AI-ready flexible buildings Data requirements and feasibility of unlocking flexibility

Page 1


RACE for Business

Research Theme: Digitalising Industry BT1

ISBN: 978-1-922746-66-5

Industry Report

Data requirements and feasibility of unlocking flexibility in commercial buildings

Citations

Liebman, A., De Nijs, F., Labouchardiere, M., Xie, Y., Salim, M., Patankar, P., Lalbakhsh, P. (2024). AI-ready flexible buildings. Prepared for RACE for 2030 CRC.

October 2024

Projectteam

Universitypartner

•Prof.ArielLiebman†

•Dr.FritsdeNijs

•Dr.MarcelLabouchardiere

•Dr.YitingXie

•MariamSalim

•PratoshPatankar

•Dr.PooiaLalbakhsh

† DeceasedNovember2023

BuildingsAlive

•Dr.CraigRoussac

•Dr.BradSchultz

Project partners

Acknowledgements

Theresearchteamwouldliketothanktheindustryreferencegroupparticipantsfortheirfeedbackandinsights onthework.

WeacknowledgetheuseofgenerativeAItoolsChatGPTandGrammarlytoimprovethetoneandclarityofthis report.

AcknowledgementofCountry

TheauthorsofthisreportwouldliketorespectfullyacknowledgetheTraditionalOwnersoftheancestrallands throughoutAustraliaandtheirconnectiontoland,seaandcommunity.Werecognisetheircontinuingconnection totheland,waters,andcultureandpayourrespectstothem,theirculturesandtotheirElderspastandpresent.

WhatisRACEfor2030?

RACEfor2030CRCisa10-yearco-operativeresearchcentrewithAUD350millionofresourcestofundresearch towardsareliable,affordable,andcleanenergyfuture. www.racefor2030.com.au

Disclaimer

Theauthorshaveusedallduecareandskilltoensurethematerialisaccurateasatthedateofthisreport.The authorsdonotacceptanyresponsibilityforanylossthatmayarisebyanyonerelyinguponitscontents.

1

Introduction

BuildingsplayasignificantroleinenergyconsumptionandtheemissionofCO2 worldwide,accountingfor approximately40%oftheworld’senergyconsumptionandcontributingtoabout35%oftotalgreenhousegas emissions(Ürge-Vorsatzetal., 2015;Magnietal., 2021;Pouranian,Akbari,andHosseinalipour, 2021;X.Zhang etal., 2020;Pervez,Ali,andPetrillo, 2021).TheNetZeroroadmapbytheInternationalEnergyAgency(2021) suggeststhattoachieveglobalnet-zeroemissionsby2050,heatingandcoolingenergyuseshoulddropfrom 100%in2020to50%by2030andfinally20%in2050.Theroadmapexpectsthat efficiency istheprimarydriver oftherequiredreductioninbuildings’totalfinalconsumption,withananticipated350MtCO2 reductionby2050 achievedthroughdigitalisationandintelligentcontrols.Deployingsuchintelligentbuildingcontrolstoretrofit existingbuildingsisonepotentiallevertoquicklyreducinggreenhousegasemissions.

Heatingandcoolinginofficebuildingsaremajorcontributorstoenergyconsumption,accountingforasignificant percentageofbuildingenergyusageandCO2 emissions.Whilethisisasignificantoutlay,buildingsalsohavethe potentialtoabsorb,store,andreleasetheheatingorcoolingintothebuildingmaterials,muchthesamewaya batterydoes.Bypre-coolingorpreheatingthebuilding,“thermalinertia”will,intheory,providetheflexibilityto usemorevariablerenewableenergy,thusloweringthecarbonemissionsofthebuilding.However,inpractice, thiswillrequiregreaterautomationandresponsivenessofthebuildingstoclimate,renewableresources,and electricitygridprices.

Thisprojectaimstodefine,test,anddocumentthedataproducedbynewautomatedsystemstodeterminethe optimaltimestoheatandcoolabuilding.ThisdecisionprocesswillalsobeautomatedbycombiningAI-driven predictivemodellingwithstrategicdataorganisationandinnovativeoptimisationtechniquestosignificantly reduceenergyusageandcarbonemissionsinofficebuildingswhilemaintainingcomfortandproductivity.

Thecoreoutputoftheprojectisasimulationmodelofanofficebuilding(modelledonanexistingofficebuildingin MonashUniversity),forwhichweoptimiseanHVACcontrolscheduletomakethebestprecoolingandpreheating decisions.Optimisingthecontrolofthebuildingautomationandmanagementsystem(BAMS)toaligntheenergy consumptionwithperiodsoflowcarbonintensitygridpower,weareabletoreducethecarbonintensityofthe building’soverallenergyrequirementsby10%.ExtrapolatingthisresulttotheentireMonashcampussuggestsa potentialannualreductionofupto23,000tonsofCO2-eifthissolutionisfullydeployedacrossallbuildings, althoughfurtherreal-worldstudiesareneededtovalidatethesimulationestimates.

Theproject’slong-termbenefitsincludegreateraccessibilityofthebuildingcontroldataforR&D,developing advancedcontrolmodelsforenergyoptimisation,significantlyreducingcarbonintensityinbuildingoperations, andsettingaprecedentforfutureinnovationsinsustainablebuildingmanagement.

Literaturereviewonbuildingautomation

Thischapterprovidesacomprehensiveintroductiontothefoundationalconceptsofbuildingautomation, emphasisingthegrowingsignificanceofdata-drivenstrategiesinbuildingenergymanagementacrossboththe designandoperationalphases.Thesesystemscanbeusedforarangeofautomationandmanagementtaskssuch asloadforecasting(M.Huetal., 2023),anomalyandfaultdetectionanddiagnosis(Leietal., 2023),predictive controlandthermalcomfort(G.HuandYou, 2023),indoorenvironmentalquality(IEQ)monitoring(Majdietal., 2022),facilityandassetmanagement(OgollaandKieti, 2022),energyefficiencyplanning(Verma,Prakash,and Kumar, 2023),securityandsafety(HsiaoandHsieh, 2023),occupancydetection(Sayed,Himeur,andBensaali, 2022),andwaterusagemanagement(AlGhamdiandSharma, 2022).

Whiletraditionalapproachesrelyonphysicalmodellingtoaccountforfactorslikeshading,passiveheating, orientation,naturalventilation,andgeometry,thedata-drivenapproachharnessesempiricaldatatoextract valuableinsights.Byanalysingdatarelatedtoheating/coolingload,lighting,electricityusage,andoccupancy profiles,amoreaccurateunderstandingofbuildingperformancecanbeachieved(PanandL.Zhang, 2020; Che-AniandRaman, 2019).Collectingandmaintainingreal-worldbuildingenergyusedata(RBEUD)thatrepresent theactualconditionofbuildingsenhancesoperationalefficiencyandcontributestovaluableknowledgeforfuture projects.Adoptingmoderndistributedcomputingplatforms,suchascloudcomputing(Mohamed,LazarovaMolnar,andAl-Jaroodi, 2016),edgecomputing(Minhetal., 2022),fogcomputing(Iftikharetal., 2022),and hybridstrategies(Himeuretal., 2022),cansimplifyandsupportthistaskbyprovidingseamlessconnectivity,data pre-processingandstorage,andpowerfulcomputingresourcesatvariouslevels.Theavailabilityofsuchdatasets playsapivotalroleininformedpolicy-makingandbridgingtheenergyperformancegap(EPG),whichrepresents thedisparitybetweenmandatedenergyconsumptionandactualusageinpractice(X.Wangetal., 2023).

2.1 TraditionalHVACControl

TraditionalHVACcontrolmethodsfocusonmaintainingindoorenvironmentalcomfortbyregulatingtemperature, humidity,andaircirculationusingsetpointandfeedbackcontrolmechanisms.Theseconventionalstrategies includemanualadjustments,whereoccupantsadjustsettingsbasedoncomfortlevels,andautomatedsystems, whichrelyonpredefinedsetpointsfortemperatureandhumiditycontrolandoperateindependentlytomaintain aspecifiedcomfortrange.

