HAL Id: hal-02180343 https://hal.archives-ouvertes.fr/hal-02180343 Preprint submitted on 12 Jul 2019 HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. How sensitive are optimal fully renewable power systems to technology cost uncertainty? Behrang Shirizadeh, Quentin Perrier, Philippe Qurion To cite this version: Behrang Shirizadeh, Quentin Perrier, Philippe Qurion. How sensitive are optimal fully renewable power systems to technology cost uncertainty?. 2019. hal-02180343
46
Embed
How sensitive are optimal fully renewable power systems to ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
HAL Id: hal-02180343https://hal.archives-ouvertes.fr/hal-02180343
Preprint submitted on 12 Jul 2019
HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.
How sensitive are optimal fully renewable power systemsto technology cost uncertainty?
Behrang Shirizadeh, Quentin Perrier, Philippe Qurion
To cite this version:Behrang Shirizadeh, Quentin Perrier, Philippe Qurion. How sensitive are optimal fully renewablepower systems to technology cost uncertainty?. 2019. �hal-02180343�
Manystudieshavedemonstratedthefeasibilityof fullyrenewablepowersystems invariouscountriesandregions.Yetthefuturecostsofkeytechnologiesarehighlyuncertainandlittleisknownabouttherobustnessofarenewablepowersystemtotheseuncertainties.Toanalyze it,webuild315long-termcostscenariosonthebasisofrecentprospectivestudies,varyingthecostsofkeytechnologies,andwemodel the optimal renewable power system for France over 18meteorological years, simultaneouslyoptimizinginvestmentanddispatch.
Ourresultsshowthatthetotalcostofa100%systemisnotthatsensitivetothepowermixchosenin2050.Certainly,theoptimalenergymixishighlysensitivetocostassumptions:theinstalledcapacityinPV,onshorewindandpower-to-gasvariesbyafactorof5,batteriesandoffshorewindevenmore.Butthe total cost will not be higher than today, and choosing a non-optimal electrical mix does notsignificantlyincreasethistotalcost.Contrarytocurrentestimatesofintegrationcosts,thisindicatesthatrenewabletechnologieswillbecomebyandlargesubstitutable.
Following Joskow (2011) and Hirth (2015), many articles have focused on the optimal proportion ofrenewable energies in the electricity mix. This literature has highlighted the existence of systemicintegration costs related to the deployment of variable renewable energies. In particular, a “self-cannibalization”phenomenonwashighlighted,linkedtothefactthatallthesolarpanelsinagivenfarmproducetheirelectricityatthesametime,justlikewindturbines.Intheabsenceofaffordablestorage,these integration costs have two consequences: (i) deployment of renewable energies leads to asignificant additional cost, rapidly increasingwith the deployment rate; (ii) the right balancemust bestruckbetweenthedifferentproductiontechnologiestominimizethisadditionalcost.
Theseresultshavedirectpolicyimplicationsinthetrade-offbetweenvisibilityandflexibility.Ontheonehand, investorswant visibility for thedevelopmentof economic sectors, suchasquantified targets intermsofinstalledrenewablecapacity.Ontheotherhand,ahighsensitivityoftheoptimalpowermixtotechnologycostsarguesforaflexibleapproach,allowingtrajectoriestobereadjustedaccordingtotheevolutionoftechnologies.Currentresultsthustendtosupporttheflexibilityapproach–attheexpenseofvisibilityforinvestors.
However, these results of increasing costs and right balancemight not holdmuch longer, due to therapiddeclineinproductionandstoragecosts. Inthespaceofsevenyears,thecostofsolarpanelshasbeenreducedbyafactorofseven,whilebatteriesnowseemtobefollowingasimilarpattern(Henze,2019).Moreover,recentwindturbinesbenefitfromaflatterproductionprofilethanoldermodels(HirthandMüller,2016).Finally,methanation,whichoffersanalternativeforseasonalstorage,isalsomakingsignificantprogress.Thesedevelopmentswillprobablystillbesignificantby2050,thepoliticalhorizonusedtodayinthedesignofpublicpolicies.
Whilethefeasibilityofa100%renewablemixhasalreadybeenhighlightedbymanystudies(Brownetal, 2018, and references therein), the question is now that of cost sensitivity: do these reductions inproduction and storage costs call into question the previous conclusions about the announced highadditionalsystemiccostofrenewableenergy?
If this phenomenon of increasing costs does not hold anymore, itwouldmean that the relationshipbetweenrenewableenergysourcesischangingfrombeingcomplementstobeingsubstitutes.Itwouldbe thenpossible to identifyoneor several “robust”energymixes, in the sense that theiroverall costdoesnotvarymuch,evenifthecostofthedifferenttechnologiesfinallydiffersfromtheinitialforecasts.In such a case, the political conclusion would shift: proving flexibility to investors through fixedrenewabletargetswouldprevailoverflexibleapproaches.
Toshedlightonthesequestions,webuildanewopen-sourcemodelcalledEOLES(EnergyOptimizationfor LowEmission Systems) and apply it to continental France. EOLESminimizes the total system costwhile satisfying power demand at each hour for a period of up to 18 years. It includes six powergeneration technologies (offshoreandonshorewind, solar, two typesofhydroandbiogas)and threestoragetechnologies(batteries,pumpedhydroandpower-to-gas).
