THE IMPACT OF CHARGING INFRASTRUCTURE ON THE LOAD SHIFT POTENTIAL OF ELECTRIC VEHICLES

Transcription

1 THE IMPACT OF CHARGING INFRASTRUCTURE ON THE LOAD SHIFT POTENTIAL OF ELECTRIC VEHICLES Anna-Lena Klngler, Fraunhofer ISI, Phone , E-mal: anna-lena.klngler@s.fraunhofer.de Tll Gnann, Fraunhofer ISI, Phone , E-mal: tll.gnann@s.fraunhofer.de Jula Mchaels, Fraunhofer ISI, Phone , E-mal: jula.mchaels@s.fraunhofer.de Abstract Plug-n electrc vehcles are the currently favoured opton to decarbonze the passenger car sector. However, a decarbonzaton s only possble wth electrcty from renewable energes and plug-n electrc vehcles mght cause peak loads f they started to charge at the same tme. Both these ssues could be solved wth coordnated load shftng (demand response). Prevous studes analysed ths research queston by focusng on prvate vehcles wth domestc and work chargng nfrastructure. Ths study addtonally ncludes the mportant early adopter group of commercal fleet vehcles and reflects the mpact of domestc, commercal, work and publc chargng. For ths purpose, two models are combned that capture the market dffuson of electrc vehcles and ther chargng behavour (ALADIN) as well as the load shftng potental of several new energy technologes (eload). In a comparson of three dfferent scenaros, we fnd that chargng of commercal vehcles does not nflct evenng load peaks n the same magntude as purely domestc chargng of prvate cars does. Also for prvate cars, chargng at work occurs durng the day and may reduce the necessty of load shftng whle publc chargng plays a less mportant role n total chargng demand as well as load shftng potental. Nonetheless, demand response reduces the system load by about 2.2 GW or 2.8% when domestc and work chargng are consdered compared to a scenaro wth only domestc chargng where a new peak mght be created n the wnter hours due to load shftng nto the nght. Introducton To attan the clmate targets, t s necessary to transform the energy system. Renewable energy sources (RES) can help to decrease greenhouse gas emssons n the electrcty sector. In the transport sector, plug-n electrc vehcles (PEVs) can be a means to reduce greenhouse gas emssons f powered by electrcty from RES. However, n a sgnfcant number, they rsk to cause addtonal load peaks that have to be balanced to ensure a stable electrcty system. Ideally, electrcty demand of PEVs and electrcty generaton of RES are coordnated e.g. by demand response. However, for ths purpose, a suffcent chargng nfrastructure s needed. Whle most studes focus on domestc chargng facltes [Dallnger et al. 2013] or nclude addtonal chargng at work of prvate passenger cars [Babrowsk et al. 2014], ths paper also consders commercal plug-n electrc vehcles (PEV) and the use of publc chargng statons. The am of ths paper s to assess the extent to whch addtonal chargng facltes contrbute to PEV market penetraton n Germany and shavng of peaks n the resdual load 1. For ths purpose, we combne two models that have been developed and descrbed earler: The model ALADIN (Alternatve Automobles Dffuson and Infrastructure) s used to determne the market dffuson of plug-n electrc vehcles and ther chargng nfrastructure. Also, the use of several types of chargng nfrastructure (domestc, work and publc) as well as dfferent vehcle user groups (prvate, commercal fleet vehcles and company cars) can be analysed. Structure and results of the model have been descrbed n several publcatons [Plötz et al. 2014, Gnann et al. 2015, Gnann 2015]. The results can be taken as an nput nto the eload (energy LOad curve Adjustment) model whch ams to analyse the load shft potental of several (new) technologes of whch electrc vehcles are one mportant technology [Bossmann 2015]. Ths combnaton permts to provde a new contrbuton n ths feld snce the potental of publc charge shftng and as well as the ncluson of commercal fleet vehcles has to the best of the authors' knowledge not been analysed before. In a case study, we apply the model to Germany and make projectons for The paper s structured as follows: Frst, we ntroduce the methods and data n Secton 2. Thereafter, the assumptons for our case study are presented n Secton 3. The results are shown n Secton 4 before we summarze and draw conclusons for electrcty supplers and polcy makers n Secton 5. 1 The resdual load equals the system load mnus the generaton of fluctuatng renewable energes.