Significantenergysavingsandpreservationofhumancomfortcanbeachievedbyoptimisingthesecontrol strategies,includingtheuseofoptimalcontrolstrategycurvesandoperationcurvesforequipmentanddevices. Moreover,theadventofintelligentcontroltechniquesanduser-adaptablecomfortcontrolsystemssignifiesa shifttowardsmoreresponsiveandenergy-efficientHVACoperation,aimingtoalignmorecloselywiththespecific thermalcomfortpreferencesofoccupants(Federspiel, 1992;Mirinejad,Welch,andSpicer, 2012).

Historically,automation,controlsystems,andmaterialscienceadvancementshavebeenthefocalpointsfor improvingHVACefficiency(Aftabetal., 2017).However,therecentintegrationofAIandmachinelearning technologiesandaglobalshifttowardsrenewableenergysourceshavesignificantlyenhancedHVACoperations. ThistechnologicalevolutionreducestheenvironmentalfootprintofHVACsystemsandrevolutionisescontrol strategiesandmethodologies. 4

2.2 AIandMachineLearningforHVACControlStrategiesandAlgorithms

RecentstudiesunderscoretheeffectivenessofleveragingpredictivecontrolalgorithmsanddeeplearningapproachesforoptimisingenergyconsumptionandenhancingthermalcomfortwithinbuildingsthroughHVAC systems(Kalaimani,Keshav,andRosenberg, 2016).Furthermore,thedevelopmentofintelligentbuildingautomationsystems,capableofdynamicallysensingoccupantcomfortneedsandaccordinglyschedulingHVAC operations,representsasignificantadvancementtowardsmoreefficientandresponsivebuildingenvironments (Aftabetal., 2017).

Buildings,particularlythroughtheirHVACsystems,canenhancegridstabilitybyadjustingtheirelectricloadsin responsetosignalsfromthegrid.ThisadjustmentcontributestomaintainingtheNetworkedEnergyManagement (NEM)frequencybalanceandoverallsystemstability.Byimplementingcontrolstrategies,HVACsystemsin bothcommercialandresidentialsettingscanofferfrequencyregulationserviceswithoutsignificantlyimpacting indoorclimatecomfort.Researchhasexploredmodelpredictivecontrolandrule-basedcontrolstrategiesthat enableHVACoperationstobedynamicallyadjusted,takingintoaccountfactorslikeoccupantcomfortand hardwareconstraints.Suchstrategiesallowforamoreresponsiveandenergy-efficientoperationofHVAC systems,providingvaluableancillaryservicestothegrid,suchasdemandpeakreductionandbalancingsupply anddemand,therebysupportingastableandefficientenergysystem.

2.3 AI-DrivenHVACControl—Buildingsthatactasbatteriesoflastresort

Theintegrationofrenewableenergysourcesintotheelectricitygridhasledtotheneedfordemandresponse strategiestomitigatethevolatilityofrenewableenergysupply.HVACsystems,particularlyincommercialand industrialsettings,havebeenidentifiedaspotentialcandidatesforfast-demandresponseapplications(Goddard, Klose,andBackhaus, 2014).Thepotentialofdemandresponseinenablingflexibilityforhigherrenewable energypenetrationandefficientresourceexploitationhasbeenhighlighted,particularlyinindustrialnear-zeroenergybuildings(Kampelisetal., 2019).Additionally,researchhasfocusedonintegratingHVACoperationinto demandresponseprograms,emphasisingtheneedforoccupant-orienteddemandresponsewithroom-individual buildingcontrol(Frahmetal., 2023).Furthermore,anintegratedapproachtoadaptivecontrolandsupervisory optimisationofHVACcontrolsystemsfordemandresponseapplicationshasbeenproposedasaneffective solutionforascalableandadaptabledemandresponseplatformforHVACsystems(Adegbenro,Short,and Angione, 2021).

Inthecontextofmicrogrids,cooperativealgorithmsutilisingHVACdemandresponsehavebeenexploredto minimisesupply-demandmismatch,therebyreducingtheneedforenergystoragedevices(J.Ma,X.Ma,and Ilic, 2019).Moreover,controlstrategiesforcoordinatingtheoperatingschedulesofmultipleHVACdevicesin residentialdemandresponsehavebeeninvestigated,emphasisingtheinseparabilityofeffectivecontrolalgorithms fromresidentialdemandresponsebasedonHVACsystems(Kouetal., 2021).Thesestudiescollectivelyunderscore thepotentialofHVACdemandresponsetoaddressthechallengesposedbyintegratingrenewableenergysources intotheelectricitygrid.

Inresponsetogridshortagesoroutages,buildingscancollectivelyactas“batteries”toprovideloadshedding orself-curtailedservicesbyadjustingtheset-pointsofHVACsystemswithinreasonableranges.Thisadaptive managementofHVACsystemsenablesbuildingstoreducetheirelectricloaddynamically,contributingtogrid stabilityduringpeakdemandtimesorinsituationswherethegrid’ssupplycapacityisstrained.Conversely, whenthegridexperiencessurplusenergyavailabilityduetofavorableconditionsforphotovoltaic(PV)orwind generation,buildingscanincreasetheirHVACconsumptionflexibly.Bydoingso,buildingscanabsorbexcess

energy,preventingpotentialwasteofrenewableresourcesandaidinginthebalancingofsupplyanddemandon thegrid.

Forinstance,researchhasproposedtheconceptofaVirtualBattery(VB)controlforcommercialHVACsystems toadjustpowerconsumptioninreal-timebyregulatingzonalairflowrates,demonstratinghowbuildingscan effectivelyrespondtogridsignals.Additionally,theintegrationofHVACandbatteryschedulinginbuildings hasbeenexploredtooptimisedemandresponse,showcasingthepotentialforbuildingstomanagepeakload demandswhileensuringthermalcomfortandleveragingbatterystorage.

Theseapproacheshighlightthepivotalrolebuildingscanplayinenhancinggridresilienceandstability.By leveragingHVACsystems’flexibilityalongsideenergystoragesolutions,buildingscaneffectivelyserveasdynamic assetsinthesmartgrid,adjustingtheirenergyprofilestosupportgridoperationswhilemaintainingoccupant comfort.

2.4 BuildingAutomationandManagementSystems

Abuildingautomationandmanagementsystem(BAMS)isasophisticatedinstallationwithinbuildingsthatplays acrucialroleinoverseeingandregulatingadiverserangeofbuildingservices.Theseservicesincludeheating, cooling,ventilation,airconditioning,lighting,shading,lifesafety,alarmsecuritysystems,andmore.Atitscore,a BAMSaimstobringautomationtotechnologically-enabledenvironmentsbyharmonisingamultitudeofelectrical andmechanicaldevices.Thesedevicesareintricatelyconnectedthroughdistributedcontrolnetworks,forming aninterconnectedecosystem.BAMSsfindwideapplicationacrossvarioussettings,rangingfromindustrial facilitiesandcommercialbuildingstobustlingmallsandevenresidentialproperties(Dominguesetal., 2016). Whilebuildingautomationandmanagementsystemstheoreticallyhavethepotentialtoprovidecomprehensive componentsandfunctionalitiesforanalysingandoperatingbuildings,thereareadditionalvitaltasksthatoften fallundertheresponsibilityoftheoperator.Inotherwords,energymanagementsystemscanbeconsidered asreactivesystems,therefore,itisnecessarytodevelopsmarterstrategiesthatarefullyfocusedonenergy improvement(Lee,Cha,andPark, 2016).Thesetasksandstrategiesincludeevaluatingbuildingperformance, detectingunusualenergyconsumptionpatterns,identifyingefficiencyimprovements,andensuringthesecurity andprivacyofend-users(Himeuretal., 2022).

Sincetheearly2000s,theapplicationofvirtualisationindatacentreshasbroughtaboutaparadigmshiftindigital systems,deliveringremarkableimprovementsinflexibility,availability,andsecurity.BAMSshaveembracedthis transformativetrendbyseamlesslyintegratingwirelesstechnologieslikeWiFi.Thisintegrationhasempowered BAMSswiththeabilitytoenableremoteaccessandmonitoring,unlockingnewopportunitiesforefficient andeffectivemanagementofbuildingsystems(S.Wang, 2009).Furthermore,thisintegrationhaslaidasolid foundationfortheimplementationofcutting-edgealgorithmsandtechnologiesthatsupporthigher-leveltasks. Thesetasksrequiretheutilisationofadvancedtoolssuchasbigdataanalyticspipelines,capableofprocessingvast amountsofinterconnectedequipmentdata.Additionally,artificialintelligenceandmachinelearningalgorithms playacrucialroleintaskssuchasloadforecasting,watermanagement,indoorenvironmentalqualitymonitoring, occupancydetection,andenergyanomalydetection.Byleveragingtheseinnovativeapproaches,building automationsystemscollaboratetoenhancetheoverallperformanceofthebuilding,ensuringoptimalefficiency, occupantcomfort,andsustainability(Debrah,Chan,andDarko, 2022).

