Demand Response and Energy Storage Integration Study

March 2016

i

Notice:ThisreportisbeingdisseminatedbytheU.S.DepartmentofEnergy(DOE). Assuch,thisdocument

waspreparedin compliancewithSection515oftheTreasuryandGeneralGovernmentAppropriationsAct

forFiscalYear2001(Public Law106-554)andinformationqualityguidelinesissuedbyDOE. Thoughthis

reportdoesnotconstitute“influential”information,asthattermisdefinedinDOE’sinformationquality

guidelinesortheOfficeofManagementand Budget’s InformationQualityBulletinforPeerReview,the

reportwasreviewedbothinternallyandexternallypriortopublication. Thisreporthasbenefittedfrom

reviewbytheStudyTeamthatincludesresearchersfromLawrenceBerkeleyNationalLaboratory(LBNL),

National Renewable Energy Laboratory (NREL), Oak Ridge National Laboratory (ORNL), and Sandia

NationalLaboratory(SNL).Inaddition,thisreportalsoincorporatesinputfrom16externalpeerreviewers

representingelectricutilityindustry,academia,andthebroaderstakeholdercommunity.

ii

AcknowledgmentsTheDemandResponseandEnergyStorageIntegrationStudywassponsoredbytheU.S.DepartmentofEnergyOfficeofEnergyEfficiencyandRenewableEnergyandOfficeofElectricityDeliveryandEnergyReliability.Thestudyrepresentsajointmulti-NationalLaboratoryefforttoexaminetheroleofdemandresponse and energy storage in electricity systems with different penetration levels of variablerenewableresourcesandtoimprovetheunderstandingofassociatedmarketsandinstitutions.

WewouldliketothankHenryKelly,CarlaFrisch,BillParks,andImreGyukfortheirsupport.Anyerrorsoromissionsaresolelytheresponsibilitiesoftheauthors.

AuthorsOfficeofEnergyEfficiencyand OokieMaRenewableEnergy

OfficeofElectricityDeliveryand KerryCheungEnergyReliability

StudyTeamLawrenceBerkeleyNationalLaboratory DanielJ.Olsen,NanceMatson,MichaelD.Sohn,CodyM.

Rose,JunqiaoHanDudley,SasankGoli,andSilaKiliccote

PeterCappersandJasonMacDonald

NationalRenewableEnergyLaboratory PaulDenholm,MarissaHummon,JennieJorgenson,andDavidPalchak

OakRidgeNationalLaboratory MichaelStarkeandNasrAlkadi

SandiaNationalLaboratories DhruvBhatnagar,AileenCurrier,andJaciHernandez

Consultant BrendanKirby

UniversityCollegeDublin MarkO’Malley

(This page intentionally left blank)

iv

ExecutiveSummary

MotivationandBackgroundDemand response and energy storage resources present potentially important sources of bulk powersystemservicesthatcanaidinintegratingvariablerenewablegeneration.Whilerenewableintegrationstudieshaveevaluatedmanyofthechallengesassociatedwithdeployinglargeamountsofvariablewindand solar generation technologies, integration analyses have not yet fully incorporated demandresponseandenergystorageresources.

Thisreportrepresentsaninitialeffort inanalyzingthepotential integrationvalueofdemandresponseandenergystorage,focusingonthewesternUnitedStates.Itevaluatestwomajoraspectsofincreaseddeploymentofdemandresponseandenergystorage:

(1) Theiroperationalvalueinprovidingbulkpowersystemservices(2) Marketandregulatoryissues,includingpotentialbarrierstodeployment.

Keyquestionsinclude:

• WhatistheregionalandtemporalavailabilityofdemandresponseresourcesinthewesternUnitedStates?

• What is the relativeoperationalvalueofdemandresponseandenergystorage inprovidingbulkpowersystemservices?

• Howdoestheoperationalvalueofdemandresponseandenergystoragevaryasafunctionofdeploymentandrenewablepenetration?

• Whatarepotentialbarrierstodeploymentincurrentmarketandregulatoryenvironments?

Theworkcontainedinthisreportstemsfromanumberofmoredetailedstudiespublishedbymembersof the study team. Links to these publications as well as related materials can be found atenergy.gov/eere/analysis/demand-response-and-energy-storage-integration-study.

Thisstudydoesnotattemptto:

• Assessallpotentialmechanismsandsourcesofdemandresponse• Assessallpossiblevaluestreamsofdemandresponseandenergystorage• Consider the costs associatedwith the deployment and integration of demand response or

energystorage• Considerenergystoragetechnologiesdeployedindistributionsystemsorbehindthemeter• Simulatecontingencyeventsorconsidertheimpactsofchangingsystemdynamics• Determinetheoptimalsizingorlocationofdemandresponseorenergystorage.

OverviewofDemandResponseandEnergyStorageDemand response and energy storage resources can be obtained from a number of differenttechnologies.Whilethesetechnologiescanprovidearangeofvaluestreamstodifferentstakeholders,for thepurposeof supportingbulkpowersystemoperations, theyhave thecommoncharacteristicof

v

beingable toshiftenergyuse in timetohelpmaintain thegeneration-loadbalance.Assuch,demandresponseandenergystoragetechnologiesareevaluatedwithacommonframeworkinthisstudy.

Demandresponseencompassesmanydifferentstrategiesbywhichcommercial,residential,municipal,and industrial electricity customers are incentivized to adjust, in the short-term, when they useelectricity (in contrast to energy efficiency and energy conservation that seek to reduce total electricload).Similarly,energystoragetechnologieslikepumpedstoragehydropower,batteries,andflywheelssaveelectricityproducedatonepoint in time for lateruse,effectivelyshiftingdemand.Bothdemandresponse and energy storage technologies can be used to provide energy services and/or ancillaryservices such as frequency regulation and contingency reserves. A key difference between demandresponseandenergystorageisthattheuseofdemandresponseisinherentlytiedtospecificend-useswith associated temporal and spatial patterns of electricity consumption. This difference also hasimplicationsfortheavailabilityofdemandresponseresourcesovertime.

SourcesofValueThis study focuses on assessing two sources of value that demand response and energy storage canprovidetobulkpowersystemoperations:energyservicesandoperatingreserves.Thevalueofenergyservices reflects the variable costs of operating thepower system,which are primarily fuel costs andvariable operations andmaintenance costs associated with committing and dispatching a generator.Operating reserves, part of a larger class of services known as ancillary services, are used to ensureelectric system balancing for short-term variations between generation and load, including during acontingency (large, sudden,andunexpected lossofgenerationor transmissioncapacity).Thevalueofoperatingreservesreflectsacombinationofdifferentcostcomponents:

(1) Incremental fixed costs associated with generation equipment required to provide operatingreserves

(2) Operationalcostsassociatedwithoperatinggeneratorsatpartload(e.g.,efficiencylosses)andresponding to short-term variations (e.g., increased wear and tear, maintenance, time spentofflineforrepairs)

(3) Opportunitycostsassociatedwithoccurrenceswhenageneratorwithholdsenergyproductioninordertosupplyoperatingreserves(e.g.,generatorlostprofits).

Threeoperatingreserveproductsareevaluatedinthisstudy:frequencyregulation,contingencyreserve,andrampingreserve.

StudyApproachandLimitationsThis study performs a set of power system simulations of thewestern United Stateswhere demandresponse and energy storage resources are deployed under differing levels of wind and solarpenetration. For simplicity, the deployment of these resources was studied without changing thegenerator composition or transmission capacity in the base scenarios. This approach may lead todifferentoperationalvaluescomparedtooneswithhigheror lowerreservemargins.Additionally, thisstudydoesnotconsidertheupfrontcapitalcostsneededtodeploytheseresourcesorthebenefitsofdeferred asset investments. The power system simulations hold capacity hour-by-hour for operating

vi

reserves but donot explicitlymodel grid resources providing frequency regulation and responding tocontingencyevents.Asa result, this studydoesnotconsider thepotentiallyhighervalueof resourcesthat canmore accurately follow frequency regulation dispatch or respond to real-time, unforecastedconditions.Themodeleddeploymentsofdemandresponseandenergystorageresourcesareevaluatedseparatelyandcomparedagainstcommonbasecases.

Theoperationalvalueofdemandresponseandenergystorage isquantified in twoways representingtwoperspectives.Inthefirstapproach,thedifferenceintotalcostforoperatingthesystemforastudyyear (i.e., totalproductioncosts)between twomodeledscenarios isused toestimate theoperationalvalueofaddingvariousamountsofdemandresponseorenergystorage.Totalproductioncostsavingsrepresentasocietalvaluederivedfromavoidedfuelandoperationsandmaintenancecostsacrosstheentiresystem.Inthesecondapproach,theoperationalvalueofdemandresponseandenergystorageisestimated by using short-runmarginal costs of production for different services calculated from thesimulations. The marginal cost calculation helps disaggregate operational value for the differentservices, hour-by-hour, and for different geographic locations. We further assume that short-runmarginalcostsareaproxy forhourlyprices forbulkpowersystemservicesandrepresent thevalueamarketparticipantmightreceive,hypothetically,ifsellingservicesintoanorganizedwholesalemarket.

ThisstudyevaluatesdemandresponseandenergystoragedeployedintheWesternInterconnectionof

FigureES-1.Studyareaincluding36balancingauthorities(smallprint)and12reservesharinggroup(largeprint)assumptionsforthe2020studyyear

vii

theUnitedStates,illustratedinFigureES-1.ThemodelingapproachisbasedontheWesternWindandSolar Integration Study Phase 2 (WWSIS-2) (Lew et al. 2013). This includes two base cases: a lowrenewablecasewith14%ofelectricityfromwindandsolarpowerandahighrenewablecasewith33%of electricity fromwind and solar power. Consistent with theWWSIS-2, the transmission capacity isincreased from the low renewable case to the high renewable case to accommodate the differentgeneration profiles. Simulations over the full 2020 study year are conducted in the production costmodel PLEXOS. Production cost models calculate various costs of system operation, including fuel,operations andmaintenance, and generator starts; however, they do not consider or include capitalcosts.Theresultsarealsobasedsolelyontheday-aheadsimulation.Thisapproachtakes intoaccountwindand solar forecasterrorsbetween theday-aheadand real-timedispatchesbyholdingadditionaloperating reserves (frequency regulation and ramping reserve) to meet anticipated increases invariabilityanduncertainty.However,thereal-timedispatchisnotsimulated.

DemandResponseandEnergyStorageDeploymentScenariosInthisstudy,wemodelonedemandresponsedeploymentscenarioandasetofdeploymentscenariosfor two general classes of energy storage technologies. The two energy storage technology classesinclude an operating reserves-only device and one that can be co-optimized for both energy andoperatingreserves.Energystoragetechnologiesareassumedtobeconnectedatthetransmissionlevel.Customer-sitedelectricenergystorage(e.g.,batteries)isnotconsideredinthisanalysis,whilecustomer-sited thermal energy storage (e.g., electricwaterheaters, building thermal capacity) is categorizedasdemandresponseresources.Thesedeploymentscenariosaremodeledindependentlybutusethesameanalysis framework. For simplification, optimal placement,optimal sizing (capacity andduration), andcapital costs of these resources are not considered. While these limitations impact the accuracy ofsimulatedoperationalvalues,theobservedtrendsareindicativeofexpectationforscenarioswithhigherpenetrationofvariablerenewableresources.

Wedevelopascenariofortheavailabledemandresponseresourcebasedonanalysisofdifferentend-uses across commercial buildings, residential buildings,municipal functions, and the industrial sector.Theestimatedresourceiscalculatedbyassessingthefractionofeachend-useelectricloadthatcanbeutilizedfordemandresponsebasedonassumptionsregardingthephysicalconstraintsoftheunderlyingend-use appliance and equipment systems. Service constraints, the deployment of suitable controlsystems, and historic rates of retail customer participation in demand response programs were alsofactored into the scenario development. End-uses included in the analysis constitute 30% of totalelectricity use in the western United States. However, after accounting for anticipated participationrates,only4%oftotalelectricityuseisassumedtobeenrolledindemandresponseprograms.Thisleadstoastudyscenariowithanannualcumulativeavailabilityof11.3TW-h,or1.4%oftotalelectricityuse.Imposingonlythephysicalandserviceconstraints,assumingfullparticipationandadoptionofnecessarycontrol systems, increases the estimated resource size by a factor of 8. In our approach, this largerquantityrepresentsthetechnicalpotentialfordemandresponseinthestudyscenario.

Thefirstclassofenergystoragetechnologymodeledisanoperatingreserves-onlydevicethatresemblesashortdurationbatterywith1hourofenergystorageat ratedcapacityand80%roundtripefficiency(similartoalithium-ionbattery).Thisfirstclassofmodeleddevicesissizedtomeet50%oftheaverage

viii

hourlyoperating reserve requirementwithineach region.One setof scenarios implements frequencyregulation-onlydevicesatatotaldeploymentof487MWinthelowrenewablecaseand597MWinthehighrenewablecase.Theothersetofscenariosimplementscontingencyreserve-onlydevicesatatotaldeploymentof1,314MWinbothrenewablecases.

Thesecondclassofenergystoragetechnologymodeledisadevicecapableofprovidingbothenergyandoperating reserves, resembling a batterywith 8hours of storage capacity at ratedpoweroutput and75% roundtripefficiency (similar toa sodium-sulfurbattery). The simulations include scenarioswherethedevicesprovideonlyenergyandalsoscenarioswherethedevicesareco-optimizedtoprovidebothenergy and operating reserves. Energy storage in these scenarioswere sized to equal about 3.3% ofaverageloadineachregion,resultinginatotalmodeleddeploymentof3,045MW.

ModelingResults

ValueofEnergyandOperatingReservesThevalueofdemandresponseandenergystorage isassessedrelativetothesystemcharacteristics inthebaselinescenarios.Inthebasesystemwithoutadditionaldemandresponseorenergystorage,theannual operational costs determined fromproduction costmodeling are about $11.5billion and$6.5billion in the low and high renewable cases, respectively. Production costs are lower in the highrenewablecasebecausetherenewablegenerationhasnodirectfuelcosts.

FigureES-2illustratesthatmostofthesimulatedproductioncostsareassociatedwiththeprovisionofenergy.Onlyabout1.0%($115million)ofthesecostsinthelowrenewablecaseand1.8%($118million)in thehighrenewablecaseareassociatedwith theprovisionofoperating reserves.Despite the lowerenergyproductioncostsinthehighrenewablecase,totalcostsforoperatingreservesarehigherduetothe greater operating reserve requirements needed to accommodate the increased variability anduncertainty.

FigureES-2.FortheWestern Interconnectionmodel,totalproductioncosts(left)andproductioncostsassociatedwithjusttheprovisionofoperatingreserves(right)

ix

Productioncostsassociatedwithoperatingreservescanbedistinguishedbygeneratoroperatingmode.Steady state costs are associated with fuel and variable operations and maintenance costs forgenerators producing electricity at a fixed operating point; startup and shutdown costs are costs forstarting and stopping a generator; and non-steady state costs are additional operations andmaintenancecoststoprovidefrequencyregulation.Whilecostcomponents likegeneratorstartupandnon-steadystateoperationarerelativelysmallcomparedtototalproductioncosts,theyaresignificantcomponents for the production costs associated with operating reserves. Avoiding these high costcomponentsisanimportantsourceofvaluefordemandresponseandenergystorage.

OperationalValueofDemandResponseFigureES-3showstheoperationalvaluesassociatedwiththedemandresponseresourcestudyscenariodeployedforthelowandhighrenewablecasesinthestudyyear.Thetotalproductioncostsavings(i.e.,the difference in production costs between scenarios with and without added demand responseresources) is shown on the left and the implied market value (i.e., marginal cost of every servicemultipliedby theamountprovidedbydemandresponseresources, aggregatedacrosseveryhourandeveryregion)isshownontheright.Inbothrenewablecasesandforbothperspectives,theoperationalvaluesarenormalizedbytheestimatedcumulativeavailabilityofdemandresponseresourcesusedforbulkpowersystemservices.

FigureES-3 (left) shows that themajorityofoperational savingsachieved through theuseofdemandresponsecomesfromavoidingsteadystatecosts.However,ahighfractionofsavingsalsocomesfromavoided generator startups and shutdowns and avoided non-steady state operational costs. The graybars in Figure ES-3 (right) show the impliedmarket value for energy shifting in terms of load sheds(positive energy value) and load recovery (negative energy value) at the assumed level of demandresponse deployment. While energy shifting constitutes the majority of gross market value, theprovisionofoperatingreservesconstitutesthemajorityofnetvalueafterconsideringthecostof load

FigureES-3.Operationalvalueofdemandresponseresourcesintermsoftotalproductioncostsavings(left)andtheimpliedmarketvalue(right)inthelowandhighrenewablecases.

x

recovery(approximatelyone-thirdofgrossmarketvalue).Fromboththeperspectiveoftotalproductioncostsavingsandtheimpliedmarketvalue,theper-unitoperationalvalueofdemandresponseislowerforthehigherrenewablecase.Thistrendstemspartlyfromthereductioninenergyproductioncostatincreased levels of renewable generation, which decreases steady state savings and implied energymarketvalue,andpartly fromthe increasedvariabilityassociatedwithwindandsolar,which leads tofewersavingsachievablefromavoidinggeneratorstartsandstopsneededtomeetthelargeroperatingreserverequirements.

Thedemandresponseresourceinthestudyscenarioshiftsupto0.8%ofaveragedailyenergyuseandreduces up to 1.9% of load during the top 100 load hours in the study year. In themodel, demandresponse resources also provide a large fraction of operating reserves, meeting 39% of frequencyregulation requirements in the low renewable case and33% in thehigh renewable case, and 22%ofcontingencyreserverequirementsinboththelowandhighrenewablecases.

System-specificissuessuchaslocalgenerationmix,loadprofiles,resourcecapabilities,andtransmissioncapacity cause significant variations in operational values by demand response resource type and byregion.FigureES-4provides the impliedmarketvalueofvariousdemand response resource types fordifferent services in themodeledNewMexico zone. Theoperational valueof eachdemand responseresourcetypedependsonthecorrelationofresourceavailabilitywithtimesofhighproductioncostsaswellastheflexibilityinschedulingenergyuse.Withincreasedpenetrationofwindandsolargeneration,certaindriversimpactnetvalueinbothdirections.Forexample,FigureES-4showsthatsomeresources,suchas residential space cooling, see increasednet value.Other resources, suchasdata centers, seedecreased net value. Greater penetration of wind and solar generation tend to reduce themarginalcostsofcontingencyreserve(greenbars)andincreasethemarginalcostsoffrequencyregulation(bluebars).Theadditionalsoaltersthemarginalcostsofenergyduringtimesofpeakloads,whichcanhaveimportant implications for the value of energy shifting (gray bars). Load sheds tend to have reducedvalue in thehigh renewablecase;however, thecostsof load recoveryalso tend tobe lower.Thenetenergyarbitragevaluecanbehigherorlowerdependingontheindividualresourcetypeandtheregion.

FigureES-4.OperationalvalueofdifferenttypesofdemandresponseresourcesinNewMexicoZoneatlow(L)andhigh(H)levelsofwindandsolargeneration,basedonmarginalcosts

xi

OperationalValueofEnergyStorage

FigureES-5showstheoperationalvalueofthefirstclassofenergystoragedevicesmodeledforthelowandhighrenewablecases. Inonesetof scenarios,energystoragedevicescanprovideonly frequencyregulation. In theother set, energy storagedevices canprovide only contingency reserve. The left ofFigureES-5showstheoperationalvalueoftheenergystoragedeployedastotalproductioncostsavings,andtherightrepresentstheimpliedmarketvaluebasedonsimulatedmarginalcosts.Inbothrenewablecasesandforbothperspectives,theoperationalvaluesarenormalizedbytheenergystoragecapacitydeployedfortherespectivescenarios.

Comparedtothelowrenewablecase,theresultsforthehighrenewablecaseshowanincreaseintheper-unit value for the frequency regulation-only storage devices but a decrease for contingency-onlystorage devices for both perspectives. The trends observed are partly from lower energy productioncosts at increased levels of renewable generation, which decrease steady state savings in thecontingency-onlyscenario,andpartlyfromtheincreasedneedforfrequencyregulation,whichleadstogreatersavingsinthefrequencyregulation-onlyscenario.

FigureES-6provides the simulation results for the secondclassofenergy storagedevicesmodeled inenergy-only and co-optimized scenarios, forboth the low renewable andhigh renewable cases.As inFigureES-5,theleftpanelprovidesthetotalproductioncostsavingsassociatedwiththeadditionofthisclassofenergystoragedevicesnormalizedbytheamountdeployedwhiletherightpanelprovidestheimpliedmarketvaluenormalizedbytheamountdeployed.Asshownintheresults,addingtheabilitytoco-optimizeenergyandoperatingreservesincreasestheoperationalvalueofenergystoragedevices.

FigureES-5.Theoperationalvalueofenergystorageintwosetsofmodelsruns,oneinwhichenergystoragecanprovideonlyfrequency regulation and another in which energy storage can provide only spinning contingency reserves. Value isdetermined by total production cost savings to the system (left), and the implied market value calculation based onsimulatedmarginalcosts(right).Make-uplossesareenergycostsassociatedwiththestorageefficiencylosses.

xii

ComparingFiguresES-5andES-6showsthattheper-unitvalueoftheco-optimizeddevicesislessthanthefrequencyregulation-onlydevices.Thiscounterintuitiveresultispartlyduetothedifferentamountsofenergystoragedeployedinthedifferentscenarios.Thetotaldeploymentofco-optimizeddevicesismuch greater in size than the frequency regulation-only devices. Because the incremental value ofenergy storage tends to declinewith additional deployment (aswith any resource, including demandresponse), the per-unit value tends to be smaller for scenarios with larger deployments of energystorage.FigureES-7showsthemodelingresultsoftheco-optimizeddevicesforthree levelsofenergystoragedeployment,withanapproximatecurvefittoillustratethisrelationship.

Furthermore,thelargeenergycapacity(8-hourdurationattheratedpower)oftheco-optimizeddevicesintroduces additional drivers that impact the modeled operational value. While the devices in thefrequency regulation-only scenario (sized tomeet 50% of the average hourly operating reserves) arenearlyfullyutilizedproviding49%ofthesystemrequirement,thedevicesintheco-optimizedscenario,althoughsignificantlylarger,provideonly27%ofthefrequencyregulationrequirement.Whenprovidingfrequencyregulation,theco-optimizeddevicescanprovideotherserviceslikecontingencyreservesandenergy shifting with the remaining capacity. When discharging energy, the energy storage devicesuppressesthemarginalcostofenergy.Displacedgenerationcanoffercapacityasoperatingreserves,which also suppresses the marginal cost of operating reserves. As a result, the larger co-optimizeddevicesmodeledprovideacombinationofserviceswhoseper-unitvalueisactuallylessthanthemuchsmallerfrequencyregulationdevices.

Figure ES-6. The operational value of energy storage (co-optimized for energy and operating reserves) in terms of totalproductioncostssavings(left)andtheimpliedmarketvalue(right)inthelowandhighrenewablecases

xiii

As demonstrated by the demand response results, there can be significant regional variation inoperationalvalueforenergystorageinboththelowandhighrenewablecases.FigureES-8providesanindicationofthisvariationbyshowingtheimpliedmarketvalueforvariousservicescalculatedbyusingsimulatedmarginal costs of production. The analysis results indicate an increase in value for energy

FigureES-7.Theper-unitoperationalvalueoftheco-optimizedenergystoragedevicesasafunctionoftotaldeploymentinthelowandhighrenewablecases.Costsavingsrepresenttotalproductioncostsavings;marketvaluerepresentstheimpliedmarketvaluebasedonmarginalcostsofproduction.

FigureES-8.Theoperationalvalueoftheco-optimizedenergystoragedevicesforthevariousservicesinthedifferentregionsbasedonmarginalcostsofproduction

xiv

storageinthepresenceofgreaterdeploymentofwindandsolarpower.Theincreaseinenergystoragevalue observed may be attributed to the greater need for operating reserves and the tendency ofrenewable generation to suppress off-peak energymarginal costs. Lower off-peak energy productioncosts increase thevalueofenergy shifting, as canbe seenby thegreater steady state cost savings inFigure ES-6. This second factor is less evident for many demand response resources that haveconstraintsonavailability.Forinstance,thereislimitedthermalinertiainbuildings,socoolingloadsmaynotbeabletoshifttohourswiththelowestenergycosts.

TheImportanceofCapacityValue

Productioncostmodelscanprovideestimatesoftheoperationalvalueofdemandresponseandenergystoragebutdonotconsiderthevalueofprovidingsystemcapacity.Theabilitytoreplaceconventionalgeneration is a potential source of value and, in some cases, may be substantially greater than theoperational value. This study does not attempt to directly quantify the value of capacity for demandresponseorenergystorageresourcesduetomodellimitationsandassumptionsusedinthedeploymentscenarios. Additionally, the issue of persistence and availability of demand response over timecomplicatesthediscussionofcapacityvalue.Productioncostmodelingdoesnotdispatchresourcesforcontingencyeventsanddonotconsiderlong-termplanningneedsforsystemadequacy.