2 Jährlche TCO Ladezustand Battere Methods and data PEV market dffuson: the ALADIN model The market dffuson model ALADIN (Alternatve Automobles Dffuson and Infrastructure, Fgure 1) s an agentbased smulaton model whch bases on a large number of vehcle drvng profles of conventonal vehcles whch are also descrbed n the followng. The model was ntroduced n [Plötz et al. 2014] and PEV market dffuson results were publshed n [Gnann et al. 2015]. Based on the ndvdual drvng behavour, the replaceablty by a battery electrc vehcle (BEV) s analysed and what share of electrc drvng (often called utlty factor) could be obtaned by a plug-n hybrd electrc vehcle (PHEV). Based on ths techncal feasblty, the utlty of four drve trans (Gasolne, Desel, BEV and PHEV) s calculated and compared. Ths utlty conssts of the total cost of ownershp for the vehcle, but also contans the cost for ndvdual chargng ponts (at home or a desgnated chargng pont at work) as an obstructng factor and a wllngness to pay more for a plug-n electrc vehcle as a favourng factor. The share of drvng profles wth PEVs as utlty maxmzng opton s consdered as ther market share for vehcle sales whch dffuse nto the vehcle stock. An enhancement for the ntegraton of publc chargng nfrastructure was ntroduced n [Gnann 2015]. After the PEV dffuson, the chargng behavour of the vehcle stock at publc chargng ponts s smulated and used to determne the energy charged n publc. Based on ths fgure, the number of proftable publc chargng ponts s calculated and constructed n the most frequented areas. Ths mght also lead to a reducton of publc chargng ponts f the amount of publc energy charged s not suffcent to cover ther cost. The new publc chargng stock s consdered n the ndvdual smulaton and may lead to a hgher utlty of PEVs. For llustraton of the model, refer to Fgure 1. The second part of the model s necessary f publc chargng nfrastructure s consdered, snce PEVs can only be charged n publc f these chargng ponts are free at the tme a PEV arrves. Thus, they nteract at publc chargng ponts whch makes a jont smulaton necessary. Ths requres to understand geographc vehcle movements. 1. Technsche Ersetzbarket/Fahrstrecke von Fahrzeugen (Smulaton Battereladestand) Gefahrene Klometer 2. Ökonomsches Potenzal von Gruppen von Fahrzeugen (TCO-Analyse) TCOa BEV TCOa PHEV TCOa ICEV TCOa DIESEL Jahresfahrstrecke Annahmen für Grafken: technsche Ersetzbarket: 1) Ladestratege am Arbetsplatz, zu Hause und am Zwetwohnstz 2) Batterekapaztät 32 kwh, Verbrauch 16 kwh/100 km; ökonomsches Potenzal: sämtlche Annahmen n (Gnann et al. 2012) Snd alle Wege mt BEV möglch? Welcher Antel der Gesamtfahrstrecke wrd bem PHEV elektrsch zurückgelegt? 1. Indvdual PEV smulaton n wth gven chargng nfrastructure Welches Fahrzeug hat de gerngsten Total Cost of Ownershp? Mnmale TCO-Opton 4. Constructon of chargng nfrastructure 3. Smulaton of chargng of PEV stock 2. Determnaton of PEV stock Fgure 1: Overvew of model ALADIN Hence, the man data source used n the model ALADIN are vehcle drvng profles,.e. all trps of vehcles n a certan tme perod (here: at least one week). These profles also contan soco-demographc nformaton of the drver, the household and vehcle n case of prvate vehcles as well as nformaton about the vehcle and the company ownng t for commercal fleet vehcles. Snce geographcal nformaton s necessary, we use the moblty panel for Stuttgart (MOPS) and drvng profles of REM2030 n the same area for commercal drvng profles [Hautznger et al. 2013, Fraunhofer ISI 2014]. 2 As the jont smulaton of several thousand vehcles profles s qute computaton ntensve, only a share of profles s consdered n the smulaton. An overvew of the full sample and the used share are presented n Table 1 2 Unlke n [Plötz et al. 2014], we do not use MOP [MOP 2010] whch contans data representatve for Germany, snce we need the geographcal nformaton n MOPS. Results were tested for representatveness for the German car sales n [Gnann 2015] and were sutable for ths purpose. Profles for company cars were selected and nformaton on garages and the wllngness to pay more for PEVs was added and tested n [Gnann 2015] and wll be used here.