TheEOLESmodel includessixpowergenerationtechnologies:offshoreandonshorewindpower,solarphotovoltaics (PV), run-of-riverand lake-generatedhydro-electricity, andbiogas combinedwithopen-cycle gas turbines. It also includes three energy storage technologies: pump-hydro storage (PHS),batteries andmethanation combinedwith open-cycle gas turbines. These technologies are shown inFigure1.
The remainder of this section presents themain equations (2.2) and the input data (2.3). A detaileddescriptionofallsets,parametersandvariablesofthemodelisavailableinAppendix4.
Theobjectivefunction,showninEquation(1), isthesumofallcostsoverthechosenperiod,includingfixed investment costs, fixed O&M costs (which are both annualized) and variable costs. For somestorageoptions,inadditiontotheCAPEXrelatedtochargingcapacityper!"#,anothertypeofCAPEXisintroduced:acapexrelatedtoenergycapacity,per!"ℎ#
where *+#, represents the installed capacities of production, :&;<=>?+@ is the volume of energystorage in MWh,'?+@ is the capacity of storage in MW, 1223456 is the annualized investment cost,C&&= andN&&= respectively represents fixed and variable operation andmaintenance costs andK+#,,Misthehourlygenerationofeachtechnology.
Adequacyequation
Electricity demandmust bemet for each hour. If power production exceeds electricity demand, theexcesselectricitycanbeeithersenttostorageunitsorcurtailed(equation3).
K+#,,M+#, ≥ SGT12SM + '(&UVK>?+@,M?+@ (3)
Where K+#,,M is the power produced by technology tec at hour h and '(&UVK>?+@,M is the energyenteringthestoragetechnologystrathourh.
Renewablepowerproduction
For each variable renewable energy (VRE) technology, the hourly power production is given by thehourlycapacityfactorprofilemultipliedbytheinstalledcapacityavailableforeachhour(equation4).
Energy stored by storage option str at hour h+1 is equal to the energy stored at hour h plus thedifference between the energy entering and leaving the storage option at hour h, accounting forcharginganddischargingefficiencies(equation5):
ThreetypesofoperatingreservesaredefinedbyENTSO-E (2013),according to theiractivationspeed.The fastest reserves are Frequency Containment Reserves (FCRs), which must be able to be on-linewithin 30 seconds. The second group is made up of Frequency Restoration Reserves (FRRs), in turndividedintotwocategories:afastautomaticcomponent(aFRRs),alsocalled‘secondaryreserves’,withan activation time of no more than 7.5 min; and a slow manual component (mFRRs), or ‘tertiaryreserves’,withanactivationtimeofnomorethan15min.Finally,reserveswithastartup-timebeyond15minutesareclassifiedasReplacementReserves(RRs).
Eachcategorymeetsspecificsystemneeds.ThefastFCRsareusefulintheeventofasuddenbreak,likea line fall, to avoid system collapse. FRRs are useful for variations over several minutes, such as adecrease inwind or PV output. Finally, the slow RRs act as a back-up, slowly replacing FCRs or FRRswhenthesystemimbalancelastsmorethan15minutes.InthemodelweonlyconsiderFRRs,sincetheyarethemostimpactedbyVREintegration.FRRscanbedefinedeitherupwardsordownwards,butsincetheelectricityoutputofVREscanbecurtailed,weconsideronlyupwardreserves.
The quantity of FRRs required to meet ENTSO-E’s guidelines is given by equation (6). These FRRrequirements vary with the variation observed in the production of renewable energies. They alsodependontheobservedvariabilityindemandandonforecasterrors:
The relationship between hourly-generated electricity and installed capacity can be calculated usingequation (7). Since the chosen time slice for the optimization is one hour, the capacity enters theequationdirectlyinsteadofbeingmultipliedbythetimeslicevalue.
K+#,,M ≤ *+#, (7)
Theinstalledcapacityofallthedispatchabletechnologiesshouldbemorethantheelectricitygenerationrequired of those technologies to meet demand; it should also satisfy the secondary reserverequirements Installedcapacity fordispatchable technologiescan thereforebeexpressedbyequation(8).
*g@@ ≥ Kg@@,M + U':g@@,M (8)
Monthly available energy for the hydroelectricity generated by lakes and reservoirs is defined usingmonthlylakeinflows(equation9).Thismeansthatenergystoredcanbeusedwithinthemonthbutnotacrossmonths.This isaparsimoniouswayofrepresentingthenon-energyoperatingconstraintsfacedbydamoperators,asinPerrier(2018).
o1!Gp ≥ Kkjq#,Mge@M∈p (9)
Where Kkjq#,M is the hourly power production by lakes and reservoir, and o1!Gp is the maximumelectricity that can be produced from this energy resource during one month. This parameter iscalculated by summing hourly power production from this hydroelectric energy resource over eachmonthoftheyeartocapturethemeteorologicalvariationofhydroelectricity,usingtheonlineportalofRTE1(theFrenchtransmissionnetworkoperator).