In the Western Interconnection model, the average production cost savings resulting from energystorage devices (co-optimized for energy and operating reserves) ranges from $33–$41/kW-year in ascenariowhere3.0GWofenergystorageisdeployed.Forthesakeofasimplefirst-ordercomparison,capacity values for non-generation resources, such as demand response or energy storage, can beapproximatedbasedonaproxygenerationresourcelikeanaturalgas-firedcombustionturbine;thoughthecapacityvaluemaynotbeequivalent.Thereisalargerangeofestimatesfortheannualizedcostofanew combustion turbine, with examples ranging from a lower value of $77/kW-year (taken from aColorado filing) to a higher value of $212/kW-year (taken from a California filing). This implies thecapacity value of an energy storage device could be considerably higher than its operational value,assumingthedevicecanprovidecapacitybenefitsandisdeployedinaregiondeficientincapacity.

Similarly, the capacity value of demand response could be significantly greater than its operationalvalue. The total production cost savings of the modeled demand response resource is about $110million per year. The total estimated reduction in peak load from utilizing the demand responseresourceisabout3.1GW(basedontheabilityfordemandresponsetoshiftenergyuseduringthetop100 hours of each balancing authority area),whichwould translate to $240–$660million per year inavoidedcapacity,assumingtheproxyresourceisanaturalgas-firedcombustionturbine.

While capacity value for non-generation resources can be quite significant, realizing this value mayrequire capacity providers to be capable of several hours of response duration—6–10 hours in someregions. Additionally, the availability and persistence of the resourcemay need to be assured acrossmultiple years to address concerns regarding resource adequacy. However, many types of demandresponseresourcesandenergystoragetechnologieshaveduration limits thatmaymake itdifficult toserveasacapacityresource.Meanwhile,otherbulkpowersystemservices,suchasoperatingreserves,

xv

requiresignificantlylessresponseduration.Thus,anon-generationresourcecapableofproviding1MWofcapacityoverseveralhoursmayalternativelybeusedtoprovidemanymoreMWsofaservice likecontingencyresponseoverthosesamehours.

MarketandRegulatoryIssuesMarket and regulatory barriers for demand response and energy storage participation in bulk powersystemservicesfallintoseveralcategories,includingissuesassociatedwitheligibility,cost,andrevenuecapture. Eligibility barriers relate to how regional reliability councils and balancing authorities definebulk power system services. These definitions can explicitly include or exclude certain classes ofresources or implicitly exclude them by defining services that only certain classes can provide. Costbarriers relate to how regional reliability councils and balancing authorities define the attributes ofperformanceandtherequiredenablinginfrastructurenecessaryforparticipationinwholesalemarkets.While these issues donot directly prevent demand response andenergy storage fromprovidingbulkpowersystemservices,theycanstronglyimpactthecostsassociatedwiththeirprovision,andtherebyindirectlylimitparticipation.

Revenue capture barriers represent the potentially limited ability of demand response and energystoragetobecompensatedappropriatelyforthevaluethattheycouldprovidetothegrid.Whilethesebarriersexistforallresources,theyaremorechallengingfornon-generationresourcessuchasdemandresponseandenergystorage.Revenuecapturebarriersalsoincludechallengesassociatedwithsmallerproviders that would require an intermediary to bring their resources to the wholesalemarket. Theintermediary, either the retail electricityproviderora thirdpartyaggregator,maynot seeabusinesscasetodoso,thuspreventingthesesmallerprovidersfromparticipatingintheprovisionofbulkpowersystemservices.

RecentNorthAmericanElectricReliabilityCorporation(NERC)standardsandFederalEnergyRegulatoryCommission (FERC) Orders (e.g., Order 755, Order 784) have sought to address a number of thesedeployment barriers. As an example, pay-for-performance market revisions can better compensatedemand response and energy storage for the capability of faster andmore accurate response whenprovidingfrequencyregulation,comparedwithconventionalgeneration.Furthermore,somebalancingauthoritieshaveaddressedcostbarriersforsmallerdemandresponseprovidersthatfacehighper-unitenablementcosts.Costscanbeprohibitiveifeachresourceisrequiredtoinvestinthesamemonitoringandcommunicationsequipmentandconnectiontoadedicatedcommunicationnetworktypicallyusedwithlargegenerators.

ConclusionsThisstudyconductsapreliminaryassessmentofthepotentialfordemandresponseandenergystorageresourcestoprovidebulkpowersystemservicesfortheWesternInterconnectioninscenarioswithlowandhighpenetration levelsofwindand solar. It alsoexamines themarket rules and regulations thatimpact the use and deployment of these non-generation resources nationally. Overall, these effortsyieldanumberofkeyfindings:

xvi

• A significant fraction of operational value attributable to demand response and energy storageresourcesare theavoided costs associatedwith generator startups/shutdownsand reduced costsassociatedwithgeneratorsmodulatingoutputwhileprovidingfrequencyregulation.Thesecostsarenominal when looking at total production costs but represent a significant fraction of costs foroperatingreserves.

• Due to the limited temporal flexibility of demand response resources, the provision of operatingreserves has more market value than energy shifting services, assuming prices are based onmarginalcostsofproduction.However,theavailabilityoftheseresourcestoprovideenergyshiftingserviceshelpstooptimizetheoperationofthebroadersystem.

• Energy storageprovides greater value in scenarioswith higher renewable penetrationdue to theincreasedneed foroperating reservesand thegreateropportunities forenergyarbitrage throughthe storageof low-cost,off-peakelectricity.Additionally, co-optimizationofenergyandoperatingreservesresultsingreatervaluethanjustenergyshiftingservicesalone.

• Marketstructurescan limittheabilityofanynewentrant, includingdemandresponseandenergystorage, to be compensated commensurate with savings they create. Existing markets do notinclude generator startup costs in price formulation, so they may not necessarily compensatedemandresponseorenergystorageforreducingthesecostsalongwithothervaluestreams.

• Marginalcostsforoperatingreservesincludelostopportunitycostsfromgenerators(forgoneprofitof selling energy). However, the lost opportunity cost component is defined as zero for demandresponseandenergystorageresourcesprovidingonlyoperatingreserves(forgoneprofitofsellingenergyiszero).Largepenetrationoftheseresourcescansaturatethemarketforoperatingreservesanddrivedownmarketclearingprices.

• While capacity value of demand response and energy storagewas not studied in detail, a simplecalculation based on the assumption of a proxy generation resource suggests that capacity valuecouldbeseveraltimeslargerthantheoperationalvalue.However,realizingthisvaluemayrequireresources toprovidemanyhoursof responseduration (e.g.,6–10hours),generally increasing thecostofprovidingtheseservices.

• While there are multiple challenges to deploying large customer demand response and energy-limited storage resources inwholesalemarkets, smaller customer resources that seek to providebulkpowersystemservicesfaceadditionalbarriers.First,theymayrequireanaggregatorthatmightnot see a business case to provide those services. Second, communications and controlrequirements imposed on individual large providers can be cost-prohibitive if applied equally tosmallerproviders.However,theserequirementscouldbemodifiedandtechnicalhurdlesovercometoreduceimplementationcostswithoutcompromisingsystemreliability.

(This page intentionally left blank)

xviii

TableofContentsAcknowledgments.......................................................................................................................................ii

ExecutiveSummary.....................................................................................................................................iv

MotivationandBackground....................................................................................................................iv

OverviewofDemandResponseandEnergyStorage...............................................................................iv

SourcesofValue.......................................................................................................................................v

StudyApproachandLimitations...............................................................................................................v

DemandResponseandEnergyStorageDeploymentScenarios.............................................................vii

ModelingResults...................................................................................................................................viii

ValueofEnergyandOperatingReserves...........................................................................................viii

OperationalValueofDemandResponse.............................................................................................ix

OperationalValueofEnergyStorage...................................................................................................xi

TheImportanceofCapacityValue.........................................................................................................xiv

MarketandRegulatoryIssues................................................................................................................xv

Conclusions.............................................................................................................................................xv

1 Introduction.........................................................................................................................................1

2 BulkPowerSystemServices................................................................................................................2

2.1 Capacity........................................................................................................................................3

2.2 Energy..........................................................................................................................................4

2.3 OperatingReserves......................................................................................................................5

2.4 ComparisonofGenerationandNon-GenerationResources.......................................................7

3 ModelingandResults..........................................................................................................................9

3.1 WesternInterconnectionandColoradoSystemModels...........................................................11

3.2 CostSavingsandMarketValue..................................................................................................14

3.3 DemandResponse.....................................................................................................................18

3.3.1 End-UseLoads....................................................................................................................18

3.3.2 DemandResponseServices...............................................................................................20

3.3.3 DemandResponseOperationalValue...............................................................................26

3.4 EnergyStorage...........................................................................................................................39

3.4.1 StorageCharacterization....................................................................................................39

3.4.2 EnergyStorageOperationalValue.....................................................................................41

3.5 OperationalValuewithVariableGeneration.............................................................................45

xix

3.6 CapacityValue............................................................................................................................52

4 MarketandRegulatoryIssues...........................................................................................................53

4.1 MarketandRegulatoryEnvironments.......................................................................................54

4.1.1 RetailEnvironments...............................................................................................................55

4.1.2 WholesaleEnvironments.......................................................................................................55

4.2 BarrierstoDeployment..............................................................................................................56

4.2.1 MarketandRegulatoryBarriers.........................................................................................57

4.2.2 IssuesthatImpactCosts....................................................................................................58

4.2.3 IssuesthatImpactRevenueCapture.................................................................................59

4.2.3.1 IssuesSpecifictoDemandResponseAggregators.............................................................61

5 Conclusion..........................................................................................................................................61

6 FutureWork.......................................................................................................................................63

References.................................................................................................................................................66

1

1 IntroductionGrid modernization and advances in technologies are enabling resources like demand response andenergy storage to support a wider array of electric power system operations. Historically, thermalgeneratorsandhydropowercombinedwiththetransmissionanddistributioninfrastructurehavebeenmostly adequate to serve customer load reliably and with sufficient power quality. While demandresponse and energy storage are potential alternatives and complements to these assets in someapplications,workisneededtobetterunderstandtheextenttowhichthesenon-generationresourcescanprovidecost-effectivetechnicalbenefitsandwhethermarketrulesandregulationsare inplacetofacilitatetheiruse.

Answeringthesequestionsischallenging.TheU.S.electricpowersectoriscomplexfromanengineeringperspectiveand froman institutionalperspective. It ishighly fragmentedacross regionswithdifferingresourcecompositionsaswellasdifferingmarketandregulatoryenvironments.Furthermore,regionalcharacteristics of the systemare changing and various technologyoptions are evolving in availability,cost,andperformance.Becausethereisgenerallynosinglesolutionthatcanmeetallthechallengesofthegrid,bothtechnicalandnon-technicalsolutionswillcompetebasedontheircost-effectivenessandtheinstitutionaldifficultyintheirimplementation.

Demand response and energy storage can have both operational value, through reducing costsassociated with using grid assets, and capacity value, through offsetting the costs of procuring newassets.Costsavingscanstemfromefficiencygainsbyfacilitatingimprovedgeneratorscheduling,moreefficient generatoroperatingpoints, andhigherutilizationof gridassets.Other value streams includeincreased system flexibility and the ability to support greater penetration of variable renewableresources. However, deployment can be limited by capital costs, integration costs (e.g., monitoring,communications,andcontrols),aswellaschallengeswithmulti-yearresourceavailabilityinthecaseofdemandresponseandenergylossesfromparasiticloadsandround-tripefficiencyinthecaseofenergystorage.

Utilizations of demand response and energy storage have long histories, but experiences arepredominantly limited to a few specific applications. Electric utilities have historically used demandresponse to manage peak prices (typically during generation shortfalls) and for curtailment inemergencysituationsduringtimesofhighlossofloadprobability(Cappers,Goldman,andKathan2010).Pumpedstoragehydropowerunitshaveextensivelyprovideddailyandweeklyloadbalancing,followingsetschedulesofpumpingwateruphillwhenelectricitydemandislowandreleasingitthroughturbineswhen demand is high. These plantswere often installed in conjunctionwith nuclear power plants tohelpmanagethesystembecausenuclearpowerplantswerenotintendedtomodulateoutputtomeettime-varying demand (Elzinga et al. 2012). However, the dramatic changes occurring in the powersystem, from generators to loads, present significant opportunities to optimize the use of demandresponse and energy storage, advance research and development in these technologies, and supportgridmodernization.

DemandResponseandEnergyStorageIntegrationStudy

2

Demand response and some energy storage resources have an inherent shortcoming with regard toresponseduration.However, the importanceof speedandaccuracyoverduration canmakedemandresponse and energy storage well-suited to the technical requirements for many power systemapplications,especiallywithgreaterdeploymentofrenewableenergyandothercleantechnologiesthatcan be limited or curtailed due to their inherent variability. Newer technology implementations forthese resources have enhanced capabilities in response speed, ramp rates, accuracy for followingsystemoperator instructionsandrespondingtofrequencydeviations,fastswitchingbetweencharginganddischarging,andlowtozerominimumloadingpoints.

This study evaluates two major aspects of increased deployment of demand response and energystorage: their operational value in providing bulk power system services, and related market andregulatoryissues,includingpotentialbarrierstodeployment.Thereportisorganizedasfollows:

Section2providesanoverviewofthethreebulkpowersystemsservicesneededbythegridtoprovidereliableelectricity: capacity, energy, andoperating reserves. It also introduceshowdemand responseandenergystoragecanprovidemanyofthoseservices.

Section3analyzesthepotentialoperationalvalueofdemandresponseandenergystorageinprovidinggridservices.ItdescribesasetofgridsimulationsofthewesternUnitedStates,wheredemandresponseand energy storage resources are deployedunder differing levels ofwind and solar penetration. Thisapproach adapts and extends state-of-the-art techniques used in renewable integration analyses tobetterunderstandthevalueofdemandresponseandenergystorage.

Section4describesmarketandregulatoryissuesassociatedwiththedeploymentofdemandresponseandenergystoragebroadly.Specifically,itdescribesthedifferentmarketandregulatoryenvironmentsatboththeretailandwholesalelevelsthatimpactthedeploymentofthesetechnologies.

Section5summarizesmajorfindingsinthereport.

Section6discussesareasforfutureworkthatcouldbringgreaterinsightintothepotentialopportunitiesfordemandresponseandenergystoragetoprovidegridservicesinevolvingmarketstructuresandactas enabling technologies for greater efficiencyof gridoperationsunder changing compositionsof thegeneratorfleet.

2 BulkPowerSystemServicesThisstudyfocusesonbulkpowersystemservicessuppliedbyelectricgeneratorsandtheiralternativeprovision by demand response and energy storage resources. Bulk power system services includecapacity,energy,andancillary services.Thereareanumberof typesofancillary services (FERC2006)(FERC2007),but thepresentwork looksonlyatasubset,calledoperatingreserves (Ela,Milligan,andKirby2011).Therelationshipsbetweencapacity,energy,andoperatingreservesareshownconceptuallyinFigure2-1.

3

Capacity, energy, and operating reserves have different units of measure. Electricity service, whenreferring to energy, is sold in units like kilowatt-hours (kWh) or megawatt-hours (MWh). However,capacityandoperatingreservesinvolvethecommitmentofresourcestoofferenergyduringsettimes.Thiscan,thereby,bemeasuredinunitsofpower(i.e.,kWorMW)timestheserviceduration,signified(inthisreport)bya‘dash’(i.e.,kW-horMW-h).Whileenergyinvolvesthephysicalbuyingandsellingofelectricity,capacityandoperatingreservesprovidetheinsurancethatenergywillbeavailablewhenandwhereitisneeded.

Requirements for bulk power system services are enforced by balancing authorities (i.e., thoseresponsible formaintaining electric generation-load balancewithin reliability standards) that overseebalancingauthorityareas,whicharecomposedofoneormoreelectricityserviceterritoriescoveredbyloadservingentities(i.e.,electricutilitiesthatprovideelectricityservicetoretailcustomers).Theseloadserving entities rely on a combination of electricity resources, such as utility-owned generation andenergystorage,utility-rundemandresponseprograms,bilateralcontracts,andpurchases inwholesalemarketstoservetheircustomers.Thefollowingsectionsdescribeeachofthethreemajorcategoriesofservices needed by the bulk power system (capacity, energy, and operating reserves) and discussimportantdifferencesintheirprovisionbygenerationascomparedwithnon-generationresourceslikedemandresponseandenergystorage.

2.1 CapacityPower systems need adequate resources to meet electricity demands with high levels of reliability.Capacity resourcesare those thatcanbecalledupon todeliverenergy (or reduce theuseofenergy),duringtimesofhighriskforunservedload(i.e.,theinabilitytobalanceelectricitysupplyanddemand).Acommonmetricforresourceadequacyistheplanningreservemargin,whichisthefractionofavailable

Figure2-1.Differenttypesofbulkpowersystemservicesincludedinthisstudyandtherelationshipbetweentheresourcesneeded to provide energy and the amount needed to provide operating reserves during times of peak load. Planningreserves are needed to meet system reliability standards overall (i.e., from year to year), while operating reserves areneededtoensurereliabilitywithinoperationaltimeframesthataremuchshorter(i.e.,hourly).

4

capacityinexcessoftheexpectedpeakelectricitydemand.Calculationsonthelevelofplanningreservemargin necessary to meet reliability requirements vary by balancing authority area, based on localconditionsandregulations(NERC2013).

In some areas served by an independent systemoperator (ISO) or regional transmission organization(RTO), the balancing authorities establish capacity markets to facilitate procurement of sufficientcapacity,attherightlocations,tomeetforecastedelectricitydemandplustheplanningreservemargin.Table2-1providessomeexamplehistoricalcapacitymarketprices.Inregionswithoutcapacitymarkets,capacity transactions occur only bilaterally through individual contracts between load serving entitiesandcapacityproviders(Griffith2008).FurtherdiscussionofcapacityvalueisprovidedinSection3.6.

Table2-1.ForwardCapacityMarketClearingPricesforISONewEnglandandPJMInterconnectioninTermsof$/kW-year.PJMhaslocationalcapacitypricesandcongestedzonestendtohavehigherpricesthantheoverallPJMsystem.Bothoperateone-yearforwardcapacitymarketsthreeyearsaheadofcontractdelivery(Kirby2013).

DeliveryYear

‘07-‘08 ‘08-‘09 ‘09-‘10 ‘10-‘11 ‘11-‘12 ‘12-‘13 ‘13-‘14 ‘14-‘15 ‘15-‘16 ‘16-‘17

ISONewEngland

$54.00 $43.20 $35.40 $35.40 $38.52 $41.16 $37.80

PJMSystem $14.88 $43.56 $37.20 $63.60 $42.84 $6.00 $10.08 $48.96 $49.68 $21.72PJMMostCongested

Zone$72.12 $81.72 $86.64 $67.92 $42.84 $81.12 $90.24 $87.48 $130.32 $79.92

2.2 EnergyThecostofcapacityisbasedprimarilyonthefixedcostsofgeneration.Thesecoverthecarryingcostsofthe capital investmentsplus fixedoperationsandmaintenance costsunderanticipated conditions.Bycontrast, the cost of energy is based primarily on the variable costs of operating the power system,which are primarily fuel costs but also include variable operations and maintenance costs. Theschedulingofenergysupply frompowerplants isbasedonthesevariablecosts,with lowercostunitsgivenpreferenceoverhighercostunitstominimizethetotaloperationalcosts (alsocalledproductioncosts). Example historical ISO/RTO energy market prices are given in Table 2-2 and hourly marginalenergycostsinthePublicServiceofColoradoareaareshowninFigure2-2.

Table 2-2. Selected Real-Time Hourly Energy Prices in Several ISO/RTOMarkets from 2002 to 2012 (Potomac Economics2014)

AverageReal-TimeHourlyEnergyPrices($/MWh) 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012ERCOT 44.3 44.6 72.8 55.2 56.4 77.2 34.0 39.4 53.2 28.3NYISO(NewYorkCity) 71.3 77.2 96.3 47.1 60.5 55.4 41.7NYISO(WestNewYork) 43.4 40.7 33.2PJM(loadweightedavg.) 41.2 44.3 63.5 53.3 61.7 71.1 39.1 48.4 45.9 35.2ISO-NE(NewEnglandHub) 48.6 52.1 79.7 59.7 66.7 80.6 42.0 49.6 46.7 36.1Energyschedulingiscomplicatedbyanumberofoperationalconstraints,likegeneratorstartuptimes,generator minimum loading points, limited transfer capacity of the transmission system (i.e.,congestion), and the need to hold operating reserves (discussed later in Section 2.3). These

5

complicationsrequirepowersystemoperatorstomakeforecastsofelectricitydemand(fromminutestomonths ahead) and then schedule resources in advanceofwhen they areneeded. Energy schedulingoccursindiscretetimeintervalsthatareasshortas5minutesinsomebalancingauthorityareasandaslongas1hour inothers.Whentheseconstraintsarebinding, likeduringtimesoftransmissionsystemcongestion, higher operational cost generatorsmay need to be scheduled compared to a lower costalternativewithoutthepresenceoftheconstraint(WoodandWollenberg2013).FurtherdiscussionofenergyisprovidedinSection3.2.

2.3 OperatingReservesPower system operators balance electric load and generation resources primarily through energyscheduling.Butbelowtheshortestenergyschedulingtimeinterval(varyingfrom5minutesto1hour),the holding of operating reserves ensures there are enough resources to meet moment-by-momentbalancing,meetchangesbetweenintervals,andrespondtocontingencies,likethesuddenlossofalargegenerator ormajor transmission line.Operating reserves are distinct from energy services;while thecostsmay involve small amountsofenergy, their real value is in the capacityheld in reserveand theabilitytorespondquicklyandreliablytomaintainbalance.

Operatingreservesincludefrequencyregulation,loadfollowingreserve,andcontingencyreserve(HirstandKirby1996);recentlyintheMid-ContinentISOandCaliforniaISO,thereisaproposedramping(alsocalled flexibility) reserve (Navid and Rosenwald 2013; Xu and Tretheway 2012). Frequency regulationresponds to random, minute-by-minute variations in aggregate system load that are too fast to befollowedbytheeconomicdispatchofenergy.Contingencyreservesrespondtosuddenbut infrequentsupplydisruptionsandmustalsobeprocuredseparatelyfromenergyscheduling.Whilebothfrequencyregulationandcontingencyreservesareheld foreveryhourof theyear,systemoperatorscontinuallyadjustfrequencyregulation(throughtheautomaticgenerationcontrolsignal)undernormalconditions

Figure2-2.Historical 2011hourly system lambda for thePublic Serviceof Colorado. System lambda is aproxy forhourlyenergyprices.Theannualaverageisabout$28/MWh,rangingbetween$0/MWhand$130/MWhthroughouttheyear.

6

butdeploycontingencyreservesonlywhenassociatedeventsoccur.Loadfollowingandtheproposedrampingreserveservicearecloselylinked.Bothensurethereissufficientoperatingrangeandrampingcapability available to meet the daily net load curve (load minus the contribution from variablegenerationlikewindandsolarpower).

Table2-3.SelectedAncillaryServiceTariffsandRequirementsinNon-ISO/RTOBalancingAuthorityAreas(OATI,Inc.2013).Notethatancillaryservicetariffsaresimilartocapacitypaymentsintheseregions.

Tariff($/kW-year)andRequirement(%ofsystempeak)† Balancingauthority Regulation SpinningContingencyAEPWestZone 31.68 1.20% 42.72 2.10%ArizonaPublicService 88.92 1.17% 75.12 3.19%DukeEnergyCarolinaPower&Light 47.52 1.20% 47.52 1.77%ElPasoElectric 37.20 0.87% 37.20 1.75%FloridaPower&Light 57.84 1.35% 61.92 0.43%IdahoPower 78.36 1.50% 78.36 2.86%PacifiCorpWest 93.60 4.24% 105.60 1.75%PortlandGas&Electric 80.40 1.30% 77.40 3.50%/2.50%*PublicServiceofColorado 80.88 1.50% 82.56 3.50%/2.50%*PublicServiceofNewMexico 103.20 1.50% 112.32 3.50%SouthernCompany 50.40 1.15% 50.40 2.00%TucsonElectric 145.20 1.29% 145.08 3.50%†Therequirementisnotnecessarilythatoftheaggregatebalancingauthority, justtherequirementimposedonusersofthebalancingauthorityareatransmissionsystem.*Somebalancingauthoritieshaveseparaterequirementsfortheloadservedbythermalandhydropowergenerators.