3 Table 1: Prepared data sets and vehcle statstcs n observaton area User group Prvate cars Company cars Fleet vehcles Data set MOPS [1] REM2030S [2] Total no of vehcles [3] 1,273,426 39, Total no of trps [3] 18,909, ,049 13,374 Total vehcle regstratons n observaton 63,772 39,391* 39,391* area (2014) [4] Total vehcle stock n observaton area 1,343,016 39, ,297 (01/01/2014) [4] Subsample szes 15, [1] [Hautznger et al. 2013]; [2] [Fraunhofer ISI 2014]; [3] reduced sample after assgnment of personal trps to vehcles and clusterng to 15mn-ntervals; [4] regstratons from [KBA 2014], * dstrbuton between fleet and company cars s an assumpton based on [Pfahl 2013]. Snce results for Stuttgart have to be extrapolated to Germany, the regstratons are multpled by 20.54, whch s equal to the nverse of the share of the regons' regstratons n Germany. Optmal vehcle chargng: the eload model The drvng, chargng and dfferent parkng profles (domestc, work, publc) as well as the total number of PEVs and ther electrcty demand from ALADIN serve as an nput for the eload (energy LOad curve ADjustment) model [Bossmann 2015]. In ths study eload s used to determne the least-cost schedulng of PEV-chargng dependng on an hourly prce sgnal. It thereby smulates the potental contrbuton of demand sde technologes resdual load smoothng (also known as demand response, DR). eload conssts of two modules. The frst module addresses the long term evoluton of the natonal system load curve, whch s drven by structural changes on the demand sde and the ntroducton of new applances (such as PEVs). By usng applance specfc load profles, such as typcal day profles, regresson based load profles etc., a yearlong load curve can be generated for all consdered applances for the base year. The load curve s then scaled accordng to the respectve annual demand n the base year. These load curves are deduced from the system load curve of the base year. The resultng remanng load curve and the applance specfc load curves are then scaled for all projecton years accordng to the yearly demand evoluton. Reassemblng the scaled remanng load and the scaled load curves gves the load curve of the projecton year. The man advantage of ths approach s ts ablty to properly take structural changes n the load curve nto consderaton by explctly modellng the man drvers for load deformaton whle preservng stochastc outlers and characterstc rregulartes from hstorc load curves. The second module of eload addresses the actve adjustment of the load curve by means of DR. In ths study, eload optmzes the natonal DR deployment to the objectve of smoothng the net load. Net load smoothng enables an effcent operaton of exstng electrcty generaton assets and grd nfrastructure as well as a reduced need for nvestment n new capactes [Boßmann et al. 2015; Boßmann 2015]. eload s an establshed model that has frequently been used n German and European studes for polcy makers and ndustral customers [Fraunhofer ISI et al. 2017]. For ths study, the methodology to represent moble storages n the model was mproved to better reflect the partcular characterstcs. Wth respect to PEVs, eload now consders techncal (battery storage sze) and organzatonal constrants (drvng and parkng cycles, chargng capacty), that are characterstc for moble storages, when determnng the least-cost schedulng of PEV chargng. When t comes to organzatonal constrants, PEVs are characterzed by the fact that ther storages are not connected contnuously to the electrc grd. Thus, a storage capacty wth a reduced avalablty and load corrdor has to be consdered when not all PEVs are connected to chargng ponts. The connecton to a chargng pont depends on the locaton of the vehcle as well as on the avalablty of chargng nfrastructure. To mplement these avalablty constrants, PEVs are dfferentated nto three storage groups n eload: connected: parked vehcles that are connected to a chargng pont and can be charged moble: all vehcles that are currently drvng and therefore dscharge dsconnected: parked vehcles that are not connected to a chargng pont, e.g. located on a parkng lot wthout chargng nfrastructure, that are not avalable for load shftng. If each of the storage groups s regarded as ts own sub-system, the followng energy exchange between the groups s possble:

4 connected: Energy can be suppled va battery chargng; addtonally, a bdrectonal energy exchange wth the moble group: the connected vehcle can swtch nto the moble group (drve off) or moble vehcles can enter the connected group (arrvng vehcles). moble: Energy s extracted from the system through drvng and thereby dschargng the battery; moble vehcles can swtch nto the connected or dsconnected groups dependng on whether they arrve at a parkng locaton wth or wthout a chargng pont. dsconnected: Only the bdrectonal energy exchange wth the moble group s possble. The battery cannot be charged or dscharged. Fgure 2 llustrates the energy flows n the dfferent storage sub-systems. chargng dschargng connected moble dsconnected Fgure 2. Energy flows between connected, moble and dsconnected PEV storages. The storage capactes and