ThesignificanceofBAMSsliesintheirabilitytoensureoccupantsatisfaction,reduceenergyconsumption,and streamlineefficientbuildingoperations.Toachievethis,BAMSsprovidecomprehensiveawarenesstorelevant managers,whichissupportedbythecollectionofhigh-resolutiondata.Thisdataisacquiredthroughadiverse

arrayofsensorsthatarestrategicallydeployedandconnectedtotheBAMSusingdifferentcommunication technologies,suchasfieldbusandIoTstandards(Pandya, 2021;Moudgiletal., 2023;Zaerietal., 2022).By harnessingthesetechnologicaladvancements,BAMSsempowerbuildingmanagerstomakeinformeddecisions andoptimisebuildingperformanceinpursuitofenergyefficiency,occupantcomfort,andsustainableoperation.

Thefieldofbuildingautomationhaslongbeendominatedbyawidearrayofproprietarysolutions,withad-hoc approachesoftenmeetingmoderateperformancerequirements.However,drivenbythegrowingmarketdemand foropensystems,evenindustryleadersaregraduallyshiftingawayfromproprietarydesigns.Recognisingthe importanceofopensystems,officialstandardsbodiesareactivelyinvolvedinensuringthatthestandardsthey developandpublishadheretotheprinciplesofopenness.Theseprinciplesincludenon-discriminatoryaccess tospecificationsandlicensing.Consequently,theadherenceofequipmenttoformalstandardsisincreasingly becomingarequirementinmanyprocurementprocesses.Standardsdirectlyrelevanttobuildingautomation systemtechnologyaredevelopedbothintheUnitedStatesandbyseveralEuropeanandinternationalstandardsbodiessuchasISO1,CEN2,andCENELEC3.Thesestandardsserveasguidelinesandbenchmarksforthe industry,promotinginteroperability,compatibility,andtheseamlessintegrationofdifferentbuildingautomation components(Kastneretal., 2005).

Kastneretal.(2005)presentageneralsystemmodelthatencapsulatesmosttypesofBAMS,byseparatingout thefunctionalityofitsindividualcomponentsacrossthreelevels:

• FieldLevel:Thefieldlevelcoverstheswitches,sensorsandactuatorsthattheBAMSisultimatelyinteracting withtoachievethedesiredreal-worldoutcomesinthebuilding.Throughthefieldlevel,theBAMSinterfaces withthephysicalworldinadistributedmanner,collectingandtransformingmeasurement,counting,and meteringdataintoaformatsuitablefortransmissionandprocessing.Additionally,thislevelfacilitates physicalcontroloverenvironmentalparametersthroughactionssuchasswitching,setting,andpositioning inresponsetosystemcommands.

• AutomationLevel:Attheautomationlevel,awiderangeofautonomouslyexecutedsequencescomeinto play.Thisleveloperatesonthedatapreparedbythefieldlevel,establishinglogicalconnectionsandcontrol loopstofacilitateefficientoperations.Theautomationlevelservesasacrucialintermediarybetweenthe fieldlevelandthemanagementlevel,providingthenecessaryintelligenceanddecision-makingcapabilities forthesmoothandefficientoperationoftheoverallsystem.

• ManagementLevel:Atthemanagementlevel,comprehensiveaccesstoinformationfromtheentiresystem ismadeavailable.Thislevelpresentsaunifiedinterfacetotheoperator,facilitatingmanualintervention whennecessary.Itprovidesverticalaccesstoautomation-levelvalues,allowingforthemodificationof parameterssuchasschedulestoadapttochangingrequirements.Furthermore,themanagementlevel playsacrucialroleinmonitoringsystemhealthandgeneratingalertsforexceptionalsituationsliketechnical faultsorcriticalconditions.Italsoencompasseslong-termhistoricaldatastorage,enablingthegeneration ofreportsandstatisticsforin-depthanalysisanddecision-making.

ItisimportanttohighlightthatdeviceswithinBAMSoftenincorporateacombinationoffunctionalitiesfrom allthreelevels.Thisamalgamationofservicesandrequirementsnecessitatesnetworkarchitecturesthatcan accommodatethisdiversity.Strictlyadheringtoathree-tiernetworkstructurewouldintroduceunnecessary

1InternationalStandardsOrganization

2EuropeanCommitteeforStandardization

3EuropeanCommitteeforElectrotechnicalStandardisation

complexitywhenitcomestosharingdevices,especiallysensors,betweendifferentfunctionaldomains(Kastner etal., 2005).

HerearethemaincomponentscommonlyfoundinaBAMS:

• Sensors:Thedevicesresponsibleforcapturingdataonvariousaspectsofthebuildingenvironment,such astemperature,occupancy,lightlevels,andairquality.SensorsprovideinputstotheBAMS,enabling effectivemonitoringandcontrol.Leveragingadvancedsensingandmeteringtechnologies,datacanbe gatheredfrommultiplemodalities,creatingacomprehensiveinformationsourcefortargetedanalysisand thedevelopmentofintelligentandsophisticatedtools.Measurementdevicesincreasetheobservabilityof thebuilding’stransientanddynamiceventsaswellasgatheractualdatarelatedtodiversefunctionalities ofthebuilding(Himeuretal., 2022).Sensorsareeitherdirectlyconnectedtocontrollersviaastandard interfaceorbymeansofafieldnetwork(Kastneretal., 2005).

• Controllers:Controllersareresponsibleforprocessingthedatareceivedfromsensorsandinitiating appropriateactions.Theymakedecisionsbasedonpredefinedalgorithmsandcontrolstrategiesto regulatebuildingsystems.ThecontrolunitisthecentralbrainoftheBAMSgenerallybuiltuponindustry standardsandprotocolssuchasBACnet4,Koinnex(KNX)5,LonWorks6,andModbus(Pricopetal., 2017).

• Actuators:Actuatorsrespondtotheoutputsignalsfromacontrollerandaccomplishesactionstooperate thefinalcontroldevice,whichmightbeavalve,damper,orswitch(S.Wang, 2009).Actuatorscontrol functionssuchasadjustingtemperature,operatingvalves,turningon/offlights,andmanagingsecurity systems.Similartosensors,actuatorsareeitherdirectlyconnectedtocontrollersorthroughafield network(Kastneretal., 2005).

• Human-MachineInterface(HMI):TheHMIistheuserinterfacethroughwhichbuildingoperatorsor occupantsinteractwiththeBAMS.Itcanbeagraphicalinterfacedisplayedonacomputer,touchpanels,or mobiledevices,allowinguserstomonitorsystemstatus,adjustsettings,andreceivealarmsornotifications. Generally,theHMIisdevelopedtomonitortheenergymanagementsystemregardingenergyconsumption (Hmidahetal., 2022).BAMS’SHMIisbuiltuponsupervisorycontrolanddataacquisitionplatformsto provideaunifiedvisualisationforallsystemsandsubsystemswhichfacilitatesthetaskoftheoperator (FigueiredoanddaCosta, 2012;Kastneretal., 2005).

• CommunicationInfrastructure:BAMScomponentsareinterconnectedthroughacommunicationnetwork, allowingdataexchangeandcoordination.Thisinfrastructuremayincludewiredorwirelessprotocols,such asEthernet,BACnet,LonWorks,Modbus,orotherindustry-specificprotocols.

• SupervisoryControlandDataAcquisition(SCADA):SCADAsystemsoffercentralisedmonitoringand controlcapabilitiesforlarge-scaleBAMSdeployments.Thesesystemsempoweroperatorsbyeliminating theneedtohandleeachpieceofequipmentlocallywithinabuildingorcomplex.Instead,theyenableremote monitoringandcontrol,allowingforthedetectionofabnormalconditionswithouttherequirementofbeing physicallyon-site.SCADAsystemsgatherdatafromvariouscontrollersandsensors,providingadvanced analytics,datafusion,andreportingfunctionalitiestoenhanceoperationalefficiencyanddecision-making (NesaandBanerjee, 2017;Kastneretal., 2005).