Operating reserves are further distinguished as either spinning or non-spinning. For conventionalgenerators, spinningrefers to resources thatareconnectedandsynchronizedwith thepowersystem,andnon-spinningreferstothosethatareavailableandreadytobeconnectedandsynchronizedwithinaspecified amount of time (usually from 10 to 30 minutes). For non-generation resources providingoperating reserves, the fundamental difference lies in response speed rather than the existence of aphysical rotating mass, as the term suggests (Kirby 2006). Frequency regulation always comes fromspinning resources,but load followingandramping reservecancome fromacombinationof spinningandnon-spinningresources.Contingencyreserves includespinningandnon-spinningcomponents,butmanyregionsrequireaminimumpercentagetobespinning(e.g.,50%forbalancingauthoritiesundertheWECC[NERC2011]and40%intheMid-ContinentISO[MISO2013]).

Innon-ISO/RTObalancingauthorityareas, transmissionproviderscharge their transmissioncustomersfor theamountofoperating reserves theydonot self-supplyorprocure through third-party suppliers(FERC 2013). Operating reserves are typically quoted on amonthly basis; costs are generally settledaccordingtothetransmissioncustomers’contributiontothetransmissionsystempeakload(seeTable2-3).Wherethereareorganizedwholesalemarkets,theISO/RTObalancingauthorityrunsacompetitivehourlymarketforthesupplyofoperatingreserves.Historicoperatingreservepricesareavailablefrommarket clearing prices posted on the individual system operator websites (see Table 2-4 for anexample). In contrast to non-ISO/RTO markets, rates paid for operating reserves vary by hour (orshorter)anddisplaystrongdailyandseasonalvariations(MacDonaldetal.2012).FurtherdiscussionofoperatingreservesisprovidedinSection3.2.

7

Table2-4.SelectedAncillaryServicePricesinSeveralISO/RTOMarketsfrom2002to2012.Contingencyreserve,asdescribedinthetext,issometimescalledspinningandnon-spinningreserve,responsivereserve,or10-minuteand30-minutereserve.(MilliganandKirby2010),updated.(Multiplynumbersintableby8.76toconvertto$/kW-year.)

AverageMarketClearingPrice$/MW-hour OperatingReserve 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

CaliforniaISO Regulation(Up+Down) 26.9 35.5 28.7 35.2 38.5 26.1 33.4 12.6 10.6 16.1 10.0Spinning 4.3 6.4 7.9 9.9 8.4 4.5 6.0 3.9 4.1 7.2 3.3Non-spinning 1.8 3.6 4.7 3.2 2.5 2.8 1.3 1.4 0.6 1.0 0.9Replacement 0.90 2.9 2.5 1.9 1.5 2.0 1.4 ElectricReliabilityCouncilofTexas(ERCOT) Regulation(Up+Down) 16.9 22.6 38.6 25.2 21.4 43.1 17.0 18.1 31.3 9.2Responsive 7.3 8.3 16.6 14.6 12.6 27.2 10.0 9.1 22.9 9.1Non-Spin 3.2 1.9 6.1 4.2 3.0 4.4 2.3 4.3 11.8 6.7 NewYorkISO(east) Regulation 18.6 28.3 22.6 39.6 55.7 56.3 59.5 37.2 28.8 11.8 10.4Spinning 3.0 4.3 2.4 7.6 8.4 6.8 10.1 5.1 6.2 7.4 6.0Non-spinning 1.5 1.0 0.3 1.5 2.3 2.7 3.1 2.5 2.3 3.9 3.830Minute 1.2 1.0 0.3 0.4 0.6 0.9 1.1 0.5 0.1 0.1 0.3 MidwestISO(day-ahead) Regulation 12.3 12.2 10.8 7.8Spinning 4.0 4.0 2.8 2.3Non-spinning 0.3 1.5 1.2 1.4 ISONewEngland Regulation+mileage 54.6 30.2 22.7 12.7 13.8 9.3 7.1 7.2 6.7Spinning 0.3 0.4 1.7 0.7 1.8 1.0 1.710-Minute 0.1 0.3 1.2 0.5 1.6 0.4 1.030-Minute 0.0 0.1 0.1 0.1 0.4 0.3 1.0

2.4 ComparisonofGenerationandNon-GenerationResourcesBulk power system services have historically been provided almost exclusively by large centrallyoperatedgenerators,andassuch,bulkpowersystemservicedefinitionsandtheassociatedattributesofperformance have evolved around their characteristics and capabilities. However, these services canalsobeprovidedbynon-generationresourceslikedemandresponseandenergystorage.

Demandresponseencompassesmanydifferentstrategiesbywhichcommercial,residential,municipal,andindustrialelectricitycustomersarepaidorincentivizedtoadjustwhentheyuseelectricity.Similarly,energystoragetechnologies likepumpedstoragehydropower,batteries,andflywheelssaveelectricityproducedatonepointforuseata latertime—maybe justmoments laterorhourstodays later.Bothdemand response and energy storage have historically been utilized by utilities to support systemoperations. However, evaluation of demand response and energy storage in the provision of amorecomprehensivesetofservicesorcomparisonbetweentheseresourcesandconventionalgeneratorsinwholesaleelectricitymarketsismorecomplexdueinparttoseveralfundamentaldifferences.

Themostsignificantdifferencebetweenthesenon-generatingresourcesandconventionalgeneratorsisthelimitedresponsedurationofbothdemandresponseandenergystorage.Generatorsdonottypicallyhave response duration limits (due to the availability of fuel supply) and can continuously provide

8

energyandoperatingreserves.Incontrast,energystorageresourceshavephysicalenergylimitsbasedonthesizeofthestoragedeviceimplemented(ratedpowerandenergycapacity)aswellasitsstateofcharge.Demandresponseresourcesalsohaveduration limits thataredrivenbyacombinationof theeffectivestoragewithin theend-useapplianceorequipmentsystemsand thecustomerwillingness tocurtail load(Kirby2006;Rubinstein,Xiaolei,andWatson2010).These limitationshaveimplicationsforallthreebulkpowersystemservicesdescribedpreviously.

For example, the limited response duration presents a scheduling problem that is largely unique todemandresponseandenergystorage.Becauseofthislimitation,bothdemandresponseloadreductionsandenergystorageoutputneedtobetimedtoprovideenergyservicesduringperiodsofhighestvalue;similarly, any load recovery or charging need to occur during periods of lowest cost, in order tomaximizevalue.Thisaddscomplexity toanalyzingdemand responseandenergy storagevalue.Theseenergylimitsalsoprovidechallengestoevaluatingtheabilityofdemandresponseandenergystoragetoprovideoperatingreserves,especiallywhenconsideringtheexpecteddurationofcontingencyeventsoractualenergyflowsthatoccurwhenprovidingfrequencyregulation.

The challenges associated with response duration are further complicated when evaluating demandresponse.Akeydifferencebetweendemandresponseandenergystorage is thatdemandresponse isinherently tied to end-use loads with associated daily and seasonal electricity consumption patterns(Piette,Kiliccote,andDudley2012).Thisaffectswhendemandresponseisavailabletoprovidecapacity,energy, andoperating reserves. Forexample, residential air conditioning is typically inuseduringhotsummerafternoonsandcanbeavailable toprovide responseduring those times,butairconditioningwill have little to no availability for response during colder seasons. Further, demand responseavailabilitymay be difficult to forecast across different time scales, from operational time frames tomulti-yearplanninghorizons.

Anotherimportantdifferencebetweennon-generatingsourcesandconventionalgeneratorsisthescaleof the individual resources and the associated challenge of measuring and verifying response;integrationofmonitoringandcommunication technologies formeasurementandverificationof thesesmall resources can be cost-prohibitive. Effective utilization of individual small demand responseresources relies on aggregation to ensure a reliable response. The aggregate response could bepredictable and reliable even if that of any individual end-use load is not (Kirby 2006). Alternatively,aggregationcanenableagreaterrangeofservicesthatcouldbeprovided,suchasfrequencyregulationfromaportfolioofend-useloadsthatareindividuallyincapableofprovidingsuchservices(Enbala2011;Callaway2009).

Asdiscussed later in Section3.3, the flexibility, duration, and locationof demand response resourcesvarygreatly.Inmanycases,flexibleend-useloadscapableofprovidingdemandresponsehaveeffectivestoragecomponents (suchas thermalandcooling inertia) thataugment their ability tobedispatchedwhen needed (Rongxin et al. 2010; Goli, McKane, and Olsen 2011). Furthermore, concentrations ofindustries,facilitytypes,andcustomerusagepatternsvaryregionally(Watsonetal.2012).Ontheotherhand,energystoragedeviceshavelimitationsdependingonthespecifictechnologyused.Someenergystoragetechnologies,suchaspumpedstoragehydropowerandcompressedairenergystorage,maybe

9

constrainedgeographically,whilemanyotherstechnologies,suchasbatteriesandflywheels,havefewersitingobstacles.

The provision of operating reserves from demand response and energy storage poses additionalchallenges inwholesale electricitymarkets. The cost of providing operating reserves in thesemarketregions is calculated by the system operator and based on the opportunity cost that results fromwithholdingenergyprovision(discussedfurtherinSection3.1).However,fornon-generationresources,these costs are difficult to calculate and, in the case of demand response, challenging to verify forindividual entities. Each end-use load is distinct, and the value of interrupting its operation imposesdifferentcosts.Thesecostscanvaryovertieandincludesomefactorsthatareultimatelysubjectivetothecustomerbeingservedbythat load.Energystoragedevicescapableofbothenergyandoperatingreserves can have opportunity costs similar to generators. However, unlike generators whoseopportunity costs can be assessed coincident with their provision of operating reserves, opportunitycostsforenergystorageresourcesareinherentlyinter-temporal;theiropportunitycostswilldependontheenergypriceswhencharginganddischargingoccurs.

Finally, the operational costs of demand response and energy storage can be quite different fromconventionalgenerators.Themainoperationalcostforenergystorageischargingenergy,plusparasiticloads and round-trip efficiency losses. For demand response, the main operational cost is lostopportunityofuseorvalueof lostservice(Woolfetal.2013). Insomeinstances, itmaynotmattertothe customer when an appliance or piece of equipment is used, and the demand response strategywouldfallwithincustomertolerances,resultinginalowopportunitycost.Atothertimes,thedemandresponse requestmaybedisruptive to thecustomer,withopportunitycosts increasingwith responseduration(CallawayandHiskens2011).

3 ModelingandResultsLarge Western Interconnection-wide renewable integration studies have evaluated many of thechallengesassociatedwithdeployinglargeamountsofvariablewindandsolargenerationtechnologies.Integrationstudiesusehigh-resolutiontimeseriesanalysiscoveringayearormoreofsystemdatawithunit commitment and economic dispatch modeling (EnerNex 2011; GE Energy 2010) to evaluateoperational impacts associated with increasing the supply of variable generation. These studies alsoexaminebenefitsofimprovedwindandsolarresourceforecastingandtrade-offsbetweeninstitutionalchanges (like increasingbalancingareacooperation)and technical changes (like installingnewflexiblegeneration).

There has been continuous advancement in the tools, analytic approaches, and data sets: improvedmeso-scalemodeling for time series weather data synchronized with load data (Potter, et al. 2008),realistic short-term resources forecasts, and improved treatment of conventional generator rampingandcyclingcosts (Lewetal.2012; IEA2013).Demandresponseandenergystorageresourcespresentimportantsourcesofbulkpowersystemservicesandcanaid in the integrationofvariablegenerationinto the grid; however, integration analyses havenot yet fully incorporated these resources explicitlyintogridsimulationmodels.

10

Themodelingapproachtaken in thisstudy isbasedontheWesternWindandSolar IntegrationStudyPhase 2 (WWSIS-2), incorporating associated recommendations from its technical review committee(Lewetal.2013).Theapproachtakendidnotattemptto:

• Assessallthepotentialmechanismsandsourcesofdemandresponse• Assessallthepossiblevaluestreamsofdemandresponseandenergystorage• Consider the costs associatedwith the deployment and integration of demand response or

energystorage• Considerenergystoragetechnologiesdeployedindistributionsystemsorbehindthemeter• Simulatecontingencyeventsandconsidertheimpactsofchangingsystemdynamics• Determinetheoptimalsizingorlocationofdemandresponseorenergystorage.

Simulationsoverthefull2020studyyearareconducted intheproductioncostmodelPLEXOS(EnergyExemplar 2013) in order to assess the value of demand response and energy storage in systemoperations.Capitalcostsassociatedwiththedeploymentofdemandresponseandenergystoragewerenot considered in the determination of value. Production cost models are utility planning tools thatmodel unit commitment and economic dispatch processes. They are commonly used in renewableintegrationanalysesbecausetheycanmimicmanyof thedecisions facedbypowersystemoperators.Further,productioncostmodelsoutputanumberofusefulmetricssuchasestimatesofdifferenttypesofoperationalcostsandpowerplantemissions.Astypicalwithproductioncostmodels,themodelcancalculate the cost of holding operating reserves, but frequency regulation dispatch and contingencyevents are not simulated. Simulation of shorter term grid dynamics is an important area of futureresearch, requiring models capable of simulating detailed generator and non-generator resourceresponse(ElaandO'Malley2012;Varadanetal.2012).

AsintheWWSIS-2study,weassumethatgridassetsarecommittedanddispatchedaccordingtoaleast-costoptimization.Inreality,therearenumerous(typicallyproprietary)contractualagreements(forbothpowerandfuel),aswellassystemoperatorbehaviorsandmarketbehaviors,thatresult indifferencesbetween modeled and actual outcomes. We also assume larger reserve sharing groups, reflectingpotential improvementsto inter-balancingauthorityareacooperationinthe2020studyyear;withinareserve sharing group, operating reservesmay come fromanymemberbalancing authority area (seethemap in Figure 3-1). Lastly,wemodel the transmission system zonally (using linearized DC powerflow)withtransferlimitsbetweenbalancingauthorityareas,greatlysimplifyingnetworkconstraints.Asa result, the modeling results do not capture localized congestion that could be relieved from thedeploymentofsmall-scaleenergystorageorenablementofdemandresponse,whichcouldprovideanadditional potential sourceof value (i.e., deferral of transmission anddistribution asset investments).AdditionaldetailsontransmissionsystemassumptionscanbefoundintheWWSIS-2.

The results are also based solely on the day-ahead unit commitment and economic dispatch. Thisapproach takes into account wind and solar forecast errors between the day-ahead and real-timedispatchesbyholdingadditionaloperatingreserves(frequencyregulationandrampingreserve)tomeetanticipated increases in variability and uncertainty. However, the real-time dispatch is not simulated.The production cost model simulations begin with the use of two scheduling models to determine

11

outageschedulingandallocatecertainlimitedenergyresources(primarilyhydropower).Theproductioncostmodelthenperformschronologicalunitcommitmentandeconomicdispatchmodelingfortheday-ahead market run. The optimization time window for the day-ahead unit commitment is 48 hours,rollingforwardin24-hourincrements.Theextra24hoursintheunitcommitmenthorizon(forafull48-hour window) are necessary to properly commit generators with high startup costs and to properlydispatchresourceswithenergyconstraints.

3.1 WesternInterconnectionandColoradoSystemModelsTo make quantitative estimates of the value of demand response and energy storage, we simulatepowersystemoperationsontwosystems,amodeloftheWesternInterconnection(seeFigure3-1)andasmallerColoradosystem.OurbaseassumptionsarederivedfromtheTransmissionExpansionPlanningPolicyCommittee(TEPPC)2020CommonCase(WECC2011)withmodificationsfromtheWWSIS-2.Thegenerationmix and electricity consumption for themodeled systems are provided Figure 3-2 and inTable3-1, respectively. Inboth systems, thehigh renewablebase casehasgreaterquantitiesofwindandsolardeployedcomparedtothelowrenewablebasecase,buteachcasehasthesameamountofconventional resources. This assumption implies greater system reliability in the high renewable caseduetoexcesscapacityandhighertotalsystemcapitalcosts(whicharenotconsideredinthisstudy).

Figure3-1.Studyareaincluding36balancingauthorities(smallprint)and12reservesharinggroup(largeprint)assumptionsforthe2020studyyear

12

Table3-1.GenerationResourceMixintheLowandHighRenewableBaseCasesfortheTwoModelSystems,aColoradoSystemModelandaWesternInterconnectionModel

ColoradoTestSystemModel WesternInterconnectionModel GeneratorType LowRenewable HighRenewable LowRenewable HighRenewableCoalSteam 6,178MW 6,178MW 30.26GW 30.26GWCombinedCycle 3,724MW 3,724MW 56.65GW 56.65GWGasTurbine/GasSteam 4,045MW 4,045MW 23.00GW 23.00GWHydropower 777MW 777MW 52.30GW 52.03GWPumpedStorage 560MW 560MW 6.30GW 6.30GWOnshoreWind 3,347MW 7,216MW 27.61GW 65.86GWSolarPhotovoltaic 986MW 2,301MW 7.07GW 20.27GWConcentratingSolarPower 0MW 0MW 4.79GW 7.19GWOther† 513MW 513MW 7.41GW 7.41GWTotal 20,130MW 25,314MW 215.39GW 268.97GW†Othergenerationincludesoil-andgas-firedinternalcombustionengines.

The operating reserve requirements utilized for each reserve sharing group in both test systems arebased on statistical analysis of load,wind, and solar variabilitywithin each of themember balancingauthorityareas(Ibanezetal.2012).Theresultingannualaveragerequirementsforfrequencyregulation,contingency reserves,and ramping reserves foreach reservesharinggroupareprovided inTable3-2.The amount of frequency regulation and ramping reserves required in the high renewable case isgenerally larger than the low renewable case due to increased variability and uncertainty. However,contingency reserve requirements are based on the size of the single largest contingency (i.e., N-1reliability criteria) which should not change between the two cases. In theWestern Interconnection

Figure 3-2. Electricity generation by source in the low and high renewable base cases of the Colorado and WesternInterconnectionmodels

13

systemmodel,therequirementforeachofthereservesharinggroupsisassumedtobe6%ofload,withatleast3%fromspinningresources.Thenon-spinningportionofthisrequirementwasnotmodeleddueto historically low opportunity costs and the abundance of non-spinning resources in the model.Additionally,we augment the generator characteristics in the TEPPCdataset to reflect part-loadheatrates and startup costs prepared by Intertek/APTECH for theWWSIS-2 (Kumar et al. 2012) to moreaccuratelyreflecttheadditionalcostsassociatedwithprovidingoperatingreserves.

Table3-2.AverageReserveRequirementsbyReserveSharingGroupasaPercentageofAverageElectricLoadinEachRegion

RegulationReserve ContingencyReserve RampingReserve Reservesharinggroup LowRE HighRE LowRE HighRE LowRE HighREColoradoSystem 1.3% 1.8% 4.5% 4.5% 0.6% 1.2%WesternInterconnection Arizona 1.1% 1.1% 3.0% 3.0% 0.6% 0.7%CaliforniaNorth 1.0% 1.0% 3.0% 3.0% 0.2% 0.3%CaliforniaSouth 1.0% 1.0% 3.0% 3.0% 0.5% 0.6%Colorado 1.1% 1.8% 3.0% 3.0% 0.7% 2.2%Idaho 1.1% 1.3% 3.0% 3.0% 0.5% 1.0%Montana 1.6% 5.6% 3.0% 3.0% 1.5% 7.6%NevadaNorth 1.2% 2.6% 3.0% 3.0% 0.5% 3.1%NevadaSouth 1.0% 1.4% 3.0% 3.0% 0.1% 1.0%NewMexico 1.2% 2.7% 3.0% 3.0% 0.6% 3.4%Northwest 1.1% 1.1% 3.0% 3.0% 0.7% 0.8%Utah 1.0% 1.2% 3.0% 3.0% 0.3% 0.8%Wyoming 1.6% 3.3% 3.0% 3.0% 1.5% 4.3%

TheColoradosystemmodelisderivedfromthecharacteristicsoftwowesternbalancingauthorityareas,PublicServiceofColoradoandWesternAdministrationColoradoMissouri.Whilemostassumptionsareinherited from the larger Western Interconnection system model, the Colorado system model isdeveloped by turning off generation and load outside of these two balancing authority areas. In theparentmodel,theColoradosystemmeetssomeof itselectric loadusinggenerationresourcesoutsideitsgeographicfootprint.Tomakeupthisdifferenceintheisolatedsystem,naturalgascombinedcycleunitsareadded into themodeledregiontomeeta15%planningreservemargin.This smallersystemenablesnumerouscontrolledexperimentstobemodeledthatwouldbeprohibitiveinthelargersystemduetocomputationaltime(Denholmetal.2013).

Apotentiallyimportantsourceofvalueforbothdemandresponseandenergystorageistheprovisionoffrequency regulation and avoiding costs associated with non-steady state operation of thermalgenerators. Our study builds on the TEPPC and WWSIS-2 datasets to better understand this valuethroughrefinedassumptions.First, thereareaddedcostsassociatedwithnon-steadystateoperation.Plants providing frequency regulation incur additionalwear and tear and heat rate degradation (PJM2012). However, actual performance data related to an individual generator’s ability to provideoperatingreservesarenotwidelyavailable.WeassumearegulationcostadderbasedonobservationsofISO/RTOmarketgeneratorbiddata(Denholmetal.2013)giveninTable3-3.

14

Table3-3.AdditionalWearandTearandHeatRateDegradationCostsforProvidingFrequencyRegulation

Generatortype Cost($/MW-h)Supercriticalcoal 15Subcriticalcoal 10Combinedcycle 6Gas/Oilsteam 4Hydropower 2Pumpedstoragehydropower 2

Whileparameterslikerampratesandoperatinglimitsdetermineagenerator’sabilitytoprovideenergyand operating reserves, only a subset of generators have the necessary equipment to follow theautomatic generation control signal for the provision of frequency regulation. We assume thatcombustionturbinesarenotinstrumentedtoprovidefrequencyregulation(WWSIS-2TechnicalReviewCommittee 2012) and that only 60% of the ramp capability of the remaining generators can providefrequencyregulation (CAISO2011).Wealsomaketheassumptionthatnuclear,geothermal,andfixedschedulehydropowerplantsarenotabletoprovideanyoperatingreserves.Whichspecificgenerators,inactuality,arecapableofprovidingfrequencyregulationisproprietaryinformation.Inpractice,theremaybefewergeneratorsthatprovidethisservicethaninourassumptions,whichwouldresultinhighervalueforfrequencyregulation.SensitivityanalysesonthecostofoperatingreservesbasedongeneratoravailabilityisprovidedinHummonetal.(2013).

3.2 CostSavingsandMarketValueCalculatingtheoperationalvalueofdemandresponseandenergystoragecanbeapproachedtwoways.First, we can compare the total costs of operating the system (i.e., production cost) between twoscenarios.Bymakingincrementalchangestothesystem,changestoproductioncostscanbeattributedtotheadditionorremovalofgridassetsortochanges insystemoperation.Differences inproductioncosts represent the total operational value to themodeled system, irrespective of how that value isdistributed to different entities. Second, we can utilize the calculated short-run marginal costs ofproductiongeneratedinPLEXOStodetermineanimpliedmarketvalueineachbalancingauthorityareaforeachserviceduringeachhour. InISO/RTOregions,marginalcostsequatetomarketclearingpricesfor individual services and indicate the expected revenue that resources would earn as marketparticipants (Hogan1998). Innon-ISO/RTOregions,marginal costsareanalogous toasystem lambda,andrelatetoaverticallyintegratedutility’savoidedcostforprovidingtheassociatedservice(BoothandRose1995).

Production costs are all costs aside from capital and fixed operations andmaintenance (fixed costs),which ultimately reduce to the sum of fuel costs and variable operations and maintenance costs(variablecosts).Itisillustrativetoassignthesebasiccostcomponentstodifferentgeneratoroperationalmodessuchasstartup-shutdown,steadystateoperation,andnon-steadystateoperation.Startupcostsarethoseassociatedwithbringinganofflineresourceonline.Forthermalgenerators,thesecostsmaydependonwhen theunitwas lastoperated.Steadystatecosts come fromholdinga singleoperatingpoint; non-steady state costs come from modulating around that operating point, as in the case ofprovidingfrequencyregulation.Figure3-3showstheproductioncostresultsfromthebasecase(i.e.,no

15

additionaldemandresponseorenergystorage) for the lowandhigh renewablecasesof theWesternInterconnection model. As shown in Figure 3-3, steady state costs constitute the majority of totalproduction costs (left), but startup-shutdown and non-steady state operation are significantcomponentsoftheproductioncostsassociatedwithoperatingreserves(right).

Short-runmarginalcostsofproductionarethechange intotalproductioncostsresultingfromasmallchangeinelectricloadoroperatingreservesrequirements.Forinstance,themarginalcostofenergyistheincrementalchangeintotalproductioncostsifthedemandforelectricitydecreases(orincreases)byoneunit(i.e.,thecosttoservethelastunitofenergy).ThesecostswouldbeequivalenttothemarketclearingpricesinanISO/RTOmarketifgeneratorsbidonlytheirtruevariablecosts(i.e.,donotengageinstrategicbidding).Ofnote,startup-shutdowncoststhatarereflectedinthetotalcostsofproductionarenotrepresentedinshort-runmarginalcosts.