chargng loads of the dfferent storage groups are calculated as the product of the overall PEV storage and the tme dependent share of connected, moble and dsconnected vehcles. For a deeper nsght nto the model, n the followng we provde a formal descrpton of the objectve functon and the relevant constrants for the least-cost schedulng of PEV chargng: The objectve functon of the lnear optmzaton problem mnmzes the costs of the load shftng actvty: h max Mn P ls,j (p j =h mn p ) j=h mn wth the shfted load from hour to hour j P ls,j, the hourly electrcty prce p and the consumpton ncrease factor cf ls, where j and,j є [h mn; h max]. Here, we only analyse the load shftng potental of PEVs, thus no consumpton ncrease s consdered for load shftng (cf. consumpton ncrease factor n [Bossmann 2015]. In the absence of a market model, the prces are equal to the net load, whch s the case n ths study. The ablty of a process to adjust ts load s frst and foremost restrcted by the load bounds,.e. the mnmal and maxmal load P mn and P max, respectvely: P mn P h + P h,ls P max wth the orgnal chargng load n each hour P h, the shfted load to and from each hour P ls,h. Addtonally, the load shftng ablty s restrcted by the storage capacty. In the case of moble storages, the storage constrants are formulated for the above descrbed groups of connected, moble and dsconnected vehcles. The capactes of the ndvdual groups are calculated from the total storage capacty n PEVs and the tme-dependent share (vsh, vehcle share) of connected, moble or dsconnected vehcles. The capacty of connected vehcles s restrcted by SFL mn vsh conn,h [ P h P ls,h ] vex conn mob,h SFL max vsh conn,h

5 wth the mnmal and maxmal storage fll levels SFL mn and SFL max, respectvely. The share of connected vehcles n each hour s referred to as vsh conn,h, and the energy that s taken out of the system wth bdrectonal exchange of energy va the transfer of vehcles (vex, vehcle exchange) form the connected to the moble state n each hour vex conn-mob,h. Whle moble storages are subject to SFL mn vsh mob,h vex conn mob,h vex mob dsconn,h P ds,h SFL max vsh mob,h Further, the amount of energy n the moble storages n the begnnng and at the end of each optmzaton nterval must be equal: h max h max h max vex conn mob,h vex mob dsconn,h P ds,h = 0 wth the share of moble (drvng) vehcles n each hour vsh mob,h and the dscharge load n each hour P ds,h. The electrcty exchange va vehcle transfer from the connected to the moble state n each hour s denoted as vex connmob,h and the transfer from the moble to dsconnected state s referred to as vex mob-dsconn,h. Dsconnected storages are subject to SFL mn (1 vsh conn,h vsh mob,h ) vex mob dsconn,h SFL max (1 vsh conn,h vsh mob,h ) Agan, the amount of energy n the dsconnected storages n the begnnng and end of each optmzaton nterval must be equal. Snce only a blateral exchange va vehcle transfer s possble n the case of dsconnected storages, the vehcle transfer s constrant by h max vex mob dsconn,h = 0 Addtonal to the storage constrants, also the blateral energy exchange va vehcle transfer s restrcted. The followng constrants for the exchange between the ndvdual storage groups ensures that the exchanged energy between the groups does not exceed the avalable storage capactes: SFL max vsh conn,h vex conn mob,h SFL max vsh conn,h SFL max vsh mob,h vex mob dsconn,h SFL max vsh mob,h SFL max (1 vsh conn,h vsh mob,h ) vex mob dsconn,h The above descrbed modellng approach results n a quanttatve load shftng potental, provdng detaled nformaton about the seasonal, weekly as well as hourly load shftng avalablty of the ndvdual applances. It generates a smoothed net load curve that may be used n an electrcty market model to quantfy the mpacts of DR on the electrcty system. We may now determne the load shft potental of electrc vehcles from dfferent user groups (prvate, commercal fleet and company cars) at dfferent chargng facltes (domestc, work and publc). Framework assumptons for case study Scenaro set up For our case study, we defne three scenaros wth dfferent avalablty of chargng nfrastructure: In scenaro S1 chargng s possble when vehcles are parked at home at 3.7 kw and prvate vehcle users pay for ther prvate chargng pont. Commercal fleet vehcles charge only commercally at 3.7 kw and pay for ths chargng possblty. We ncrease the avalablty of chargng nfrastructure for prvate users by allowng chargng at work at 3.7 kw n scenaro S2 whle users have to pay for the prvate chargng pont and the one at work. Commercal fleet vehcles are charged as n the frst scenaro. In scenaro S3 also publc chargng at 3.7 kw s possble for prvate and commercal

