segmented dynamic model: Case study of Norway 1960–2015
Nina Holck Sandberg
a,∗, Igor Sartori
b, Magnus I. Vestrum
a, Helge Brattebø
aaIndustrialEcologyProgramandDepartmentofEnergyandProcessEngineering,NorwegianUniversityofScienceandTechnology(NTNU),7491 Trondheim,Norway
bSINTEF,DepartmentofBuildingandInfrastructure,P.O.Box124Blindern,0314Oslo,Norway
a r t i c l e i n f o
Articlehistory:
Received31October2015
Receivedinrevisedform25April2016 Accepted31May2016
Availableonline11June2016
Keywords:
Dynamicmodelling Energyanalysis Dwellingstock Historicaldevelopment Scenarioanalysis Energyefficiency Energymix Userbehavior
a b s t r a c t
Asegmenteddynamicdwellingstockmodelisprovenusefulforunderstandingthedevelopmentand changesofageingbuildingstocks,whichishighlyrelevantforrenovationmeasuresandestimatesof energyuseandemissionsinaggregatedbuildingstocks.Inthispaper,suchamodelisdevelopedfurther fordetailedanalysesofdwellingstockenergydemandandexemplifiedfortheNorwegiandwelling stock1960–2015.Thedwellingstockmodelsimulatesthedevelopmentinstocksizeandcomposition andiscombinedwitharchetype-specificenergyintensitiestoestimatethetotalenergydemand.After calibratingthemodelresultswithstatistics,themodelisusedtoexplorethephenomenaandcausesof historicalchanges.Alarge-scaleimprovementoftheenergyefficiencyoftheNorwegiandwellingstock hastakenplacethroughrenovationandconstructionofnewdwellings.Ahistoricalshifttomoreefficient energycarriersandheatingsystemshashadaneffectonenergysavingsinthesystem,ofthesamesize astheeffectoftheimprovedenergyefficiencyofthestock.However,thetotalaverageenergysavings perm2areoffsetbychangesinuserheatinghabits.Asignificantdecreaseinaveragedeliveredenergy intensityperm2isonlyobservedaftertheintroductionofheatpumps.
©2016TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCC BY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
Thebuildingsectorisimportantforfuturemitigationofgreen- housegas(GHG)emissions[1],asbuildingsare responsiblefor about40%oftheenergyconsumptionintheEU[2].Toquantifythe energysavingpotentialofthestockandtoensurethatanypoten- tialsavingswillbeobtained,energyanalyses,scenariomodels,road mapsandactionplansareimportantpolicytools.
Arangeofmodelsinvestigateandanalyseenergyuseinbuilding stocks[3–6].Suchanalysesarebasedonmodelsforthedevelop- mentofthedwellingstockintermsofthenumberofdwellings, andtheircharacteristics(type,age,sizeandtechnicalstandard).
Themodelleddwellingstocksizeissubsequentlymultipliedwith theaverageenergyintensitypersquare meter tofindthetotal energydemand.Thismeansthatgoodestimatesforthetotalenergy demanddependondetailedandreliablemodels,forboththestock andtheaverageenergydemand.
∗Correspondingauthor.
E-mailaddresses:[email protected](N.H.Sandberg),
[email protected](I.Sartori),[email protected](M.I.Vestrum), [email protected](H.Brattebø).
AdetailedoverviewofexistingmodelsispresentedinVásquez etal.[7].Theenergyanalysesusedtomodelfutureenergycon- sumptionofspecificbuildingtypesarecommonlyverydetailed andwellgrounded(e.g.[3–6,8]).However,traditionalstockmod- elsappliedforscenariomodellingandforecastingofenergyuseof dwellingstocksoftenuselinearorsimplifiedassumptionsregard- inghowconstruction,demolitionandrenovationactivitieschange over time.By combiningbasic linearassumptions forthestock developmentwithadetailedenergyanalysis,thereliabilityofthe finalresultswillbelimitedbythesimplificationsofthestockmodel.
Theuncertaintyoftheinputparametersindwellingstockmodels andtheireffectonthefinalresultsoftheenergyanalysisarerarely discussed.Toachievereliableandvalidresultsfromabuildingstock energymodel,aproperdwellingstockmodelshouldbecombined withadetailedenergyanalysis.
Incontrasttotraditionaldwellingstockmodels,whichareoften basedonaccounting(e.g.[9–12]),dynamicdwellingstockmod- elsaimatdescribingthedevelopmentindwellingstocksizeand compositionbyuseofmass-balancetime-consistent calculation principles.Construction,demolitionandrenovationactivitiesare basedontheunderlyingdriversand parametersin thesystem;
thepopulation’sdemandfordwellingsandtheagingofdwellings leadingtoneedforrenovationandfinallydemolition[13].Vásquez
http://dx.doi.org/10.1016/j.enbuild.2016.05.099
0378-7788/©2016TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/
4.0/).
etal.[7]foundthatthedynamicsofadwellingstocksystemisof largeimportancefortherecommendedfuturestrategiesforenergy savingsinthedwellingstocks.
Sandbergetal.[14,15]madeafirstattemptatmakingenergy scenariosfordwellingstocksbyusingadynamicdwellingstock model for Norway. Their results clearly showed that the sim- plifications in the linear models commonly used for dwelling stockdevelopmenthadlargeimplicationsontheresultingenergy demand in the dwelling stock and on potential energy reduc- tions.However,thesestudiesalsorevealedaneedforcombining moredetailedenergyanalyseswithdynamicdwellingstockmod- eling.Thedynamicdwellingstockmodelthathadbeendeveloped througharangeofpublications[13–18]haduntilthenexamined thedevelopmentofatotaldwellingstock,regardlessofthestock compositionofdifferentdwellingtypes.
Toimprovethequalityoftheenergyanalysis,more detailed informationaboutthedwellingstockcompositionwasrequired.