4Anobject-orientedpeer-to-peernetworkprotocol

5KNX,isapopularopenstandardforhomeandbuildingautomation.Twistedpair(TP),powerline(PL),radiofrequency(RF),and Ethernet(IP)areamongthein-housecommunicationpossibilitiescoveredbyKNX(Shikhlietal., 2022)

6Astandardisedbussystemusedincentralisedanddecentralisedbuildingautomationcontrol(Merz,Hansemann,andHübner, n.d.)

• EnergyManagementSystem(EMS):AnEMScomponentwithintheBAMSfocusesonoptimisingenergy consumptionandefficiency.Itanalysesenergydata,identifiesareasforimprovement,andimplements strategiestoreduceenergyusageandcosts.

• IntegrationGateways:Integrationgatewaysplayacrucialroleinensuringseamlessinteroperabilitybetween variousbuildingsystemsandprotocols.Thesegatewaysserveascommunicationbridges,enablingthe exchangeofdataandinformationbetweentheBAMSandsubsystemslikeHVAC,lighting,security,lifesafety, andaccesscontrol.Byadoptingthegatewayapproach,controlapplicationsondifferentnetworkscan utilisetheirnativeprotocolstocommunicatewithoneanother,whilethegatewaytakescareofestablishing thesemanticconnection.Thisapproachallowsfortheabstractionandconcealmentofthecomplexities associatedwithspecificprotocols,astheyarehandledbehindthescenesbythegateway.Asaresult, integrationgatewaysfacilitatesmoothandefficientcommunicationbetweendiversebuildingsystems, promotinginteroperabilityandstreamlinedfunctionality(Kastneretal., 2005).

• DatabasesandDataStorage:BAMSsystemsoftenincludedatabasestostorehistoricaldata,eventlogs,and configurationinformation.Thesedatabasessupportreporting,trendanalysis,andlong-termperformance monitoring.

• AlarmManagementandNotification:TheBAMSincludesmechanismstodetectandrespondtoabnormal conditionsorfaults.Itcangeneratealarmsandnotifications,alertingbuildingoperatorsormaintenance personnelofpotentialissuesrequiringattention.

Methodology

Itmaybepossibletouseabuilding’sthermalmasstoachievedemandflexibility,ifitisknowninadvancewhen theflexibilityisrequired.Themethodologyweusetorealisethisflexibilityisacombinationofmachinelearning andoptimisationtechnology,toultimatelyderiveacontrolset-pointschedulethatreduces(orincreases,as required)theenergyconsumptionofthebuildingduringtheflexibilitywindow.

Figure 3.1 showsthehigh-leveloverviewofthecomponentsthatmakeupthe(ideal)methodology.Inthedata collectionphase,sensorreadingsarecollectedfromthebuilding’ssensorsandstoredinalong-termstorage database.Subsequently,themachinelearningphaseconsistsoffittingapredictivemodeltothereadingsinthe database.Thepredictivemodeltakesthetheexternal(uncontrollable)conditionstogetherwiththecontrol signalasinputfeatures,andpredictsanexttemperature(oratemperaturetrajectory).Thispredictivemodel isusedintheoperationalphasebyanoptimisationmodelthattakesasinputthecurrentstateoftheexternal conditionstogetherwiththepredictivemodel,andproducescontrolset-points(oracontrolset-pointschedule) tobedeployedonthebuilding.

Datacollection phase

Machinelearning phase

Operational phase

Timestamp,T_in,T_out,T_set,T_fcast,kW 2023-08-01,20.3,8.2,21.0,[8.5,9.0,…],2.0 2023-08-02,20.5,8.6,21.0,[9.0,9.2…],0.0

<now>,21.5,11.0,21.0,[…],2.5

Machinelearning

Surrogatemodel ofBuildingHVAC

<now>,21.5,11.0,21.0,[…],2.5

Database Evaluation

ComparisonoftCO2,$, kWh,DRcompliance

Using

BAMSControl BuildingHVAC

Forecastsofweather, occupancy;set-point schedule

Sensors Act

Actualweather, occupancy

Optimise

Optimisation

Forecastsof CO2,wholesale, weather,occupancy; set-pointschedule,DR

Sensors Act Act

BAMSControl BuildingHVAC

Sensors

Actualweather,occupancy, CO2,wholesale

Figure3.1: Idealisedmachine-learningandoptimisationpipelinetotransitionfroma‘normal’buildingtoa‘flexible’building. AnexistingbuildingwithBAMSisinstrumentedtocollectoperationaldata(datacollectionphase).Asurrogatemodelisfitto thisdatatomatchthebuilding’sthermalresponseunderdifferentweatherandoccupancymodes(machinelearningphase). Intheoperationalphase,thereal-timeBAMSset-pointscheduleisoptimisedusingthesurrogatemodelandadditional forecastinformationsuchasthecarbonintensityandwholesaleelectricityprice(blueinputs).

Thismethodologyhasbeenshowntoworkusingsimulatedbuildingmodels(Lametal., 2020),demonstrating optimalflexibleloadresponsescalculatedforthecontrolofupto20buildings.However,thepracticalapplication ofthismethodologyintherealworldfacessignificanthurdles.Crucially,thequalityoftheoptimisedcontrol scheduledependsinlargepartontheaccuracyandcompletenessoftheinputdatausedtofitthebuildingmodel. However,traditionallybuildingoperatorshavehadlittleneedtokeepaccuratehistoricsensorlogsacrosslong horizons,resultinginlongleadtimesnecessarytocollectenoughreal-worlddatatomaketheapplicationfeasible.

Becauseofthelargecostinvolvedinmaintainingahigh-accuracydatacollectionframework(includingthe maintenanceandmonitoringofsensorsfordrift),animportantquestiontoanswerishowmuchdataisneededto makethismethodologywork.Thisresearchaimstoanswerthisquestionviaasimulationapproach:by simulating thedatacollectionphaseatdifferentlevelsofaccuracyandgranularity,wecanobtaina(best-case)estimationof theamountofdatanecessarytoreconstructa(sufficiently)accuratepredictivemodel.

3.1 Buildingsimulationsoftwareanddata

3.1.1 EnergyPlus

EnergyPlusistheofficialbuildingenergysimulationprogramfromtheU.S.DepartmentofEnergy.Itmodels heating,cooling,lighting,ventilation,waterusage,andotherenergyflowswithinbuildings.Akeystrengthisits abilitytoperformintegratedsub-hourlysimulationsofinterconnectedheattransferpathsacrosstheentire building.EnergyPlusaccuratelyaccountsforwindowgains,shading,daylighting,renewables,andconstruction materials.Itsdetailedmodellingallowstheoptimisationofenergy-efficientbuildingandHVACdesignswhile consideringwateruseandutilitycosts.Therobustprogramiswidelyusedbytheglobalbuildingenergyanalysis community(Fumo,Mago,andLuck, 2010).

ThemethodologyproposedinthisstudyispredicatedondataderivedfromEnergyPlussimulationsofHVAC controlandbuildingthermalresponse,withaparticularemphasisonthesetpointsrepresentingthethermal zonesofthetargetbuilding.Incontrasttocollectingreal-timebuildingcontroldatafromBAMS,EnergyPlus simulationmodelsofferanefficientmeansofstreamliningthedatacollectionandinterpretationprocesses.For intricatecommercialbuildings,thesubstantialvolumeandcomplexinterconnectionsofHVACsystems,electrical equipment,andothercomponentsnecessitateextensivedatasciencecapabilitiesforcleaning,interpretation,and analysis.EnergyPlusenergysimulationmodelsconsolidatethedataintoaunifiedformatatacontrollablescalefor subsequentdeep-learningmodeldevelopmentwhilecomparativelyreflectingthesalientphysicalpropertiesofthe buildingunderinvestigation,includingsize,internalandexternalmaterials,andbasicHVACsystemconfiguration.

Nonetheless,asatrade-offbetweenresourceallocation,time,andaccuracy,thisstudyconsiderstheintegration ofEnergyPlussimulationmodelsforAI-drivencontrolasareasonablemethodologyforachievingflexiblebuilding goals.