6 fleet vehcles. In all scenaros, the ndvdual chargng nfrastructure (domestc, commercal and at work) s pad by the vehcle holder and hence used whenever possble,.e. all vehcles are charged whenever they are at domestc, commercal chargng spots and, f avalable, at work. All users pay for publc chargng through the publc chargng prce whch contans the costs for publc chargng nfrastructure ncludng a subsdy from publc authortes. For ths reason, vehcles are only charged n publc f the battery s half depleted (to return home) and, n case of PHEVs, drvng wth electrcty from publc chargng statons s cheaper than drvng wth conventonal fuel. Framework condtons for PEV market dffuson For the smulaton of the PEV market dffuson, we need a varety of techncal and economc assumptons. For BEVs, we consder battery szes of 40 kwh and a depth of dscharge of 90 % whch results n an average electrc drvng range of 180 km (2015) to 210 km (2030). The PHEV contans a battery wth 10 kwh capacty allowng 42 km (2015) to 50 km (2030) electrc drvng dstance at 80 % depth of dscharge. All other vehcle parameters can be found n [Gnann 2015] whle battery and energy prces are shown n Table 2. Here, we consder an average scenaro wth a conservatve exponental decrease of battery prces and a slght ncrease of fuel prces based on the New Polces scenaro of the World Energy Outlook [Pfahl 2013, IEA 2012]. Electrcty prces ncrease untl 2025 due to the EEGsupplement, but proft from economes of scale for renewable energes n 2030 [Schlesnger et al. 2011, BCG 2013, Lepzger Insttut für Energe GmbH 2012, McKnsey 2012]. The nvestment horzon as well as holdng tme for frst vehcles s 3.8 years for commercal vehcles and 6.2 years for prvate vehcles [DAT, 2011, VCD 2008]. An nterest rate of 5 % s assumed [Pfahl 2013]. Table 2: Battery and energy prces (all prces wthout VAT n 2015) Battery and energy prces unt Ref. Battery prce /kwh [1] Gasolne prce /l [2] Desel prce /l [2] Electrcty prce prvate /kwh [3] Electrcty prce commercal /kwh [3] [1] (Pfahl 2013); [2] (IEA 2012, MWV 2013); [3] (Schlesnger et al. 2011, BCG 2013, Lepzger Insttut für Energe GmbH 2012, McKnsey 2012) Furthermore, we assume nfrastructure costs to be very low n the begnnng decreasng 5 % per year untl 2030 (see [Gnann 2015, Table 4.2]). We dstngush chargng nfrastructure accordng to the four types of accessblty (at home, at work, commercal and publc) and consder also for prvate users f they own a garage. The nvestment horzon s assumed to be 15 years [Plötz et al. 2013]. Prevous smulatons showed that chargng ponts always need to be subsdzed [Gnann 2015]. Thus, we use a subsdy as suggested n Scenaro "home, work & publc" n [Gnann 2015, p. 116] whch assumes that publc chargng ponts receve a subsdy of 85% of ts annual cost n 2015 lnearly decreasng to 0% n Framework condtons for optmal vehcle chargng The smulaton of optmal vehcle chargng s subject to techncal restrctons. To be consstent wth the assumptons n the ALADIN model, the average storage capacty n each year s assumed for the ndvdual PEVs. We assume that the electrcty demand has to be suppled on each day,.e. loads cannot be shfted from one day to the next. As long as the vehcles electrcty demand s suppled on tme for the next trp, t s assumed that the chargng load of each vehcle can be controlled for the entre tme nterval n whch the PEV s connected to the grd. Possble user nterferences or user preferences are not consdered here, whch results n the techncal potental of optmal loadschedulng. Results of case study PEV market dffuson Let us now turn to the smulaton results for the PEV stock. Fgure 3 llustrates the results of the smulatons for PEVs n the three scenaros from 2015 to 2030.

7 share of PEV stock PEV stock mllon S1 S2 S Fgure 3: Total stock of plug-n electrc vehcles n three scenaros We can observe the followng: The PEV stock rses to four mllon vehcles n the home chargng scenaro (S1), 4.7 mllon when home and work chargng are allowed (S2) and to 4.7 mllon PEVs n the home, work and publc chargng scenaro (S3). Thus, wth the avalablty of addtonal work chargng nfrastructure, the PEV stock may ncrease by about 700,000 vehcles. However, addtonal chargng nfrastructure n publc has no effect on the PEV stock. Ths s a surprsng result whch has been ntensvely dscussed n [Gnann, 2015]. Interestng for the load shft potental s also the share of prvate and commercal fleet vehcles as well as company cars and the number of BEVs and PHEVs. These shares are shown n Fgure 4 and Table % 80% 60% 40% 20% 0% S1 S2 S3 S1 S2 S fleet PHEV fleet BEV prvate PHEV prvate BEV Fgure 4: PEV stock n the three chargng scenaros dstngushed by user group and PEV type Here, the PEV stock n 2020 and 2030 s decomposed nto sx vehcle user groups: prvate BEVs (green dashed), prvate PHEVs (blue squared), fleet BEVs (green dotted) and fleet PHEVs (blue dashed) and company BEVs and PHEVs whch have a share below 1% and are thus not shown at all. We can observe that the PEV stock s domnated by PHEVs (80% n 2020 and 70% n 2030) whch s mportant snce ther battery szes decrease the load shft potental. Further, t s clearly vsble that fleet vehcles are mportant to consder n 2020 (about 80% wth prvate chargng only (S1) and about 55% f chargng at work s also possble (S2)), but also n 2030 when ther stock shares are stll around 25-30%. Ths s also an mportant fndng as fleet vehcles drve sgnfcantly more and more regularly than prvate cars [Gnann, 2015]. Hence, the shape of ther load curve could be dfferent. Ths wll be analyzed n the next secton.