Based on the same principles, Sandberg et al. [19] developed a segmented dynamic model that allowedfor segmentation of the dwelling stock in dwelling types and construction periods (cohorts).Renovationactivitywasmodeledwithinthemodel,as inSartorietal.[17],attemptingtoestimatetheneedformainte- nanceandupgradingofpreviousconstruction.Therenovationrate wasthusaresultfromthemodelratherthananinputtoit.One ofthemainfindingsinSandbergetal.[19]wasthattherenova- tionrates(shareofthestocktoberenovatedperyear)commonly assumedintraditional scenariomodelsandaction plansarefar abovewhatcanbeexpectedbasedonthe“natural”needforren- ovationofdwellingsduetoagingprocessesinthebuildingstock.
Thisfindingwasshowntoberobustinathoroughscenarioanalysis [20].
Themethodologyandalgorithmofthedynamicdwellingstock modelisexplainedindetailinSartorietal.[21]andlaterappliedina comparativestudyfor11EuropeancountriesinSandbergetal.[22]
(bothinthisissue).Thesimulatedfuturerenovationratestowards 2050inthe11countriesareintherange0.6-1.6%,andthusnever ashighasthelevels2.5-3%thatotherstudiescommonlyassumeto bepossibleandnecessaryforreachingmitigationgoals[3,6,23].
Modelsfor assessing the energy demandin dwelling stocks commonlycoveraspecificyearorpotentialfuturedevelopment [3,11,24–28].However,toevaluatethereliabilityandapplicabil- ity of themodel, it should alsobe calibrated against historical development.Further,historicalmodelscanbeusedtounderstand whathasbeentheimportantfactorsforthehistoricaldevelopment, whichisinterestinginitselfandimportantfordescribingpossible futuredevelopmentpaths.Totheknowledgeoftheauthors,thisis rarelydoneinliterature.OneexceptionisNässenandHolmberg’s studyonthehistoricalimprovementoftheSwedishdwellingstock resultingfromrenovationandnewconstruction[29].Theyfound thatthecalculatedenergydemandperm2inbuildingswithoneor twodwellingswasreducedby11%between1975and2000.41%
ofthisreductioncouldbeattributedtonewconstruction,and59%
toimprovementsoftheexistingstock.Changesinuserbehaviour wasnottakenintoconsideration.
SubstantialchangeshavetakenplaceintheNorwegiandwelling stock system during the period since 1960: strong population growth,changingenergystandardofthedwellingsthroughreno- vationandconstruction,changingenergymix,heatingsystemsand outdoorclimate,aswellaschangesinlifestyleanduserbehavior.
Inthispaper,thesegmenteddynamicstockmodelfromSand- bergetal.[19]isdevelopedfurtherfordetailedanalysesofdwelling stockenergydemand.Themodelisexemplifiedforacasestudyon thehistoricaldevelopmentinNorwaysince1960.Theapplicability ofthedwellingstockmodelforenergyanalysesisexploredthrough calibrationofthemodelresultsagainstrecordedhistoricaldata.
Throughascenarioanalysis,theimportanceofdifferentcausesof
thehistoricalchangesonenergydemandisexamined.Finally,we explorewhatthesituationwouldhavebeen,ifsomeoftheimpor- tantchangesinthesystemhadnottakenplace.Thesephenomena arerarelyreportedanddocumentedinliterature;thereforethis studyprovidesnewinsightbothmethodologicallyandempirically.
2. Methods
2.1. Analyticalmethods
ThemodelisconceptuallyoutlinedinFig.1,whichshowshow differentvariablesarerelatedtoeachother.Furtherdetailsofthese relationshipsarenotincluded,asthiswouldmakethefiguretoo complex.Themainprinciplesofthemodelareexplainedbelow, andamoredetaileddescriptionofthemodelanditsmathematical frameworksispresentedinAppendixB(Supplementarymaterial).
Themodelconsistsoftwoparts;thefirstpartisthebuilding stockmodelandthesecondisthebuildingstockenergymodel.
Thecoreofthebuildingstockmodelisthepopulation’sdemand fordwellings,SD,andthedistributionofthestockovervarious dwellingstocksegments,SDs.Asegmentisdefinedbythedwelling typeandconstructionperiod(cohort),e.g.SingleFamilyHouses fromthe1970s.Thedemandfordwellingsisestimatedforeach year,basedonthedevelopmentintheunderlyingdriversinthe system:populationsize,P,numberofpersonsperdwelling,PD,and shareofdwellingsbeingofeachdwellingtype,W.
Demolitionactivityinacertainyearisestimatedbyapplying a demolition probabilityfunction onconstruction activityfrom allpreviousyears.Constructionactivityisestimatedusingmass- balance consistentcalculation principles;i.e. what needstobe constructedtoreplacedemolished buildingsand tomeet stock changes according to changing demand. No otheradditions or subtractionstothebuildingstockthannewconstructionanddemo- lition(e.g.changeoffunction)areincludedinthemodel.
Whiledemolitionofa dwellingcanhappenonlyonce,reno- vationcanhappenseveraltimesduringabuilding’slifetime.The renovationactivityinacertainyear,Dren,isestimatedbyapplying arenovationprobabilityfunctiontotheconstructionfromallpre- viousyears.Themodelallowsforcyclicrepetitionsofthisfunction, describedbytherenovationcycle,RC,whichrepresentstheaver- agetimespanbetweenrenovationsofacertaindwelling.Thecyclic renovationprobabilityfunctionislinkedtothelifetimeprobabil- ityfunction,preventingadwellingfrombeingdemolishedshortly aftergoingthroughrenovation.Thedefinitionoftherenovation activityiscase-specificandtherelatedrenovationcycledescribes theaveragetimespanbetweenrenovationsofthedefinedtype.The renovationactivityisindependentofthemassbalanceanddoesnot affectthedwellingstocksizeoritsdistributiontosegments.
Thenumberofdwellingsdemolished,Ddem,constructed,Dnew, andrenovated,Dren,each yearareoutputs fromthemodel,and hencealsothedemolition,constructionandrenovationrates.
Thismodeldiffersfrompreviousversionsbydistributingthe segmentstoarchetypesaccordingtotheirrenovationstate,e.g.
SingleFamilyHousesfromthe1970sbeingintheiroriginalstate withoutsignificantenergy-renovationimprovements.SDs,risthe archetypedefined by segment,s, and renovation period, r.The actualenergyperformanceofeacharchetype−ofnewconstruc- tionandofdwellingsthatgothroughrenovationduringdifferent timeperiods−isscenariospecific.Thismeansthatthedistribution torenovationperiodsdoesnotdeterminetheenergystandardof thedwellings.