3.1.2 OpenStudio

OpenStudioisacross-platformsuiteofsoftwaretoolsdevelopedbytheNationalRenewableEnergyLaboratory (NREL)tofacilitatewhole-buildingenergymodellingusingtheEnergyPlussimulationengine.Akeycomponent istheOpenStudioApplicationSuite,whichprovidesauserinterfaceandgeometryeditorforcreating,editing, running,andvisualisingEnergyPlusmodels.

GiventheinherentcomplexityofconstructingathermalmodeldirectlywithintheEnergyPlusenvironment,this studyutilisedOpenStudiotostreamlinetheprocessofbuildingthefoundational3DmodelviaSketchUpSoftware andconfiguringtheHVACsystemforsubsequentEnergyPlussimulations.ByleveragingOpenStudio’sintegrated

toolset,thedifficultiesassociatedwithestablishinganEnergyPlusmodelfromtheoutsetweremitigated,enabling amoreefficientworkflow.

Ultimately,OpenStudioservesasacomprehensivesoftwareframeworkdesignedtofacilitateandexpeditebuilding energymodellingworkflowsbyprovidingauser-friendlyinterfaceanddevelopmentenvironmentforharnessing therobustsimulationcapabilitiesofEnergyPlus.Asanopen-sourcetool,itbenefitsfromanactivedeveloper communitycontributingtoplugins,enhancements,andtechnicalsupport.

3.1.3 Climatedata

Inordertosimulatearangeofdifferentpotentialweatherconditions,EnergyPlusneedsaccesstoaweatherdata filethatrepresentsthetypicalmeteorologicalyear.WemakeuseoftheexistingdatasetofEnergyPlusweather datafilesforAustraliaproducedbyCSIRO(Ren,Tang,andJames, 2021).Thisdatasetisbasedonhistorical weatherdatadrawnfromtheyears1990to2015,containinghourlyobservationsforanentireyear.Sinceweare modellingabuildinglocatedinClayton,Victoria,theclosestdatafileis‘MelbourneRO’,whichistheonethatwe usedinoursimulations.

Amajorchallengeinperformingvalidatedstudiesintooptimisedbuildingcontrolisthefactthatvirtuallyno datasetforclimateandweatherdatacontainsforecastdatainadditiontothetruesignal.Withoutaccessto forecastdata,studiesmusteitherassumeperfectforecastsorattempttoproducereasonablyrealisticforecast errors.Unfortunately,duetothecomplexityinweatherprediction,successfullycreatingrealisticforecasterrors isunlikely.Becauseofthis,forourstudywewillalsorestrictourselvestoperfectforecasts,whichwillresultin over-estimationofthetruebestperformance,butrepresentsthebest possible outcome.

3.2 Simulatingamulti-purposeuniversitycampusbuilding

FortestingthemethodologyproposedinFigure 3.1 wedevelopedanEnergyPlusmodelofamulti-purposecampus buildinghousingbothofficeworkers,lecturehallsandbreakoutspacesforcollaborativework.Wepatternedour buildingmodelontherecentlydelivered‘WoodsideBuildingforTechnologyandDesign’onMonash’Clayton campus.Thisbuildinghasseveralpropertiesthatmakeitparticularlysuitableforthisstudy:ithasmultiple thermalzonesacrossseveralopenandconfinedspaces,ithousesbothresearchlabsandstudentlecturehalls, andithasahighthermalcapacitanceowingtoitsPassivhausdesign(Grimshaw, 2020).

3.2.1 EnergyPlusmodeldesignparametersandfloorplan

TheprojectemployedSketchupandOpenStudioforconstructingtheEnergyPlusmodel,utilisingthereal-world geometricdimensions(120x42x25metres)andfundamentalmaterialdata.CompliancewithAIRAHStandard 189.1-2009,theBuildingCodeofAustralia(NCCVolume1),AustralianStandard1668.2,andtheAIRAHTechnical Handbookwasmaintainedtoestablishparametersforboththeexteriorandinteriorbuildingmaterials,thetype ofbuilding,defaultelectricalequipment,occupancy,andpredefinedschedulesforeachthermalzone.

Forthisproject,threeEnergyPlussimulationmodelsweredeveloped,eachprogressivelyincreasingthenumber ofthermalzonestomoreaccuratelyrepresentthecomplexityofthethermaldynamicsandreal-timeHVAC controloperations.Takingthe35thermalzonesmodelasanexample,thebuildingissegmentedintofivefloors. Eachfloorispartitionedintoaprimaryspaceflankedbysixsmaller,interconnectedspaces,asshowninFigure 3.2

Whilethetruearchitecturalcomplexityofthereal-worldbuildingismoreintricate,thisprojectemploysthe building’stotalannualenergyconsumptiondatatoensuretheapproximationmatchesintermsofthesimulated energyconsumption,thenumberofthermalzones,HVACconfigurations,andbuildingmaterialconfiguration. Themodelthusservesasacloserepresentationoftherealbuilding’sthermalbehaviourandenergydynamics.

Figure3.2: MostgranularversionoftheEnergyPlussimulationmodel,showingthesubdivisionofeachfloorinto7zones. Onelargeopenspace(left)and6equally-sizedzones(right),ofwhichonezoneisentirelyintheinterior.

3.2.2 SimulatingtheHVACSystem

TheHVACsystemwithinthebuildingsimulationincorporatesacomprehensiveclimatecontrolsolutiontailored forbothefficiencyandcomfort.Thesystemfeaturespackagedrooftopairconditioners,designedforeaseof installationandmaintenance,providingreliableandcentralisedcoolingthroughoutthebuilding.Additionally,the simulationincludesavariableairvolume(VAV)systemwithreheatcapabilities,ensuringprecisetemperature controlacrossdifferentzonesandtheabilitytoadjustairflowbasedonoccupancyandthermaldemand,thereby reducingenergyconsumption.Servicehotwatersystemsareintegratedintothedesigntomeetthebuilding’s hotwaterneedsforvarioususessuchasrestroomsandkitchenfacilities.Thismulti-facetedapproachtoHVAC ensuresthatthebuildingmaintainsoptimalindoorairqualityandcomfortwhilestrivingforenergyefficiency.

TostreamlinethethermalsimulationprocesswithintheEnergyPlusmodelandplacegreateremphasisonthe applicationofdeeplearningandoptimisationmethodsforHVACcontrol,thisprojectconsolidatesallthebuilding’s thermalzonesintoaunifiedHVACsystemasdetailedinFigure 3.3

3.2.3 SimulatingtheOccupancySchedule

Anaccuraterepresentationofoccupancypatternsiscrucialforpredictingbuildingenergyconsumptionin EnergyPlussimulations.Occupancydirectlyimpactsheating/coolingloadsthroughheatgainsfromoccupantsas wellasequipmentandlightingusageschedulestiedtooccupancy.

Universitybuildingsoftenexhibithighlyvariableandintricateoccupancyschedules,necessitatingtheconsideration ofcyclicacademiccalendars,seasonalvariations,weekdayandweekenddifferences,andotherpertinentfactors.

ToaccuratelycapturethesecomplexitieswithintheEnergyPlusmodellingframework,thisstudyemployeda stochasticapproachtogenerateoccupancydataprofilesgroundedinexistingspatial-temporaloccupancypattern studies(Juetal., 2023).

Specifically,theoccupancymodellingprocessaccountedforthedynamicnatureofuniversityfacilitiesbyincorporatingthefollowingelements:cyclicacademiccalendarstoreflecthigheroccupancyduringinstructional periodsandloweroccupancyduringbreaks,seasonalfluctuationstocapturevariationsinoccupancypatterns

Figure3.3: TheEnergyPluscomponentsusedinthesimulatedbuilding’sHVACsystem.

acrossdifferenttimesoftheyear,anddistinctionsbetweenweekdayandweekendoccupancylevelstoaccount forthediverseactivitiesandschedulesinherenttouniversitycampuses.

Byintegratingthesefactors,thegeneratedoccupancydataaimedtoprovideacomprehensiveandrealistic representationofthespatio-temporaldynamicsgoverningoccupancywithinthecontextofauniversitysetting. ThisapproachfacilitatedamoreaccurateassessmentoftheenergyconsumptionprofilesandHVACsystem requirementswithintheEnergyPlussimulations,therebyenhancingthereliabilityandapplicabilityofthemodelling outcomes.

