide a better standard of living to the peo-le living inside the house, e.g., informing the people about theirurrent energy consumption can help them to be more energy effi-ient. Related approaches to provide semantic access control to a

10 http://idi.fundacionctic.org/muo/muo-vocab.html.

le ontology.

system may be found in the literature. Pan et al. [18] proposed aSemantic Access Control Enabler (SACE), a middleware-based sys-tem that has been designed and implemented to enable SemanticAccess Control on the Web. Toninelli et al. [19] proposed a seman-tic context aware policy model that adopts ontologies and rules toexpress context and context-aware access control policies and sup-ports policy adaptation. Ionita and Osborn [20] proposed a modelthat regulates access control on ontologies defined in the SemanticWeb.

Since the main issue of this paper is on power consumptioninformation sharing, a very basic solution is adopted for our SEIPFimplementation: currently the SEIPF uses the Authentication andAuthorization mechanism provided by the WoD architecture tocontrol the access to information.

5.3. Machine understandable format

Exposing the power consumption information in an open,neutral and semantic format will allow multiple services to useinformation according to their application goals and, in future, willallow intelligent negotiation between automated software agents.To achieve the aforementioned task, RDF is adopted as an openformat that has embedded semantic information which allowsinformation to be machine understandable. The SEIPF publishesall information pertaining to an appliance, including its operatingstates, related current power consumption and general properties,as a pure RDF structure.

To provide reasoning support over the RDF response receivedfrom the SEIPF, a basic set of general concepts are defined througha vocabulary. The vocabulary is encoded in the SimpleDomoticDataontology, which is very similar to the E.P. ontology but it has beencreated separately for following reasons:

1398 D. Bonino et al. / Energy and Buildings 43 (2011) 1392–1402

nsum

1

2

bts

1

2

6(

tatii

pimHtW

sumption level of an appliance/device in the house based onthe current state of the device. It takes the device identifier andreturns the power consumption information about the device asan instance of SimpleDomoticData ontology. The device identi-

Table 1List of parameters.

Parameter Values Example Detail

command info info To request an appliancepower consumption

device device id lamp9 The identifier of thedevice

room room id livingroom To request informationabout all devices in thelivingroom

devicecategory device type Lamp To ask the power

Fig. 4. An excerpt of the power co

. The E.P. ontology was developed to model only the power con-sumption of an appliance or group of appliances. The actualappliance modeling is provided through the DogOnt ontology.Whereas, in the SimpleDomoticData ontology, we model theappliance, its properties and its power consumption in currentstate. It can be extended to include other appliance properties.

. Providing the response based on the DogOnt and the E.P. ontolo-gies would have required external applications to understandthe more complex structure of DogOnt with additional informa-tion about the domotic system, which in this case is not alwaysneeded. Therefore, to provide easy and simple integration a sim-plified SimpleDomoticData vocabulary is preferred to encode theresponse.

The simplified ontology and its general concepts are explainedelow, and are basically the minimum set of classes and proper-ies, extracted from the E.P. and the DogOnt ontologies, that allowemantic publishing of power information:

. Device: This class indicates the appliance for which power con-sumption is inquired.• hasConsumption: This property points links this instances of

the Device class to the Consumption class (defined below)instances describing its current power consumption value.

• hasState: This property defines the actual state value of theinstance of State class, as a string.

. Consumption: This class encodes the power consumption infor-mation of the appliance in the current state.• hasUnit: This property defines the unit of power for the power

consumed by the appliance, according to the MeasurementUnits Ontology.

• value: This property shows the power consumption of a device,encoded as a real number.

An example of such encoding is shown in Fig. 5.

. Semantic Energy Information Publishing FrameworkSEIPF) architecture

The SEIPF can be installed on any centralized residential gatewayhat uses the DogOnt ontology to model the domotic structure ofn environment. It comprises a core Publishing Unit that provideshe power consumption details pertaining to different appliancesn the house and other appliance properties and is explained moren Section 6.1.

The SEIPF is integrated with the WoD architecture. The WoDrovides an interface over the web based on the REST over HTTP

nteraction paradigm to access the domotic system of an environ-ent. The interface is implemented through an OSGi bundle namedttpAccess. The HttpAccess bundle was extended to add new func-

ionalities to query the SEIPF. The integration of the SEIPF and theoD archiecture provides functions that enable a requesting entity

ption information about a device.

to acquire information in Linked Data format about the generaldomotic structure of the environment and identification of dif-ferent devices installed in the environment. The requesting entitycould be a third-party application, a service or an automated agent,etc.