Inthebuildingstockenergymodel,averagefloorareaperseg- mentandarchetypespecificenergyneedintensitiesareappliedto thenumberofdwellingspersegmenttoobtaintheenergyneed persegment.Finally,theheatpumpcontribution,deliveredenergy
Fig.1.Conceptualoutlineofthebuildingstockmodelandthebuildingstockenergymodel.Hexagonsrepresentinputvariables,rectanglesstocksandovalsflows.Allinputs andoutputsaretime-dependent.
anduseofvariousenergycarriersareestimatedpersegmentand forthetotalstock.
Inthisstudythemodelresultsarecalibratedagainststatistics ontotaldeliveredenergyinthesystem,since1960.Thesimulated deliveredenergyisexpectedtodifferfromthestatistics,especially farbackintimewhentheheatinghabitsdifferedsubstantiallyfrom whatisassumedinthetechnicallyestimatedenergyneedinten- sity(calculatedaccordingtocurrentstandardsandmethods).The adaptationfactorfAis definedastheaggregated measuredover calculatedannualenergydemand,henceannualdeliveredenergy valuesfromstatisticsoverannualvaluescalculatedbythemodel.
Theadaptationfactorthereforeincludeschanginguserbehavior (heatinghabits)anduncertaintyinmodelresults.
2.2. DataandassumptionsfortheNorwegiancase
The presented model is generic and can be applied to any dwellingorbuildingstockandtoanytimeperiod.Inthefollow- ingwepresentacasestudyofhistoricaldevelopmentinenergy demand in the national aggregated Norwegian dwelling stock.
Themaininputstothemodelaredescribedinthissection.More detailedinformationaboutthedataandassumptionsispresented inAppendixC(Supplementarymaterial).
Duetothelonglifetimeofdwellings,thedwellingstocksys- temchangesslowlyandthecompositionofthestockdependson activitiesfarbackintime.Alongtimehorizonisthereforeneeded whenworkingwithdwellingstockmodels.Thiscasestudycovers theperiod1800–2050inordertocapturelong-termchangesinthe
Table1
Cohortdefinitionandaverageheatedfloorareaperdwellingineachsegment.
SFH00–01 SFH02 SFH03–04 SFH05–07 TH00–01 TH02–07 MFH
1960 Shareofdwellingstock 51% 5% 0% 0% 21% 2% 22%
Shareoffloorareastock 64% 6% 0% 0% 17% 2% 11%
Shareofdeliveredenergy(heating+dhw) 68% 5% 0% 0% 17% 1% 9%
1990 Shareofdwellingstock 25% 12% 20% 0% 10% 11% 20%
Shareoffloorareastock 29% 11% 0% 0% 16% 3% 11%
Shareofdeliveredenergy(heating+dhw) 37% 14% 22% 0% 9% 9% 9%
2015 Shareofdwellingstock 14% 8% 15% 12% 6% 17% 28%
Shareoffloorareastock 18% 11% 22% 16% 5% 15% 15%
Shareofdeliveredenergy(heating+dhw) 23% 11% 22% 12% 6% 13% 13%
Norwegiandwellingstock.Theenergyanalysis,however,canstart atanypointintime,astheenergydemandinacertainyeardepends onthecurrentstocksizeandcomposition,butdoesnotdepend directlyontheenergydemandinpreviousyears.Inthispaper,the analysisofthehistoricaldevelopmentindeliveredenergyinthe Norwegiandwellingstockissimulatedfortheperiod1960–2015.
Detailedstatisticsonenergyuseinhouseholdsisavailableforthis periodandisusedformodelcalibrationandreference.
Asetofbuildingsrepresentedinthenationaldwellingstocksof arangeofEuropeancountrieshasbeendescribedindetailinthe IntelligentEnergyEuroperesearchprojectsTABLEandEPISCOPE [30].For theNorwegiancase, three dwellingtypesSingle Fam- ilyHouses(SFH),TerracedHouses(TH)andMultiFamilyHouses (MFH)and7cohorts(cohort1–7)resultin21type/cohortcom- binations for which both a realexample building as well as a syntheticaveragebuildinghasbeendescribed.Thissegmentation ofthedwellingstockisalsousedinthepresentstudy,however, duetothewaythemodelworks,theinitial1800stockneedsto beaseparatecohort,definedascohort0.Thecohortdefinitionand theaverageheatedfloorareaperdwellingforeachsegment(taken fromstatistics[31])arelistedinTable1.
Thesegmentsdefinedbydwellingtypeandcohortarefurther distributedtoarchetypesaccordingtotheirrenovationperiod,r.
Therenovationperioddefines ifand whena dwellinghasgone throughitsmostrecentrenovation.Dwellingsintheiroriginalstate anddwellingsexposedtorenovationpriorto1980areplacedin renovationperiod1,sincethecommonrenovationmeasuresuntil 1980tolittledegreeincludedenergy-efficiencymeasures.Further, itisassumedthatsince1980technologyhasbeenavailablesothat inclusionofenergy-efficiencymeasureswaspossiblewhenevera dwellingwasrenovated.Dwellingsrenovatedsince1980arethere- foreplacedinrenovationperiod2.Thebaselineassumption,used inmodelcalibrationandinsomeofthescenarios,isthatrenova- tionsinrenovationperiod2correspondtostandardrenovationas definedintheTABLEproject[30].
Fig.2showsthetimeseriesfortheinputparametersPopulation, P,(leftfigure,leftaxis),sharelivinginSFHorTH,W,(leftfigure,right axis)andpersonsperdwelling,PD,(rightfigure).Furtherdetails abouttheseinputs,dataprocessingandassumptionsintheseg- menteddwellingstockmodelinthecaseofNorwayaredescribed inSandbergetal.[19]andAppendixC(Supplementarymaterial).
ThelifetimeprobabilityfunctionisassumedtofollowaWeibull distribution, defined by the parameters average lifetime per dwellingandtheinitialperiodafterconstructionwheretheproba- bilityofdemolitioniszero,asexplainedindetailinSandbergetal.