3.3 Optimisationalgorithmsforflexiblebuildingcontrol

Optimisationreferstotheprocessoffindinganassignmentofvaluestodecisionvariables x thatmaximises(or minimises)theobjectivefunction f (x),possiblysubjecttosomeconstraintsonthekindofvaluesthat x are allowedtotake.Informallystated,theoptimisationproblemforflexiblebuildingcontrolcanbestatedasfollows: minimise energyconsumptionduringflexibilityrequest subjectto occupantcomfortconstraints plantoperationalconstraints

Inthismodelsketch,occupantcomfortconstraintsrelatetomaintainingcomfortableindoortemperatureand humidityinoccupiedzonesofthebuilding,whileplantoperationalconstraintsrefertoanyrestrictionsthatmight applytothecontrolsignal(forexample,toavoidexcessivewearontheHVACplantduetofrequentswitching).

Therearemultiplewaystoimplementsuchanoptimisationtask;thechoiceofapproachdependsontheruntime andaccuracyrequirements.Exhaustivesearchmethodscanfindasolutionthatisprovengloballyoptimal,atthe

(a) Packagedrooftopairconditioner
(b) PackagedrooftopVAVwithreheat
(c) ServiceHotWater
(3.1)

costofimposingsomemandatorystructureonthekindofconstraintsthatcanbeencoded,whiletrial-based searchmethodsareopen-endedinthekindsoffunctionstheycanoptimiseover,atthecostofbeingunableto provideaccuracyestimatesoroptimalityguarantees.Inthiswork,wefocusontheexhaustivesearchmethods, becausewewanttomeasuretheaccuracyofthepredictionmodels,ameasurewhichwouldbeaffectedby findingsub-optimalsolutionsduetotheoptimisationmethod.

WeusetheMiniZinclanguageandcompilertorapidlyprototypetheconstraintoptimisationprograms.MiniZinc isdevelopedatMonashasadomain-specificlanguagethatcantranscribeanoptimisationprogramspecifiedata highlevelofabstractionintoalow-levelformulationthatcanbeinterpretedbyoneofthemanycommercial andopen-sourcesolversavailable.AlthoughMiniZinccancompiletodifferentsolverparadigms(i.e.,Constraint Programming,Mixed-IntegerProgramming,andSatisfiabilitysolvers),wewillexplicitlytargetMixed-Integer Programmingsolversbecausewewillneedtodealwithfloating-pointquantities(i.e.,temperature)whichare generallynothandledefficientlyintheothersolvertypes.

SeveralrecentworkshaveexploredtheinclusionofneuralnetworksintoMixed-IntegerProgramming(MIP) models,withAndersonetal.(2020)givingamathematicaltreatmentofanefficientformalismtocapturethe operationsofaReLU-basedneuralnetworkinaMIP.

4

Simulationexperiments

Computersimulationmodelsforheatingandcoolingloadshavebeenaroundforalongtime,withoneofthe earliestproposedbyMortensenandHaggerty(1988).Thissimplemodelmirrorsanelectriccircuitwithasingle resistance,capacitor,andacurrentsource.Figure 4.1 demonstratestheresponseofanexampleinstanceofsuch athermalmodel.Thethermalresponseisanexponentialdecayacrossthefullrangeoftemperature,however,it isapproximatelylinearinthethermalcomfortrangearoundtheusualset-point θset ≈ 21°C.Thiscomfortrange isannotatedbyadead-bandaroundtheset-pointrangingfrom θ∆ to θ∆+ .Temperaturesinsidethisdead-band rangeareassumedtobe‘sufficientlycomfortable’fortheoccupantofthezone(orotherwiseacceptableforthe thermaldemandsofthezone,incasethemodelisforadifferentkindofzonelikearefrigerator).

Figure4.1: Thermalresponsecurvesofthesimplestthermalsimulationmodelundercooling(‘off’)andheating(‘on’) operation,fordifferentexternalconditions.

Asimpleflip-flophysteresiscontrollerforsuchadead-bandsystemwillkeepthetemperaturewithintheindicated boundsatalltimes,whileallowingforsomemodicumofoperatorflexibility.Figure 4.2 (left)showsasimulationof anaggregateof200thermalloadswithsmallrandomvariationinthermalparameters,operatingunderhysteresis control.Thefirst10loadsareshowninlightgreydashedlines,demonstratinghowthedirectionoftemperature changeinvertsatthedead-bandedges.At t =400,theflexibilityofthearrayiscalledupon,bymanuallyoverriding allthehysteresiscontrollerstothe‘off’or‘cooling’setting,independentlyoftheircurrentmodeortemperature. Asaresult,theloaddemandoftheaggregateimmediatelydropsto0,however,itreboundsrelativelyquicklyas thedevicesreachthelowerdead-bandedge.

Bycontrast,Figure 4.2 (right)demonstratesanoptimalcoordinatedflexibilityresponse.Underthiscondition,the thermalzonesareoptimisedfor minimum energyusageatalltimes,butwithanadditionalconstrainttoensure thateverydeviceisatmaximum θ∆+ at t =400.Asaresultofthiscontroloperation,thedemandforpower rapidlyrampsuptomaximumatthetimeleadinguptothedemand-responseevent.However,oncetheevent hits,demandcanstayat0foroverhalfanhourwithoutbreachinganyofthedevices’comfortbounds.Notethat bothcontrolmechanismsguaranteethecomfortlimitsatalltimes.

Hypothesis1. Optimal(minimum)energyoperationofthermalzonesandoptimal(maximum)flexibilityoperationofthermalzonesare orthogonal concernsthatcan both betargetedatthesametime.

Figure4.2: Flexibilityresponseofanaggregateof200buildingsunderreactivecontrol(aone-timedeactivationofheatpumps),versusaplannedoptimaltrajectoryforthesametrajectory.

Hypothesis2. Controllingforoptimalflexibilityrequiresproactive,planningcontroloverafuturehorizon, involvingconcernsof forecasting andpredictive what-ifmodels

Thisoptimalcontrolmethodologycanequallybeappliedtominimisethecarbonintensityoftheenergyconsumed, ifaforecastofcarbonintensityisavailable.Figure 4.3 demonstratesthisprinciplewiththesameaggregateof thermalloads,butnowoptimisingthenumberofdeviceson(pt,i =1),timesthecarbonintensity Ct attime t, acrosstheentireplanninghorizon h

Suchanoptimisationmodelisreadilysolvableifthetransferfunctionofzone i, fi canbeencodedlinearly (e.g.,multi-linearregressionmodel,orneuralnetworkwith ReLU activationfunctions).Inthecaseofthese simplethermalmodelswhichcanbeencoded directly,optimisingsuchamodeltakeslessthanasecondon amoderndesktopPC,foracontrolhorizonseveralhoursout.AsTable 4.1 shows,comparedtooptimising foronlyenergyconsumption( h t=1 n i=1 pt,i),optimisingforcarbonintensityreducesthecarbonimpactby approximately 11%,atthecostof 2 4%morepower(forthisparticularinstance,onaper-unitbasisbecausethe powerconsumptionoftheheatersisnotdimensioned).

(4.1)

Proportion heating / Carbon intensity (blue)

Figure4.3: Optimalflexibilityresponseofanaggregateofthermalloadsagainstacarbonintensitysignal.

Table4.1: Effectofoptimisingforminimumgridcarbonemissionsversusminimumbuildingenergyconsumption.

Solutionmetric:

Objective:Totalcarbon(p.u.)Totalenergy(p.u.)

Minimisecarbon 2637 43840 6

Minimiseenergy 2966 53753 9

5

Discussion—ChallengesandOpportunities

Buildingshaveapotentialroletoplayinourfutureelectricitygrid,ifasufficientvolumeofHVACloadcan beconvertedintoflexibledemand.Thisprojectinvestigatedanddemonstratedthesoftwareanddatasideof themethodologyrequiredtoenableAI-readyflexiblebuildings.Nevertheless,manychallengesremainbefore widespreadreal-worldadoptionofthetechnologyispossible:

• Noteverybuildingissuitableforthismethodology.Onlybuildingswithsufficientthermalmassandsensor instrumentationaresuitablecandidates.

Recommendation: Buildingenergyefficiencycomesbeforeenergyflexibility.Existingbuildingstock shouldbeimprovedandupgraded,includingenhancingtheirsensorcoverage,andlong-termdatacollection shouldbeestablishedwherepossible.