The request can be made with certain parameters (using theHTTP GET method) to retrieve the power consumption informationfrom the SEIPF. The parameters and their possible values are shownin Table 1.

Access control is provided through the Authentication andAuthorization bundle of WoD. To provide an authentication whichis scalable and has a web wide scope, WoD adopts Open ID [21]as an authentication mechanism for the requesting entity. Open IDis a decentralized standard to authenticate the requesting entity.Currently, Authorization is provided through a type-based policybut as stated earlier, we are working on a separate ontology-basedpublishing policy that will be independent of WoD.

6.1. Publishing unit

The publishing unit is the core logic unit of the SEIPF (Fig. 6). Itaccepts the request through the HttpAccess bundle. Based on therequest, it queries the XML-RPC bundle of Dog to determine thecurrent state of an appliance or set of appliances. Then it determinesthe power consumption of appliances by quering the E.P. ontologyand DogOnt ontology (available in the HouseModel Bundle of Dog).The publishing unit then sends a pure RDF response (based on theSimpleDomoticData ontology) to the requesting entity.

It has three methods, which can be accessed through the HttpAc-cess bundle by providing different parameters depicted in Table 1:

1. getPowerInfo: This function provides the current power con-

consumption ofLamp-type devices.

query SPARQL query – To query the DogOntand E.P. ontologiesdirectly.

D. Bonino et al. / Energy and Buildings 43 (2011) 1392–1402 1399

Fig. 5. An excerpt of the energy consumption information about a device in the SimpleDomoticData format.

ramew

2

3

6

iomidottaeTe

application goals. Applications can be standalone providing thefeedback on the current power consumption of devices or theycould be mash-up applications that consume data from differentdata sharing sources to provide feedback.

Fig. 6. Publishing f

fier is mentioned through the device parameter and the commandparameter must be set to value info. An excerpt of the responseis shown in Fig. 5.

. getRoomPowerInfo: This function provides the power consump-tion of all appliances present in a given room inside the house.The room is specified through the room parameter.

. getDeviceTypePowerInfo: This function provides the power con-sumption of all appliances belonging to a single device category.By device category we mean the type of the device. For exam-ple, Lamp is the device category for all lamps inside the house.The name of the device category is passed as a devicecategoryparameter.

.1.1. SPARQL endpointIn addition to predefined queries exposed through the WoD

nterface, the SEIPF also allows direct querying of the underlyingntologies, for applications that need to reason about the wholeodel, and not just to use the power data. We terefore provide an

nterface to send arbitrary queries, using the Semantic Web stan-ard query language, SPARQL. The SPARQL endpoint provides thentological level access to the requesting entity. To ensure safety,he access to this point is granted to highest levels of Authoriza-

ion, only. SPARQL queries are forwarded to query the DogOntnd the E.P. ontologies directly, for extracting general device prop-rties and the power consumption of the device. As defined inable 1, the query parameter can be used to access the SPARQLnd-point.

ork architecture.

7. Implementation and experiments

The core publishing unit of the SEIPF is implemented as a sepa-rate OSGi bundle inside the Dog. We adopted the Eclipse Equinox11

OSGi framework to implement this bundle. As stated earlier, theSEIPF use the HttpAccess bundle (inside the WoD architecture)to provide access over the web. The HttpAccess bundle uses thejetty web server12 to expose its services. Access control is pro-vided through the Authentication and Authorization bundle (insidethe WoD architecture). It uses the OpenID4Java13 library to for-ward the authentication credentials to the appropriate externalOpenID server. Requesting entities are authorized according todifferent levels, based on the type of entity accessing the frame-work.

The goal of the SEIPF is to provide residential gateways the abil-ity to expose the power consumption of different devices in anopen, effective and semantic format, which in turn enables exter-nal applications to consume information according to their own

11 http://www.eclipse.org/equinox/.12 http://www.mortbay.org/jetty/.13 http://code.google.com/p/openid4java/.

1400 D. Bonino et al. / Energy and Buildings 43 (2011) 1392–1402

eDcaigstmTsslpi

dattsu

vtqegaF

aovhoatRaaiit

Fig. 7. BTicino and KNX domotic demo cases.