[19]andSartorietal.[21](thisissue).Thisisinlinewiththerec- ommendationsinSereda[32].Theaveragelifetimeofdwellings isestimatedto125years,inlinewiththefindingsinBohneetal.
[33]andtheinitialperiodafterconstructionwithnodemolitionis assumedtobeequaltoonerenovationcycle,Rc.
BasedondatafromtheDirectorateofCultural Heritage[34]
theshareofbuildingsfromeachconstructionyearthatisassumed
nevertobedemolishedisestimatedto5%forSFHandTHand9%
forMFH.
Thedefinitionoftherenovationactivityinthemodeliscase- specific.Inthisstudy,weexplorethedynamicsofrenovationsthat havethepotentialforincludingenergy-efficiencymeasuresthat leadtoalargedecreaseintheenergydemand.Theimplementation ofthesemeasuresarecostlyandnotlikelytotakeplaceifadwelling isnotgoingthrougharenovationinanycase.Hence,suchmeasures couldbeimplementedwhenthedwellingisrenovatedduetoits
“natural”ageingprocessandneedformaintenanceandupgrading.
Inthisstudy,weestimatethetotalrenovationactivityresulting fromthisageingprocessofthedwellingstockinNorway;deep renovationoffacadeswhichareassumedtooccurinrenovation cycles,Rc,of40years.ThisisinlinewithfindingsinKristjansdottir etal.[35].
InTABLE[30],thetechnicalstandardofeachexamplebuilding andsyntheticaveragebuildinghasbeendescribedindetailfortyp- icalbuildingsintheiroriginalstate,afterastandardrenovationis carriedout,andafteranadvancedrenovationiscarriedout.The energyneedintensityperdwellingtype, cohortand renovation variantinTABLEisshowninFig.3.Thesevaluesincludeenergy needforspaceheatinganddomestichotwater(dhw),excluding electricalappliances.
Inthisstudy,archetypesaredefinedbythedwellingtype,cohort andrenovationperiod.Theactualrenovationvariant,anditscor- respondingenergystandard,chosenforeachrenovationperiod,is decidedtobescenariospecific.Thisaddsflexibilitytothemodel, sinceenergyintensitiesareaconsequenceofbuildingcodes,energy savingmeasures and otherfactors that maychange over time.
Inthisstudy,energyneedintensityvalues(inkWh/m2/year)for eacharchetypearetakenfromTABULA,andreflectenergy-related changesintheNorwegiandwellingstocksince1960.
Thehistorical developmentin theenergy mixin Norwegian householdsisknownfromstatistics[36–38].Estimationsonshare oftheenergybeingusedforheatinganddhwandsystemefficien- ciesperenergycarrierasafunctionoftimeareusedtoestimatethe overallweightedaveragesystemenergyefficiencies,aspresented inTable2.
Finally,theenergyneedperarchetype,ENss,r,iscorrectedfor energycontributionfromheatpumpsandconvertedtodelivered energyperarchetype,DEss,r,usingtheweightedaveragesystem efficiency.Themodeliscalibratedforchangingoutdoorclimateand changingelectricload,asdescribedinAppendixB(Supplementary material).
Asummaryofthedefinitionofkeytermsthatareusedinthe modelandanalysisisgivenAppendixA,attheendofthispaper.
2.3. Scenariodescription
ABaselinescenarioandsixadditionalscenarioswillbeexplored.
The dwelling stock model is the core of all thescenarios, and themodelisrunusingtheinputsdescribedinAppendixC(Sup-
Fig.2. Left:developmentintotalpopulation,andinpersonslivingineachofthetwodwellingtypes.Right:developmentinpersonsperdwellingforthetotalstockaswell asforeachdwellingtype.
Fig.3.Energyneedintensitiesperdwellingtype,cohortandrenovationvariant.
Table2
Energymix(spaceheatinganddhw)andweightedaveragesystemefficiencyspecifiedfordifferentsegmentsandyears.
Year 1960 1982 1990 2012
Dwellingtype All All All SFH TH MFH
Cohort All All All 00–03 04–05 06–07 00–03 04–05 06–07 00–03 04–05 06–07
Shareenergycarrier(%) Total 100 100 100 100 100 100 100 100 100 100 100 100
El 27 62 67 72 73 81 82 86 87 67 82 72
Bio 31 15 17 21 23 18 14 12 11 9 5 5
Oil 42 23 15 6 4 0 4 1 0 12 5 2
Districtheating 0 0 1 0 0 1 0 2 2 12 8 21
Weightedaveragesystemefficiency 0.64 0.83 0.84 0.87 0.87 0.91 0.92 0.94 0.94 0.92 0.96 0.96
plementarymaterial)forallscenarios. Thenumber ofdwellings constructed,demolishedandrenovatedeachyeardonotchange between the scenarios. Further, the conversion factor used to convertfromOsloclimatetonationalaverageacrossdifferentNor- wegianclimatezonesisappliedinallscenarios.
TheBaselinescenarioaims atreproducingtherealhistorical developmenttrendsin thebestpossible way.Theenergy need intensitiespersegmentandvariantareasdefinedinFig.3,reno-
vationactivitysince1980isassumedtoshifttheenergyefficiency fromvariant1tovariant2.Theenergyneedisconvertedtodeliv- eredenergybeforethenationalclimateconversionfactorandthe heatingdaydegrees(HDD)trendlineareappliedtothesimulated deliveredenergyforheatingandbeforetheadaptationfactorare appliedtotheresultingdeliveredenergy forspace heatingand dhw.Theelectricloadisalsoincludedandassumedtofollowthe estimatedtrendline.
Table3
Scenariodefinition.