Foroperatingreserves,theshort-runmarginalcostsofproductionarecomposedoftwocomponents:alost opportunity cost and any additional operational costs. Power system markets and operationscalculatelostopportunitycostsforgeneratorsthatprovideoperatingreserves.Thelostopportunitycostis the difference between the systemmarginal cost for energy and the generator’smarginal cost forenergy; in other words, forgone profit. Generators may also incur additional operational costs (e.g.,additional fuel and operations and maintenance) for providing operating reserves, as discussed inSection3.1withourassumedvalues inTable3-3.Thus, thesumof the lostopportunitycostandanyadditionaloperationalcostsforthegeneratorservingthelastunitofoperatingreservessetsthesystemmarginal cost foroperating reserves (Isemonger2009). Thedistributionofmarginal costswithineachregionforthevariousbulkpowersystemservicesoverthestudyyearareprovidedinFigure3-5forthebasecasesintheWesternInterconnectionsystemmodel.

Figure3-3.ProductioncostsintheWesternInterconnectionmodeldisaggregatedbydifferentoperationalmodes.Ontheleftare total costs for combined energy and operating reserves, and on the right are costs associated with just operatingreserves.Thecostsofoperatingreservesiscalculatedbytakingthedifferenceinproductioncostsbetweenthebasecasesandcaseswheretheoperatingreservesrequirementsaresettozero.

16

Figure3-4illustrateswhattheimpliedmarketsizewouldbeintheWesternInterconnectionifresourcesare paid at the marginal costs of production. In other words, the market size is equal to the totaldemandforenergyandoperatingreservesforeachregionduringeachhourmultipliedbythemarginalcostforeachservice inthathourforthatregionandsummedovertheentirestudyyear. Itshouldbeemphasizedthatthis implicitlyassumesthatthemarginalcostsgeneratedbyaproductioncostmodelareequal to theprices inamarket settinganddoesnotconsider thepotentially significant impactofgenerator bidding strategies and other factors that could cause market prices to deviate from agenerator’sactualvariablecosts.

Finally,powersystemmarketandoperations(whichproductioncostmodelsattempttomimic)arebuiltaroundanumberofhardandsoftconstraints.Hardconstraintsarethosethatsystemsoperationsmustobey,whereassoftconstraintsarethosethatsystemoperationscanviolateifmeetingtheconstraintismore expensive than a penalty price. Often, penalty prices are set sufficiently high such that softconstraints are rare. In ISO/RTO markets, these penalty prices are set higher than what marketparticipantsareallowedtobid.However,thespecificimplementationsofmarketsandoperationsdifferamong balancing authorities. Even in the case of hard constraints, the modeling software may notalways find a solutionwithin the allowable tolerance. In these cases, the various constraintsmay berelaxedsequentially, inorderofpresumedsystemcriticality,untilasolution is reached.Theseevents,oftenreferredtoasscarcity(Hogan2012),haveanumberofeffectsonmarketprices,whicharedifficulttoquantifywithoutintroducingassumptionsthatcanbiasthefinalresults.

Figure 3-4. Implied market size for the Western Interconnection model if resources are paid at the marginal cost ofproduction for energy and operating reserves. On the left is the total market size for combined energy and operatingreserves and on the right is themarket size associatedwith just operating reserves. The black lines (Net) show the netmarketsizeaftersubtractingouttheproductioncosts(showninFigure3-3).

17

Figure 3-5. Marginal cost of energy, ramping reserve, contingency reserves, and frequency regulation in the WesternInterconnectionmodelinthebasecasesforboththelow(L)andhigh(H)renewablecases.Verticallinesshowthe10%and90%valuesinthedistribution,theboxesprovidethequartilevalues(i.e.,25%,median,and75%),andthehorizontal linesshowtheaveragevalues.

18

3.3 DemandResponseTherearetwobroadcategoriesofdemandresponseimplementation,distinguishedbywhetherornotthedemand response resources are explicitly dispatchedby thepower systemoperator (FERC2010).Undersomecontrolstrategies,retailelectricitycustomersrespondtoretailpricingtariffs.Forinstance,customerscouldchangetheirconsumptionbasedontime-varyingelectricityprices(ZhouandBotterud2013). In other strategies, demand response resources are treated similarly to generators or energystorage (Hummon et al. 2013) and the load response is characterized by operational constraintsincluding maximum response duration, ramp rates, and timing of load recovery (when the energyutilizationlostduringaloadshedincreasesloadatanearlierorlatertime).Whilethereareadvantagesand disadvantages to different approaches, we utilize the second approach to explore the use ofdemand response for the provision of a broader range of grid services. However, the underlyingmechanismsbywhichretailelectricitycustomersareenrolledindemandresponseprogramstoprovidetheseservicesisoutsidethescopeofthisstudy.AnumberofalternativeprogramsandimplementationstrategiesarediscussedinChuang(2009).

3.3.1 End-UseLoadsToestimatethepotentialoperationalvalueofdemandresponse,weneedtoestimatetheavailabilityofdemandresponseresourcesandimplementthatestimateinproductioncostmodels.Wefirstbeginbygenerating regional profiles for different loads that may be able to shift energy use and provideoperating reserves. We start with projected hourly profiles generated by TEPPC for each balancingauthority area in the year 2020 and approximate the relative contributions from appliances andequipment systems in different economic sectors. In the resource assessment, we include end-usesacross commercial buildings, residential buildings, municipal functions, industrial non-manufacturing,andindustrialmanufacturing.Whilewediscussallfiveofthesesectorsinthisreportforcompleteness,the data from the industrial manufacturing resource assessment was not completed in time for theproductioncostmodeling.Theseend-use loads,while importantdemand response resources,arenotanalyzed for their potential operational value in our simulations. End-uses included in the resourceassessmentaregiveninTable3-4,andtheassociatedelectricitydemandsaregiveninFigure3-7.

Table3-4.End-UsesIncludedintheDemandResponseResourceAssessmentacrossDifferentEconomicSectors

CommercialBuildings ResidentialBuildings MunicipalFunctions IndustrialFacilities

Indoorlighting Spacecooling Freshwaterdistribution Agriculturalwaterpumping

Spacecooling Spaceheating Roadandgaragelighting Coldstorageinrefrigeratedwarehouses

Spaceheating Waterheating Wastewaterpumping Datacenterserversandequipmentcooling

Ventilation Manufacturing††Representing28manufacturingprocesses.FurtherbreakdowngiveninTable3-6.

Forend-useloadsthataresensitivetoweather,wenormalizetheresourcedatasetstotheyear2006usingalinearregressiononhistoricalload,temperature,andrainfalldata.Asthemostrecentwindandsolarresourcedataisfromtheyear2006,thisnormalizationcapturessomeofthecorrelationsbetweenelectricitydemandandwindandsolargeneration(Lewetal.2013).

19

Figure 3-7. Electricity consumption by various end-use types included in the demand response resource assessment. Thewhitebarsignifiestheremainingelectricityconsumptionbyend-usesnotincludedintheassessment.

Thereareanumberofadditionalresourcesthatarenotincludedinthisassessment.First,projectingthechangingcompositionofend-useloadsisoutsidethescopeofthisproject(e.g.,electricvehiclescouldbeasignificantdrawonfutureelectricityproduction).Second,therearemanysmallerloadsthatcouldbe aggregated (e.g., miscellaneous building plug loads) for demand response, but their individualcharacteristicsvarysignificantly,complicatingtheiranalyses.Generally,existingdatasourcesfortheseloadslackthenecessaryfidelitytodeterminetheirpotentialcontribution,preventingtheir inclusioninthis study. Lastly, demand response strategies can be coupled with on-site generation to provideservices jointly. These may include back-up generators, combined heat and power systems, andmicrogrids.Whiledistributedgenerationmayplayarole inprovidingbulkpowersystemservices, thisareaofinvestigationisoutsidethescopeofthisstudy.

The availability of different types of end-use appliance and equipment systems (potentially used fordemand response) varies significantly across theWestern Interconnection. Each region has differentdriversforelectricitydemanddependingonfactorssuchaspopulationgrowth,localclimate,consumerbehavior,andcommercialandindustrialeconomicactivity.Forinstance,Figure3-8showsthelocationsofmanufacturingfacilitiesforTextileMillProductsandRubberandMiscellaneousPlasticsProductsandassociated clustersof economicdevelopment.Quantificationof thedemand response resource startswith understanding the regional distribution of different end-use appliances and equipment systemsandtheirpatternsofuse.Aflexibilityanalysis(discussedinthenextsection)isthenappliedtoelectricityconsumption data of each end-use type to estimate response capabilities for different bulk powersystemservices(Olsenetal.2013;Starke,Alkadi,andMa2013).

20

3.3.2 DemandResponseServicesTheanalysisdescribed in theprevious sectionestablishes the loadprofilesof end-useappliancesandequipment systems assumed to be capable of participating in demand response. However, only afraction of the loadmay be available for shedding or shifting to provide bulk power system services.Commondefinitionsfortheseservicesdonotexistacrossallregions;butforthisstudy,wemakesomegeneralizationsbasedonphysicalrequirements(seeTable3-5).Combinedwithflexibilityanalysesofthedifferentend-uses,thesegeneralizedproductdefinitionsdeterminetheassumedeligibilityofpotentialdemandresponseresourcesfordifferentservices(seeTables3-6and3-7).Thismethodologyalsohelpsdetermine associated strategies for the demand response resources to participate in power systemoperations(Maetal.2013).

Theproductioncostmodeltreatsenergyandoperatingreservesdifferently.Demandresponseenergyprovision involves the shifting of energy use from one energy scheduling interval to another. This isaccomplished explicitly in the model, including constraints like the load shed duration (i.e., themaximumnumberofhoursofashed)andtimingofloadrecovery.Dependingonthedemandresponsestrategy, loadrecoverymayoccurprior to the loadshedormayoccur following the loadshed.Thesetwostrategies,usingtheanalogywithenergystorage,correspondwithpre-chargingandrechargingtheeffective storage capability utilized for demand response. During load recovery, some fraction of theavoided energy use will be consumed. The timing and profile of load recovery has importantimplications for value to thepower systemaswell as to the retail electricity customersproviding thedemand response service. For instance, thermal loadsmustmaintain temperaturewithin tolerances,

Figure 3-8. Red dots shows plants locations in the western United States affiliated with the Textile Mill Productsmanufacturingsubsector(left)andtheRubberandMiscellaneousPlasticsProductsmanufacturingsubsector(right).

21

limitingthetimedifferencebetweenaloadshedandaloadrecoveryandlimitingthepotentialbenefitsofarbitragingbetweentimesofdifferentenergycosts(Hummonetal.2013).

Table3-5.GeneralizedBulkPowerSystemProductDefinitionsUsedtoAssesstheCapabilitiesofDifferentTypesofDemandResponseResources.Actualrequirementsimplementedindifferentregionsmaydifferfromtheseassumptions.

PhysicalRequirements BulkPowerSystemProduct

GeneralDefinition

Howfasttorespond

LengthofResponse

Timetorespondfully

Howoftencalled

FrequencyRegulation

Responsetorandomunscheduleddeviationsinschedulednetload

30secondsEnergyneutralin15minutes

5minutesContinuouswithinbidperiod

SpinningContingencyReserve

Rapidandimmediateresponsetoalossinsupply 5minutes ≤30

minutes ≤10minutes Lessthanonceperday

Ramping(Flexibility)Reserve

Loadfollowingreserveforlargeun-forecastedwindandsolarrampsbeyondthatfordailyload

5minutes 1Hour 20minutesContinuouswithinbidperiod

Energy Shiftenergyutilizationfromonetimeperiodtoanother 5minutes ≥1Hour 10minutes

1-2timesperday,4-8hournotification

Table 3-6. Assumed Eligibilities for Residential and Commercial Buildings, Municipal Functions, and Industrial Non-ManufacturingEnd-UsestoProvideDifferentTypesofBulkPowerSystemServices

Foroperatingreserves,theconstraintsonthedemandresponseresourceareimplicitintheeligibilityforprovidingservices.Productioncostmodelsdonotsimulatetheactualutilizationofoperatingreserves.Theymerelyreservesufficientrampcapabilityandoperatingrangefromgridresourcestoensurethatoperatingreserverequirementscouldbemetintheresultingunitcommitmentandeconomicdispatch.Operating reserves aremostly compensated by the amount of capacity offered in each time period,rather than theamountofenergydelivered (or curtailed in the caseofdemand response)when thatcapacityisdispatched.However,thefrequencyofuseinoperatingreservescanhavesignificantimpactson the utilization of demand response. For instance, spinning contingency reserve may require

BulkPowerSystemServices

ResourcesFrequencyRegulation RampingReserve

ContingencyReserve Energy

CommercialBuildings Cooling X X X XHeating XLighting X X Ventilation X X X

ResidentialBuildings Cooling X X X XHeating X X X XWaterheating X X X X

MunicipalFunctions Outdoorlighting X X X Freshwaterpumping XWastewaterpumping X

IndustrialNon-Manufacturing Datacenters X XAgriculturalpumping X XRefrigeratedwarehouses X

22

resourcesthatcanrespondalmostinstantly,ramptofullresponseinlessthan10minutes,becalledasfrequentlyasdaily,andmustbecapableofsustainingacurtailmentforupto30minutes.Onlydemandresponse resources able to do so can qualify as a spinning contingency reserve resource in ourassessment. The actual frequency of use (also referred to as dispatch-to-contract ratio) variessignificantlybyoperatingreserveproduct.

Table 3-7. Assumed Eligibilities for IndustrialManufacturing Processes to Provide Bulk Power System Services Sorted byStandardIndustrialClassification(SIC)

BulkPowerSystemServices Industry Frequency

RegulationRampingReserve

ContingencyReserve EnergySIC DominantProcess

FoodandKindredProducts 20 Packaging X20 Chiller X X X X

TextileMillProducts 22 Wrapping X22 Weaving X

Apparel,FinishedProductsfromFabricsandSimilar 23 Wrapping X23 Weaving X

LumberandWoodProducts,ExceptFurniture 24 Sawing X24 Planning X

FurnitureandFixtures 25 Sawing X25 Planning X

PaperandAlliedProducts 26 Chipper X26 DewateringPress X X X X

Printing,PublishingandAlliedIndustries 27 Chipper X27 DewateringPress X X X X

ChemicalsandAlliedProducts 28 Electrolysis X X X X28 Compressor X X X X28 Grinding X

PetroleumRefiningandRelatedIndustries 29 CatalyticCracking X X X X

RubberandMiscellaneousPlasticProducts 30 Mixing X X X X30 Mill X

LeatherandLeatherProducts 31 Mixing X X X X31 Mill X

Stone,Clay,Glass,andConcreteProducts 32 ElectricFurnace X X X X32 Crushing X

PrimaryMetalIndustries 33 Electrolysis X X X X33 CrushingandClassifying X

TransportationEquipment 37 MetalCutting X37 FinalAssembly X X X X

23

Basedonassumedproductdefinitionsanddispatch-to-contract ratios, theability forend-use loads toprovidedemandresponsecanberepresentedbythreefiltersthat,incombination,estimatethesubsetandfractionofloadthatisavailabletoberesponsive.Thesefiltersincludesheddability,controllability,andacceptability.Sheddabilityrelatestophysicalconstraintsoftheend-useequipmentandisequaltothe percentage of the load which could be shifted or shed by a demand response strategy.Controllabilityreferstothepercentageof loadthathasthecontrols inplacenecessarytoachievethisshiftorshed.Acceptabilityrelatestoend-userattributeslikebuildingoccupantcomfortandemployeework schedules. This last filter is particularly difficult to assessbecause it refers to thepercentageofend-use load that reflects theend-userswillingness toacceptanactualorperceived reduction in thelevelofservicetoparticipateasapowersystemresource.InFigure3-9,anexampleapplicationofthese

Figure3-9.Exampleflexibilitycalculationforrampingreserveprovidedbycommercialbuildingventilation.Thecombinationofsheddable,controllable,andacceptablefiltersdeterminetheoverall flexibility (top)andwhenmultipliedbytheelectricload,yieldamaximumavailableresponse(bottom).Theoverallflexibilityisonlyasmallfraction(i.e.,4–6%)oftheoverallelectricend-useload.

24

threefiltersforcommerciallightingloadisgiven.ThetopofFigure3-9showsthethreefiltersandtheirvaluesovera24hourperiod.ThebottomofFigure3-9showstheapplicationofthesefilterstoanend-useloadprofile.Theresultingfractionofend-useloadthatisabletobeutilizedfordemandresponseisthedifferencebetweenthesolidanddotted lines.Additionaldiscussionof these filters isgiven in thetextboxonpages34-35(Olsenetal.2013).

Figure 3-11. Estimated resource (from partial participation) and technical potential (with full participation) for demandresponse toprovideoperating reserves in regionsof theWestern Interconnection.Theverticalaxis is limited to400% forclarity, though inmany cases the technical potentials are higher. Percentages are based onmodel assumptions and notactualoperatingreserverequirements.

Figure 3-10. Average max daily energy shifting potential in regions of the Western Interconnection for the estimatedresource(frompartialparticipation)andthetechnicalpotential(withfullparticipation).Horizontallinesindicatetheaverageenergyshiftingpotentialduringthetop100loadhours.

25

The flexibility determination of different types of loads is derived from a number of existing studies,engineering-based analyses of the appliance and equipment systems, historical participation data indemand response programs, and survey data. Our 2020 scenario for the available demand responseresource is based on expectations of communications and control systems in place due to factorsoutside of demand response programs (e.g., the “Internet-of-things”), as well as extrapolations ofexisting demand response program participation rates to new programs. The technical potential ofdemandresponseinthisstudyassumesfullparticipation(i.e.,applicationofonlythesheddabilityfilterto the end-use load profiles). Figures 3-10 and 3-11 summarize the estimated resource (from partialparticipation)andtechnicalpotential(withfullparticipation)fordemandresponseprovidingenergyandoperatingreserves.

Theestimates fordemand response resources, fromabottom-upcalculationmethodology,arebasedon characterizationof applianceandequipment systems that compriseof justunderhalf of the totalelectricity use in U.S. regions of the Western Interconnection (49% of TEPPC 2020 forecast). Theremaining51%ofelectricityuserepresentsotherend-uses(suchasmiscellaneousbuildingplug loads)that are difficult to characterize and are excluded from the present work. The estimated demandresponse resource in 2020, based on assumptions using an extrapolation from current rates of retailcustomerparticipation(i.e.,6%oftotalelectricityuse),indicatesthattheresourcescouldprovideaboutone-thirdoftheoperatingreserverequirementinascenariowith33%windandsolargenerationinthewestern United States, and shift 1.0% of the average daily energy use in the 2020 study year. Thetechnical potential, which represents full participation (i.e., 49% of total electricity use), is about 10times largerandcouldshiftabout9%of theaveragedailyenergyuse.Notethat inthepowersystemsimulations, the demand response resources from industrial manufacturing are not included. Thedemandresponseresourceinthesimulatedscenariosrepresents4%oftotalelectricityuseandcanshiftupto0.8%ofaveragedailyenergyuse.Eachofthedifferentdemandresponseresourceshasdifferent

Figure3-12.Averagehourly availabilities for ramping reserves fromdifferent types ofdemand response resources in theColoradosystemmodelcomparedwiththeaveragesystemload

26

hourly, daily, weekly, and seasonal availabilities. For instance, the availability of commercial andresidentialheatingtendstopeakbetween7a.m.and9a.m.,butavailabilityofmunicipallightingtendstopeakbetween8p.m.and4a.m.Theavailabilities foragriculturalpumping,datacenters,municipalpumping, and wastewater pumping remain fairly flat over the course of a day. For a seasonalperspective,agriculturalpumping,commercialcooling,residentialcooling,andrefrigeratedwarehouseshavepeakavailabilities inthesummer;municipal lighting,commercialheating,andresidentialheatinghavepeakavailabilities inthewinter;anddatacenters,residentialwaterheating,commercial lighting,commercial ventilation, municipal pumping, and wastewater pumping loads have fairly constantavailabilities over the year, with regular weekday-weekend usage patterns. The average hourlyavailabilities for ramping reserves from a few estimated demand response resources in the ColoradosystemmodelareshowninFigure3-12,andcorrespondingmonthlyaveragesforoperatingreservesareshowninFigure3-13.

3.3.3 DemandResponseOperationalValueThissubsectionexaminestheoperationalvalueofdemandresponseinthelowrenewableandhighrenewablecasesoftheWesternInterconnectionmodel.Inthelowrenewablecase,14%ofelectricityproductioncomesfromwind(≈10%)andsolar(≈4%)power;inthehighrenewablecase,33%ofelectricityproductioncomesfromwind(≈25%)andsolar(≈8%)power.Theresultspresentedreflecttheimplementationofdemandresponseresourcesestimatedfromcommercial,residential,municipal,andindustrialnon-manufacturingend-usesintheproductioncostmodel(industrialmanufacturingend-usesarenotmodeled).End-usesincludedinthemodelingconstitute30%oftotalelectricityuse,whereas49%oftotalelectricityusewasassessed.Afteraccountingforparticipationassumptionsindemandresponseprograms,only4%oftotalelectricityuseisavailabletosupportpowersystemoperations.

Figure3-13.Cumulativemonthly availabilities forrampingreserves fromdifferenttypesofcommercialbuildingdemandresponseresourcesintheColoradosystemmodel

27

Ultimately,anannualcumulativeavailabilityof11.3TW-hfordemandresponse(1.4%oftotalelectricityuse)wasusedinthemodeledscenarios.ThissubsectionlooksatgeneralfindingsfromoursimulationsandSection3.5looksinmoredetailatthedriversofvalueunderincreasinglevelsofwindandsolarpower.

Thevalueofdemandresponsecanbeexaminedfromtheperspectiveofsavingsinproductioncostsorfromtheperspectiveof impliedmarketvalue,asdiscussedatthebeginningofSection3.1.Productioncostsavingsrefertodifferencesinfuelandoperationsandmaintenancecostsbetweenscenarioswithandwithout added demand response resources. Impliedmarket value is associatedwith the sum ofpayments(andcharges)toproviders(andconsumers)forenergyandoperatingreservesineveryhourandacrossallregions.Paymentsarebasedoncalculatedmarginalcostsinthemodelingframeworkandweequatethesemarginalcoststomarketclearingprices.

Inthemodeling,wetreatdemandresponseresourcessimilarlytoenergystoragedeployedonthebulkpower system. The model co-optimizes demand response to provide both energy and operatingreservesalongsideconventionalgeneration.Anexampledispatch forwaterheatingdemandresponseresourcesandtheassociatedhourlymarginalcostsfordifferentbulkpowersystemservicesisgiveninFigure3-14.AsindicatedbythedashedlineinthetopofFigure3-14,demandresponseresourcesthat

Figure3-14.ExampledispatchofwaterheatingdemandresponseresourcesintheColoradomodelsystem(top)andassociatedhourlymarginalcostsforvariousbulkpowersystemservices(bottom)

28

canprovidemorethanoneserviceareconstrainedsuchthat thetotalcapacityprovided ineachhourcannot exceed the single largest individual availability for energy, frequency regulation, contingencyreserve,orrampingreserve.ThebottomofFigure3-14givesthehourlymarginalcostsforthedifferentgridservicesgeneratedbytheproductioncostmodelthatareusedinthemarketvaluecalculations.Aswith energy storage, energy transactionswith demand response include both a load shed (similar todischarging)andaloadrecovery(similartocharging).Whencalculatingmarketvalue,demandresponseresourcesareassumedtobepaidthemarginalcostofenergywhensheddingloadandareassumedtopaythemarginalcostofenergywhenrecoveringthatload.Inthisstudy,weassumethatenergyshiftingwithdemandresponseisnet-energyneutral.Inotherwords,therearenolosses(orgains)fromshiftingenergyfromonetimeperiodtoanother.Inactuality,end-useloadsmayincuranefficiencypenaltysuchthattheenergyusedforloadrecoveryisgreaterthantheamountusedforloadshedding,similartotheeffectofroundtripefficiencyforenergystorage.Otherend-useloadsmayexperiencetheopposite,i.e.,a conservation effect. There is insufficient information at this time to make either assessmentcomprehensivelyacrossourresourceassessment.