• Applicationofthismethodologyrequiresadeliberatededicationofbuildingmanagerstoup-frontdata collection.However,itmayonlybecost-effectivetoapplythismethodologyonceabuildingisbeing refurbishedforenergyefficiency.Afterrefurbishing,itsthermaldynamicswillhavechanged,andhistoric datawillnotberelevantanymore.

Recommendation: Futureworkshouldlookintoalgorithmstosafelytransfersimulation-trainedthermal modelstoreal-worldbuildings,togetherwithrapidmodeladaptionmechanisms.

• Currentmethodologyassumesbespokethermalmodelsandoptimisationformulationswillbecreatedfor eachbuilding.Thesemustbealignedtotheavailablesensordataandavailableactuators.Wideadoption ofthismethodologycanonlybeachievedifitcanbesimplifiedintoaproduct.

Recommendation: Futureworkshouldlookintoautomaticmodelextractionfromschematics,for examplebymakinguseofbuildingontologytomapconceptsinBAMStothemethodology.

Byfittingamachinelearningmodelofabuildingthatcapturesenoughofthethermaldynamics,wecanapply optimisationalgorithmstoimprovetheloadprofileofabuildingtotargetacontrolsignalsuchasgridcarbon intensity.Wehavedemonstratedthismethodologyonahigh-accuracysimulationmodelrunninginEnergyPlus, provingtheconceptispossible.Withsufficientforecastinformationandhighthermalmodelaccuracy,thecarbon intensityofthebuilding’sheatingandcoolingrequirementsmaybereducedbyupto11%.However,real-world validationexperimentsarenecessarytovalidatethefindingsofthisstudyundermorerealisticconditionssuchas uncertaintyovertheforecastdataandsensordrift,tounderstandhowmuchoftheidealcarbonreductioncan beachievedinpractice.

References

Adegbenro,Akinkunmi,MichaelShort,andClaudioAngione(2021).“AnIntegratedApproachtoAdaptiveControl andSupervisoryOptimisationofHVACControlSystemsforDemandResponseApplications”.In: Energies 14.8. issn:1996-1073. doi: 10.3390/en14082078

Aftab,Muhammad,ChienChen,Chi-KinChau,andTalalRahwan(2017).“AutomaticHVACcontrolwithreal-time occupancyrecognitionandsimulation-guidedmodelpredictivecontrolinlow-costembeddedsystem”.In: EnergyandBuildings 154,pp.141–156. issn:0378-7788. doi: 10.1016/j.enbuild.2017.07.077

AlGhamdi,RayedandSunilKumarSharma(2022).“IoT-BasedSmartWaterManagementSystemsforResidential BuildingsinSaudiArabia”.In: Processes 10.11,p.2462.

Anderson,Ross,JoeyHuchette,WillMa,ChristianTjandraatmadja,andJuanPabloVielma(2020).“Strongmixedintegerprogrammingformulationsfortrainedneuralnetworks”.In: MathematicalProgramming 183,pp.3–39. doi: 10.1007/s10107-020-01474-5

Che-Ani,AdiIrfanandSudharshanNRaman(2019).“Thermalenvironmentaccuracyinvestigationofintegrated environmentalsolutions-virtualenvironment(IES-VE)softwarefordouble-storyhousesimulationinMalaysia”. In: JournalofEngineeringandAppliedSciences 14.11,pp.3659–3665.

Debrah,Caleb,AlbertPCChan,andAmosDarko(2022).“Artificialintelligenceingreenbuilding”.In: Automation inConstruction 137,p.104192.

Domingues,Pedro,PauloCarreira,RenatoVieira,andWolfgangKastner(2016).“Buildingautomationsystems: Conceptsandtechnologyreview”.In: ComputerStandards&Interfaces 45,pp.1–12.

Federspiel,CliffordC.(1992).“User-AdaptableandMinimum-PowerThermalComfortControl”.PhDthesis.

Figueiredo,JoãoandJoséSádaCosta(2012).“ASCADAsystemforenergymanagementinintelligentbuildings”. In: EnergyandBuildings 49,pp.85–98.

Frahm,Moritz,ThomasDengiz,PhilippZwickel,HeikoMaaß,JörgMatthes,andVeitHagenmeyer(2023).“Occupantorienteddemandresponsewithmulti-zonethermalbuildingcontrol”.In: AppliedEnergy 347,p.121454. issn: 0306-2619. doi: 10.1016/j.apenergy.2023.121454

Fumo,Nelson,PedroMago,andRogelioLuck(2010).“Methodologytoestimatebuildingenergyconsumption usingEnergyPlusBenchmarkModels”.In: EnergyandBuildings 42.12,pp.2331–2337. issn:0378-7788. doi: 10.1016/j.enbuild.2010.07.027

Goddard,Gary,JosephKlose,andScottBackhaus(2014).“ModelDevelopmentandIdentificationforFast DemandResponseinCommercialHVACSystems”.In: IEEETransactionsonSmartGrid 5.4,pp.2084–2092. doi: 10.1109/TSG.2014.2312430

Grimshaw(2020). MonashUniversityWoodsideBuildingforTechnologyandDesign:CaseStudy url: https: //grimshaw.global/sustainability/monash-case-study/ (visitedon10/14/2024).

Himeur,Yassine,MariamElnour,FodilFadli,NaderMeskin,IoanPetri,YacineRezgui,FaycalBensaali,andAbbes Amira(2022).“AI-bigdataanalyticsforbuildingautomationandmanagementsystems:asurvey,actualchallengesandfutureperspectives”.In: ArtificialIntelligenceReview,pp.1–93.

Hmidah,NawalAbdunasseer,NuzulAzamHaron,AidiHizamiAlias,TeikHuaLaw,AbubakerBasheerAbdalwhab Altohami,andRajaAhmadAzmeerRajaAhmadEffendi(2022).“TheRoleoftheInterfaceandInterface ManagementintheOptimizationofBIMMulti-ModelApplications:AReview”.In: Sustainability 14.3,p.1869.

Hsiao,Chung-JungandShang-HsienHsieh(2023).“Real-timefireprotectionsystemarchitectureforbuilding safety”.In: JournalofBuildingEngineering 67,p.105913.

Hu,GuoqingandFengqiYou(2023).“Multi-zonebuildingcontrolwiththermalcomfortconstraintsunder disjunctiveuncertaintyusingdata-drivenrobustmodelpredictivecontrol”.In: AdvancesinAppliedEnergy 9, p.100124.

Hu,Maomao,BruceStephen,JethroBrowell,StephenHaben,andDavidCHWallom(2023).“Impactsofbuildingloaddispersionlevelonitsloadforecastingaccuracy:Dataoralgorithms?Importanceofreliabilityand interpretabilityinmachinelearning”.In: EnergyandBuildings 285,p.112896.

Iftikhar,Sundas,MuhammedGolec,DeeprajChowdhury,SukhpalSinghGill,andSteveUhlig(2022).“Fogcomputing basedrouter-distributorapplicationforsustainablesmarthome”.In: 2022IEEE95thVehicularTechnology Conference:(VTC2022-Spring).IEEE,pp.1–5.

InternationalEnergyAgency(2021). NetZeroby2050.Aroadmapfortheglobalenergysector.Tech.rep.Paris: IEA. url: https://www.iea.org/reports/net-zero-by-2050

Ju,Yi,ZheWang,XinyuanJu,BinCao,ChenChen,andBorongLin(2023).“Understandingoccupancypatterns ofuniversitylibrariesinthepost-pandemicera”.In: EnergyandBuildings 291,p.113138. issn:0378-7788. doi: 10.1016/j.enbuild.2023.113138

Kalaimani,RachelK.,SrinivasanKeshav,andCatherineRosenberg(2016).“Multipletime-scalemodelpredictive controlforthermalcomfortinbuildings”.In: ProceedingsoftheSeventhInternationalConferenceonFuture EnergySystemsPosterSessions.e-Energy’16.Waterloo,Ontario,Canada:AssociationforComputingMachinery. isbn:9781450344173. doi: 10.1145/2939912.2942355

Kampelis,Nikolaos,NikolaosSifakis,DionysiaKolokotsa,KonstantinosGobakis,KonstantinosKalaitzakis,Daniela Isidori,andCristinaCristalli(2019).“HVACoptimizationgeneticalgorithmforindustrialnear-zero-energy buildingdemandresponse”.In: Energies 12.11.