Scenario number
Scenario name
Renovation after1980
New construction
Energymix and efficiencies
HDD correction
Thermal adaptation factor
Electricload
0 Baseline Shifttovariant2 Segmentspecificvariant1 Changing overtime
Trendline Trendline Trendline
1 Noenergy
efficiencythrough renovation
Stillvariant1 Segmentspecificvariant1 Changing overtime
Trendline Trendline Trendline
2 Noenergy
efficiencyinnew built
Shifttovariant2 Noimprovementafter1970 Changing overtime
Trendline Trendline Trendline
3 Fixed1960energy
mix
Shifttovariant2 Segmentspecificvariant1 1960 energymix andeffi- ciencies
Trendline Trendline Trendline
4 Fixed1960climate Shifttovariant2 Segmentspecificvariant1 Changing overtime
1960value Trendline Trendline
5 Fixed1960thermal
adaptationfactor
Shifttovariant2 Segmentspecificvariant1 Changing overtime
Trendline 1960value Trendline 6 Fixed1960electric
load
Shifttovariant2 Segmentspecificvariant1 Changing overtime
Trendline Trendline 1960value
Toexploretheimportanceofthevariousenergy-relatedparam- eters,thesearevariedbetweenthescenarios,oneatatime.The scenariosaredefinedinTable3.Inaddition,someofthescenarios willbecombinedtoexplorethecombinedeffectofchangingmore parameterssimultaneously.
2.3.1. Uncertaintyofinputparameters
Forthishistoricalanalysis,manyoftheinputparametersare takenfromofficialstatistics,andtheiruncertaintyisthereforelow.
Thisisthecasefortheparameterspopulation,personsperdwelling, shareofdwellingsbeingindifferenttypes,averagefloorareaper segment,sharehavingheatpumpandtheenergymix.
Thedwellingstockmodeliscalibratedagainststatisticsforthe historicalperiodunderstudy.However,althoughthenumberof dwellingsinthestockiswellcalibrated,thereissomeuncertaintyin thedynamicsofthesystem,intheparametersrelatedtodemolition andrenovation,duetolackofempiricaldata.Further,althoughthe averageutilityfloorareaperdwellingineachsegmentisbasedon statisticsandoflowuncertainty,thereisuncertaintyrelatedtothe conversionfromutilityfloorareatoheatedfloorarea.
Finally,theenergyneedintensityofeacharchetypeisestimated throughtheTABLEmethodology,whichusesadetaileddescription ofthetechnicalstandardofeachsyntheticaveragebuildingand renovationvarianttoestimatetheenergyintensities.Thereis a fairlyhighuncertaintyrelatedtohowwellthegivensegmentand renovationvariantsrepresentstherealaveragevaluesinthestock.
Finally,thereishighuncertaintyintheassumedcontributionfrom heatpumpsandintheheatingsystemefficiencies.
Theuncertaintyofthevariousinputparametersisevaluatedin Table4.Theparametersaregroupedaccordingtothepartofthe modelthattheyinfluence:i)thedwellingstocksizeandcompo- sitionmeasuredinnumberofdwellings,ii)thefloorareastock measuredinsquaremetersofheatedfloorarea,andiii)thedeliv- eredenergydemandmeasuredinGWh.Thesourceoftheinput dataandtheevaluationoftherelateduncertaintyislisted.When theuncertaintyisregardedashigh,theinputparameterwillbe includedinasensitivityanalysis,whereinputparametersarevar- iedtotheirlowandhighvariants,oneatatime.Thelowandhigh variantswillbe+/−10%forallparametersexceptthestartyearof renovation,where+/−10yearswillbeused.Finally,theresulting effectsonthemodelresultsareevaluated.
3. Resultsanddiscussion
3.1. Evolvementofdwellingstocksizeandcomposition
Fig.4showsthesimulatedhistoricalevolvementtheNorwegian dwellingstocksizeandcomposition,measuredinsquaremeters ofheatedfloorarea.Thesimulateddwellingstockevolvementis appliedtoallthehistoricalscenarios.Thestockhasgrownby121%
from115millionm2in1960to254millionm2in2015,intermsof heatedfloorarea.Theleftpartofthefigureshowsthedevelopment instockcompositionofdwellingtypesandrenovationperiods.Ren-
Fig.4.Dwellingstockdevelopmentinmillionm2.Stockdistributedtotypesandrenovationperiods(left)andtoenergyneedlevels(right).
Contributionfromheatpump NS3031 High −10% +10%
Energymix Statistics Low
Efficiencies NS3031 High −10% +10%
Electricload Statistics(NorwayandSweden) High −10% +10%
Fig.5. Simulatedtotaldeliveredenergywithandwithoutintroductionofheat pumpscomparedwithstatistics.
ovationperiod 1consistsof alldwellingsintheiroriginalstate anddwellingsrenovatedbefore1980.Regardlessoftheirenergy standardwhenconstructed,newconstructioninallyearsisadded totherenovationperiod1band.Renovationperiod2consistsof dwellingsrenovatedafter1980.Fig.4demonstrateshowtheNor- wegiandwellingstockhasbeenandstillisdominatedbysingle familyhouses(SFH),buttheshareofthefloorareastockbeingin multifamilyhouses(MFH)hasincreasedovertime.
The development of the stock composition in terms of the energystandardispresentedintherightpartofFig.4,whereseg- mentswithsimilarenergyneedintensitiesaregroupedtogether.
Eachcolorrepresentsa certainrange of energyintensities.The shadedyellowandgreenareasrepresenttheestimatedtotalfloor areathathasreachedthisenergyintensitylevelafterrenovation.
Therehasbeenastrongtrendofanimprovingaggregatedenergy standardofthestock,boththroughenergy-efficientnewconstruc- tion and renovation. Until1970,all dwellings had a calculated energyneedintensityabove170kWh/m2 (accordingtocurrent calculationmethod).Accordingtothemodelresults,thisshareis reducedto14%ofthefloorareastockin2015.
3.2. Evolvementofdwellingstockenergydemandandthe adaptationfactor
Thesimulateddevelopmentintotaldeliveredenergycorrected fortemperaturevariationsiscomparedwithenergystatisticsin Fig.5.Despitetheyearlyfluctuations,thesimulatedtotaldelivered energywasratherstableatalong-termyearlylevelofabout50
GWhfrom1960to2000.Giventhestronggrowthinpopulation andfloorareaduringthisperiod,theseresultsconfirmtheaggre- gatedimprovementincalculated(technical)specificenergyneed intensities(kWh/m2/year).