The energy that can be produced by biogas is limited, since the main resources of this energy aremethanization (anaerobicdigestion)andpyro-gasificationofsolidbiomass.Bothprocessesare limitedby several constraintsandaccording to theADEME“visions2030-2050” report (2013)electricity frombiogasproducedbythesetwoprocessescanbeprojectedas15TWhperyearfrom2030on(Gs\etj?
pj/ ),whichispresentedinequation10.
Ks\etj?,MuvwxMyz ≤ Gs\etj?
pj/ (10)
Run-of-river power plants represent another source of hydro-electricity power. River flow is alsostrongly dependent on meteorological conditions and it can be considered as a variable renewable
AsshowninFigure1,tworenewablegastechnologiesareconsidered;biogasandmethanation.Bothofthemproducerenewablemethane,whichcanbeused ingaspowerplants. Inthemodel, the latter isconsideredtobeanopencyclegas turbine (OCGT)dueto itshighoperational flexibilityandequation(12)showstherelationshipofthepowerproductionfromthesetwomethaneresources;
Ktj?,M = K,eps,M,eps (12)
WhereK,eps,M is the power production from each renewable gas resource, andKtj?,M is the powerproductionfromtheOCGTpowerplantwhichusesthesetworesourcesasfuel.Itisworthmentioningthat theefficiencyof this combustionprocess is considered inboth the15("ℎ# of yearlyelectricityproductionfrombiogas,andthedischargeefficiencyofthemethanationprocessasdefinedinequation(5).
The maximum installed capacity of each technology depends on land-use-related constraints, socialacceptance, the maximum available natural resources and other technical constraints; therefore, atechnologicalconstraintonmaximuminstalledcapacity isdefined inequation (13)where.+#,pj/ is thiscapacity limit, taken from the development trajectories for the French electricity mix for the period2020-2060(ADEME,2018):
*+#, ≤ .+#,pj/ (13)
Storage-relatedconstraints
Topreventoptimization leading toaveryhighamountof storedenergy in the firsthour representedandalowoneinthelasthour,weaddaconstrainttoensurethereplacementoftheconsumedstoredelectricityineverystorageoption(equation14):
The economic parameters for the power production technologies are taken from the EuropeanCommission Joint Research Center (2017) study of scenario-based cost trajectories to 2050, whileenergytechnologyreferenceindicatorprojectionsfor2010-2050(JRC,2014,havebeenusedforOCGTgaspowerplants.Valuesattributedtotheeconomicparametersofpowerproductiontechnologiesfor2050aresummarized inTable1. It isworthmentioningthat thegridentrycostof€25.9/kWforeachpower plantmandated by RTE (2018) has been added to the capital expenditure values of each VREtechnology, and the annuities (annualized CAPEX) are the results of these calculations. MoreinformationaboutthecostscenariosandthecostestimationmethodologyusedintheJRC’s2017studycanbefoundinAppendix1.
*Foroffshorewindpoweronmonopilesat30kmto60kmfromtheshore, foronshorewindpower, turbineswithmediumspecific capacity(0.3kW/m2)andmediumhubheight(100m)andforsolarpower,anaverageofthecostsofutilityscale,commercialscaleandresidentialscalesystemswithout tracking are taken into account. In this cost allocation,we consider solar power as a simple average of ground-mounted,rooftopresidentialandrooftopcommercialtechnologies.Forlakeandreservoirhydrowetakethemeanvalueoflow-costandhigh-costpowerplants.
13
Forthestoragetechnologies,the“CommercializationofEnergyStorageinEurope”reportpreparedbyFCH-JU(2015)andaveryrecentarticlebySchmidt(2019)about long-termcostprojectionsofstoragetechnologieshavebeenusedrespectivelyforpumpedhydrostorageandLi-Ionbatterystorageoptions.“The potential of Power-to-Gas” study by ENEA consulting (2016) has been used for methanationstorage. Using these three studies the 2050 cost projection of storage technologies are presented inTable2.ThecostofmethanationismadeupofthecostofelectrolysisunitsandtheSabatierreaction1.
Thecarbondioxiderequiredformethanationisassumedtocomefromcapturingandtransportingtheexcess carbondioxide resulting from themethanizationprocess (for theproductionofbiogas).About30%oftheproductofbio-methaneproductionfrommethanizationbyanaerobicdigestionisgasphasecarbon dioxide (Ericsson, 2017). According to ZEP (2011) on%&| transport, the cost of transportingcarbondioxidealonga200kmonshorepipelineis€4/5%&|.
Considering a 100km long onshore pipeline (considering maximum 100km of distance between themethanationunitsandthebiogasproductionunits),the%&|transportcostforthemethanationstorageis€1/MWh(Seeappendix5),tobeaddedtothegasstoragecostwhich is€2/MWh(accordingtoCRE(2018) - French energy regulation commission), the variable cost of the methanation storage is€3/="ℎ#.
Variable renewable energies’ (offshore and onshore wind and solar PV) hourly capacity factors havebeenpreparedusingtherenewables.ninjawebsite1,whichprovidesthehourlycapacityfactorprofilesofsolarandwindpowerfrom2000to2017atthegeographicalscaleofFrenchcounties(départements),followingthemethodselaboratedbyPfenningerandStaffell(2016)andStaffellandPfenninger(2016).These renewables.ninja factors reconstructed from weather data provide a good approximation ofobserveddata:Moraesetal.(2018)findsacorrelationof0.98forwindand0.97forsolarpowerwiththein-situobservationsprovidedbytheFrenchtransmissionsystemoperator(RTE).