8 electrcty demand of PEVs [MWh] Table 3: PEV stock n the three dfferent chargng scenaros dstngushed by user group and PEV type PEV STOCK S S S S S S Prvate PHEV 111, , ,000 1,838,000 2,606,000 2,501,000 Prvate BEV 13,000 30,000 31, , , ,000 Commercal PHEV 342, , , , , ,000 Commercal BEV 69,000 69,000 69, , , ,000 Optmal vehcle chargng We frst take a look at the load curves of a regular Tuesday for all three scenaros shown n Fgure 5 3. As the PEV stock and the chargng optons dffer especally n scenaro S1 compared to S2 and S3, ther load curves vary too. In scenaro S1, we fnd the typcal load curve of prvate electrc vehcles wth a lot of chargng overnght when vehcles are parked at home. Commercal chargng, however, more often occurs durng the day when commercal fleet vehcles are often parked at the company. Ther chargng also ncreases the evenng peaks, yet a lot of chargng s performed durng the day when a substantal amount of electrcty from RES s avalable. Turnng to the possblty of addtonally chargng at work n S2, we fnd that an addtonal load peak occurs n the mornng hours when people arrve at ther workplace (green coloured area). Agan, ths may be favourable to match demand and supply of RES electrcty even wthout load shftng (see also [Babrowsk et al. 2014]). If publc chargng ponts are added n S3, only a small part of the electrcty demand s covered by these addtonal chargng optons (purple coloured area) whch also occurs n the evenng (peak) hours. 3,000 2,500 2,000 1,500 1, S1 S2 S3 domestc chargng commercal chargng work chargng publc chargng Fgure 5: Electrcty demand of PEVs wthout DR on a Tuesday dependng on the avalable chargng nfrastructure n the scenaros. Consderng demand response, chargng of electrc vehcles n summer tme s prmarly shfted nto mdday hours due to substantal PV-based power generaton and a correspondngly low level of the net load. In the wnter season, vehcle chargng s partally shfted nto nght tme hours, especally at days wth low solar generaton. As descrbed n the methodology secton, the ALADIN model provdes tme-dependent shares of PEVs n the three storage groups connected, moble and dsconnected whch serve as organzatonal constrants wthn the eload model. The hourly shares n each group are calculated from ndvdual drvng profles and the vehcle stock n each year. Fgure 6 depcts the hourly share of PEVs n the connected state for Tuesdays for prvate cars (left panel) and commercal vehcles (rght panel) n the dfferent scenaros. Snce PEVs can only be charged n the connected state, the fgure llustrates that the possblty to shft chargng loads to the mddle of the day s greatly enhanced n the scenaros S2 and S3. The reason s the possblty to charge prvate passenger cars at the workplace, where they are parked durng the day. In the scenaro S3, the opton of chargng n publc places enhances the avalablty of prvate cars further and addtonally allows a hgher share of connected commercal PEVs durng the day. 3 We chose a Tuesday for comparson snce most vehcle profles start on Monday and we thus may see some untypcal effects.

9 Chargng load [GW] Share of connected PEVs Share of connected PEVs 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% S1 S2 S Hour of the day 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Hour of the day Fgure 6: Average share of connected PEVs on Tuesdays n the three scenaros over the course of the day; prvate passenger cars (left) and commercal fleet (rght) The dfferent chargng optons between the scenaros lead to sgnfcant dfferences n the electrcty consumpton when the demand response opton s consdered. Fgure 7 compares the load profles of uncontrolled chargng (black lne) and optmzed chargng (demand response (DR)) n scenaro S1 (blue lne), S2 (green lne) and S3 (dotted lne). The fgure llustrates that the addtonal chargng opton at work leads to a new load peak n summer at mdday, when load s shfted towards hours wth substantal PV-based power generaton and a correspondngly low net load. In the wnter season the load s shfted prmarly nto nght hours, and the effect of the chargng opton at work s consderably smaller. If publc chargng ponts are added n scenaro S3, the electrcty consumpton remans almost the same. Hence, the avalablty of electrc vehcles durng hours wth a low net load s already suffcent when chargng at home and at work s possble S1 - Uncontr. S1 - DR S2 - DR S3 - DR Summer Wnter Fgure 7: Electrcty consumpton of PEVs on Tuesdays n summer and wnter, wth uncontrolled chargng (uncontr.) and wth demand response (DR) To address the effect of the demand response opton on the net load, optmal vehcle chargng s compared to uncontrolled chargng for the scenaro S3 n Fgure 8 (left panel - uncontrolled chargng, rght panel - demand response). The net load curve s charactersed by a valley durng mdday hours wth substantal PV power producton, especally n the summer months. In Wnter, the mdday valley s less pronounced, and the lowest net load occurs n the nght hours. Consequently, the optmal chargng of electrc vehcles s generally scheduled at mdday, and addtonally durng the nght hours n wnter. Both prvate and commercal vehcles are able to shft substantal amounts of ther chargng load, due to overall hgh shares of vehcles n the connected state n the scenaro S3 (for the avalablty, see Fgure 6).