Kastner,Wolfgang,GeorgNeugschwandtner,StefanSoucek,andHMichaelNewman(2005).“Communication systemsforbuildingautomationandcontrol”.In: ProceedingsoftheIEEE 93.6,pp.1178–1203.

Kou,Xiao,YanDu,FangxingLi,HectorPulgar-Painemal,HeliaZandi,JinDong,andMohammedM.Olama(2021). “Model-BasedandData-DrivenHVACControlStrategiesforResidentialDemandResponse”.In: IEEEOpen AccessJournalofPowerandEnergy 8,pp.186–197. doi: 10.1109/OAJPE.2021.3075426

Lam,Edward,FritsdeNijs,PeterJ.Stuckey,DonaldAzuatalam,andArielLiebman(2020).“LargeNeighborhood SearchforTemperatureControlwithDemandResponse”.In: PrinciplesandPracticeofConstraintProgramming Ed.byHelmutSimonis.Cham:SpringerInternationalPublishing,pp.603–619. isbn:978-3-030-58475-7.

Lee,Donghwan,GichunCha,andSeungheePark(2016).“Astudyondatavisualizationofembeddedsensorsfor buildingenergymonitoringusingBIM”.In: Internationaljournalofprecisionengineeringandmanufacturing 17, pp.807–814.

Lei,Lei,BingWu,XinFang,LiChen,HaoWu,andWeiLiu(2023).“Adynamicanomalydetectionmethodof buildingenergyconsumptionbasedondataminingtechnology”.In: Energy 263,p.125575.

Ma,Jie,XiandongMa,andSuzanaIlic(2019).“HVAC-BasedCooperativeAlgorithmsforDemandSideManagement inaMicrogrid”.In: Energies 12.22. issn:1996-1073. doi: 10.3390/en12224276.

Magni,Mara,FabianOchs,SamueldeVries,AlessandroMaccarini,andFerdinandSigg(2021).“Detailedcross comparisonofbuildingenergysimulationtoolsresultsusingareferenceofficebuildingasacasestudy”.In: EnergyandBuildings 250,p.111260.

Majdi,Ali,AliJawadAlrubaie,AliaHaiderAl-Wardy,JamelBaili,andHiteshPanchal(2022).“Anovelmethodfor IndoorAirQualityControlofSmartHomesusingaMachinelearningmodel”.In: AdvancesinEngineering Software 173,p.103253.

Merz,Hermann,ThomasHansemann,andChristofHübner(n.d.). BuildingAutomation:CommunicationSystems withEIB/KNX,LONandBACnet,2009

Minh,QuyNguyen,Van-HauNguyen,VuKhanhQuy,LeAnhNgoc,AbdellahChehri,andGwanggilJeon(2022). “EdgeComputingforIoT-EnabledSmartGrid:TheFutureofEnergy”.In: Energies 15.17,p.6140.

Mirinejad,Hossein,KarlaConnWelch,andLucasSpicer(2012).“Areviewofintelligentcontroltechniquesin HVACsystems”.In: IEEEEnergytech,pp.1–5. doi: 10.1109/EnergyTech.2012.6304679

Mohamed,Nader,SanjaLazarova-Molnar,andJameelaAl-Jaroodi(2016).“CE-BEMS:Acloud-enabledbuilding energymanagementsystem”.In: 20163rdMECInternationalConferenceonBigDataandSmartCity(ICBDSC) IEEE,pp.1–6.

Mortensen,R.E.andK.P.Haggerty(1988).“Astochasticcomputermodelforheatingandcoolingloads”.In: IEEE TransactionsonPowerSystems 3.3,pp.1213–1219. doi: 10.1109/59.14584

Moudgil,Vipul,KasunHewage,SyedAsadHussain,andRehanSadiq(2023).“IntegrationofIoTinbuildingenergy infrastructure:Acriticalreviewonchallengesandsolutions”.In: RenewableandSustainableEnergyReviews 174,p.113121.

Nesa,NashreenandIndrajitBanerjee(2017).“IoT-basedsensordatafusionforoccupancysensingusingDempster–Shaferevidencetheoryforsmartbuildings”.In: IEEEInternetofThingsJournal 4.5,pp.1563–1570.

Ogolla,WalterandRaphaelKieti(2022).“ApplicationofBuildingManagementSystem(BMS)inCommercial PropertyManagementinKenya”.In: UnderstandingAfricanRealEstateMarkets.Routledge,pp.188–200.

Pan,YueandLimaoZhang(2020).“Data-drivenestimationofbuildingenergyconsumptionwithmulti-source heterogeneousdata”.In: AppliedEnergy 268,p.114965.

Pandya,Darshit(2021).“ASurveyofDescriptionMethodsforFieldbusSystems”.In: AdvancedInformation NetworkingandApplications:Proceedingsofthe35thInternationalConferenceonAdvancedInformation NetworkingandApplications(AINA-2021),Volume3.Springer,pp.200–210.

Pervez,Hamza,YousafAli,andAntonellaPetrillo(2021).“Aquantitativeassessmentofgreenhousegas(GHG) emissionsfromconventionalandmodularconstruction:Acaseofdevelopingcountry”.In: JournalofCleaner Production 294,p.126210.

Pouranian,Fatemeh,HabibollahAkbari,andSMHosseinalipour(2021).“Performanceassessmentofsolarchimney coupledwithearth-to-airheatexchanger:Apassivealternativeforanindoorswimmingpoolventilationin hot-aridclimate”.In: AppliedEnergy 299,p.117201.

Pricop,Emil,JaouharFattahi,NicolaeParashiv,FlorinZamfir,andEliesGhayoula(2017).“Methodforauthenticationofsensorsconnectedonmodbustcp”.In: 20174thInternationalConferenceonControl,Decisionand InformationTechnologies(CoDIT).IEEE,pp.0679–0683.

Ren,Z.,Z.Tang,andMJames(2021). Typicalmeteorologicalyearweatherfilesforbuildingenergymodelling.User Guide.Tech.rep.CSIRO. url: https://agdatashop.csiro.au/future-climate-typical-meteorological-year

Sayed,AyaNabil,YassineHimeur,andFaycalBensaali(2022).“Deepandtransferlearningforbuildingoccupancy detection:Areviewandcomparativeanalysis”.In: EngineeringApplicationsofArtificialIntelligence 115,p.105254.

Shikhli,Siraj,AmirShikhli,AnwarJarndal,ImadAlsyouf,andAliCheaitou(2022).“TowardsSustainabilityin Buildings:aCaseStudyontheImpactsofSmartHomeAutomationSystems”.In: 2022AdvancesinScienceand EngineeringTechnologyInternationalConferences(ASET).IEEE,pp.1–8.

Ürge-Vorsatz,Diana,LuisaF.Cabeza,SusanaSerrano,CamilaBarreneche,andKseniaPetrichenko(2015).“Heating andcoolingenergytrendsanddriversinbuildings”.In: RenewableandSustainableEnergyReviews 41,pp.85–98. issn:1364-0321. doi: 10.1016/j.rser.2014.08.039

Verma,Anurag,SuryaPrakash,andAnujKumar(2023).“AI-basedbuildingmanagementandinformationsystem withmulti-agenttopologyforanenergy-efficientbuilding:Towardsoccupantscomfort”.In: IETEJournalof Research 69.2,pp.1033–1044.

Wang,Shengwei(2009). Intelligentbuildingsandbuildingautomation.Routledge.

Wang,Xia,JiachenYuan,KairuiYou,XianruiMa,andZhaojiLi(2023).“UsingRealBuildingEnergyUseDatato ExplaintheEnergyPerformanceGapofEnergy-EfficientResidentialBuildings:ACaseStudyfromtheHot SummerandColdWinterZoneinChina”.In: Sustainability 15.2,p.1575.

Zaeri,Narges,ArazAshouri,HBurakGunay,andTareqAbuimara(2022).“Disaggregationofelectricityandheating consumptionincommercialbuildingswithbuildingautomationsystemdata”.In: EnergyandBuildings 258, p.111791.

Zhang,Xingxing,FilippoPellegrino,JingchunShen,BenedettaCopertaro,PeiHuang,PuneetKumarSaini,and MarcoLovati(2020).“ApreliminarysimulationstudyabouttheimpactofCOVID-19crisisonenergydemand ofabuildingmixatadistrictinSweden”.In: AppliedEnergy 280,p.115954.

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.