To prepare hourly capacity factor profiles for offshorewind power,we first identifiedall the existingoffshore projects around France using the “4C offshore” website2, and using their locations, weextracted the hourly capacity factor profiles of both floating and grounded offshorewind farms.Wethenaveragedthemostremarkableprojectsforeachoffshorewindfoundationtechnology(floatingandgrounded)foreachyearfrom2000to2017.TheSiemensSWT4.0130hasbeenchosenastheoffshorewind turbine technology because of recent increase in the market share of this model and its highperformance.Thehubheightofthisturbineissetto120meters.
Appendix 2 provides more information about the methodology used in the preparation of hourlycapacityfactorprofilesofwindandsolarpowerresources.
Totestthesensitivityoftheoptimalpowertothecostsofvarioustechnologies,weconsidertherangeofuncertaintyindicatedinTable3.Forpowergenerationtechnologies,uncertaintyappliestothefixedcosts,definedascapitalcostsandfixedoperationandmaintenancecosts.Forstoragetechnologies, itapplies to themain cost component of each of them; fixed costs formethanation (similar to powergenerationtechnologies)andenergy-relatedCAPEXforbatteries.Forwindtechnologies,thechoiceofa+/-25%uncertaintyrangeratherthan+/-50%comesfromtheexpertelicitationsurveybyWiseretal.(2016).No variation in the cost of hydro and biogas is accounted for, the former because it is a maturetechnology with low uncertainty and the latter because in the model the amount of biogas used isdeterminedbytheavailabilityconstraint,notbyitscost.
AllthecombinationsofvariationspresentedinTable3wouldgive405differentcostscenarios(5Z×3å).Outofall theseoptions,weselect315scenarioswhichprovidehigher internal consistency. Indeed,afutureinwhichoffshorewindwouldbemoreexpensivethanexpectedandonshorewindcheaperthanexpected (or vice-versa) is not realistic, so we select only the scenarios in which the costs of thesetechnologies canonlydifferby25%atmost.This leads tosevendifferentoffshoreandonshorewindpower cost scenario combinations. Multiplying by five solar power cost scenarios and three costscenariosforeachstoragetechnology(7×5Z×3|),weobtain315futurecostscenarios.
Our results indicate that theoptimalenergymix ishighly sensitive to costuncertainty.Offshorewindoftenreacheseitherzeroinstalledcapacityorthemaximumallowedvalue,whiletherangeofonshorewind and PV capacities is approximately five-fold across the cost scenarios (Figure 3a). StoragetechnologiesalsodemonstratesuchhighsensitivitywiththeexceptionofPHSwhosecapacityisalwaysfixedbythemaximumallowedvalue.Batterycapacityrangesfrom7.6tomorethan279K"ℎ#,nearlyfourtimesthecapacity inthereferencecostscenario(Figure3d1),andmethanationrangesfrom7to33.5("ℎ,morethantwicethecapacityinthereferencecostscenario(Figure3d2).
This analysis also highlights some patterns of substitutability and complementarity betweentechnologies.Obviously,eachoption isparticularly influencedby itsowncost,butalsoby thecostofother technologies. In particular, a higher cost ofmethanation entailsmuchmore offshorewind and
16
vice-versa. Indeed, electricity fromoffshorewind suffers from a higher LCOE than other VREs but itsproduction is more stable, generating less need for storage. Conversely a higher cost of batteriesreducessolarcapacity:batteriesareespeciallyinterestingwhenenergymustbestoredforafewhours,sotheycomplementsolartechnology.
Finally, the system LCOE and the average power price are much more influenced by the cost ofgeneration technologies thanby thatof storage technologies1;Keeping the reference investmentcostscenarioforpowerproductiontechnologies,changingtheinvestmentcostofbatteryandmethanationstorageoptions from the lowest storage investment cost scenario (both -50%) to thehighest storageinvestmentcost scenario (both+50%)changes theoverall systemLCOE from€46/MWhto€51/MWh,whilechangingthe investmentcostof threeVREpowerproductiontechnologies fromthe lowestcostscenariotothehighestcostone(keepingthestorageoptionsatthereferencecostscenario),changestheoverallsystemLCOEfrom€37/MWhto58€/MWh.
1 Schlachtberger et al. (2018) find nearly no effect of storage cost variation on the final cost of the electricitysystem,whichisinaccordancewithourconclusions.
Globally,thepreviouscostsensitivityanalysisconfirmsthattheoptimalpowermixishighlysensitivetotechnologycosts.Thus,adecisionmakermightbetemptedtofavoraflexiblepolicyoveramorerigidone,attheexpenseofvisibilityforinvestors.However,ahighcostsensitivityoftheoptimalpowermixdoes not imply a high cost for choosing a non-optimal mix. In the case of highly substitutabletechnologies,asmallchangeincostwillleadtoastrongshiftintheoptimalmix,butchoosingonemixortheotherwouldnotchangetotalcostmuch.