To demonstrate such use cases, we have implemented twoxperimental applications. We integrated the SEIPF with theog and ran tests in our department Lab, using two demoases (shown in Fig. 7) equipped with a BTicino MyOpen and

KNX domotic plant, respectively. In the absence of a realnhabited house, we used the emulation capabilities of the Dogateway to simulate the behavior of devices configured in theample houses. In fact, Dog simulates domotic environmentshanks to the DogSim [22] library, that exploits a state machine

odel describing the dynamic behavior of each device class.hanks to DogSim, we are able either to interact with fullyimulated environments, or to environments that include someimulated devices and some real domotic devices: we call “emu-ation” this last case, in which we may interact with a reallant by emulating some new devices before they are actually

nstalled.We marked different buttons to emulate the functioning of three

evices for our experiment namely: a computer, a coffeemakernd a lamp. The experiments were run with arbitrary input condi-ions, and no special assumption has been made over user behavior;herefore the attained results may be used to validate the mea-uring and publishing framework, but are not suitable to inferser-related conclusions.

The first experiment exploits a standalone application that pro-ides the current power consumption of the emulated devices. Theesting application was built as a separate Java application thatueries the SEIPF to acquire the current power consumption ofmulated devices and uses the Google Chart Tools14 to provideraphical feedback on the current power consumption of appli-nces. A snapshot of the response of the application is shown inig. 8.

The second experiment acts as a data sharing application. Itccumulates the power consumption of individual emulated devicever time. The application shares the data with the Pachube15 ser-ice. Pachube is a convenient, secure and scalable platform thatelps applications and services connect to and build the Internetf Things. It stores, shares and discovers realtime sensors, energynd environment data from objects, devices and buildings aroundhe world. Pachube provides most of its functionality through aEST based API and can be used to send real time sensor, energynd environment data from anywhere around the globe. We built

Java application that periodically queries the SEIPF for polling

nstantaneous power consumption of appliances at regular timentervals. The application uses polling to acquire data over time andhen sends the power consumption data to the Pachube server. A

14 http://code.google.com/apis/charttools/.15 http://www.pachube.com/.

Fig. 8. Current power consumption of emulated devices.

snapshot of the power consumption data accumulated over timeon Pachube is shown in Fig. 9.

The two experimental applications prove the feasibility of theframework as well as provide a step towards defining an open,standard and semantic powered format that will allow differentapplications to use the power consumption data according to theirown application specific goals.

8. Related works

Several works can be found in the literature that approachthe residential energy consumption problem from many differ-ent viewpoints. These different approaches can be classified onthe basis of the tackled aspects, e.g., estimation of consumption,gathering of available information, processing of measured con-sumption data, to cite the most relevant. In the first category,Fumo et al. devised a simplified methodology to estimate hourlyelectrical and fuel energy consumption of a (residential) build-ing by applying a series of predetermined coefficients to monthlyenergy consumption data from electrical and fuels utility bills[23]. This approach can exploit the publishing framework intro-duced in this paper to gather precise data on consumed energyand fuel and may exploit the easy-to-access information exposedby the SEIPF to increase the granularity of the estimation itprovides.

Pérez-Lombard et al. worked on a review of available infor-mation concerning energy consumption in buildings [24], inparticular related to HVAC systems. This work can be easilyintegrated in the SEIPF approach, contributing to first identify rel-evant information to be gathered and, second, receiving back asa benefit, more granularity on the same information plus inte-gration with other information sources available in the home forwhich explicit billing data or statistical consumption data is stilllacking.

In the information processing area many approaches can befound, which are more strictly related to the proposed SEIPF: Seem,for example, introduced a solution for the detection of abnormalenergy consumption values in buildings [25] that exploits intelli-gent data analysis. In his paper, Seem describes a novel method

for detecting abnormal energy consumption in buildings based ondaily readings of energy consumption and peak energy absorp-tion. In this context SEIPF can act as data source allowing fordirect collection of measurements and providing the basis for fineranalysis based on shorter intervals of the order of minutes or