Inthesameperiod,however,therewasadoublingoftheactual energydemand,recordedinthestatistics,from22GWhin1960 to 44 GWh in 2000. Various factors affected the development intotalenergy demandduringthisperiod.Thegrowthinnum- berofdwellingsinthestockandtheincreaseduseofhousehold applianceswouldleadtoanincreaseinthetotalenergydemand.
However,thisincreasewasdeceleratedbyimprovedenergystan- dard in the stock, increase in the share of energy carriers and heatingtechnologieswithahighersystemefficiencyandalarger shareofthestockbeingapartmentswithsmallerfloorareaand lowerenergyintensities.Allthesefactorsareincludedinthemodel results.ThechangingadaptationfactorfA,showninFig.6,i.e.mea- suredovercalculatedenergy,canexplaintheremainingdifference betweenthemodelresultsandthestatistics,includingthechanges inheatinghabits.
From2000onwards,therehasbeena decreaseinthesimu- latedtotaldeliveredenergy,whereastheactualenergydemandin dwellings,accordingtothestatistics,wasratherstable.Heatpumps wereintroducedtodwellingsin2004[39].Intheperiod2004–2015 thishasledtoastrongerdecreaseinthesimulatedtotaldelivered energy(bluecurveinFig.5).TheyellowcurveinFig.5showsthe estimateddeliveredenergyifnoheatpumpswereusedanddirect electricitywasusedinsteadtomeetthesameheatingenergyneed.
Thetendencywouldthenhavebeenacontinuationofthelong-term trendobservedearlier,whichindicatesadiscontinuityintheuser behaviorafterheatpumpsareinstalled.Whentheadditionalcost ofincreasingtheindoortemperatureisverylowduetouseoflocal energysources(e.g.heatpump)orduetoveryhighenergyeffi- ciencyofthebuilding,theindoortemperatureisoftenincreased.
Thisiscommonlyreferredtoasthecomfortfactor[40].
Fig.6showstheestimatedhistoricaladaptationfactorforheat- inganddhw,fA,obtainedbydividingtheyearlystatisticsbythe modelresults,representingtheratioofmeasuredovercalculated deliveredenergy demand.Theelectric loadiskeptoutside this adaptationfactorasitsevolvementisalreadyestimateddirectly.
Asplittrendlineisusedtocoverthechanginguserbehaviorafter theintroductionofheatpumpsfrom2004.
ThefAtrendlinestartsatalevelof0.4in1960andendsatalevel of1.1in2015.Themainreasonfortheloweradaptationfactors backwardsintimeisdifferentheatinghabits.Backwardsintime, areassuchaskitchenandlivingroomswereheatedliketoday,oth- erswereheatedmuchless(bedroomsandbathrooms)andsome werenotheatedatall(basementandloft,thatonlylaterwerecon-
Fig.6. Historicaladaptationfactordevelopment.
vertedtousefuland heatedfloorarea).Further,theremightbe someuncertaintyinhowwelltheenergyneedintensitiesinTABLE reflecttherealaveragesofeachsegment.Thisuncertaintyisalso accountedforintheadaptationfactors.
3.3. Scenarioresults
The scenario results for the development of total delivered energyarepresentedinFig.7.Bydefinition,allscenarioshavethe samestartingpointin1960andthereafterthedevelopmentdiffers betweenthem.Thediscontinuityobservedbetween2003and2004 isduetotheuseofthesplitfAtrendline.
TheredcurveinFig.7showstheresultsofthebaselinescenario thatrepresentsthemodelresultswhereweareaimingatrepro- ducingtherealdevelopmenttrend,asexplainedinSection2.3.The baselineresultscorrespondswellwiththestatisticspresentedin Fig.5.
ThebluedottedlinesinFig.7representthescenarios1–4,where 1)theenergyefficiencystandard ofrenovatedbuildings, 2)the energyefficiencyofnewconstruction,3)theenergymixand4)the outdoorclimatearekeptatthe1960level,oneatatime.Thesesce- nariosshowthatiftheseparametershadnotchangedsince1960, theenergydemandinthedwellingstockwouldhaveincreasedto ahigherlevel.Theeffectofcombiningallscenarios1)-4)isshown inthesolidblueline.Ifalltheseparameterswerestillatthe1960 level,the2015deliveredenergywouldhavebeen94TWh,which ismorethanadoublingofthebaselineresult.Consequently,signif- icantenergyintensityreductionsthedwellingstock,represented bytheshadedblueareainFig.7,hascertainlyalreadytakenplace since1960.
ThedottedgreenlinesinFig.7representthescenarios5and6, where5)thethermaladaptationfactorand6)theaverageelectric loadarekeptatthe1960level,oneatatime.Theeffectofcombining thescenarios5and6isshowninthesolidgreenline.Ifthethermal
Fig.7.Scenarioresults.
Fig.8. Segmentedresultsforthebaselinescenario.
adaptationfactorandtheelectricloadwerestillatthe1960level, the2015deliveredenergywouldhavebeen19TWh,59%lessthan thebaselineresult.Thesescenariosshowthatenergysavingsinthe stockhavebeenoffsetbythechangesintheadaptationfactorand theelectricload.
A“1960Frozenscenario”whereallparameters1)-6)arekeptat the1960level(notshowninFig.7)wouldleadtodeliveredenergy of38TWhin2015,18%lessthanthebaselineresults.Thismeans thatintotal,thefactorsleadingtoanincreaseinenergydemandhas hadalargereffectthantheenergysavingsthathavebeenachieved inthesystemintheperiod1960–2015.
Segmentedresultsfor thebaseline scenarioarepresentedin Fig.8,illustratingthedistributionofthetotalenergydemandto differentpartsofthedwellingstock.Forbetterreadabilityofthe figure,somesegmentsaregroupedtogether.Fig.8demonstrates thatthemajorshareoftheenergydemandhasbeeninsinglefamily houses,andthesinglefamilyhousesconstructedbefore1955play themostimportantrole.