10 Vehcle chargng load [GW] Resdual load [GW] Commercal vehcles Prvate vehcles Resdual load Summer Wnter Summer Wnter Uncontrolled chargng Demand response Fgure 8: Average vehcle chargng and net load n 2030 wth chargng optons at home, at work and n publc (Scenaro S3). In scenaro S3, the descrbed deployment of load shftng can facltate peak shavng of the system load by about 2.2 GW or 2.8% compared wth a scenaro wthout load shftng. Thus, the surplus of electrcty by RES can be reduced by 1.8 TWh or 30% (see also Table 4) resultng n 1.8 TWh of RES surplus used for chargng of PEVs. In contrast, uncontrolled chargng of electrc vehcles at work, at home and n publc rases electrc load by more than 2 GW, n partcular at current peak hours (at 10 am and 7 pm, see left part of Fgure 8). Table 4 lsts the reducton of RES surplus electrcty and the peak load reducton for all three scenaros. The reducton of RES surplus s sgnfcantly lower, when only chargng at home s possble for the prvate PEVs (0.2 TWh less than n scenaros S2 and S3). Addtonally, the peak load ncreases n the scenaro S1. Ths stems from the lmted possblty of load schedulng, whch results n large amounts of chargng load beng shfted towards the nght hours n wnter and thus creatng a new load peak. The addtonal publc chargng ponts n scenaro S3 only have a small effect on the RES surplus reducton as well as on the peak load reducton. Nevertheless, the results of scenaros S1 n comparson to S2/S3 show that addtonal chargng nfrastructure at work does not only promote the dffuson of prvate PEVs but also enables vehcle chargng durng mdday hours when solar electrcty generaton s hghest. Table 4: Summary of load shftng potental n the three dfferent chargng scenaros Type S1 S2 S3 Reducton of RES surplus electrcty 1.6 TWh 1.8 TWh 1.8 TWh Reducton of peak load (System load curve) GW 2.2 GW 2.2 GW Summary, dscusson and concluson The am of ths study was to analyse the load shftng potental of plug-n electrc vehcles when takng dfferent accessblty types of chargng nfrastructure (domestc, work and publc chargng optons) as well as dfferent types of vehcle user groups nto consderaton. We found that the market projectons for plug-n electrc vehcles are domnated by commercal fleet vehcles (at least n 2020) and plug-n hybrd electrc vehcles (also untl 2030). Ths largely nfluences the load curve of electrc vehcle chargng, snce commercal fleet vehcles charge more often durng the day when electrcty from RES s avalable. Apart from that, addtonal chargng at work also shfts chargng of prvate cars nto the day creatng an addtonal mornng peak. Addtonal chargng optons n publc nether have meanngful effect on the electrcty demand of PEVs nor on load shftng potentals. Demand response can reduce the total amount of RES surplus electrcty by 1.6 TWh (26%) f only domestc chargng optons are avalable and by 1.8 TWh (30%) wth addtonal chargng at work. We have to remnd the reader that both results are based on smulaton models whch cannot provde perfect projectons of the future. Chargng of users may be dfferent, e.g. hgher n publc, yet large research projects and reports support the hgh mportance of domestc chargng, meanngful mportance of chargng at work and unsure relevance of publc chargng [Ecotalty and INL 2012, Bruce et al. 2012, IEA 2017]. Furthermore, the here presented approach decomposes the load for dfferent electrc vehcles to groups but nevertheless represents the PEVs wthn each group on an aggregated level. Ths aggregaton may be crtczed snce t can possbly overestmate the potental of demand response. Yet, t enables the consderaton of avalablty and electrcty consumpton wth a hgh tme resoluton and therefore provdes an added value to the dscusson of the load shftng potental provded by moble storages. Further research could nclude the evaluaton of load shftng potentals for ndvdual PEVs.

11 In our pont of vew, an nterestng fndng for electrcty supplers s the hgh share of commercal chargng durng the day and the possblty to shft loads by addtonal chargng optons at work. Snce provdng addtonal chargng optons at work mght be less costly than equppng a lot of chargng ponts wth technology for demand response, sellng ths equpment could be an nterestng busness model wth addtonal gans for electrcty supplers as well. For polcy makers, ths underlnes the necessty to facltate chargng at work. Support programs for settng up chargng nfrastructure at work could support the current legslatve smplfcaton for companes to provde chargng facltes for ther employees. Further research could nclude publc fast chargng optons whch have been neglected n ths study. Also, optmzng on based on prce sgnals may be another nterestng opton n new modellng attempts. Moreover, the comparson of the DR potental of PEVs could be compared to the potental of other