D. Bonino et al. / Energy and Buildings 43 (2011) 1392–1402 1401

chube

ssom

awottObbemattits

udmdDhMam[t

sitdTRwtaan

Fig. 9. A snapshot of the graph obtained on Pa

econds. At the same time, data exposed by the SEIPF can be con-umed by other detection services allowing for the integrationf several analysis toolkits in the same home/building environ-ent.Google PowerMeter [26] is a free energy monitoring tool that

llows you to view your home’s energy consumption from any-here online. It assumes that the device must make an SSL-secured

utbound TCP/IP connection to Google so that it can periodicallyransmit data to Google via HTTPS. Typically, a device should usehe device owner’s home Internet connection to transmit data.ur approach is different from that of Google PowerMeter. Ourasic assumption is that all devices in the house are controlledy a centralized residential gateway. The publishing frameworkxposes device information in a machine understandable RDF for-at, which can be semantically post-processed by any third party

pplication, service or automated agent. The information is exposedhrough a proper authentication mechanism and the resident ofhe house is provided the complete control over information thats exposed through the framework. Incidentally, a simple applica-ion using the SEIPF could act as a client of the Google PowerMeterystem.

Weiss et al. [27] proposed an interactive feedback system thatses a smart electricity meter to provide consumption feedback forifferent household devices. It also provides a set of API to com-unicate with the smart electricity meters. The SEIPF approach is

ifferent as it can be installed on any residential gateway that usesogOnt to define the domotic structure of a house. The SEIPF alsoas the ability to expose different information related to devices.oreover, the SEIPF exposes information in pure RDF format which

llows any application to consume the power consumption infor-ation according to its own application requirements. By contrast,

27] provides a predefined custom format, and exposes consump-ion information, only.

Sheth et al. [28] proposed that the sensor data retrieved fromensor networks is annotated with semantic metadata to increasenteroperability as well as provide contextual information essen-ial for situational knowledge. In particular, annotating sensorata with spatial, temporal, and thematic semantic metadata.he SEIPF publishing framework exposes information in a pureDF along with the option to view ontologies to understand the

hole structure of information instead of annotating informa-

ion. The framework also provides a mechanism to authenticatend authorize third party applications, services or automatedgents, and provides the consumer information in a restricted man-er.

by posting power consumption information.

9. Conclusion and future works

This paper presents a Semantic Energy Information PublishingFramework (SEIPF) that can be installed on a residential gateway(Dog2.0) to publish power consumption information of differentappliances in a house environment. A new modular E.P. ontology tomodel the power consumption of different appliances in the houseis proposed. Modularity allows the E.P. ontology to be plugged inthe DogOnt ontology that models the domotic plant in a house.The SEIPF exposes the power consumed by different appliances andtheir properties in a pure RDF format. The goal of our approach isto make power consumption information machine understandableto support distributed applications such as intelligent negotiation.This will enable today third-party applications, services or agents(given Authorization) to access the power consumption of a houseor a building and help build standalone or data sharing applicationsto evolve a energy aware and energy efficient society.

We plan to extend the SEIPF by taking into account the timebehavior, enabling it to publish energy figures over time inter-vals, thus reducing requirements over the polling intervals of clientapplications. In the future, the SEIPF could also help us to buildsystems where energy consumption can be co-ordinated betweendifferent consumers. The semantic nature of the exposed data willhelp building applications where automated intelligent negotia-tion and consumption coordination can take place, evolving in toan intelligent energy grids.

References

[1] International Energy Agency, Energy Consumption by Sector, Technical Report,International Energy Agency, 2005.

[2] P. Lalanda, J. Bourcier, Towards autonomic residential gateways, in: IEEE Inter-national Conference on Pervasive Services (ICPS 2006), IEEE, 2006, pp. 329–332.

[3] C. Escoffier, J. Bourcier, P. Lalanda, J. Yu, Towards a home application server,in: IEEE Consumer Communications and Networking Conference, 2008, pp.321–325.

[4] S. Chung, W. Chen, MyHome: a residential server for Smart Homes, in:Knowledge-Based Intelligent Information and Engineering Systems, Springer,2010, pp. 664–670.

[5] J. Bourcier, A. Chazalet, M. Desertot, C. Escoffier, C. Marin, A dynamic-SOA homecontrol gateway, in: IEEE International Conference on Services Computing,2006. SCC’06, 2006, pp. 463–470.

[6] D. Bonino, F. Corno, DogOnt—ontology modeling for intelligent domotic envi-

ronments, in: A. Sheth, S. Staab, M. Dean, M. Paolucci, D. Maynard, T. Finin, K.Thirunarayan (Eds.), International Semantic Web Conference, Number 5318 inLNCS, Springer-Verlag, 2008, pp. 790–803.