Figure3-15comparestheproductioncostsavingsassociatedwithdemandresponseresources(left)forthelowrenewablecase(14%)andthehighrenewablecase(33%),aswellastheimpliedmarketvalueifdemandresponseresourcesparticipatedinamarketenvironment(right).Theoperationalvaluesshownarenormalizedby themaximumcumulativeavailability fromdemand response resources that canbeused for bulk power system (see text box on pages 34-35 for more detail on defining a metric tomeasure demand response value). The total reduction in production cost is $117 million in the lowrenewable caseand$107million in thehigh renewable case,withamaximumcumulativeavailabilityfromdemand responseof 11.3 TW-h, leading to savingsof $10.4perMW-hof availability in the lowrenewablecaseand$9.5perMW-hinthehighrenewablecase.Furthermore,thisfiguredisaggregatescost savings into several components. Savings in steady state costs represent the increased use of

Figure3-15.Operationalvalueofdemandresponseresourcesintermsofproductioncostsavings(left)andimpliedmarketvalue if resourcesarepaidatthemarginalcostsofproduction(right)inthe lowandhighrenewablecasesoftheWesternInterconnectionmodel.

29

generators with lower variable costs (i.e., fuel) and improved heat rates from generators operatingclosertotheirpeakefficiencies.Savingsinstartup-shutdowncostsrepresentthereducedneedtostartand stop generators to meet system requirements, thus avoiding the associated startup fuel costs,increases in operations and maintenance, and any additional wear and tear. Lastly, savings in non-steady state costs represent decreased need for generators to modulate output when providingfrequencyregulation.Theassumedunitcostsfornon-steadystateoperationsareprovidedinTable3-3.

OntherightofFigure3-15,theimpliedmarketvalueisshownwithandwithouttheneedtopurchaseenergy(atthesystemmarginalcost)forloadrecovery.Thenetvalue,ifthepurchasecostissubtractedout, isrepresentedbyablackhorizontal line.Thecostof loadrecovery isasubstantialfraction,aboutonethird,ofthegrossmarketvalueforthemodeleddemandresponseresources.Whileenergyshiftingconstitutes the majority of gross market value, the provision of operating reserves constitutes themajorityofnetmarket value. Total production costs aredominatedby steady-stateoperational costs(see left side of Figure 3-3); however, the operational savings achieved through the use of demandresponse (see left side of Figure 3-15) comes from a high fraction of avoided generator startups andshutdowns as well as avoided non-steady state operational costs. One consequence (which will beshown more explicitly in the discussion of energy storage value in Section 3.4) is that the value ofdemand response (orenergy storage)declines rapidlywith the levelofdeployment.This results fromthe small overall opportunity to avoid these specific high operational cost items and the relativelyinfrequentspikesinthecostsforenergy.

In addition to examining the total operational value of demand response resources, the study alsoexamines the operational value of resources by region and by type (see Figures 3-16 through 3-19).Implied market values based on marginal costs are used to disaggregate the value of the discreteservices;productioncostsarenotdistinctonaregionalbasisduetosignificantinter-balancingauthorityareatransactionsthatoccurinthemodeling.Ingeneral,theregionaldifferencesinvaluesfordemandresponse resources mostly follow the regional patterns of operating reserve marginal costs. Forinstance,examiningFigure3-5,theaveragemarginalcostoffrequencyregulationissixtimeshigherinNewMexicocomparedwithArizonaleadingtosimilarly largedifferences in impliedmarketvalue.Themodeled operational values by region are a good indicator of how the availabilities of each type ofdemand response resource correlatewith times of high system costs. Thosewith higher correlationstendtohavehighervaluesandthosewithlowercorrelationstendtohavelowervalues.

Anumberofotherobservationscanbedrawn fromthedatapresented inFigures3-16 through3-19.Across regions, cooling loads consistently have low net values for energy transactions (gray bars infigures).Bothcommercialandresidentialcoolingend-useshaverestrictionsregardingthetimingofloadshedding and load recovery (i.e., limited flexibility) due to the assumed limits on thermal inertia ofbuildings and residences.While cooling loads and their associateddemand response availabilities arecorrelatedwithtimesofhighenergycosts,theyarelimitedintheirabilitytotakeadvantageoflowcostoff-peakenergyforloadrecovery.Fortheseend-useloads,thevalueofloadsheddingisnearlycanceledoutbythecostsofloadrecovery.Eventhoughtheenergyarbitragevalueofcoolingload-baseddemandresponse is low, theproduction costmodel still utilizes these resources in order to reduce generator

30

startups.Fromamarketperspective,thesedemandresponseresourcesmayearnlimitedrevenues,buttheycanstillhelploweroverallproductioncosts.

31

Figure3-16.Operationalvalue(impliedmarketvalue)basedonmarginalcostsofproductionfordifferenttypesofdemandresponseresourcesintheWesternInterconnectionmodelatlowandhighlevelsofwindandsolargeneration

32

Figure 3-17. Operational value (implied market value) based onmarginal costs of production for different types ofdemandresponseresourcesintheWesternInterconnectionmodel,atlowandhighlevelsofwindandsolargeneration

33

Figure3-18.Operationalvalue(impliedmarketvalue)basedonmarginalcostsofproductionfordifferenttypesofdemandresponseresourcesintheWesternInterconnectionmodel,atlowandhighlevelsofwindandsolargeneration

34

Figure3-19.Operationalvalue(impliedmarketvalue)basedonmarginalcostsofproductionfordifferenttypesofdemandresponseresourcesintheWesternInterconnectionmodel,atlowandhighlevelsofwindandsolargeneration

35

TextBox:MeasuringDemandResponseValue

Unlike power plants that have standard measures, there appears to be no consistent and clearmeasure for demand response. Themeasurebywhichdemand response is valued (i.e., dollar perwhat)hasimportantimplicationsforunderstandingtherelativevalueacrossdifferentresourcetypesandscenarios.Forconventionalgenerators(andenergystorage),thedenominatorisoftenkW-year.This gives the value per kW of a generator summed over the year. The size of conventionalgeneratorsandenergystorage,intermsofkW,iswell-definedandprovidesagoodmeasureforthedevicecost(inthecaseofenergystorageitrequiresknowledgeofthenumberofhoursofstorage).Asimilar metric could be used for demand response; however, there are several importantconsiderations.For instance, thecapacity rating fordemandresponsecould bebasedon thepeakenabledcapacityortheaverageloadoftheenabledend-useresource.

As discussed, thedemand response resource (!) isdefinedas the flexibility (!) times the end-useloadprofile(!)andtheflexibilityisthesheddability(!)timestheparticipationrate(!):

!!,!(!) = !!,!(!) ∙ !!(!) = !!!,!(!) ∙ !!,!(!)! ∙ !!(!)

whichistime-dependent(!)andspecifictoeachtypeofload(!)andeachtypeofbulkpowersystemproduct(!).Theparticipationrateisdefinedas:

!!,!(!) = min !!!,!(!),!!,!(!)!

wheretheparticipationrateequalstheminimumofcontrollability(!)andacceptability(!).Wecanthendefinethepeakenabledcapacity(!!)as:

!!!,! = max!

!!!,!(!) ∙ !!(!)! = max!

!!!,!(!)!

Wecanalsodefinetheaverageenabledend-useload(!!)as:

!!,!!!!!! = 1! ∙! !!!,!(!) ∙ !!,!(!)!

!!!

!!!= 1! ∙!

!!,!(!)!!,!(!)

!!!

!!!

where (!) is the number of time intervals (i.e., 8,784 hours in the year 2020). Figure 3-20 showsthese values for commercial building cooling (top) and commercial building lighting (bottom) end-uses.Coolingloadshaveprominentpeaksinbothloadanddemandresponseavailability,leadingtotheenabledpeakcapacitybeingmuchlargerthantheaverageenabledend-useload.Theoppositeistrueforlightingloads.

Eachof thesemetrics leads to difficulties incomparing valueacrossdifferentend-use types.Thus,thisstudyusesmaximumcumulativeavailability(!)fordemandresponse:

!!,! = !max!,!

!!!,!(!) ∙ !!(!)! !!!

!!!

36

Figure3-20.Maximumhourly availabilities for energy andoperating reservesusing commercial building cooling loads(top)andcommercialbuildinglightingloads (bottom) intheNorthwestregion,showingacomparisonofpeakenableddemandresponsecapacityandaverageelectricloadfortheenableddemandresponseresource

Themaximumcumulativeavailabilityisthemaximumavailabilityforallbulkpowersystemproductsateachhourandsummedoverthestudyyear.Thisprovidesanassessmentofhowtheavailabilityofdifferent demand response resources matches system needs. Within an individual region, thoseresources more correlated with high production cost times will have greater values in terms of$/MW-h of availability; those less correlated will have lower values. However, none of the threediscussedmeasures provides a good indication of the costs associatedwith enabling the demandresponse resource. Enablement costs may have large fixed costs that scale with the number ofenabled sites, rather than the total size or timing of the resource availability. Additional work isneededtobetterassessdemandresponseenablementcosts.

37

In contrast, other end-use loads, such asmunicipal freshwater andwastewater pumping, havemoreenergy scheduling flexibility and are assumed to be capable of shifting energy throughout a 24-hourperiod.These loadsprovide,onaverage, twiceasmuchvalue for loadsheddingthantheypay in loadrecoverybecausetheyarecapableofarbitragingbetweenthehighestandlowestenergycosthoursofthe day. Figure 3-21 provides an illustration of how different resources vary in their ability to takeadvantageofenergyarbitrageopportunities.Thefigurecomparesthevalueofenergyshiftingachievedin the simulation to a theoretical value derived from the difference between the two highest hourlyenergy costs and the two lowest hourly energy costs daily. Two-hour arbitrage value is chosen forcomparisonbecausenoresource is foundtobecapableofexceedingthisbenchmark.Municipal loadsprovide nearly themaximum value,whereas commercial and residential cooling loads provide only asmallfractionofit.Ofnote,ourresultsarebasedonstaticestimatesofenergyschedulingflexibility,sopricesignalscouldaltertheratesandpatternsofcustomerparticipation,changingflexibility.

The modeled demand response resources provide greater value from the provision of operatingreservesovertheshiftingofenergyuse.Thecalculatednetmarketvalueforoperatingreservesisabout3timesgreaterthanthatforenergyshifting.Thegreatervaluestemspartlyfromthefactthatoperatingreserves are held every hour of the year, but energy shifting is a daily activity with significantoperationalconstraints.However,thisvaluewillbeimpactedifactualdispatch-to-contractratiosdiffersignificantly with study assumptions (i.e., frequency regulation mileage). As shown in Figure 3-22,resources capable of providing operating reserves are typically utilized close to their maximumavailabilities.However,energyshifting-onlyresourceshaveutilizationfactorsoflessthan20%.Overall,thedemandresponseresourcesmodeledprovideabout25%ofthetotaloperatingreservesrequiredinthe Western Interconnection for the study year 2020; in some regions, the share for specificrequirementscanbeashighas50%orlessthan1%(seeTable3-8).Table3-8alsoshowsthatramping

Figure3-21.Comparisonoftheaveragenetvaluethatresourcescanprovideforenergyshiftingtoatheoreticalvaluebasedonthedifferencebetweenthetwohighestandtwolowesthourlymarginalcostsforenergyforeachday.A2-hourenergyarbitragereferencewasselectedasnoresourceisfoundtobecapableofexceedingthisbenchmark(i.e.,they-axishasnoresourcegreaterthan100%).

38

reserveprovisionsbydemand responseare low in themodeled scenarios, reflecting the lowvalueoframping reserves comparedwith frequency regulation and contingency reserves. Ramping reserve isassumedtohave lessstringentrequirementsthanotheroperatingreserves(seeTable3-5)andwouldgenerallybelessexpensivetoprocure(Hummonetal.2013).

Table3-8.FractionofDemandResponseProvisionintheLowandHighRenewableCasesoftheWesternInterconnectionModel.Foroperatingreserves(frequencyregulation,contingencyreserve,andrampingreserve),thefractionprovisioniscalculatedbydividingtheamountofserviceprovidedbythetotalamountneeded.Forenergy,thefractionprovisioniscalculatedbydividingtheamountofenergyshiftedthroughdemandresponsebythetotalelectricityusage.

Freq.Regulation ContingencyReserve RampingReserve Energy

LowRE HighRE LowRE HighRE LowRE HighRE LowRE HighREArizona 48% 46% 15% 14% 1.2% 0.8% 0.3% 0.3%CaliforniaNorth 49% 48% 24% 24% 16% 13% 0.1% 0.1%CaliforniaSouth 41% 42% 36% 36% 7.6% 4.9% 0.2% 0.2%Colorado 30% 16% 14% 14% 0.1% 0.4% 0.1% 0.1%Idaho 37% 32% 20% 20% 1.8% 1.2% 0.3% 0.3%Montana 22% 4.6% 7.6% 6.6% 0.2% 0.4% 0.1% 0.1%NevadaNorth 26% 12% 24% 22% 8.6% 1.6% 0.2% 0.2%NevadaSouth 37% 34% 19% 17% 1.9% 0.6% 0.3% 0.3%NewMexico 49% 24% 23% 22% 0.1% 0.4% 0.2% 0.2%Northwest 36% 35% 15% 15% 11% 9.0% 0.1% 0.1%Utah 38% 43% 23% 19% 0.1% 0.1% 0.1% 0.1%Wyoming 9.3% 5.5% 6.0% 4.8% 0.4% 0.2% 0.1% 0.1%Lastly, thereareonlyafewdemandresponseresourcestypes(reflected inFigures3-16through3-19)forwhichthenetvalueofdemandresponsechangesappreciablybetweenthelowandhighrenewablecases.Onaverage, fromtheperspectiveofbothproductioncost savingsor impliedmarketvalue, thedifferencebetweenthelowandhighrenewablecasesis lessthan10%.Formostservices,theaverageoperational value is similar between the low and high renewable cases. However, the average

Figure3-22.UtilizationofdemandresponseresourcesintheWesternInterconnectionmodelbasedontheratioofdemandresponseprovisionedtothemaximumcumulativeavailability.

39

operationalvalueforspinningcontingencyreservesgenerallydeclines.Thisresultsfromthechangesintheutilizationofconventionaldispatchablegenerationunderincreasedlevelsofwindandsolarpowerinthemodeledscenarios.Duetoincreasedvariability,moreconventionaldispatchablegenerationmaybeutilizedatpart-load,increasingthesupplyofoperatingreservesavailableinherentlythroughenergyscheduling. The role of various drivers of operational value under changing levels of wind and solarpenetrationisdiscussedfurtherinSection3.5.

3.4 EnergyStorageEach energy storage technology, including batteries, flywheels, compressed air, and pumped storagehydropower, has different capabilities and cost structures. While there are numerous technicaldifferencesamongandwithinthesetechnologytypes,weconsiderthegeneralcharacteristicsofenergystorageratherthanattemptingtodeterminethevaluesofindividualenergystoragetechnologies.

3.4.1 StorageCharacterizationInthisstudy,weexaminetwogeneralclassesofenergystoragetechnologies; thefirst isanoperatingreserves-onlydevice,which resemblesahigh-power, shortdurationbattery.Weassume thedevice isnotramp-constrainedandcanprovideitsfulloutputrangeforoperatingreservesinstantly.Foradeviceproviding spinning contingency reserves, we assume that the device simply provides up to its fulldischarge capacity without incurring any operational costs, and we do not consider the energycomponentofacontingencyevent.Forfrequencyregulation,wealsoassumethedevicecanprovideuptoitsfullcapacityandthattheserviceisnet-energyneutralineachone-hoursimulationinterval.

Evenifthefrequencyregulationserviceisnet-energyneutralovertime,therewillbeenergylosses.Thiswillproduceanetconsumptionofenergybythestoragedevice,whichistheproductoftwofactors:thefraction of capacity used to provide energy (sometimes referred to as the dispatch-to-contract ratio)(Kempton2005)andthedeviceefficiency.Thefirstfactordependsontheamountofenergythatflowsthrough thedevicewhen called toprovide frequency regulation. This energy ismultipliedby the lossratetoproducetheamountofenergyconsumedbythestoragedevice.Adispatch-to-contractratioof14.2%ischosen(Ferreira2013)andanefficiencylossrateof20%wasselected,basedonanetround-triptripefficiencyof80%(similartoa lithium-ionbattery)(Akhiletal.2015). Inotherwords, foreachhour,astoragedeviceproviding100MWofregulationconsumes2.8MWhofenergy.Thecostofthismake-up energy is assumed to be at the marginal cost of energy and loss calculations are handledoutside the production cost model. This approach was taken because deployment of provisionedoperatingreservesisnothandledinthemodelingsoftware,asdiscussedinDenholmetal.(2013).

The second general class of energy storage technologies is one that can provide both energy andoperating reserves, a device that resembles a high-energy, long duration battery. The simulationsincludescenarios inwhich thedevicesprovideonlyenergyandscenarios inwhich thedevicesareco-optimizedtoprovidebothenergyandoperatingreserves(whichallowsfordisaggregationofthevalueof the two classesof services). Themodeleddevices canprovidebothenergyandoperating reservessimultaneously,butonlyuptotheratedcapacityofeachdevice.Weassumea75%round-tripefficiency(similartoasodium-sulfurbattery)(Akhiletal.2015)andthatthedevicescanrampoveritsentirerangewithinasingleenergyscheduling interval (withnominimumgeneration levelandtheabilitytoswitch

40

between charging and discharging within a single scheduling interval). We also assume constantefficiency as a function of load and state of charge, nominimum up or down times, and 8 hours ofenergystorageattheratedcapacity.Lastly,nofixedorvariableoperationsandmaintenancecostswereassignedtoeitherclassofenergystoragedeviceforsimplicity.

Two sets of scenarios are investigated to analyze the value of energy storage, and how the valuechanges as a function of the amount deployed and the level of renewable penetration. As with thedemand response simulations (seeSection3.3),weevaluateaWestern Interconnectionmodelwherevariousamountsofenergystoragearedeployed inthedifferentregions.Energystoragedeviceswereaddedwithout replacementofexistinggenerators in the regions.However,due to the longmodelingrun times, only a limited number of scenarios could be evaluated with the entire WesternInterconnectionmodel.Asaresult,wealsoanalyzeamorecomprehensivesetofmodelingscenariosinthe Colorado systemmodel (see Section 3.5) to examine, by extension, the general trends in energystoragevalueasafunctionoftheamountdeployedandthelevelofrenewablepenetrationobservedintheWesternInterconnectionmodel.

Figure 3-23 shows the modeled deployment of energy storage in each region. For the operatingreserves-onlyscenarios,modeleddevicesaresizedtomeet50%oftheaveragehourlyoperatingreserverequirements within each region (see Table 3-2) and are distributed to the various regions in theWestern Interconnection based on these requirements. While somewhat arbitrary, this size is largeenough to represent a significant contributionwithin the system, but not so large as to saturate theoperating reserves requirement. Requirements differ between frequency regulation and contingency

Figure3-23.Deploymentofenergystorageinthedifferentscenarios.Thetotaldeploymentofregulation-onlydevicesis487MWinthe lowrenewablecase(L)and597MWinthehighrenewablecase(H).Contingency-onlydevicestotal1,314MWandco-optimizedenergyandoperatingreservedevicestotal3,045MWinboththelowandhighrenewablecases.

41

reserves and lead to a totalmodeled deployment in theWestern Interconnectionmodel of 487MW(597MWinthehighrenewablecase)and1,314MW,respectively.

Fortheenergy-onlyandco-optimizedscenarios,largerenergystoragedeploymentsarechosenequaltoabout3.3%ofaverageloadineachregion,resultinginatotalmodeleddeploymentof3,045MW.Thisissubstantially larger than the operating reserve-only scenarios to account for the potential additionalrevenue streams associated with energy transactions. In addition, we examine two other levels ofdeployment to understand the impact of energy storage penetration on its value: one inwhich eachmodeledenergystoragedeviceis50%smallerandoneinwhicheachis50%larger.

3.4.2 EnergyStorageOperationalValueThis subsection examines the operational value of energy storage in the low renewable and highrenewablecaseoftheWesternInterconnectionmodel,asdescribedinSection3.1.Weapplyasimilarapproach as used to calculate the operational value of demand response in Section 3.3.3. We firstcalculate the total production cost in the base system without the deployment of additional energystoragedevices.Wethenaddtheamountsandtypesofenergystorage(summarizedinFigure3-23)tothevariousregionsandcalculatethedifferenceinproductioncosts.Anyreductioninoperationalcostisattributedtothedeploymentofenergystorage,representingitssystemoperationalvalue.Inaddition,weusethemarginalcostsofenergyandoperatingreservestoestimatetherevenuethattheresourcescouldexpect to receive inamarketenvironment.Asdiscussedpreviously inSection3.2, thisassumesthat the market clearing prices correspond to the marginal costs calculated by the production costmodel.

Thediscussionbeginswithanassessmentofthescenarioswithoperatingreserves-onlyenergystoragedevicesandisfollowedbyanassessmentofthescenarioswithlargerdevicesthatarecapableofbeingco-optimizedforenergyandoperatingreserves.

Figure3-24providestheoperationalvalueforenergystoragedevicesprovidingonlyoperatingreservesfor the low renewable case and the high renewable case. The bar charts on the left represent theproductioncostsavingsnormalizedbytheamountofenergystoragedevicesdeployedforthespecificscenarios (regulation-only and contingency reserve-only) in each case. In Figure 3-24 (low renewablecase,left),introductionofthefrequencyregulation-onlydevicesreducetotaloperationalcostsby$16.7million.Dividingthisvaluebythetotalinstalledstoragecapacity(487MW)producesanaveragevalueofabout$34/kW-year.Thefigurefurtherbreaksthissavingsintothecategoriestrackedbytheproductioncostmodel.Steadystatesavingsrepresentsthefuelandoperationsandmaintenancecostsavoidedbymoreefficientdispatchthatresultswhenthermalplantsarenotrequiredtoprovideoperatingreserves.For frequency regulation, the steady state benefits are somewhat offset by the required make-upenergyassumedinthisanalysis.Startupandshutdownsavingsrepresentstheabilityofenergystorageto reduce costs associatedwith plant starts. For regulation-only devices, the largest net cost savingsstem from avoided non-steady state operation costs associated with thermal plants following aregulation signal. This includes both additional maintenance costs as well as heat rate degradationcaptured in the regulation cost described in Table 3-3. For the scenarios with devices that providecontingencyreservesonly,theper-unitvalueislessthanthatoffrequencyregulation-onlydevices.The

42

calculatedsavingsfromuseofcontingencyreserves-onlydevicesismostlyassociatedwithreducedpart-loadoperationandmoreefficientoverallcommitmentanddispatchofthegeneratorfleet.

Figure3-24 (right) represents the impliedmarketvaluebasedonoperatingreservesbeingpaidat thehourlymarginal costs of production (as calculated by the production costmodel). For the regulationdevices,weincludethesavingsachievedthroughtheprovisionofreservesaswellastheassumedcostofmake-upenergy,withthenetmarketvaluerepresentedbytheblackline.Theimpliedmarketvaluesarelessthanthetotalproductioncostsavingsduetotwofactors.First,thecostsofgeneratorstartupsarenotcapturedinmarginalcostsofproduction.Second,theadditionofenergystorageintothesystemwillgenerallyreducemarginalcosts(i.e.,marketclearingprices),whichcanhaveasignificantimpactonthecalculatedvalue(EPRI2013).Thehighrenewablescenariosshowamodestincreaseinthevalueofenergystoragecomparedtothelowrenewablescenarios.Thisincreasecanbeattributedtothelarger

Figure 3-24. The operational value of energy storage providing either frequency regulation only or spinning contingencyreservesonly.Value isbasedonproductioncostsavingstothesystem(left)and impliedmarketvaluebasedonmarginalcostsofproduction(right).

43

operating reserves requirements in the high renewable case and the lower cost of energy used formake-up.ThesefactorsarediscussedinmoredetailinSection3.5.

Figure3-25provides the results for theenergy-only and co-optimizedenergy storage scenarios.As inFigure3-24,theleftpanelprovidestheproductioncostsavingsassociatedwiththeadditionofenergystorageforthelowandhighrenewablecases,normalizedbytheamountdeployed,andtherightpanelprovidestheimpliedmarketvalue.

Figure3-25.Theoperationalvalueofenergystorage(co-optimizedforenergyandoperatingreserves)intermsofproductioncostssavings(left)andimpliedmarketvalue(right),inthe lowrenewable(top)andhighrenewable (bottom)casesoftheWestern Interconnection model. The ‘trade off’ represents the reduction in implied market value achieved from energyarbitragealonedue toenergy storage capacityused for theprovisionofoperating reserves, leading to greaternet valuethroughco-optimization.

44

Theproductioncostmodel recordshowmuchenergyandoperating reservesareprovidedby theco-optimizeddevicesintheenergy-onlyandco-optimizedscenarios.Usingthehourlyenergystorageplantdispatchdataandthecalculatedmarginalcosts,wecanacquireasenseofthetradeoffsamongpossibleservicesenergystoragecanprovide.ThisisdemonstratedinthechartsontherightsideofFigure3-25.Byparticipatinginenergytransactions,aco-optimizedenergystoragedeviceforegoessomeofitsabilitytoacquirevaluefromprovidingoperatingreserves,andviceversa.Comparingtheenergy-onlyandco-optimizedscenarios,theco-optimizeddevicesgainaboutdoublewhatthey lose inthetrade-offso, inaggregate,theyprovidemoreoperationalvalue.