The questionwe aim to answer in this subsection is the following: “Ifwe decide nowa trajectory ofrenewablecapacities for the futurebasedoncurrentcostestimates,could itentailahighover-cost ifour assumptions of technology costs are wrong”? To answer that question, we use the installedcapacities of generation and storage technologies optimized for the reference cost scenario, andwecalculatethesystemLCOEforthis“rigidcapacity”acrossour315costscenarios(Figure4).ThesystemLCOE is necessarily equal to or higher than that of the “flexible capacity”, the difference being the“regret”frombasingtheoptimizationonthewrongcostassumptions.
In most cases the regret is remarkably low given the wide range of cost scenarios considered: theaveragevalue is4% i.e.€2/="ℎ#, the thirdquartile is6%,and the regret isbelow9% in95%of thescenarios. A close examination of the 14 cost scenarios with the largest regret (more than 2 billion€/year, i.e. around 10%) shows that all but one concern scenarios in which the cost of onshore oroffshorewind, or both, is lower than expected. Hence the regret in this scenario stems from havinginstalled too little windpower. The only exception is a scenario in which PV and batteries are 50%cheaper than in the reference scenario, onshore at the reference cost, offshore 25% cheaper andmethanation50%moreexpensive.Inthiscaseonly,theregretstemsfromhavingnotinstalledenoughPVandbatteries.
TheaveragesystemLCOEfortherigidcapacity(blackvertical line inFigure4)equalsthesystemLCOEunder the reference cost scenario (red vertical dashed line). This result is due to the symmetricdistributionof technology cost shocks and the linear natureof themodel, and canbeunderstood asfollows:startingfromthesystemoptimizedoverthereferencecostscenario,atechnologycostshockby+25%entailsadecrease insystemLCOEby thesameamount (inabsolutevalue)asa technologycostshockby-25%,sotheaveragesystemLCOEisthesameastheonewithoutuncertainty.
Ourresultsshowthattheoptimalpowermixvariessignificantlyfromoneyeartoanother,bothintermsofelectricityproduction, installedcapacity, storagevolumeandstoragecapacity (Figures5and6andAppendix3).The largestvariationsbetweenminimumandmaximum installedcapacityareassociatedwith onshore and offshore wind power. In particular, offshore capacity ranges from zero to 20 GW,
20
which is themaximumvalue allowed1.High values foroffshorewindare reachedeither forweather-yearswithahighcapacityfactorforoffshorewind(asin2015)orforweather-yearswithalowcapacityfactor for onshorewind (as in 2016). In comparison, installed solar capacity ismore stable (between100.5GW and 122.2GW), due to a less volatile capacity factor (Figure 6c). Biogas always reaches themaximum allowed power generation and hydro the maximum allowed capacity. As far as storagecapacity is concerned, pumped hydro storage (PHS) also always reaches its maximum value whilebatteries andmethanation vary a lot acrossweather-years (Figures 6d1 and6d2). In comparison, thesystemLCOEandaveragepowerprice(thedualvariableoftheadequacyconstraint,i.e.equation3),aswellasthesumofVREcurtailmentandstoragelossesaremuchmorestable(Figures6eand6f).
These results show that if the aim is to find anoptimal energymix, running amodel on a randomly-chosenweather-yearcanbeverymisleading.Theoptimalmixof renewables ishighly sensitive to thechosenweather-year.ThisconclusionisconsistentwiththoseofCollinsetal.(2018)andZeyringeretal.(2018). As theweather of future years cannot be predicted, the best approachwould be to run themodeloverseveralweather-years,asinour18-yearsimulation.
Theselectionofarepresentativeyearcouldbemadeusingseveralcriteria.Wechosetoselecttheyearwith a capacity factor closest to our 18-year optimal mix.We used the capacity factor because it isinvariablewith respect to technologycosts,onwhichweperformthesensitivityanalysis.Tomeasurethedistancetothe18-yearoptimalmix,wecomputethesumofabsolutedifference1ofthethreeVREs.Usingthisapproach,2006istheclosestyeartotheoverall18-yearlongperiod,withasumofabsoluteerrorvaluesof1.5%(TableA.4).Welaunchedthemodelwiththeoptimalinstalledcapacitiesfoundfor2006 over all otherweather-years to test the adequacy of this installed capacitywith respect to theother17weather-years,andwedidnotobserveanyoperationalinadequacy.
Figure 7 shows the energy mix of the chosen representative year (2006) and the whole 18-yearmodelling.Thereisaveryclosematchbetweenthepercentageofeachenergysourcefortheoverall18-year-longoptimizationandtherepresentativeyear.Onshorewindpowerisclearlydominantwithsolarpowerandoffshorewindpowerasthesecond-andthird-biggestsourcesofenergyrespectively.
The system LCOE for power generation and storage ranges from €36 to €65/="ℎ#, depending ontechnology costs, with an expected value between €50 and €52/="ℎ#, depending on whether thepowersystemisoptimizedbeforeorafterthearrivalofinformationabouttechnologycosts.AccordingtothelatestquarterlyreportfromtheFrenchenergyregulator(CRE,2018),35%ofatypicalelectricitybill represents electricity production, hence from a bill varying between €160 and €170/="ℎ#, €56-€60/="ℎ# representsproduction.Hencethecostofa100%renewableelectricitysystemforFrancein2050wouldbelowerthanorsimilartothatofthecurrentpowersystem.