[7] D. Bonino, E. Castellina, F. Corno, The DOG gateway: enabling ontology-basedintelligent domotic environments, IEEE Transactions on Consumer Electronics54 (4) (2008) 1656–1664.

1 Build

[

[

[

[[[[[

[

[

[

[

[

[

[

[

402 D. Bonino et al. / Energy and

[8] D. Cook, M. Youngblood, E. Heierman III, K. Gopalratnam, S. Rao, A. Litvin, F.Khawaja, MavHome: an agent-based smart home, in: Proceedings of the FirstIEEE International Conference on Pervasive Computing and Communications,2003 (PerCom 2003), 2003, pp. 521–524.

[9] IEA, World Energy Outlook 2009 Fact Sheet: Why is our Current Energy PathwayUnsustainable? Technical Report, International Energy Agency, 2009.

10] T. Berners-Lee, Keynote on web architecture, http://www.w3.org/DesignIssues/LinkedData.html, 2006.

11] S. Meyer, A. Rakotonirainy, A survey of research on context-aware homes,in: ACSW Frontiers’03: Proceedings of the Australasian Information SecurityWorkshop Conference on ACSW Frontiers 2003, Australian Computer Society,Inc., Darlinghurst, Australia, 2003, pp. 159–168.

12] G. Demiris, M.J. Rantz, M.A. Aud, K.D. Marek, H.W. Tyrer, M. Skubic, A.A. Hussam,Older adults’ attitudes towards and perceptions of ‘smart home’ technologies:a pilot study, Informatics for Health and Social Care 29 (2004) 87–94.

13] I. Herman, Semantic Web Activity Statement, Technical Report, W3C, 2001.14] W3C, RDF Primer, Technical Report, W3C, 2004.15] T. Gruber, What is an ontology?, 1992.16] W3C, OWL: Ontology Web Language, Technical Report, W3C, 2004.

17] F. Razzak, D. Bonino, F. Corno, Mobile interaction with smart environments

through linked data, in: IEEE International Conference on Systems, Man, andCybernetics, IEEE, 2010, pp. 2922–2929.

18] C.-C. Pan, P. Mitra, P. Liu, Semantic access control for information interopera-tion, in: SACMAT’06: Proceedings of the Eleventh ACM Symposium on AccessControl Models and Technologies, ACM, New York, NY, USA, 2006, pp. 237–246.

[[

[

ings 43 (2011) 1392–1402

19] A. Toninelli, R. Montanari, L. Kagal, O. Lassila, A semantic context-awareaccess control framework for secure collaborations in pervasive comput-ing environments, in: The Semantic Web Conference—ISWC, 2006, pp. 473–486.

20] C.M. Ionita, S.L. Osborn, Specifying an access control model for ontologies forthe semantic web, in: Secure Data Management, 2005, pp. 73–85.

21] D. Recordon, D. Reed, OpenID 2.0: a platform for user-centric identity man-agement, in: Proceedings of the Second ACM Workshop on Digital IdentityManagement, 2006, pp. 11–16.

22] D. Bonino, F. Corno, DogSim: a state chart simulator for Domotic Environments,in: 8th IEEE International Conference on Pervasive Computing and Communi-cations Workshops (PERCOM Workshops), IEEE, 2010, pp. 208–213.

23] N. Fumo, P. Mago, R. Luck, Methodology to estimate building energy consump-tion using energyplus benchmark models, Energy and Buildings 42 (2010)2331–2337.

24] L. Pérez-Lombard, J. Ortiz, C. Pout, A review on buildings energy consumptioninformation, Energy and Buildings 40 (2008) 394–398.

25] J.E. Seem, Using intelligent data analysis to detect abnormal energy consump-tion in buildings, Energy and Buildings 39 (2007) 52–58.

26] Google, Google power meter, 2009.27] M. Weiss, F. Mattern, T. Graml, T. Staake, E. Fleisch, Handy feedback: connect-

ing smart meters with mobile phones, in: Proceedings of the 8th InternationalConference on Mobile and Ubiquitous Multimedia, ACM, 2009, pp. 1–4.

28] A. Sheth, C. Henson, S.S. Sahoo, Semantic sensor web, IEEE Internet Computing12 (2008) 78–83.