3.4. Importanceofsegments
Theimportanceof differentsegmentsof thestockis further exploredinTable5wherethesharesofthenumberofdwellings, totalfloorareaanddeliveredenergyforspaceheatinganddhw forthesamegroupsof segmentsare listedforthethree obser- vationyears1960,1990and2015.In1960,single-familyhouses constructedbefore1955dominatedboththedwellingstock,the floorareastockandtheenergydemand,withsharesof54%,66%and
69%,respectively.In2015,thisgroupofsegmentsstillaccountsfor 26%oftheenergydemandeventhoughtheshareofthedwellings andfloorareaisreducedto16%and20%,respectively.Intotal,49%
ofthecurrentdwellingsaresingle-familyhousesandtheyaccount for64%ofthefloorareaand67%oftheenergydemand.Thetotal share ofterracedhousesof thedwellingstock,floorareastock andtheenergydemandhasbeenratherstableatabout23%,19%
and18%,respectively,althoughterracedhousesconstructedbefore 1955naturallyplayedamuchmoreimportantrolein1960thanin 2015.Theshareofmulti-familyhousesofthedwellingstockand floorareastockhasincreasedfrom17%and8%in1960to29%and 16%in2015,respectivelyandtheshareoftheenergydemandhas increasedfrom7%in1960to14%in2015.
FromtheinformationinTable5weconcludethatthelargest future energy-savingpotentialintheNorwegiandwellingstock isstillinthesegmentofsingle-familyhousesconstructedbefore 1955.However,thetotaldeliveredenergydemandin2015iswell distributedovermanydwellingsegmentgroups.Thisindicatesthat futureambitiousenergy-savingpolicieswillhavetotargetrenova- tionmeasuresinavarietyofdwellingtypesandagecohorts,and thechoiceofstrategiesandtechnologiesshouldreflectthis.
Finally,thedevelopmentinaverageenergyintensityperm2and theaverageenergydemandperpersonareillustratedinFig.9.The yearlyresultsofthebaselinescenarioaredividedbythecorre- spondingsimulatedtotalnumberofsquaremetersinthestockand population,respectively.Improvedenergyefficiencyofthestock, thechangingenergymixandthehigheraverageoutdoortemper- atureleadtoadecreasewhilethechanginguserbehaviorleadto
Table5
Sharesofdifferentsegmentsinthedwellingstock,floorareastockanddeliveredenergy(spaceheatinganddwh)in1960,1990and2015.
SFH00–01 SFH02 SFH03–04 SFH05–07 TH00–01 TH02–07 MFH
1960 Shareofdwellingstock 51% 5% 0% 0% 21% 2% 22%
Shareoffloorareastock 64% 6% 0% 0% 17% 2% 11%
Shareofdeliveredenergy(heating+dhw) 68% 5% 0% 0% 17% 1% 9%
1990 Shareofdwellingstock 25% 12% 20% 0% 10% 11% 20%
Shareoffloorareastock 29% 11% 0% 0% 16% 3% 11%
Shareofdeliveredenergy(heating+dhw) 37% 14% 22% 0% 9% 9% 9%
2015 Shareofdwellingstock 14% 8% 15% 12% 6% 17% 28%
Shareoffloorareastock 18% 11% 22% 16% 5% 15% 15%
Shareofdeliveredenergy(heating+dhw) 23% 11% 22% 12% 6% 13% 13%
Fig.9.Averageenergyintensityperm2andperperson.
anincreaseintheenergyintensities.Insum,thelong-termtrend from1960to2003showsthatthesefactorshavemoreorlessneu- tralizedeachotherinthecaseofaverageenergyintensityperm2, whichwasataratherstablelevelof160–175kWh/m2/yr.How- ever,astheaveragefloorareaperpersonincreasedintheperiod, theaverageenergyintensityperpersonincreasedcorrespondingly from5700kWh/person/yrin1960to7500kWh/person/yrin2003.
Aftertheintroductionofheatpumps,therehasbeenadecreasein bothintensitiesfrom2003to2015.Thesimulatedaverage2015 energyintensitiesperm2andperpersonare142kWh/m2/yrand 6500kWh/person/yr,respectively.
Thetotalsavingsduetochangingenergymix,energyefficiency ofthestockandwarmerclimatearehowevercancelledoutbythe changinguserbehaviorandtheincreasedfloorareademandper person.Thisis inlinewithfindings inotherstudies wherethe theoreticaleffectofenergyefficiencymeasures(ormeasuresfor otherenvironmentalissues)ishigherthantherealobservedeffect [41–45].
3.5. Sensitivityanalysis
Thesensitivityanalysisinvestigatestheimpactonfinalresults whenchangingtheinputparameterswithhighuncertainty.The parameterswithhighuncertaintyidentifiedinTable4arechanged oneatatime,by±10%or±10years,whileallotherparametersare keptasinthebaselinescenario.Thesensitivityresultsfor2015,rel- ativetothebaselineresults,arepresentedinFig.10.Moredetailed resultsfrom thesensitivity analysis,including a corresponding graphfor1960aswellasthefulltimeseries,arepresentedand discussedinAppendixD(Supplementarymaterial).
Thesensitivityanalysisrevealslargedifferencesintheimpacts ofvariationsindifferentinputparameters.Someparametersarea directfactorintheequationusedtocalculatetheenergyneedfor heatinganddhw.A10%changeintheseparameterswillleadto acorrespondingcloseto8%changeinthemodelresults,or10%
changeintheenergyfor heatingand dhwexcludingel-specific load.Inthesensitivityanalysis,thisisthecasefortheparameters shareheatedareaandenergyneedintensities.Thisdemonstratesthe importanceofhavingbothagooddwellingstockmodelandrealis-
ticenergyintensitiestogetreliableresultsindwellingstockenergy models.
Further,variationsintheefficienciesofheatingsystemshave correspondinglarge impacts onthe resulting deliveredenergy.
Thereishighuncertaintyinthesystemefficiencyforwoodand fueloil,whereastheuncertaintyinthesystemefficienciesforelec- tricityanddistrictheatingislow.Backin1960,woodandfueloil coveredalargeshareofthehouseholdenergyuse.Therealeffi- cienciesoftheheatingsystemsusedbackthenarealsoofhigher uncertaintythanthecurrentequipmentthatisbetterdocumented.