TheseresultscontrastwiththoseofKrakowskietal. (2016)whofindanannualizedcostofmorethan€60bn/yr. intheirscenario100RES2050(cf. theirFig.23)vs.€21bn/yr. inours.Theexplanationdoesnot stem from their investment cost assumptions, which are similar to ours (cf. their Table 1). Oneexplanationmight be that they take a higher discount rate, but theydonot disclose it sowe cannotverifythishypothesis.Partialexplanationsare(i)aslightlyhigherpowerdemand(cf.theirFig.7:about460("ℎ#/yr. vs. 422); (ii) a slightly lower capacity factor for onshorewind (28%) andoffshorewind(50%); (iii) the fact that they assume a perfect correlation between onshore and offshore windproduction,which artificially limits the complementarity between these technologies.Moreover, theybasetheirwindproductionprofilesonobservedpowergenerationin2012,whichneglectsthefactthatadvanced turbines generate electricity more constantly than those installed in the past (Hirth andMüller,2016).
Villavicencio (2017),who does not specify the time horizon considered, finds even higher annualizedcost:morethan€180bn/yr.for100%renewables,i.e.morethan8timesourresult.Severalfactorsmayexplainthishugedifference.First,hetakesarealdiscountrateof7%/yr.Thisismuchhigherthanours,whichcorrespondstotheraterecommendedforsocio-economicanalysisinFrance(4.5%).Second,hisinvestmentcost forPV ismuchhigherthanours:€3.6/"#,whilethecurrent investmentcostatutilityscaleisaround$1/"# (Lazard,2018).ThisexplainswhyPVdoesnotappearinhisreferencescenario(F1)with100%renewables.Third,totaldemandishigherthanours(512("ℎ# vs.422("ℎ#).
for 2050will be lower than that estimated by these studies. In the remainder of this subsection,weaddressseveralfactorsinturnwhichcouldpushourestimatesupordown.
4.2 Modellimitations
Factorswhichcouldpushcostsup
Costofthetransmissionanddistributionnetwork
OursystemLCOEincludesstorageandconnectionofpowergenerationtothegrid,butnotthecostofthetransmissionanddistributionnetwork.Currentlythiscostaccountsfor27%ofthetypicalelectricitybill,i.e.about€45/="ℎ#.Calculatingthiscostforthevariouspowersystemsconsideredinthepresentstudywouldexceed the scopeof thepresentarticle,but several recent studies indicate that thecostdifferentialacrossscenariosfeaturinggreaterorlesserpercentageofrenewableswouldbelimited.
• According to the RTE systems and network perspectives study (2018), for a 71% renewableelectricitymix(theso-calledWattscenariofor2035)inFrance,theextranetworkcostswouldbeintheorderof€1bn/yr., less than5%of the totalproductioncost.However, therelationship isnotlinearanditcannotbeeasilyextrapolatedforhigherproportionsofrenewables.
• AccordingtotwostudiesbyADEME(2015,2018),thecostofrenovatingtheFrenchnetwork,whichisplanned to takeplacebefore2030,willbeat leastoneorderofmagnitudemore than thecostrequiredtostrengthenthegridforafullyrenewablepowernetwork.
Ouroptimalscenariocorrespondingtothereferencetechnologycostsincludesabout80GWofonshorewind,12GWofoffshorewindand110GWofPV.TheavailabilityoflandforPVdoesnotappeartobeproblematical since the amount of suitable land is much higher than required (Cerema, 2017). Foroffshore wind, WindEurope’s “high” scenario for 2030 forecasts 11 GW, roughly equivalent to ouroptimal scenario corresponding to the reference technology costs. Here again, reaching this capacitydoesnotseemproblematical.
Somestudiesusehigherdiscountratesthanours,e.g.7%inVillavicencio(2017),asmentionedabove.ThiswouldincreasetheannualizedLCOE,andespeciallythecostofcapital-intensivetechnologies.Whilehigher ratesmaywellbeusedbyprivatecompanies,4.5% isalreadymuchhigher thanboththerate-free real interest rate available on financial markets, and expected GDP growth over the next fewdecades. Using a higher rate in a socioeconomic analysis means than future generations would bepenalizedwhencomparedtocurrentones,whichcanhardlybedefendedonethicalgrounds.
Perfectweatherforecasts
Our optimization has been conducted on the assumption that the weather is known for the wholeperiod.With imperfect weather forecasts, the cost would be higher, but such an optimization for acountry-scale system would be computationally challenging. Gowrisankaran et al. (2016) haveperformedsuchanoptimization just forsolarenergy,ona limitedgeographicalscale,andhavefoundthat“intermittencyoverall isquantitativelymuchmore important thanunforecastable intermittency.”However,whetherthisconclusionwouldholdforacomplex,multi-energysystemisanopenquestion.