flexble technologes, such as heat pumps or statonary storage battery that are currently nstalled by households to store ther self-generated PV power. Also, an optmzaton on prce sgnals nstead of peak shavng could be nterestng for further research as well. Acknowledgments Parts of the analyss were executed wthn the project Bewertung des Lastverschebepotenzals von Elektrofahrzeugen n Deutschland unter Berückschtgung dfferenzerter Haltergruppen und Ladenfrastruktur m Verglech zu anderen flexblen Verbrauchern (fundng code A ) whch receves fundng from the Stftung Energeforschung Baden-Württemberg. For further nformaton see: We also performed a part of ths work wthn the framework of the Proflregon Mobltätssysteme Karlsruhe whch s funded by the Mnstry of Economc Affars, Labour and Housng n Baden-Württemberg and as a natonal Hgh Performance Center by the Fraunhofer-Gesellschaft. References Babrowsk, S., Henrchs, H., Jochem, P., and Fchtner, W. (2014). Load shft potental of electrc vehcles n Europe. Journal of Power Sources, 255(0): BCG (2013): Trendstude Kompetenzntatve Energe des BDI. Stude der Boston Consultng Group m Auftrag des Bundesverbandes der Deutschen Industre (BDI) BCG: München Boßmann, T. (2015). The contrbuton of electrcty consumers to peak shavng and the ntegraton of renewable energy sources by means of demand response. Dssertaton, KIT Karlsruhe. Boßmann, T., Elsland, R., Klngler, A-L., Catenazz, G., Jakob, M. (2015). Assessng the Optmal Use of Electrc Heatng Systems for Integratng Renewable Energy Sources, Energy Proceda, Vol. 83, pp , Bruce, I., Butcher, N., and Fell, C. (2012). Lessons and Insghts from Experence of Electrc Vehcles n the Communty. In Proceedngs of Electrc Vehcle Symposum 26 (EVS 26), Los Angeles, US. Dallnger, D., Schubert, G., and Wetschel, M. (2013). Integraton of ntermttent renewable power supply usng grdconnected vehcles - A 2030 case study for Calforna and Germany. Appled Energy, 104(0): DAT (2011): DAT Report (2011). Ostfldern: Deutsche Automobl Treuhand (DAT). Ecotalty and INL (2012). The EV Project Q Report. Techncal report, Ecotalty Inc. and Idaho Natonal Lab, US. Fraunhofer ISI. REM2030 Drvng Profles Database V Fraunhofer Insttute of Systems and Innovaton Research ISI, Karlsruhe, July Fraunhofer ISI, IREES, TEP Energy (2017). FORECAST/eLOAD webste. Last accessed Gnann, T. (2015). Market dffuson of plug-n electrc vehcles and ther chargng nfrastructure. Dssertaton, KIT Karlsruhe.

12 Gnann, T.; Plötz, P.; Kühn, A.; Wetschel, M.: Modellng Market Dffuson of Electrc Vehcles wth Real World Drvng Data German market and Polcy optons. Transportaton Research Part A, Vol. 77, July 2015, pp Hautznger, H., Kagerbauer, M., Mallg, N., Pfeffer, M., Zumkeller, D. Mkromodellerung für de Regon Stuttgart - Schlussbercht - INOVAPLAN GmbH, Insttute for Transport Studes at the Karlsruhe Insttute of Technology (KIT), Insttut für angewandte Verkehrs- und Toursmusforschung e.v., Karlsruhe, Helbronn, 2013 Internatonal Energy Agency (IEA). World Energy Outlook 2012, Internatonal Energy Agency (IEA): Energy Technology Perspectves Catalysng Energy Technology Transformatons, IEA/OECD 2017, ISSN: KBA. German Motor Transport Authorty. Vehcle regstratons 2013 and vehcle stock (01/01/2014) dstngushed by vehcle regstratons areas (FZ1, FZ5), 2014 Lepzger Insttut für Energe GmbH (2012): Entwcklung der Prese für Strom und Erdgas n Baden-Württemberg bs 2020, Endbercht, Lepzg McKnsey (2012): De Energewende n Deutschland Anspruch, Wrklchket und Perspektven. McKnsey & Company MOP. Mobltätspanel Deutschland Techncal report, Project conducted by the Insttute for Transport Studes at the Karlsruhe Insttute of Technology (KIT). Avalable at: Pfahl, S.. Alternatve Antrebskonzepte: Stand der Technk und Perspektven De Scht der Automoblndustre. In Alternatve Antrebskonzepte be sch wandelnden Mobltätsstlen: Tagungsbeträge vom 08. und 09. März 2012 am KIT, Karlsruhe, KIT Scentfc Publshng, pages , Eds. Jochem, P. and Poganetz, W.-R. and Grunwald, A. and Fchtner, W., Plötz, P., Gnann, T., and Wetschel, M. Modellng market dffuson of electrc vehcles wth real world drvng data part : Model structure and valdaton. Ecologcal Economcs, 107(0): , Plötz, P., Gnann, T., Wetschel, M., and Kühn, A. (2013). Market evoluton scenaros for electrc vehcles - detaled verson. Commssoned by acatech - German Natonal Academy of Scence and Engneerng and Workng Group 7 of the German Natonal Platform for Electrc Moblty (NPE). Fraunhofer ISI, Karlsruhe, Germany. Schlesnger, M., Lndenberger, D., und Lutz, C. (2011): Energeszenaren 2011, Basel/Köln/Osnabrück, Prognos AG, Energewrtschaftlches Insttut an der Unverstät zu Köln (EWI), Gesellschaft für wrtschaftlche Strukturforschung (GWS) VCD (2008). CO 2-baserte Denstwagenbesteuerung. Berln: Verkehrsclub Deutschland (VCD).