AcomparisonofFigures3-24and3-25showsthattheper-unitvalueoftheco-optimizeddevicesislessthan the frequency regulation-only devices. This counter-intuitive result is partly due to the differentamountsof energy storagedeployed in thedifferent scenariosmodeled. The totaldeploymentof co-optimizeddevicesismorethanfourtimesgreaterthanthefrequencyregulation-onlydevices.Becausetheincrementalvalueofenergystoragetendstodeclinewithadditionaldeployment(aswithdemandresponseandotherresourcesaddedtoanunconstrainedsystem),theper-unitvaluetendstobesmallerfor larger deployments of energy storage devices. Furthermore, the large energy capacity (8-hourduration at the rate power) of the co-optimized devices introduces an additional driver that impactsvalue.Whilethedevicesinthefrequencyregulation-onlyscenarioarenearlyfullyutilizedproviding49%oftherequirement(thescenariodeployedenoughtomeet50%oftherequirement),thedevicesintheco-optimized scenario, while significantly larger, provide only 27% of the frequency regulationrequirement. Despite the lower provision of frequency regulation by the co-optimized devices, anindicator that the frequency regulation requirement is not saturated by the devices, the averagemarginalcosts for frequencyregulationare lower thanthat in the frequencyregulation-onlyscenario.When providing frequency regulation, the co-optimized devices can provide other services likecontingency reserves and energy shifting with the remaining capacity.When discharging energy, theenergy storage device suppresses the marginal cost of energy, which in turn also suppresses themarginalcostofoperating reserves (note that themodeldoesnotcapture thiseffect for themakeuplosseswhenenergystorageprovidesfrequencyregulation).Asaresult,thelargerco-optimizeddevicesmodeledprovideacombinationofserviceswhoseper-unitvalueisactuallylessthanthemuchsmallerfrequencyregulation-devices.

Consequently,quantifyingthevalueofenergystoragecannotbedisentangledfromtheassumedlevelofdeployment.Figure3-26showstheresultsoftheco-optimizedscenarioforthethreelevelsofenergystorage deploymentmodeled, with an approximate curve fit to illustrate this relationship. The curveshows the operational value based on production costs savings as well as the implied market valueassuming resources are paid at themarginal costs (i.e.,market clearing prices). From this result, it isevidentthattheoperationalvaluesofenergystorageindifferentregionswouldchangesubstantiallyiftheassumeddeploymentlevelsdifferedfromourmodeledscenarios.

45

3.5 OperationalValuewithVariableGenerationSections 3.3 and 3.4 provide an indication of the operational value of demand response and energystorage in a large system (i.e., theWestern Interconnectionmodel). However, they do not provide a

Figure3-26.Theoperationalvalueofenergystorage(co-optimizedforenergyandoperatingreserves)asafunctionoftotaldeployment in the lowandhigh renewable casesof theWestern Interconnectionmodel. The value calculated frombothperspectivesdeclinewithgreaterdeploymentlevels.

Figure3-27. Impliedmarket valueofenergystorage(co-optimizedforenergyandoperatingreserves) inthe lowandhighrenewablecases,assumingpricesarebasedonthemarginalcostsofproduction

46

detailed understanding of why the value of these resources changes with increased penetration ofvariablegeneration.

Theoperationalvaluesofbothdemandresponseandenergystoragedevices(capableofprovidingbothenergy and operating reserves) are largely driven by two primary factors: the required amount ofoperating reserves in a system and the associated cost, and the time-varying cost of energy in thatsystem,includingthedifferencebetweenon-peakandoff-peakmarginalcostsofenergy.Duetosystem-wide effects, there are competing drivers whose net effect can increase or decrease the value ofdemand response or energy storagewithin a power system. As a result,while generalizations drawnfrom a limited set of scenarios cannot be applied to specific market environments, the analysis canidentifyimportantdriversofvalueandtheirexpecteddirectionalityinimpact.

Toexplorethese issues, thestudyusesasetofscenarioswith increasing levelsofvariablegenerationandcalculatesthechangestototalproductioncostsandimpliedmarketvaluebasedonmarginalcostsofproductionassociatedwithenergyandoperatingreservesgeneratedfromtheproductioncostmodel,asdiscussedinSection3.2.ThescenariosarebasedontheColoradosystemmodeldescribedinSection3.1.Itsrelativelysmallsizeallowsforsimulationofmanycombinationsofgeneratorcompositionsandcharacteristicswithsignificantlyshortercomputationaltimes.Wefirstexaminetheimpactofrenewablepenetration levelsonthecostofoperatingreserves,akeysourceofvalue forbothdemandresponseand energy storage. Scenarios vary from 15% to 35% renewables (by energy) in approximately 5%increments.ReserverequirementsforeachlevelofrenewablegenerationarebasedonthemethodsinIbanezetal.(2012).Wethencalculatethecostofsystemoperationwithandwithoutholdingreserves(thedifferencebeing the total cost to thesystem forprocuring reserves)aswellas themarginal cost(i.e.,price)ofeachoperatingreserveproduct.

Table3-9.FractionofRenewableGenerationfromWindandSolarResourcesintheColoradoSystemModelandtheAssociatedModelRequirementsforOperatingReserves

GenerationfromRenewables

WindGeneration(GWh)

SolarPVGeneration(GWh)

CumulativeOperatingReserveRequirement Regulation(GW-h)

Contingency(GW-h)

Ramping(GW-h)

15% 10,705 (13.6%) 1,834 (2.3%) 1,050 3,548 502

20% 13,838 (17.4%) 2,556 (3.2%) 1,134 3,548 600

25% 18,097 (22.8%) 3,168 (4.0%) 1,281 3,548 769

30% 21,433 (27.0%) 3,750 (4.7%) 1,364 3,548 855

35% 23,752 (29.9%) 4,260 (5.4%) 1,422 3,548 918

Table3-9summarizesthequantityofoperatingreservesrequiredinthevariousscenariosandTable3-10 gives cost values associated with those reserves. For these scenarios, the frequency regulationrequirementincreasesfrom1.1TW-hto1.4TW-handtherampingreserverequirementincreasesfrom0.5TW-hto0.9TW-h.Despitethelargeincreasesinoperatingreserverequirements,thereisarelativelymodestincreaseinthecostsofprovidingtheseoperatingreserves.Therelativelysmallincreaseincostsispartlyduetodecreasesinenergyproductioncostinasystemwithincreasedlevelsofwindandsolar

47

power as well as energy schedule changes that reduce conventional generation output, releasingcapacitytoprovideoperatingreserves.

Table3-10.TotalCostsandMarginalCostsAssociatedwithOperatingReservesatDifferentLevelsofWindandSolarGenerationintheBaseColoradoSystemModelwithoutAdditionalDemandResponseorEnergyStorage

GenerationfromRenewables

TotalProductionCost($M)

OperatingReserveTotalCosts OperatingReserveMarginalCosts($/MW-h)

TotalCosts($M)

UnitCosts($/MW-h)

RegulationMean/Median

ContingencyMean/Median

RampingMean/Median

15% 1,427 27.4 (2.0%) 5.4 15.48/13.81 6.15/3.32 1.62/0

20% 1,309 29.1 (2.3%) 5.5 16.31/14.61 6.14/3.09 2.05/0

25% 1,170 32.3 (2.8%) 5.8 16.95/14.58 5.88/2.81 2.21/0

30% 1,072 32.1 (3.1%) 5.6 16.81/14.52 5.31/2.61 2.3/0

35% 1,003 31.2 (3.2%) 5.3 16.53/14.52 5.04/2.51 2.18/0

Figure 3-28. Example days in the Colorado systemmodel showing the contingency reserves requirement (gray solid line)comparedtothemarginalcostofcontingencyreserves(greensolidline).Themarginalcostofcontingencyreservefallswhenthere is sufficient excess operating range and ramping capability (purple solid line) to meet the operating reserverequirementsolelythroughenergyscheduling.

Energyschedulingaloneleadstosomeresidualoperatingrangeandrampingcapabilitythatprovidesaninherentlevelofoperatingreserves.Figure3-28illustrateshowtheinherentcapabilityofthesystemtoprovide (spinning) contingency reserve (purple line) varies over three model days. Generators havevarious constraints such as startup times, minimum operating points, and minimum up times.Consequently, even if therewerenouncertainty innet load, theoptimal schedulingof generators tofollownet loadwould still result in somehourswithunusedspinningcapacity (Hummonetal.2013).Any remaining need for operating reservesmust bemet through a re-dispatch of the system (i.e., achange from the optimal energy schedule), which subsequently incurs additional costs (i.e., lost

48

opportunity costs). The level of inherent capability is an important factor in the value of demandresponseandenergystorage.AsshowninFigure3-28,themarginalcostofcontingencyreserves(greenline)tendstobelowwhenthenetrequirement(dashedblackline)isalsolow,signifyingthatthesystemcansupplyalargefractionofcontingencyreserveswithoutincurringlostopportunitycosts.

Understanding inherent capability is particularly important to understand how changes in thecomposition of the generator mix, especially in cases with increasing levels of variable renewablegenerationasdiscussedinthissection,impactthecostsofoperatingthesystem.Asystemthatisabletoaccommodate variable generation will necessarily have more inherent operating range and rampingcapabilityduetoenergyschedulingalone.SimulationsoftheColoradosystemmodelwiththeadditionof an energy storage device demonstrate this effect in more detail. We add a 100-MW frequencyregulation-onlydevicetothesystemandcalculateitsvalueasafunctionofrenewablepenetration(withscenariosupto55%renewablegeneration).

Figure3-29.Valueofa100-MWfrequencyregulation-onlyenergystoragedeviceasafunctionofrenewablegeneration intheColoradosystemmodel.ProductioncostsavingsandimpliedmarketvaluebasedonmarginalcostsofproductionfortheColoradosystemareshown.

Figure 3-29 provides the results for the frequency regulation-only device, with operational value interms of production cost savings and implied market value. It shows a modest increase and thendecreaseinoperationalvalueswithincreasingpenetrationofwindandsolarpower.Thisbehaviorstemsfromacompetitionbetweenthe increasing frequencyregulationrequirementathigher levelsofwindandsolarpenetration(whichleadstoincreasedoperationalvalueoftheenergystoragedevice)andtheincreasinglevelofinherentcapability(whichleadstodecreasedoperationalvalueoftheenergystoragedevice). The impactof these factors canbe seenmore clearly inTable3-10,whichprovides the totalcostsandmarginalcostsofoperatingthebasesystematdifferent levelsofwindandsolargenerationwithoutadditionaldemandresponseorenergystorage.Notethattheinherentcapabilityleadstozeromarginalcostsforrampingreserveformorethanhalfofallhours.

49

Aswith previous comparisons, Figure 3-29 also shows a lower quantity for the impliedmarket valuecomparedwiththeoperationalvaluebasedonproductioncostsavings.Thisresult isdriven largelybysuppressionofthemarginalcostsoffrequencyregulation(asshowninFigure3-30)andotheroperatingreserve products due to saturation from excess capacity. In the scenario with the lowest amount ofrenewable penetration, the energy storage device provides about 82% of the system’s regulationrequirement and reduces the averagemarginal cost of regulation by about 60%, from $16/MW-h to$6.5/MW-h.

InFigure3-29,theproductioncostsavingsarerelativelyflatwithchangingpenetrationofwindandsolarpower,but themarketvalue increasessignificantlybetweenscenarios inwhichwindandsolarpowerincreasefrom15%to44%ofgeneration.Thecontrastbetweenthetwomeasuresofoperationalvaluestemfromseveraldrivers.Athigherlevelsofwindandsolarpower,themodeledfrequencyregulationrequirementisalsolarger.Becausethesizeoftheenergystoragedeviceisconstantbetweenscenarios,at100MW,theenergystoragedeviceprovidesaprogressivelysmaller fractionoftherequirementasrenewable penetration increases. By providing a smaller fraction, there is an associated smallersuppressionofmarginalcostsforfrequencyregulation.Thissuppressionoccursbecauseenergystorage(as well as demand response) provides the regulation service at zero lost opportunity cost. This isdemonstratedbyFigure3-30,whichshows theaveragemarginalcost for frequency regulationbeforeandaftertheadditionofthefrequencyregulation-onlydevice.

The second issue evaluated is the relationship between renewable generation andmarginal costs ofenergy. The introduction of any zero marginal cost source of generation into an otherwise staticgenerationmixwill tend to reduce operational costs during the period of generation. The impact onenergy costswill largely dependon the timing of the renewable generation. A renewable generationsourcethatgeneratespredominantlyduringpeakperiodscouldacttoreducepeakmarginalcostsandreduce potential operational value for demand response or energy storage providing energy shifting.

Figure3-30.Averagemarginalcostof frequencyregulationasafunctionofrenewablepenetrationintheColoradosystemmodelatdifferentlevelsofrenewablegenerationandtheimpactofdeployingenergystorage

50

Alternatively, renewable generation during off-peak periods will decrease off-peak marginal costs,reducingthecostsofenergystoragecharging(orloadshifting)andincreasepotentialoperationalvalue.Ultimately,itistheoverallimpactofrenewablegenerationonon-peakandoff-peakmarginalcoststhatwillaffectthepotentialopportunitiesfordemandresponseandenergystoragetoprovideoperationalvaluetothesystem.

SomeindicationofenergyshiftingvalueintheColoradosystemmodelcanbederivedfromthemarginalcostdurationcurvesforenergyillustratedinFigure3-31.Curvesareshownforthreelevelsofrenewablepenetrationandtwonaturalgasprices(atabout$4.1/MBtuand$8.2/MBtu).Thehighergaspriceleadsto largerdifferentials in themarginal costs for energy. In the lowest renewable case (blue lines), thisdifferential formanyhoursof theyear isbasedon thedifferencebetween theoperatingcostof coalunits (generatingatabout$20/MWh)andnaturalgascombinedcycleunits (generatingatabout$25–$35/MWhinthelowgaspricescenarioorabout$45–$70/MWhinthehighgaspricescenario).Ifenergystorage isused for shiftingenergyproducedby coalunits todisplaceenergyproducedbynaturalgasunits in the low renewable, low natural gas price scenario (solid blue lines), the net operational costsavingswill be small, especially considering losses associatedwith the round trip efficiencyof energystoragedevices.Thenetoperationalsavingswillbegreaterinthelowrenewable,highnaturalgaspricescenario(dashedblue line),buttherearerelativelyfewhoursduringwhichenergyfromoff-peakcoalunits is available for charging energy storage devices.When energy from coal units is not available,energystoragedevicesmustchargewithenergyfromnaturalgascombinedcyclegenerationtodisplacehigher cost, lower efficiency peaking units (e.g., natural gas- or oil-fired combustion turbine (CT),internalcombustion,orsteamunits).Theseunitshaveawiderangeofmarginalcosts,asillustratedbythelumpinessontheleft-handsideofthepricedurationcurves.However,therearestillrelativelyfewhours of very high marginal costs for energy during which there are significant opportunities for

Figure3-31.MarginalcostdurationcurveforenergyintheColoradosystemmodelforrenewablepenetrationsof16%,32%,and45%andtwonaturalgasprices(about$4.1/MMBTuand$8.2/MMBTu).

51

operationalcostsavingsfromenergyshifting.Theincreasedpenetrationofrenewablesresultsinloweroverallmarginal costs forenergy,butwithanapparentgreater impactonoff-peakmarginal costs forenergy(rightsideofcurvesinFigure3-31).Movingfromthe16%casetothe32%casegreatlyincreasesthe number of hours off-peak coal is available (hours during which marginal costs are less than$20/MWh).At thehigher renewablepenetration case (45%), there aremorehoursof zero-costwindandsolaravailable to chargeenergy storagedevices,presentinggreateropportunities foroperationalcostsavingsfromenergyshifting.

TheimpactofrenewablepenetrationonthevalueofenergystorageisfurtherexaminedwiththeresultsshowninFigure3-32,whichillustratestheoperationalvalueofa300-MWco-optimizedenergystoragedeviceintheColoradosystem.Theleftpanelshowsthepotentialoperationalvalueasmeasuredbythereductioninproductioncosts.Itshowsthegeneralincreaseinvalueassociatedwithbothsteady-stateoperation (largely associatedwith energy shifting) and non-steady-state operation (largely associatedwithprovidingfrequencyregulation)asrenewablepenetrationlevelsincrease.

Figure3-32.Comparisonofoperationalcostsavingsandimpliedmarketvaluefora300-MWco-optimizedenergystoragedeviceintheColoradosystemmodel.Operationalvaluesareprovidedfor5levelsofwindandsolargeneration,from16%to55%ofenergy.

Therightpanel inFigure3-32providesthesystemimpliedmarketvalue,assumingtheenergystoragedevice is paid at themarginal cost of energy and operating reserves (and pays themarginal cost ofenergy when charging). These results demonstrate how the energy shifting value of energy storageincreasesduetoreductions inthecostofchargingenergy(betterenergyarbitrage). Ineachhour, therevenue or cost associated with energy shifting is calculated by multiplying the amount ofcharging/discharging by the marginal price. In our model system, as variable renewable penetrationincreases,there isgenerallyadecrease invaluefromsourcing(discharging)energyastherearefewerhoursofhighpriceenergy(shownbythe“Gen”barinthefigure).However,thereisagreaterreductionin the cost of charging, illustrated by the negative values that decrease as a function of renewablepenetration.Thisproducesanoverall increase innetmarketvalue fortheco-optimizeddevices (black

52

lines). Additionally, the provision and value of regulation increases as a function of renewablepenetration as shown earlier (Figure 3-29), despite a suppression ofmarginal costs compared to thebasecasewithoutenergystorage.

3.6 CapacityValueTheproductioncostmodelingrevealsanumberof importantdriversofoperationalvalue fordemandresponse and energy storage resources.However, it does not consider their potential capacity value,which is the ability of demand response and energy storage to offset theneed for other new capitalinvestments.Electricsystemplannersutilizeavarietyofapproachestoassesshownewinvestmentsingridassetscancontributetocapacityneedsandthemonetaryvalueofthatcapacity.Intheshort-run,thevalueofcapacitycanvarysubstantiallydependingonwhetherthesystemhasexcessorisdeficientinmeetingitstargetplanningreservemargin(i.e.,thelevelofcapacitynecessaryformeetingreliabilitystandards).Inthelong-run,thevalueofcapacitymaytendtowardwhatisoftencalledthecostofnewentry(orcostofbestnewentrant).Acomprehensiveanalysisofcapacityvalueisoutsidethescopeofthisreport.Thestudyapproachassumesacapacityvalueofzerobecausenogeneratorswerereplacedorretiredfromthebasecaseswiththedeploymentofdemandresponseandenergystorageresources.However, roughestimatescanbemadetoassess therelativemagnitudesofoperationalandcapacityvalues.

Asasimplification,thepotentialcapacityvalueofdemandresponseorenergystoragecanbebasedona proxy resource, like a natural gas-fired combustion turbine used commonly for meeting peakelectricitydemands. There is a large rangeof estimates for the annualized costof anewcombustionturbine,basedonequipmentcosts,location,andfinancingterms.Examplesofthisrangeincludealowervalue of $77/kW-year (PSCO 2011) and a higher value of $212/kW-year (CAISO 2012). Strategicplacementofenergystorageorenablementofdemandresponse,suchasinareasoftransmissionanddistributioncongestionorwhere itwouldbedifficulttositetraditionalsourcesofpeakinggeneration,offeradditionalvalueforthesenon-generationresources.Thevalueofthisflexibilitycouldbedifficulttoquantify,thoughitmightbereflectedbyahighervalueofcapacity.

Alternatively, ISO/RTO capacitymarkets prices can be used as points of comparison; however, theseprices may not be directly applicable to the outcomes in this report because the present analysisneglects the roleof scarcitypricingand strategicbiddingon thepartof ISO/RTOmarketparticipants.These mechanisms provide an additional measure of capital cost recovery through transactions inwholesaleenergyandancillaryservicemarkets,whichare influencedbymanyfactors.Asanexample,marketvaluesforcapacityinrecentyearshaveaveraged$34/kW-yearinthePJMRTOto$82/kW-yearinthePJM’smostcongestedareas(seeTable2-1).Thesemarketvaluesarelessthanthefullannualizedcostofanewcombustionturbine,partlybecausePJMhasutilizeddemandresponsetoprovidecapacityalready.

Energystoragedevicessuchaspumpedstoragehydropowerandundergroundcompressedairenergystoragetypicallyhavemanyhoursofenergystorageattheratedcapacity(e.g.,8ormorehours).Theamountofcapacitytheyprovidetothesystemtendstobeclosetotheirpeakoutput,afterconsideringtheirexpected forcedandunforcedoutage rates. Energy storagedeviceswith fewerhoursof storage

53

capabilitymay result in reducedavailability for theprovisionof energyduringhoursofpeakdemand(Sioshansi, Madaeni, and Denholm 2013). Alternatively, energy-limited storage devices (i.e., thosecapable of providing only operating reserves) may also contribute to system capacity planning byprovidingoperating reservesduringpeak times (Kirby2006).However,detailedcalculationsof lossofloadprobabilitiesandotherreliabilitymetricsarenecessaryforverificationandareoutsidethescopeofthisstudy.

Comparing the capacity value of demand response resources to conventional generation is alsochallenging. Even if demand response resources are capable of responding during peak times, theunderlying demand response programs may have limits with regards to response duration and themaximum number of calls. System planners would also need to be sure that the demand responseresources would persist and be available over many years. Additionally, other issues such as localvoltageandsystemfrequencystabilityrequirementswouldneedtobemet.Theseconsiderationswouldreduce their capacity value relative to generatorsorenergy storagewithno such restrictions.On theother hand, the ability of demand response to provide operating reserves during peak times cancontributetoaddressingsomesystemcapacityissues.

IntheWesternInterconnectionmodel,thevalueofenergystoragedevices(co-optimizedforenergyandoperatingreserves)rangesfrom$33to$41/kW-yearinproductioncostsavings,assuminganadditionalenergy storage deployment level of 3.0 GW. Utilizing the simplistic assumption of a proxy resourcediscussedabove,theestimatedcapacityvalueforthisclassofenergystoragedevicesiscomparableinmagnitude to the operational value but can be several times greater. Using a similarly simplisticassumption,thecapacityvaluefordemandresponseresourcescouldalsobeseveraltimeslargerthanits operation value. The total production cost savings of demand response deployed in the model(assuming1.4%ofaverage load isavailable fordemandresponse) isabout$110millionperyear.Thetotal estimated resource reduction in peak load is about 3.1 GW (based on the ability for demandresponse to shift energy use during the top 100 hours of each balancing authority area), whichtranslates to roughly $240million to $660million per year in avoided capacity (based on the proxycombustionturbinecosts).

4 MarketandRegulatoryIssuesIn addition to technical issues, non-technical issues impact the potential deployment of demandresponse and energy storage resources through the adoption and application of market rules andregulations that may (1) limit or outright forbid these resources from providing certain bulk powersystemservices, (2) increasethecosts forprovidingtheseservices, (3)decreasetherevenuepotentialforproviders,or(4)decreasetheabilitytocaptureprofits.Whileengineering-basedanalysesandpowersystem modeling can quantify economic and environmental benefits from utilizing these resources,market design and the regulatory process impact the cost-benefit assessment and introduce otherbarriers with differing perspectives from various stakeholders. Technical cost-benefitmetricsmay beonlyoneof several importantdecision factors. Identifyingandunderstanding theothernon-technicalissues that impact the development of demand response and energy storage resources can beimportanttosatisfycost-effectivepublicpolicyandsocietalobjectives.

54

4.1 MarketandRegulatoryEnvironmentsThe U.S. electric power sector is highly heterogeneous; the characteristics of wholesale and retailelectricitymarkets,andtheregulatorystructureinwhichelectricityprovidersoperate,stronglyimpactthe financial opportunities for providers of resources like generation, demand response, and energystorage.Withintheseenvironments,thereareanumberofinfluentialentities,institutions,andresourceoptions, as illustrated in Figure 4-1. At the wholesale level, some regions are served by an ISO/RTObalancingauthority,andinothers,wholesaletradingoccursonlythroughpowerexchangesandprivatebilateral transactions. At the retail level, electric utilities have experienced varying degrees ofrestructuring,leadingtodifferentownershipmodelsforgridassets,governancestructuresforelectricityproviders,andrelationshipsbetweenelectricityprovidersandretailcustomers(FERC2012).