Factorswhichcouldbringcostsdown
Demand-sidemanagement
Our model does not feature price-elastic electricity demand or flexibility in the power consumptionprofile,becausethiswouldhaverequireddebatableassumptions.Moreover,thedemandprofile,takenfromADEME(2015),isalreadyflatterthanthecurrentone.Includingthesefeatureswouldreducetheneedforstorageandtherelatedenergylosses.
Interconnectionwithneighboringcountries
Manystudieshaveshownthatinterconnectionswithneighboringcountriescansignificantlyreducethecostofafullyrenewablesystem.Forinstance,Annan-PhanandRoques(2018)haveshownthatpowerpricevolatilitycanbereducedbycross-borderexchangeswithneighboringcountries.Indeed,thisleadsto benefits from the differences both in climatic and weather conditions between the countriesconcerned.
26
Spatialoptimizationofrenewableenergycapacities
Asmentioned above, we do not optimize the quantity of renewables at every location but only theaggregate capacity,which is thus scaledup compared to thevalueobserved in2017.A lower systemcostwouldbeobtainedbyoptimizingtheirlocation,whichwouldpresumablyleadtogreatercapacityinwindier or sunnier locations, although this effectwould bemitigated by the need to obtain a flatteraggregategenerationprofile.Yetthiswouldmakethemodelcomputationallyintractableandmightleadtounrealisticconcentrationsofonshorewindinsomelocations.
Neithervehicle-to-gridnorsecond-handbatteries
Wehavenotconsideredvehicle-to-gridi.e.thepossibilitythatelectricvehiclebatteriescouldbeusedtoprovideflexibilityintheelectricitysystem.Yetthestoragecapacityofelectricvehiclesmaybehugeby2050:TheFrenchTSORTE (2018)estimates itat900("ℎ#,about tentimesthebatterycapacities inourreferencecostscenario.MobilizingevenasmallpartofthiscapacityforpowerstoragewouldbringdownthesystemLCOE,butwehavepreferrednottoincludethisoptionbecausetheimpactonbatterylifetimeisstillbeingdebated.Anotherpossibilityistorecycleusedcarbatteriesasstationarybatteries,but again, we believe that modeling this option would require precise assumptions on batterydegradation.
4.3 Conclusion
Inthisarticle,wehavestudiedthesensitivityofoptimalfullyrenewablepowersystemstotechnologycost.Tothatend,wehavedevelopedEOLES,amodeloptimizinginvestmentanddispatchinthepowersector, and applied it to the study of fully renewable power systems in France. We built 315 costscenariosbycombiningassumptionsaboutthelong-termcostofthekeypowergenerationandstoragetechnologies.
Ourresultsindicatethateventhoughthetechnologiesinvolvedinafullyrenewablepowersystemarevery different, they are by and large substitutable. For instance, if batteries are 50%more expensivethanexpected,theoptimalenergymixincludesfewerbatteriesandlessPV,butthisiscompensatedforby additional wind power, with a very limited impact on the system LCOE. On the contrary, if windpoweris25%moreexpensivethanexpected,theoptimalmixobviouslyincludeslessofthistechnology,butthisiscompensatedforbymorePVandstorage.
Overall, the impact of storage cost should not be overestimated: even in a 100% renewable powersystem,storage(batteries,PHSandmethanation)accountsforonly14.5%ofthesystemcost,vs.85.5%forelectricitygeneration.Wereourmodeltoincludedemand-sidemanagement,interconnectionswith
27
neighboring countries, vehicle-to-grid or second-hand batteries, the share of storage in overall costwouldbeevenlower.
Across all cost scenarios, the system LCOE, including generation and storage, ranges from €36.5 to€65.5/="ℎ#, depending on the cost scenario, with an average value of €50/="ℎ#. This is cheaperthantoday’svalue.Andsettingacapacity target inadvance forevery technologywouldonly increasethesystemLCOEby€2/MWhaveragedoverthe315costscenarioscomparedtotheoptimummix,evenifcostsvaryby+/-25%forwindand+/-50%forsolarandstorage. Intermsofpolicy implications, thisresultcallsforprovidingvisibilitytoinvestors,evenifitentailsreducingflexibilitythepolicydesign.
Finally, our analysis shows that theoptimalpowermix is highly sensitive to the chosenweather-yearand to the cost assumptions. In the literature, many analyses of the powermix are still based on auniqueweather-year, chosen fordataavailability rather than representativeness.Our result thuscallsforcautionoversuchconclusionsontheoptimalpowermix,whentheyarebasedonalimitednumberofweather-yearsorcostscenarios.
This work could be extended in many directions, for example including the other power generationtechnologies that entail low direct CO2 emissions: CO2 capture and storage and nuclear power. TheircostandthepossibilityofstoringmassivequantitiesofCO2beingveryuncertainintheFrenchcontext,wedecidednottoincludetheminthepresentstudy,buttheycouldbeconsideredinfuturework.