Consequently,thereissomeuncertaintyinthemodelresultsinthe beginningoftheperiodunderstudy,butasthemodelresultsare calibratedagainststatisticsandtheadaptationfactorappliedtothe variousscenarios,thiswillnotinfluencetheconclusionsfromthe historicalscenarioanalysis.From1960to2015theuseofelectricity forheatinghasincreasedandiscurrentlydominatingtheenergy mix.Theshareofdistrictheatinghasalsoincreased.Thecurrent uncertaintyofaverageheatingsystemefficienciesisthereforelow.
Variations in renovationcycleRcand start year of renovation alsoresultinvariationsintheestimateddeliveredenergy.A±10%
inRcand±10yearsinthestartyearofrenovationleadtoa±3- 5% change in the simulated delivered energy for heating and dhw.±10%variationintheshareof energyneedcoveredbyheat pump lead to a minor change of±1% in the simulated deliv- eredenergyforheatinganddhw.Whenvaryingtheelectricload by±10%,thetotaldeliveredenergychangeby±2%.
Finally,variationsof10%inthelifetimeparametersaveragelife- timeand shareof never demolishedbuildings donot leadtoany significantchangeinthemodelresults.Thisisaninterestingfind- ing asthese parameters areconsidered tobe highlyuncertain.
However,adifferencelargerthan±10%iswellpossibleforthese parameters.
Therearesomesignificantuncertaintiesinthemodel,butthe presentedsensitivityanalysisdemonstratesthatthemodelresults arenotverysensitivetochangesinthemostuncertaininputparam- eters.Further,weclaimthatitisbettertouseadetailedmodel, whereallimportantparametersarespecifieddirectlybasedonthe bestavailableinformation,ratherthantomakemoresuperficial andlinearassumptionsthatcannotbere-examinedinthesame
Fig.10.Sensitivityanalysis:Totaldeliveredenergy,relativetothebaselinescenarioin2015.
wayandarelesslikelytorepresentreality.Whenusingthedetailed model,alluncertaintiescanbeidentifiedandtheirrelativeeffecton thefinalresultsandconclusionscanbestudied,andthereadercan evaluatethereliabilityofthemodelresultsmoreeasily.Actually, theseaspectsaresomeoftheobviousstrengthsofmodelling,also insituationswhereonehastoacceptuncertaintyininputvalues andassumptions.
4. Conclusions
Thedevelopmentof theNorwegiandwellingstocksystemis highlydynamicandstronglydependsonthehistoryofthesystem.
Thereisingeneralaneedformodelsdescribingthedevelopment indwelling stocksin a realisticand detailedway. Thedynamic buildingstockmodeloutlinedinthisstudyisfoundusefulforthis purpose,andthemodelresultscorrespondverywellwiththeactual dwellingstockdevelopmentinthestatistics.Themodelisgeneric andcanalsobeappliedforothercountries,orfordifferentkindsof buildingstocks.Similarenergy-relatedinputdataaree.g.readily availableforthecountriesinvolvedintheEPISCOPEproject[30].
Thescenarioanalysisshowsthatimportantenergysavingshave already taken place through theshift to more efficient energy carriersandheatingsystems,andthishasimprovedalottheover- alltechnicalenergyefficiency oftheNorwegiandwellingstock.
Theeffectofthechangingenergymixisatthesamelevelasthe combinedeffectofenergyefficiencythroughrenovationandnew construction.Higheroutdoortemperatureshavealsoalreadylead toreductionsintheenergydemand.Thetotalsavingsduetochang- ingenergymix,energyefficiencyofthestockandwarmerclimate are,however,cancelledoutbythechanginguserbehavioranda growingstock,i.e.increasedfloorareademandperperson.
Theaverageestimatedenergydemandperm2wasrathersta- blefrom1960to2000whereas theenergy demandperperson increased.Inthesameperiod,therewasalsoalargeincreasein thepopulationandheatedfloorarea,leadingtoastrongincrease inthetotalenergydemand.
Thishistorical analysis hasshown that single-family houses havedominated thesysteminthepast,both in termsof num- berofdwellings,numberofsquaremetersheatedfloorarea,and
energydemand.Thelargestpotentialforfurtherenergysavings intheexistingstockisfoundinoldsingle-familyhouses.However, futureambitiousenergy-savingpolicieswillhavetotargetavariety ofdwellingtypesandagecohorts.
The sensitivity analysis showed that the model results are directly related to the input parameters that determine the dwellingstocksize andtheaverageenergy needintensity.This confirms our hypothesis that detailed and reliable models are neededbothforthedwellingstockdevelopmentandfortheenergy demand.Further,thesensitivityanalysisshowedthatthemodel resultswerelesssensitivetotheinputparametersthat wereof highestuncertainty,atleastwithintherangeofvariationthatwas usedinthesensitivityanalysis.Therefore,weconcludethatthe modelresultsarerobusttochangesintheseinputparametersand thattheoverallresultsofthismodelareofrelativelylowuncer- taintywhenappliedtothehistoricalanalysis.
Thedynamicbuildingstockenergymodelhasproventobesuit- ablefor explainingvarious importantissues relatedtothepast evolvingenergydemandintheNorwegiandwellingstock.Impor- tantchangesinthesystemarealsoexpectedduringthecoming decades,andwhenanalyzingfuturedevelopmentpathwaysthese canbetterbeunderstoodandimplementedinadynamicmodel compared to in a linear model. After now having successfully appliedand calibratedourdynamic segmentedbuildingenergy modeltothehistoricalenergydemandfrom1960to2015,wealso believethesamewayofmodelingmayserveasanexcellenttool foranalyzingpossibleeffectsoffutureenergy-relatedpolicies.This wouldberelevantforbothstrategiesandsolutionsinexistingand innewbuildings,andhow thesemayhelpreachambitiousand neededgreenhousegasemissiontargetstowards2050.
Acknowledgements
This paper is published as a result of participation in the EPISCOPEresearchproject(EnergyPerformanceIndicatorTrack- ingSchemes for theContinuousOptimisation ofRefurbishment Processes in European Housing Stocks), with co-funding from the ‘Intelligent Energy − Europe’ Programme, contract No.
IEE/12/695/SI2.644739).