In each of the various environments, different types of grid investments (e.g., generation, demandresponse, and energy storage) have different regulatory treatments for cost recovery and return-on-investment(i.e.,profit).Thesecompensationmechanismsareeithercost-basedormarket-based.Undercost-basedmechanisms,resourceprovidersearnaregulatedrateofreturnontheircapitalinvestments,either as part of an electric utility’s regulated asset base or as part of tariffs approved by the FERC.Under market-based mechanisms, the prices for services are determined through an organizedwholesale electricitymarket or through a bilateral contract resulting from a competitive solicitation.Unlikecost-basedmechanisms,market-basedmechanisms(whilesubjecttoregulatoryoversight)donot

Figure4-1.Relationshipsbetweeninfluentialentities,institutions,andresourceoptions(adaptedfromCappers,MacDonald,andGoldman[2013])

55

necessarily reflect the outcomes of an explicit cost of service analysis. Generally, grid resourceinvestmentshavereceivedcost-basedormarket-basedtreatmentsbutnotbothsimultaneously.

4.1.1 RetailEnvironmentsAt the retail level, electricutilitiesaredifferentiatedprimarilyonhow theyareownedandgoverned.Publically-ownedutilitiesareownedandgovernedbythemunicipalitiesthattheyserve.Cooperatively-owned utilities are owned by their member ratepayers and governed by boards of directors. Lastly,investor-ownedutilitiesareownedby their shareholdersandgovernedby corporateentities. For thisreport,wefocussolelyonthisthirdtype,investor-owned-utilities.Here,anelectedorstate-appointedgroupofutilitycommissionersregulatetheinvestor-ownedutilityinanumberofways,includinghowitsets retail electricity rates and earns profit through capital investments (The Regulatory AssistanceProject2011).

SomestatesintheUnitedStateshaveundergoneelectricityrestructuring,buttheimplementationsaredifferent.For instance, inNewJerseyandTexas, investor-ownedutilitieshavebeenrequiredtodivesttheir generation assets. In these states, retail electricity customers are able to choose amongcompetitive providers. In New Jersey, incumbent electric utilities continue to serve customer load asdefaultserviceprovidersandtherebyhavesupplyobligations.However,inTexas,theincumbentutilitiesareonlytransmissionanddistribution(i.e.,wires-only)companiesanddonotservecustomersdirectly.Bycontrast,inotherstateslikeColoradoandWisconsin,electricityserviceprovidersarepredominantlyvertically integratedutilitiesthatowngeneration, transmission,anddistributionaswellasserveretailcustomer load. In still other states, retail electricity restructuring is only partial. For instance, inCalifornia,someoftheinvestor-ownedutilitieshavedivestedsomeoftheirgenerationassets,withtheremaining balance of their supply obligations met through bilateral contracts (i.e., power purchaseagreements). Further, only a limited number of California electricity customers have the option ofselectingacompetitiveprovider.

Differenttypesofelectricutilitieshavedifferentopportunitiesforinvestinginandprofitingfromsupplyresources. For instance, a vertically integrated utility can propose generation, demand response, orenergystoragesolutionstoitsstateregulatorycommissionforcost-recoveryandreturn-on-investmentthrough retail electricity rates. On the other hand, a restructured utility acting as default serviceprovidertypicallysignsa fullservicesupplycontract (i.e.,capacity,energy,andancillaryservices)withthirdparties.As such, theseutilitiesdonotmakedetaileddecisions regarding thecompositionof theunderlying supply portfolios. These decisions are generally made by the power marketers andindependent power producers that service those supply contracts. Furthermore, utilities that provideonlydistributionservicehavenoresponsibilitieswithregardstoelectricitysupplyprocurement.

4.1.2 WholesaleEnvironmentsWholesalemarketsaredifferentiatedprimarilybywhetherornottheyarecentrallyadministeredbyanISO/RTObalancingauthority.InareasservedbyanISO/RTO,therearetransparentpricesformanybulkpower system services, and resource providers can participate in wholesale markets if they satisfystandardizedrequirements.Thisallowsresourceslikegeneration,demandresponse,andenergystorageto readily monetize the resource providers’ provision of bulk power system services but does not

56

provide future income certainty due to the lack of long-term contracts (which are separate from theISO/RTOmarket). Outside ISO/RTO balancing authority areas, wholesale transactions occur primarilybetweenindividualbuyersandsellers.Thistransactionlimitspricetransparencybutprovidescontractedincome certainty, typically for multiple years. While energy pricing indices are available at powerexchanges,thesearelimitedtoafewtypesofenergyproductsanddonotincludeancillaryservices.Incontrast to ISO/RTO regions, the value of bulk power system services is largely derived through costsavingsinternaltoverticallyintegratedutilities(i.e.,theutilities’avoidedcostforthealternativesupplyoption),ratherthanamarketprice.

In both ISO/RTO and non-ISO/RTO regions, most electricity supply is planned in advance to ensureadequacy—days,months,andyearsaheadofwhenneeded.Closer to real time,balancingauthoritiesoverseeunitcommitmentandeconomicdispatchprocessestoensurethatelectricity flowswithinthephysical limitsof thepower system (asdiscussed in Section3) and that anyenergy imbalances (fromissues like loadforecastingerrors,generatorsnotfollowingschedules,andcontingencyevents)canbehandledwithinacceptablereliabilitystandards.WhilethebasiccontrolalgorithmsaresimilarbetweenISO/RTOandnon-ISO/RTObalancingauthorityareas,establishedprocesseshavedifferent implicationsintermsofwholesaletransactionsandfinancialsettlementsforserviceproviders.

In ISO/RTO balancing authority areas, system operators run organized wholesale electricity markets,bringing together multiple buyers and sellers, and establishing hourly (or shorter) locational basedmarketclearingpricesforenergyandancillaryservices.WhereasanISO/RTObalancingauthority isanindependent entity that solely operates electricity markets, a non-ISO/RTO balancing authority istypicallyaverticallyintegratedutility.Inthesenon-ISO/RTOregions,electricutilitiesscheduleresourcesto balance supply and demand, including their provision of ancillary services. Any residual energyimbalancesorotherdeficiencies inancillary servicecommitmentsaremadeupbyassetsoperatedbytheelectricutilityservingasthebalancingauthority.SomeISO/RTOmarketsco-optimizetheprovisionofenergyandancillaryservicesfromthesameresourcesbutverticallyintegratedutilitiesoutsidetheseregionsmayrelyondedicatedgenerationbuilttoprovidespecificservices(WWSIS-2TechnicalReviewCommittee2012).

Inadditiontoenergyandancillaryservices,balancingauthoritiesalsoadministercapacityrequirementsto ensure there will be sufficient resources available, at the right locations, to meet forecastedelectricity demand plus a reserve margin (see Figure 2-1 in Section 2.1). Some ISO/RTOs haveestablishedcapacitymarketstoallowmarketparticipantstosellanyavailablecapacityinexcessoftheirneeds. These markets can also help address concerns in regions where energy and ancillary servicemarkets may not provide sufficient incentives for participants to develop andmaintain resources tomeet system adequacywithin their planning horizons. In non-ISO/RTO balancing authority areas andthoseISO/RTOareaswithoutcapacitymarkets,capacitytransactionsoccurbilaterally.

4.2 BarrierstoDeploymentAnumberof studieshaveexaminedvariousdeploymentbarriers to theutilizationof specific typesofgrid resources (FERC 2009; Cappers, MacDonald, and Goldman 2013;Sioshansi, Denholm, and Jenkin2012;Bhatnagaretal.2013),thoughthesestudieshavetakenonlyalimitedlookacrossthesemultiple

57

typesofresourcestoassessbarriersinanintegratedfashion.Manyofthebarrierstodemandresponseandenergystorageatthewholesalelevelaresimilarandcouldbeaddressedjointly.However,someofthe barriers at the retail level are distinct and require different approaches. A holistic andcomprehensive framework may be useful for understanding how generation and non-generationresources, likedemandresponseandenergystorage,areabletocompeteside-by-sideasprovidersofbulkpowersystemservices.

Thereareabroaderrangeofissuesthatcanadverselyaffectinvestmentdecisionsinthedeploymentofdemandresponseandenergystorageresources;however,weattempttoconfinethediscussioninthefollowingsubsectionstoonlythoseissuesthat(1)relatespecificallytobulkpowersystemservicesand(2) are distinct to demand response and energy storage compared with generation resources. Someissuesthatfalloutsidethescopeofthesetwocategoriesarehighlightedbelow.

One issuearea is thatdemandresponseandenergystoragecanhaveanumberofapplicationsatthecustomer-levelandatthedistribution-level,separatefromthebulkpowersystem.Forinstance,energystoragecanprovideback-uppowerto increasethereliabilityandresiliencyofelectricitysupplytothecustomer.Additionally,demandresponseandenergystoragecouldbeusedtohelpalleviatecongestionand losses in the distribution system. These applications, in addition to the provision of bulk powersystemservices,canimprovetheoverallvaluepropositionforinvestment(Akhiletal.2015).However,assessingthesevaluesstreamsareoutsidethescopeofthisstudy.

Anotherissueareaisthatsomebarriersarecommontobothgenerationandnon-generationresources.For instance, current markets do not cover all services needed by the bulk power system. As anexample, there is no market-based compensation mechanism for providers of primary frequencyresponse(Elaetal.2013).Infact,theremaybeadisincentiveforgeneratorstoprovidetheservice(Eto2010).Generation,demandresponse,andenergystoragecouldprovidethisservice(i.e.,virtualinertia).However,theinabilitytocapturerevenuefromprovidingthisservicebyitselfisnotnecessarilyabarrierthatisdiscriminatorytonon-generationresources.

The followingdiscussion focusesonlyon thebarriers thatmaybediscriminatory todemandresponseand energy storage. These barriers are primarily a result of market rules and regulations designedaround the specific characteristics of resources that have historically provided bulk power systemservices(i.e.,generation).Alevelplayingfieldwouldtakeasystemsperspectiveinstead,focusingontheessentialphysicalcharacteristicsandqualitiesrequiredfortheprovisionofbulkpowersystemservices.

4.2.1 MarketandRegulatoryBarriersMarket and regulatory barriers to demand response and energy storage participation in bulk powersystem services fall into several categories. To start, regional reliability councils and balancingauthoritiesestablishbulkpowersystemproductdefinitionsthatmayexplicitlyincludeorexcludecertainclassesofresources,orexcludethemimplicitlybydefiningservicesthatonlycertainclassescanprovide.Even if demand response and energy storage resources are eligible to provide bulk power systemservices,other issuescan impact thevaluepropositionofusing theseresourcesbyaffecting thecostsand revenue forproviding services. Furthermore, smallerproviders faceanadditionalbarrierbecause

58

theyrequireanintermediarytobringtheirresourcestothewholesalemarket.Theintermediary,eitherthe retail electricity provider or a third party aggregator, may not have a profit motive to do so,preventingthesesmallerprovidersfromparticipatingintheprovisionofbulkpowersystemservices.

AnumberofactionsbytheNERCandFERChavesoughttoeliminatemanyofthebarriersfacingdemandresponse and energy storage in market and regulatory environments. For instance, a number ofdecisionshave targeted issuesassociatedwitheligibility (FERC2008;FERC2011).Yet, thereare somespecificimplementationsofbulkpowersystemservicedefinitionsthatremainasbarriers.Forinstance,theWECC, the regional reliability council for theWestern Interconnection,hasnot yetadoptedNERCstandardsanddoesnotallowdemandresponseandsometypesofenergystoragetoprovideservicesclassified as spinning (see Section 2.2). Also, ISO New England does not allow demand responseresources to provide frequency regulation, though they have administered a pilot program toinvestigatetheirperformance.

Other FERC actions have addressed issues associated with revenue capture. Many energy storagetechnologies can ramp substantially faster and deliver more accurate response than thermal powerplants (Makarov et al. 2008), and demand response resources can ramp nearly instantly (Kirby et al.2008), depending on communications and control equipment (Kiliccote et al. 2012). By contrast,generatorshavemechanicalandthermalinertiathatlimitstheirabilitytorampandquicklyrespond.Inresponse to FERC Order 755, several ISO/RTO balancing authorities have implemented performance-based payments (FERC 2011). Following this order, FERC Order 784 extends the treatment of fasterresourcestonon-ISO/RTObalancingauthorityareasandalsoalleviatesbarrierstothird-partyprovisionofoperatingreservesinnon-ISO/RTOareas(FERC2013).

Whilemany barriers are being addressed by policymakers, the following subsections discuss severalremainingissuesthatimpacttheabilityofdemandresponseandenergystoragetoprovidebulkpowersystemservicesalongsideconventionalgeneration.

4.2.2 IssuesthatImpactCostsRegional reliability councils and balancing authorities define the attributes of performance and therequiredenablinginfrastructurenecessaryforparticipationinwholesalemarkets.Whiletheseissuesdonotdirectlypreventdemandresponseandenergystoragefromprovidingbulkpowersystemservices,they can strongly impact the costs associated with their provision and thereby indirectly limitparticipation.

Demand response and some energy storage technologies have inherent shortcomingswith regard toresponseduration(i.e.,energy-limited)thatimpactcosts.Forinstance,ashortdemandresponseeventformanytypesofend-usesmaynotbenoticeable,butalongermulti-houreventmayleadtosignificantdisruption in service quality. Achieving the longer response, while meeting service demands, mayrequireacquisitionofmorecustomersandenablementof substantiallymoreend-use loads to spreadthe impact of interruptions. Similarly, some energy storage technologies such as batteries, flywheels,and above ground compressed air energy storage have capital costs that are roughly proportional toresponseduration;deviceswithlongerresponsedurationswillcostmore.

59

Requirementsaroundresponsedurationtypicallyhavelittleimpactongeneratorsbutcansubstantiallyincrease costs for demand response and energy storage. As an example, NERC reliability rules onlydictate that contingency reserves are restored within 105 minutes, though various markets requireresources to be capable of sustaining the response from30minutes to up to 2 hours.Dependingonregion,contingencyreservesaredeployedasoftenaseverycoupleofdaysorasseldomaseverycoupleofweeks;theaveragedurationis10minutes,butreservesrarelylastlongerthan30minutes(Kirbyetal.2008).Whiletheessentialqualitiesoftheservicemaynotchange,the2-hourrequirementcancostfourtimesmorethana30-minuterequirementfordemandresponseandenergystorageresourceproviders.

Similarly,energy-limitedresourcesarecapableofprovidingfrequencyregulationsincetheserviceis,inprinciple, energy-neutral over a set time interval. However, demand response and energy storageresources need sufficient ability to sustain the response for themaximumenergy utilization risk in asinglescheduling interval(e.g.,5minutesto1hour).Allthingsbeingequal, largerscheduling intervalsleadtolargerenergyutilizationsandlargerassociatedcostsfordemandresponseandenergystoragetoprovidethisservice,whichisnotthecaseforconventionalgenerators.

For the case of demand response, individually smaller resource providers face high per-unit costs ifbalancingauthoritiesrequirethesamemonitoringandcommunicationsequipmentandconnectiontoadedicated communicationnetwork typically usedwith largegenerators. Theseenabling infrastructurerequirements ensure that balancing authorities can monitor and verify, in real time, a resource’scompliancewithdispatchinstructions.However,utilizationoflargeaggregationsofsmallresourcesmayhave different reliability concerns compared with those of a single large generator. The aggregatedresponseofthevariousloadsandnotthatofanyoneindividualend-usemaybemoreimportanttothepower system (Kirby 2006). Some ISO/RTOs have relaxed these requirements for demand responseresources. For instance, PJM allows demand response providing contingency reserve (i.e., 10-minutesynchronizedreserve)tosubmit,withintwobusinessdaysoftheeventday,historicalmeteringdata(ata 1-minute scan rate) for measurement and verification (PJM 2013). Additionally, ERCOT hasimplemented rules allowing aggregation of small providers without expensive real-time monitoring(ERCOT2014).

4.2.3 IssuesthatImpactRevenueCaptureInadditiontocostissuesassociatedwithparticipationinbulkpowersystemservices,potentialdemandresponseandenergystorageresourceprovidersmayalsohavevaryingabilitiestocapturerevenue. Inthiscontext,revenuemaystemfrommarket-basedsalesofbulkpowersystemsservices,ortheymaystemfromavoidedcostsavingsinternaltoanelectricutility.Thus,withinthisbroaddefinition,revenuereflects a value stream that may be monetized in order to recoup costs of implementing demandresponseorenergystorageresourcesaswellastoprovideprofitstoresourceproviders.

In some cases, ISO/RTO market rules can impact revenue for demand response and energy storageproviders.Forinstance,someISO/RTObalancingauthoritiesrequireresourcestobidjointlyintoenergyand operating reservemarkets. Demand response and energy storage resources that are capable ofmeetingperformanceattributesforoperatingreservesmaynotbecapableofsustainingtheresponsefor one or more hours as an energy resource. In order to participate, they may need to hedge

60

themselves through some sort of financial instrument (with a concurrent economic loss). If they areunable to take this risk, theywill need towithdraw from theoperating reservesmarket (Todd2008).This issue has been addressed in the PJMRTO,New York ISO andMid-Continent ISO, specifically forlimitedenergystorageresources(PJM2013;NYISO2013;MISO2011),andisunderconsiderationintheCaliforniaISOforbothlimitedenergystorageanddispatchabledemandresponse(CAISO2012).

Furthermore, demand response and energy storage resourcesmay have differing abilities to providesymmetric response for frequency regulation (i.e., the same amount of up and down movement),limitingrevenuetotheamountbasedonthemorelimitedability.Thisissuehasbeenalleviatedinsomebalancingauthorityareas.Forinstance,theNewYorkISOhasmodifieditsautomaticgenerationcontrolsignal to dispatch limited energy storage resources with consideration of the state of charge (i.e.,decreased capability in one directionwhen reaching a fully charged or fully discharged state) (NYISO2013).Similarly,theMid-ContinentISOhasadjustedits5-minutereal-timeenergydispatchtomaximizeregulation capacity for energy storage resources (Chen et al. 2011). Other balancing authorities,including the California ISO and ERCOT, have separate up and down regulation services, therebyallowingresourcestobidasymmetrically.

Even ifmarket rulesenable revenue capture fordemand responseandenergy storage, theprices forbulk power system services can be volatile, leading to increased risk for potential providers. This isparticularlychallengingforprovidersthatcanonlyofferoperatingreserves.First,therearenoorganizedlong term contractingmechanisms for operating reserves to allowproviders ameans tomanage thisrisk.Second,entryofdemandresponseandenergystorageresourcesinoperatingreservemarketscancollapsemarketclearingpricesforthoseservices.CurrentISO/RTOelectricitymarketsaredesignedsuchthat generators are mostly profit-neutral to the provision of energy or operating reserves. In otherwords, the market guarantees these resources recover lost opportunity costs when they reservecapacityforoperatingreservesratherthangenerateelectricity(i.e.,provideenergy).Evenifgeneratorsbid zero costs for operating reserves, they would still receive their lost opportunity cost if they areeconomicasenergyresources.This isnotnecessarilythecaseformanydemandresponseandenergystorageresourcesseekingtoprovideonlyoperatingreserves.Fortheseresources,thelostopportunitycost component could be taken as zero by unit commitment and economic dispatch processes (notethat this is different from the lost opportunity cost associated with deferred electricity usage bycustomers).

Demandresponseandenergystorageresourcescanalsoimpactwholesaleenergymarketpriceswhenshiftingenergy.Allsupplyresourcesthatsuccessfullybidintoenergymarketswilltendtosuppressthemarket clearing price and reduce revenue for all providers, in accordance with general economics.However, energy-shifting resources face the dual problem of elevating prices during charging anddepressing prices during discharging, thereby reducing the value of energy arbitrage. This issue isparticularlyacute forpumpedstoragehydropowerplants.For this technology,projecteconomics relyon economies of scale. These systems cannot be sized optimally when considering the elasticity ofenergy prices in a specific location or at specific times. While historical energy price differentialsbetweenpeakandoff-peaktimesmayprovidesufficientrevenuefor investment intheseplants, theirentranceintothemarketcansubstantiallyerodethatrevenueopportunity.

61

4.2.3.1 IssuesSpecifictoDemandResponseAggregatorsLargeelectricitycustomersseekingtoactasdemandresponseproviderscanregisterdirectlywiththeISO/RTOor negotiate individual contractswith their electricity providers.However, smaller electricitycustomers cannotmeetminimum response size requirements for ISO/RTOmarket participation (e.g.,between 100 kW and 1 MW) nor obtain individualized contracts. These smaller customers mustparticipate in a demand responseprogramofferedby their retail electricity provider or go through athird-party aggregator to facilitate entry into ISO/RTO markets. Even if there are sufficient revenuestreamsfordevelopmentofsmalldemandresponseresources,theseintermediariesmusthaveaprofitmotivetoprovidetheaggregationservices.Otherwise,itfallssolelyonstateregulatorycommissionstointerveneandrequireelectricutilitiesundertheirauthoritytooffersuchprograms.

Development of small demand response resources can conflict with current investor-owned utilitybusiness models. For vertically integrated utilities, demand response can provide operational valuethrough savings in fuel and operations and maintenance costs and reducing the need to purchaseservices from the wholesale market, particularly at times of high market prices. However, currentregulationspreventaninvestor-ownedutilityfromretainingthosesavingsasprofit.Similarly,ifdemandresponseresourcesenabletheelectricutilitytoincreaseoff-systemsalesofbulkpowersystemservices,theymaybeabletokeeponlya fractionoftherevenuesasprofits,atbest.For instance,thestateofWisconsindoesnotpermit theutility to retainanyof these revenues (WisconsinAdministrativeCode2012),whilethePublicServiceofColoradoallowutilitiestoretainupto20%ofnetproceeds(ColoradoPUC2009). Lastly, theremaybeanumberof soft costsassociatedwithdevelopingdemand responseresources (i.e., customer acquisition and backend services), for which an investor-owned utility mayreceivecostrecoverybutnotarateofreturn.Manyofthesebusinessmodelissuesaresimilartothoseforenergyefficiencyinvestments(Goldmanetal.2010).

For other potential demand response providers that are not vertically integrated utilities, issues aredifferent.Wires-onlyutilities(i.e.,thosethatdonotserveretailcustomers)havenofinancialincentivestoofferdemandresponseprogramsandaredriven-onlybyregulatoryinterventionthatcompelsthemto do so. Competitive retail providers may not see value in investing in the enablement of theircustomerstoprovidedemandresponse.Theshort-termcontracts forretailelectricityservicemaynotbelongenoughtoachievesufficientreturn-on-investmentthresholdsforenablingcustomerstoprovidedemandresponse.Alternatively,athird-partywhoisnottheretailelectricityprovidercanserveastheaggregatorofsmalldemandresponseresources.However,insomecases,stateregulatorycommissions,likeinWisconsin,haveprohibitedthird-partyaggregators.Inothers,regulatorshaverequiredutilitiestooutsourcedemand responseprograms toa third-party, like inColorado. In still others, likeTexasandNew Jersey, thirdpartyaggregators areallowed tooperate freely and sell services in their respectiveISO/RTOwholesalemarkets(Cappers,MacDonald,andGoldman2013).

5 ConclusionThis study assesses the potential for demand response and energy storage resources to provide bulkpower system services and examines themarket rules and regulations that impact the use of theseresources.Inthisreport,weprovidefindingsfromsimulationsofaWesternInterconnectionmodeland

62

a smaller Colorado test model. The simulations explore scenarios with deployment of new demandresponseandenergy storage resourcesunderboth lowandhigh levelsofwindand solarpower. Themarket and regulatory assessment takes a national view. It looks across different environments fordemand response and energy storage deployment with specific examples from select wholesale andretailmarkets,representativeofthediversityoftheelectricutilitysectorintheUnitedStates.

Overall,theseeffortsyieldanumberofkeyfindings.

• A significant fraction of operational value attributable to demand response and energy storageresourcesare theavoided costs associatedwith generator startups/shutdownsand reduced costsassociatedwithgeneratorsmodulatingoutputwhileprovidingfrequencyregulation.Thesecostsarenominal when looking at total production costs but represent a significant fraction of costs foroperatingreserves.

• Due to the limited temporal flexibility of demand response resources, the provision of operatingreserves has more market value than energy shifting services, assuming prices are based onmarginalcostsofproduction.However,theavailabilityoftheseresourcestoprovideenergyshiftingserviceshelpstooptimizetheoperationofthebroadersystem.

• Energy storageprovides greater value in scenarioswith higher renewable penetrationdue to theincreasedneed foroperating reservesand thegreateropportunities forenergyarbitrage throughthe storageof low-cost,off-peakelectricity.Additionally, co-optimizationofenergyandoperatingreservesresultsingreatervaluethanenergyshiftingservicesalone.

• Marketstructurescan limittheabilityofanynewentrant, includingdemandresponseandenergystorage, tobe compensated commensuratewith the savings they create. Existingmarketsdonotinclude generator startup costs in price formulation, so they may not necessarily compensatedemandresponseorenergystorageforreducingthesecostsalongwithothervaluestreams.

• Marginalcostsforoperatingreservesincludelostopportunitycostsfromgenerators(forgoneprofitof selling energy). However, the lost opportunity cost component is defined as zero for demandresponseandenergystorageresourcesprovidingonlyoperatingreserves(forgoneprofitofsellingenergyiszero).Largepenetrationoftheseresourcescansaturatethemarketforoperatingreservesanddrivedownmarketclearingprices.