InthisJRCreport,historic installedcapacityofeachtechnologyfor2015, learningraterelatedtoeachtechnologyandthecapitalinvestmentcostofeachtechnologyin2015hasbeentakenasinputvalues,andusingthreedifferentfutureinstalledcapacityscenarios,threedifferentfuturecosttrajectoriesareproposed. Equation (A1) shows themainmethodology used in the cost projection using the learningratemethod:
%ëI5+ = %ëI5z ∙ Ç_Çì
î (A1)
This log-linear relation relates the future cost (%ëI5+) of a technology to the existing cost (%ëI5z),existinginstalledcapacity(%z)andthefutureprojectedinstalledcapacity(%+)ofitusingtheexperienceparameteri.ThelearningrateLR isrelatedtotheexperienceparameterasit isdescribedinequation(A2);
Scenario Baseline This scenario is used to cover the lower end of RES-E deployment. It is based on the
"6DS" scenario of the Energy Technology Perspectives published by the InternationalEnergyAgencyin2016.Itrepresentsa"businessasusual"worldinwhichnoadditionaleffortsare takenonstabilizing theatmosphericconcentrationofgreenhousegases.By2050, primary energy consumption reaches about 940 EJ, renewable energy suppliesabout30%ofglobalelectricitydemandandemissionsclimbto55GtCO2.
Diversified The "Diversified" portfolio scenario is taken from the "B2DS" scenario of theInternational Energy Agency's 2017 Energy Technology Perspectives and is used asrepresentative for themid-range deployment of RES-E found in literature. To achieverapiddecarbonizationinlinewithinternationalpolicygoals,allknownsupply,efficiencyandmitigationoptionsareavailableandpushedtotheirpracticallimits.Fossilfuelsandnuclear energy participate in the technology mix, and CCS is a key option to realizeemission reduction goals. Primary energy consumption is comparable to 2015 levels(about580EJ),theshareofrenewableelectricityintheglobalsupplymixis74%whileemissionsdeclinetoabout4.7GtCO2by2050.
ProRES The"ProRES"scenario resultsare themostambitious in termsofcapacityadditionsofRES-E technologies. In this scenario the world moves towards decarbonization bysignificantlyreducingfossilfueluse,however,inparallelwithrapidphaseoutofnuclearpower.CCSdoesnotbecomecommercialandisnotanavailablemitigationoption.Deep
The used economical parameters for the power production technologies are taken from the 2050projectionsofthisstudyforthediversifiedscenarioasanaverageandmorerealisticscenario.
1 System LCOE (levelized cost of electricity) is an economic assessment of the average total cost to build andoperateanelectricitysystemoveritslifetimedividedbytotalelectricityconsumptionoverthatlifetime.
TableA.5presentsthesetsandindicesoftheEOLESmodel,TableA.6theparameters.Throughoutthepaper,everyenergyunit(e.g.MWh)orpowerunit(e.g.MW)isexpressedinelectricity-equivalent.Forinstance, someenergy is stored in the formofmethane, tobe transformed later intoelectricityusingopen-cyclenaturalgasplantswith45%efficiency.Inthiscase,whenweindicatethat45="ℎ# isstoredinthenaturalgasnetwork,itmeansthat100MWhofmethaneisstored,whichwillallow45="ℎ# ofelectricitytobegenerated.
The cost of transporting carbon dioxide along a 200km onshore pipeline is€4/5%&|, for 100km lingpipeline, this transporting cost can be assumed around €2/5%&|. Given that each mole of carbondioxideweighs 44 grams, andwe can produce onemole ofmethane fromonemole of%&|with anefficiency of 80% and eachmole ofmethane can produce 802.3kJ of thermal energy, considering anOCGTcombustionefficiencyof45%(JRC2014):
Z+ÇúùZzzzzzztÇúù
×ååtÇúùZpekÇúù
×ZpekÇúùz.upekÇöü
×ZpekÇöüuz|.êq†
×Zq†+M
z.zzz|||vvvvuq°M+M×
Zq°M+M
z.åwq°M#k#,×Zzzzq°M#k#,
Z¢°M#k#,=
0.5486+Çúù
¢°M#k#,
Multiplying this transport cost by €2/5%&|, the %&| transport cost for methanation becomes€1.0972/="ℎ.
InTableA.8wesummarizetheyearlypowerproductionLCOE(thelevelizedcostofelectricityproducedfromeach power plantwithout considering any future load curtailment or other losses) and averagesellingprice foreachgeneration technologyandLCOS (levelizedcostof storage;cf. Jülchetal.,2015)andtheaveragesellingpriceofeachstoragetechnologyforweather-year2006.
For all power production technologies, the averagemarket selling price is higher than the LCOE, thedifference being very low for the VRE resources (offshore wind ~€0.10/="ℎ#, onshore wind~€0.15/="ℎ# and solar PV ~€0.40/="ℎ#) while for the other technologies this difference ismuchgreater, especially for biogas (€44.40/="ℎ#). The profitability of hydro is due to the capacityconstraint,while for biogas it is due to the production constraint, since these constraints generate ascarcityrent.
While theprofitabilityanalysis is straightforward forall thepowerproduction technologies, it ismorecomplicatedforthestoragetechnologiessincetheybuyelectricityfromthemarket,andtherearelossesrelated tocharginganddischarging inefficiencies.Equation (17) shows theprofitabilitycriteria for thestoragetechnologiesinthecalculationofunitprofit:
electricity bought by the storage units, charged and sold on the electricity market (accounting forstorage-related inefficiencies), which can be considered as the net present value of each storagetechnologyperunitofpowersold.