• While capacity value of demand response and energy storagewas not studied in detail, a simplecalculationbasedontheassumptionofaproxygenerationresourcesuggeststhatitscapacityvaluecouldbeseveral times larger than itsoperationalvalue.However, realizing thisvaluemayrequireresources toprovidemanyhoursof responseduration (e.g.,6–10hours),generally increasing thecostofprovidingtheseservices.

• While there are multiple challenges to deploying large customer demand response and energy-limited storage resources inwholesalemarkets, smaller customer resources that seek to providebulkpowersystemservicesfaceadditionalbarriers.First,theymayrequireanaggregatorthatmightnot see a business case to provide those services. Second, communications and controlrequirements imposed on individual large providers can be cost prohibitive if applied equally tosmallerproviders.However,theserequirementscouldbemodifiedandtechnicalhurdlesovercometoreduceimplementationcostswithoutcompromisingsystemreliability.

63

6 FutureWorkThis study represents an initial effort toward a broad set of research goals to quantify the potentialresource availability of demand response providing bulk power systems services and the operationalvalueofdemandresponseandenergystorageundervarying levelsofwindandsolargeneration.Thiseffortfurtherseekstounderstandpotentialmarketandregulatorybarriersthatinhibitgenerationandnon-generationresourcesfromcompetingside-by-sideasserviceproviders.AssummarizedinSection5,thepresentworkprovidesanumberofinsightstowardthesebroadergoals.

Thereareanumberofareasforfuturework.

• While this studyexamines theoperational valueofdemand responseandenergy storage, itdoesnotlookforoptimaldeploymentsofdemandresponseandenergystorageortheeconomicviabilityof developing those resources. For instance, additional work is necessary to quantify the costsassociated with different types of demand response resources providing different types of bulkpowersystemservices.Thisstudysimplyimplements,intheproductioncostmodel,theestimatedamountofselectdemandresponseresources.Inanoptimalsolution,morevaluableresourceswillhavegreaterincentivestooffertheircapabilitiestothepowersystemandtherebymayhavehigheravailabilitiesanddifferenttechnicalattributes.

• The unit commitment and economic dispatch of the demand response resources, through theproduction cost modeling, has not been tested against detailed examinations of the resources’capabilities. Because we are modeling the aggregate response capabilities of many individualproviders, it is not clear howmany customers need to be enrolled in order tomeet aggregatedresponseutilizationandwhethertheinitialassumptionsonavailabilitymakeefficientuseofthoseenrolled.

• This report focuses exclusively on bulk power system services including energy transactions andoperatingreserves(i.e.,frequencyregulation,contingencyreserve,andrampingreserve).Therearea number of other applications for demand response and energy storage such as frequencyresponseandvoltagesupport.Further,thereareanumberofapplicationsonthecustomersideandservicesondistributionsystems,includingrelievingcongestiononboththedistributionnetworkandlocaltransmissionsystem,whichareomittedinthisassessment.Acomprehensiveevaluationofgridservicesandtheirprovisionbydemandresponseandenergystorageisneededtoquantifytheirfullvalue.

• Therearemanypotentialdemandresponseresourcesomitted inthisassessment.First,projectingthechangingcompositionofend-useloadsisoutsidethescopeofthisreport.Forinstanceelectricvehicles could be a significant demand for future electricity production. Second, there aremanysmallerloadsthatcouldbeaggregated(e.g.,appliances,miscellaneousbuildingplugloads,andpoolpumps)forflexibleresponse,buttheirindividualcharacteristicsvarysignificantly,complicatingtheiranalyses. Furthermore, demand response strategiesmay be coupled with on-site generation likecombined heat and power systems to provide bulk power system services. Lastly, the demandresponse resource assessment for industrial manufacturing process end-use loads was notcompletedintimetoconducttheproductioncostmodeling.

64

• This study provides an initial estimate of available demand response resources for the WesternInterconnectionfortheprovisionofoperatingreserves.Acompletenationalinventoryofpotentialdemandresponseresources(residential,commercial,industrial,agricultural)bylocationforenergyandeachoftheancillaryserviceswouldbeveryusefultoaidregulators,utilities,anddevelopersofgridresources.Theinventoryshouldincludethecostofenablingresponse.

• In addition to examination of theWestern Interconnection, this work could be implemented forother U.S. grids. For instance, expansion of the analysis to the Eastern Interconnection couldleveragemodeled scenarios developed through Eastern Renewable Generation Integration Study(Bloometal.2015).

• Methods to increase cost-effective demand response participation could be investigated to helpreducethedramaticorderofmagnitudedifferencebetweendemandresponsetechnicalpotentialandtheestimatedresource.

• There are several potential extensions of the grid simulations. Results presented here are frommodelingday-aheadunitcommitmentandeconomicdispatchprocesses.Futurework isnecessaryto investigate operations closer to real-time and sub-hourly scheduling. This will enable greaterinsights into forecast errors and the utilization of the ramping reserve product. Additionally,resources beyond demand response and energy storage, such as combustion turbines and smartinverters,canbesimulatedtoassesstheircapabilitieswithasimilarframework.

• There are numerous questions regarding wholesale market design that are not included in theanalysis.Forinstance,wehavenotinvestigatedalternativemarketdesigns,likeperformancebasedratesforfrequencyregulationperFERCOrders755and784.Additionally,wehavenotexaminedindepththeroleofscarcitypricingandtheintersectionwithcapacitymarkets.

• The results provided in this report represent only a few example scenarios; as such, the resultscannot be applied universally. A variety of generator compositions could occur under increasingrenewable penetration scenarios. Specifically, increased penetration of variable renewablegenerationwillresultindecreasedcapacityfactorsandmarketrevenuesofexistinggenerators.Thiscould incentivize additional plant retirements, changing the overall system composition. Morecomprehensiveevaluationofdifferentgenerationmixes, includingcombinationsofwindandsolarpenetrationlevels,willprovideadditional insightsintotheevolvingvalueofdemandresponseandenergystorage.

• The values of demand response and energy storage are driven greatly by the cost of providingoperating reserves, which itself depends greatly onmany assumptions regarding the operationalflexibility of the generation fleet, in particular the assumed ramp rates and the fraction of fleetavailable to provide operating reserves. In addition, a large fraction of the marginal cost offrequencyregulationinthesesimulationsisderivedfromtheassumedcostofgeneratorsoperatingatanon-steadystatewhileprovidingregulationreserves.Becauseperformancedatarelatedtoanindividual generator’s ability to provide reserve are notwidely available, reproducing the cost ofoperatingreservesinaproductioncostmodelinvolvessignificantuncertaintywithoutbetterdata.

• Productioncostmodels,likethePLEXOSmodelusedinthisstudy,cancalculatethecostofholdingoperating reserves, but they do not simulate explicitly frequency regulation dispatch andcontingency events. Simulation of shorter term grid dynamics is an important area of future

65

research,requiringmodelscapableofsimulatingdetailedgenerator,demandresponse,andenergystorageresponse.Thismaybecomeparticularly relevant ifvariablegeneration likewindandsolarpowerbecomesalargecontributortoelectricitysupply.

• Theoptimalsizingandlocationofdemandresponseandenergystoragewerenotconsideredinthisstudy. Additionally, generator retirements were not considered in the scenarios with higherrenewablepenetration.Expandingthestudytofactorinthesebusinessoperatingprinciplesorwithscenarios that have similar levels of reliability (e.g., without excess capacity) can produce morerealisticandmeaningfulresults.

66

References AESO.2013."EnergyStorageInitiativeIssueIdentification."

Akhil,A.,G.Huff,A.B.Currier,B.C.Kaun,D.M.Rastler,S.B.Chen,A.L.Cotter,D.T.Bradshaw,andW.D.Gauntlett.2013.DOE/EPRI2013ElectricityStorageHandbookinCollaborationwithNRECA.SAND2013-5131.

Akhil,A.,G.Huff,A.B.Currier,B.C.Kaun,D.M.Rastler,S.B.Chen,A.L.Cotter,D.T.Bradshaw,andW.D.Gauntlett.2015.DOE/EPRIElectricityStorageHandbookinCollaborationwithNRECA.SAND2015-1002.

Bhatnagar,D.,A.B.Currier,J.Hernandez,O.Ma,andB.Kirby.2013.MarketandPolicyBarriersfor

EnergyStorageDeployment.SAND2013-4747.

Bloom,A.,A.Townsend,D.Palchak,J.King,E.Ibanez,C.Barrows,M.Hummon,andC.Draxl.2015.EasternRenewableGenerationIntegrationStudy.NREL/TP-6A20-64472.

Booth,W.C.,andJ.L.Rose.1995.UsingHourlySystemLambdatoGaugeBulk-PowerPrices.FortnightlyMagazine.

CAISO.2012."2011AnnualReportonMarketIssuesandPerformance."

CAISO.2012."CaliforniaISONon-GeneratorResourceEnergyManagementProjectImplementationPlan-Version2.1."

CAISO.2011.SummaryofPreliminaryResultsof33%RenewableIntegrationStudy.CPUCLTPPDocketNo.R.10-05-006.

Callaway,D.S.,andI.A.Hiskens.2011."AchievingControllabilityofElectricLoads."ProceedingsoftheIEEE99(1).

Callaway,D.2009."TappingtheEnergyStoragePotentialinElectricLoadstoDeliverLoadFollowingandRegulation,withApplicationtoWindEnergy."EnergyConversionandManagement50(5).

Cappers,P.2012.inserttext.LBNL.

Cappers,P.,C.Goldman,andD.Kathan.2010."DemandResponseinU.S.ElectricityMarkets:EmpiricalEvidence."Energy35(4):1526-1535.

Cappers,P.,J.S.MacDonald,andC.Goldman.2013.MarketandPolicyBarriersforDemandResponse

ProvidingAncillaryServicesinU.S.Markets.LBNL-6155E.

Cappers,P.,Mills,A.,Goldman,C.,Wiser,R.andEto,J.H.2011.Massmarketdemandresponseand

variablegenerationintegrationissues:Ascopingstudy.LBNL-5063E.

67

Chen,Y.,M.Keyser,M.H.Tackett,andX.Ma.2011."IncorporatingShort-TermStoredEnergyResourceintoMidwestISOEnergyandAncillaryServiceMarket."TransactionsonPowerSystems(IEEE)26(2):829-838.

Chuang,A.2009."Demand-sideintegrationforcustomerchoicethroughvariableservicesubscription."IEEEPower&EnergySocietyGeneralMeeting,2009.

ColoradoPUC.2009."DecisionNo.C09-1446:OrderAddressingPhaseIandECAIssues[DocketNo.09AL-299E]."

Denholm,P.,J.Jorgenson,M.Hummon,T.Jenkin,B.Kirby,O.Ma,M.O'Malley,andD.Palchak.2013.TheValueofEnergyStorageforGridApplications.NREL/TP-6A20-58465.

Ela,E.,andM.O'Malley.2012."StudyingtheVariabilityandUncertaintyImpactsofVariableGenerationatMultipleTimescales."IEEETransactionsonPowerSystems27:1324-1333.

Ela,E.,M.Milligan,andB.Kirby.2011.OperatingReservesandVariableGeneration.NREL/TP-5500-51978.

Ela,E.,M.Milligan,andM.O'Malley.2011."AFlexiblePowerSystemOperationsSimulationModelforAssessingWindIntegration."PowerandEnergySocietyGeneralMeeting.NREL/CP-5500-50641.

Ela,E.,V.Gevorgian,A.Tuohy,M.Milligan,andM.O'Malley.2013."MarketDesignsforthePrimaryFrequencyResponseAncillaryService-PartI:MotivationandDesign."TransactionsonPowerSystems(IEEE)PP(99):1-11.

Elzinga,D.,J.Dillon,M.O'Malley,andJ.Lampreia.2012."TheRoleofElectricityStorageinProvidingElectricitySystemFlexibility."InElectricityinaClimateConstrainedWorld.IEA.

Enbala.2011."PresentationtoDOEWorkshopOctober25-26,2011."Presentedat2011DepartmentofEnergyLoadParticipationinAncillaryServicesWorkshop.

EnergyExemplar.2013.PlexosIntegratedEnergyModel.AccessedNovember2013.www.energyexemplar.com.

EnergyInformationAdministration.2010.StatusofElectricityRestructuringbyState.AccessedJanuary2014.www.eia.gov/electricity/policies/restructuring/index.html.

EnerNex.2011.EasternWindIntegrationandTransmissionStudy.NREL/SR-5500-47078.

EPRI.2013."QuantifyingtheValueofHydropowerintheElectricGrid."

ERCOT.2014."RequirementsforAggregateLoadResourceParticipationintheERCOTMarket."

Eto,J.2010.UseofFrequencyResponseMetricstoAssessthePlanningandOperatingRequirementsfor

ReliableIntegrationofVariableRenewableGeneration.LBNL-4142E.

68

FERC.2009.ANationalAssessmentofDemandResponsePotential:StaffReport.FERC.

FERC.2012.EnergyPrimerAHandbookofEnergyMarketBasics.FERC.

FERC.2010."NationalActionPlanonDemandResponse."

FERC.2008."Order719,WholesaleCompetitioninRegionswithOrganizedElectricMarkets."

FERC.2011."Order745,DemandResponseCompensationinOrganizedWholesaleEnergyMarkets."

FERC.2011."Order755,FrequencyCompensationinOrganizedWholesalePowerMarkets."

FERC.2012."Order764,IntegrationofVariableRenewableEnergyResources."

FERC.2013."Order784,Third-PartyProvisionofAncillaryServices;AccountingandFinancialReportingforNewElectricStorageTechnologies."

FERC.2006."Order888,PromotingWholesaleCompetitionthroughOpenAccessNon-DiscriminatoryTransmissionServicesbyPublicUtilities;RecoveryofStrandedCostsbyPublicUtilitiesandTransmittingUtilities."

FERC.2007."Order890,PreventingUndueDiscriminationandPerferenceinTransmissionService."

—.2013.ProformaOpenAccessTransmissionTariff.AccessedNovember2013.www.ferc.gov/industries/electric/indus-act/oatt-reform.asp.

Ferreira,S.R.,Rose,D.M.,Schoenwald,D.A.,Bray,K.,Conover,D.,Kintner-Meyer,M.,Viswanathan,V.2013.ProtocolforUniformlyMeasuringandExpressingthePerformanceofEnergyStorage

Systems.SAND2013-7084.

GEEnergy.2010.WesternWindandSolarIntegrationStudy.NREL/SR-550-47434.

Goldman,C.,M.Reid,R.Levy,andA.Silverstein.2010.CoordinationofEnergyEfficiencyandDemand

Response.LBNL-3044E.

Goli,S.,A.McKane,andD.Olsen.2011."DemandResponseOpportunitiesinIndustrialRefrigeratedWarehousesinCalifornia."ACEEESummerStudyonEnergyEfficiencyinIndustry.LBNL-4837E.

Griffith,M.2008."CapacityMarketsDemystified."FortnightlyMagazine(PublicUtilitiesFortnightly).

Hirst,E.,andB.Kirby.1996.Electric-PowerAncillaryServices.ORNL/CON-426.

Hogan,W.1998.CompetitiveElectricityMarketDesign:AWholesalePrimer.HarvardUniversity.

Hogan,W.2012.ElectricityScarcityPricingThroughOperatingReserves.Whitepaper.

Hummon,M.,P.Denholm,J.Jorgenson,D.Palchak,B.Kirby,andO.Ma.2013.FundamentalDriversof

theCostandPriceofOperatingReserves.NREL/TP-6A20-58491.

69

Hummon,M.,P.Denholm,J.Jorgenson,D.Palchak,O.Ma,D.Olsen,M.D.Sohn,etal.2013.GridIntegrationofAggregatedDemandResponse,Part2:ModelingDemandResponseina

ProductionCostModel.NREL/TP-6A20-68492.

Ibanez,E.,G.Brinkman,M.Hummon,andD.Lew.2012.SolarReserveMethodologyforRenewable

EnergyIntegrationStudiesBasedonSub-HourlyVariabilityAnalysis:Preprint.NRELReportNo.CP-5500-56169.

IEA.2013."StateoftheArtinRenewableIntegrationStudies."

Isemonger,A.G.2009."TheEvolvingDesignofRTOAncillaryServiceMarkets."EnegyPolicy37:150-157.

ISO/RTOCouncil.n.d.AccessedDecember30,2013.http://www.isorto.org/site/c.jhKQIZPBImE/b.2604471/k.B14E/Map.htm.

Kiliccote,S.,P.Price,M.A.Piette,G.C.Bell,S.Pierson,E.Koch,J.Carnam,H.Pedro,J.Hernandez,andA.K.Chiu.2012.FieldTestingofAutomatedDemandResponseforIntegrationofRenewable

ResourcesinCalifornia'sAncillaryServicesMarketforRegulationProducts.LBNL-5556E.

Kirby,B.,J.Kueck,T.Laughner,andK.Morris.2008."SpinningReservefromHotelLoadResponse."TheElectricityJournal21(10):59-66.

Kirby,Brendan.2006.DemandResponseforPowerSystemReliability:FAQ.ORNL/TM-2006/565.

Kumar,N.,P.Besuner,S.Lefton,andD.Agan.2012.PowerPlantCyclingCosts.NREL/SR-5500-55433.

Lew,D.,G.Brinkman,E.Ibanez,A.Florita,M.Heaney,B.-M.Hodge,M.Hummon,etal.2013.TheWesternWindandSolarIntegrationStudyPhase2.NREL/TP-5500-55588.

Lew,D.,G.Brinkman,N.Kumar,P.Besuner,D.Agan,andS.Lefton.2012."ImpactsofWindandSolaronFossil-FueledGenerators."PowerandEnergySocietyGeneralMeeting.IEEE.

Ma,O.,N.Alkadi,P.Capper,P.Denholm,J.Dudley,S.Goli,M.Hummon,etal.2013."DemandResponseforAncillaryServices."IEEETransactionsonSmartGridPP(99).

MacDonald,J.,P.Cappers,D.Callaway,andS.Kiliccote.2012."DemandResponseProvidingAncillaryServices:AComparisonofOpportunitiesandChallengesintheUSWholesaleMarkets."Grid-InterOp2012.LBNL.

Makarov,Y.V.,J.Ma,S.Lu,andT.B.Nguyen.2008.AssessingtheValueofRegulationResourcesbasedontheirTimeResponseCharacteristics.PNNL-17632.

Marks,G.,E.Wilcox,D.Olsen,andS.Goli.2013.OpportunitiesforDemandResponseinCalifornia

AgriculturalIrrigation:AScopingStudy.LBNL-6108E.

70

Milligan,M.,andB.Kirby.2010."UtilizingLoadResponseforWindandSolarIntegrationandPowerSystemReliability."WindPower2010.NREL/CP-550-48247.

Milligan,M.,E.Ela,B.-M.Hodge,B.Kirby,andD.Lew.2011."Cost-CausationandIntegrationCostAnalysisforVariableGeneration."TheElectricityJournal24(9):51-63.

MISO.2013."BusinessPracticesManualEnergyandOperatingReserveMarkets."

MISO.2011."BusinessPracticesManual:EnergyandOperatingReserveMarkets."

Navid,N.,andG.Rosenwald.2013.RampCapabilityProductDesignforMISOMarkets.Whitepaper.

NERC.2004."FrequencyResponseStandardWhitepaper."

—.2013.PlanningReserveMargin.http://www.nerc.com/pa/RAPA/ri/Pages/PlanningReserveMargin.aspx.

—.2011.WECCStandardBAL-002-1-ContingencyReserves.AccessedNovember2013.www.nerc.com/files/BAL-002-WECC-1.pdf.

NYISO.2013."AncillaryServicesManual."

NYISO.2013."AncillaryServicesManual."

OATI,Inc.2013.OpenAccessSame-timeInformationSystem.AccessedNovember2013.www.oati.oasis.com.

Olsen,D.,M.D.Sohn,N.Matson,J.Dudley,S.Goli,S.Kiliccote,M.Hummon,etal.2013.GridIntegrationofAggregatedDemandResponsePart1:LoadAvailabilityProfilesandConstraints

fortheWesternInterconnection.LBNL-6417E.

Olsen,D.,S.Goli,D.Faulkner,andA.McKane.2012.OpportunitiesforAutomatedDemandResponsein

WastewaterTreatmentFacilitiesinCalifornia-SoutheastWaterPollutionControlPlantCase

Study.LBNL-6056E.

Piette,M.A.,S.Kiliccote,andJ.H.Dudley.2012.FieldDemonstrationofAutomatedDemandResponse

forBothWinterandSummerEventsinLargeBuildingsinthePacificNorthwest.LBNL-6216E.

PJM.2013."Manual12:BalancingOperations."

PJM.2012."Manual15:CostDevelopmentGuidelinesRev.20."

PJM.2013."PJMManual11:EnergyandAncillarySericesMarketOperationsRev.62."

PotomacEconomics.2014.StateoftheMarkets.AccessedJanuary2014.http://www.potomaceconomics.com.

71

Potter,C.W.,D.Lew,J.McCaa,S.Cheng,S.Elchelberger,andE.Grimit.2008."CreatingtheDatasetfortheWesternWindandSolarIntegrationStudy."WindEngineering32(4):325-338.

PSCO.2011."2011ElectricResourcePlan.VolumeIITechnicalAppendix."

Rongxin,Y.,P.Xu,M.A.Piette,andS.Kiliccote.2010."StudyonAuto-DRandPre-CoolingofCommercialBuildingswithThermalMassinCalifornia."EnergyandBuildings42(7):967-975.

Rubinstein,R.,L.Xiaolei,andD.Watson.2010.UsingDimmableLightingforRegulationCapacityand

Non-SpinningReservesintheAncillaryServicesMarket:AFeasibilityStudy.LBNL-4190E.

Sioshansi,R.,P.Denholm,andT.Jenkin.2012."MarketandPolicyBarrierstoDeploymentofEnergyStorage."EconomicsofEnergyandEnvironmentalPolicy1(2):47-63.

Sioshansi,R.,S.H.Madaeni,andP.Denholm.2013."ADynamicProgramingApproachtoEstimatetheCapacityValueofEnergyStorage."TransactionsonPowerSystems(IEEE)PP(99):395-403.

Starke,M.,N.Alkadi,andO.Ma.2013.AssessmentofIndustrialLoadforDemandResponseacrossU.S.

RegionsoftheWesternInterconnection.ORNL/TM-2013/407.

TheRegulatoryAssistanceProject.2011.ElectricityRegulationintheU.S.:AGuide.TheRegulatoryAssistanceProject.

Todd,D.,B.Caulfield,M.Helms,M.Starke,B.Kirby,andJ.Kueck.2009.ProvidingReliabilityServicesthroughDemandResponse:APreliminaryEvaluationoftheDemandResponseCapabilitiesof

Alcoa,Inc.ORNL/TM2008/233.

Todd,D.,Caufield,M.,Helms,B.,Generating,A.P.,Starke,I.M.,Kirby,B.andKueck,J.2008.Providingreliabilityservicesthroughdemandresponse:ApreliminaryevaluationoftheDemandResponse

CapabilitiesofAlcoaInc.ORNL/TM-2008/233.

Varadan,S.,G.Freddo,H.Todus,J.Thiemsuwan,KeChen,KhoiVu.,D.Hawkins,andS.Shen.2012."ANewApproachtoStudyingtheImpactofIntermittentRenewableResources."PowerandEnergySocietyGeneralMeeting.IEEE.

Watson,D.,N.Matson,J.Page,S.Kiliccote,M.A.Piette,K.Corfee,B.Seto,etal.2012.FastAutomated

DemandResponsetoEnabletheIntegrationofRenewableResources.LBNL-5555E.

WECC.2011.AssumptionMatrixfrothe2020TEPPCDataset.Accessed2013November.www.wecc.biz/library/StudyReport/Documents/Assumptions%20Matrix%20for%20the%202020%20TEPPC520Dataset.pdf.

WisconsinAdministrativeCode.2012."WisconsinAdministrativeCode."PSC§117.05.

—.2012."WisconsinAdministrativeCode."PSC117.05.

72

Wood,A.J.,andB.F.Wollenberg.2013.PowerGeneration,Operation,andControl.JohnWiley&Sons,Inc.

Woolf,T.,E.Malone,L.Schwartz,andJ.Shenot.2013.AFrameworkforEvaluatingtheCost-

EffectivenessofDemandResponse.LBNL.

WWSIS-2TechnicalReviewCommittee.2012.(November).

Xu,L.,andD.Tretheway.2012.FlexibleRampingProducts,SecondRevisedDraftFinalProposal.CAISO.

Zhou,Z.,andA.Botterud.2013."PriceResponsiveDemandforOperatingReservesandEnergyinElectricityMarkets."PowerandEnergySocietyGeneralMeeting.IEEE.

(This page intentionally left blank)

DOE/EE-1282 • March 2016

For more information, visit: energy.gov