EVS27 Barcelna, Spain, Nvember 17-20, 2013 Pwer Management fr Private and Semi-Private EV Charging Andreja Žitnik 1, Brut Mehle 1 Etrel d..., Ukmarjeva ulica 2, inf@etrel.si Abstract The aim f this paper is t prpse a cncept f smart charging equipment [1] and algrithms which can be used t fulfil different requirements f electric vehicle users and the pwer grid. Tw different aspects are cnsidered: private charging (charging at hme, multi-dwelling buildings r business premises) and semiprivate charging (in frnt f supermarkets etc.) n clusters f charging statins. The technical cncept f the suggested smart charging slutin is presented in detail. Keywrds: charging, lad management, ptimizatin, pwer management, smart grid 1 Intrductin Past experience and studies n the behaviur f EV users indicate that a large majrity f EV charging will take place n private (hme, wrk, multi-dwellings) r semi-private (supermarkets, park&ride) lcatins. This is ne f the main reasns why a special emphasis shuld be placed n the ptimizatin f private and semi-private charging. As a result f higher levels f EV penetratin (> 10 %) [2], EV charging can cause prblems in lcal grids, especially in less develped grids and in areas with higher-than-average EV penetratin levels, where an extra lad n specific substatins is expected. The grid peratrs will have t upgrade any bttlenecks and at the same time try t influence the EV users' charging behaviur t ptimize the grid perfrmance. The technical slutins shuld balance user expectatins and grid requirements already in their initial develpment phase. Private charging is characterized by a smaller number f EVs that are charging n the private netwrk behind the grid cnnectin pint/energy metre. The main issue in this case is the limited nminal pwer f the cnnectin, which must cver all ther cnsumers, cnnected t the building's electrical installatin. Semi-private charging is characterized by the charging f several EVs n a cluster f public r semi-private charging statins under the cntrl f a Charging Statin Operatr (CSO). The CSO can ptimize the peak EV charging demand n its cluster(s) by managing the pwer accrding t the cnditins f the grid and its cntracts with ther business entities. 1.1 Private charging (hme charging, multi-dwelling buildings, business premises and ther private parking areas) Electric vehicle is a new type f electricity cnsumer that brings abut the issues f grid's physical limitatins, ecnmics f charging and safety. A single EV with average cnsumptin can use a cnsiderable share f the maximum capacity f the grid cnnectin pint. The maximum capacity can be achieved in terms f cntracted maximum pwer r in terms f physical limitatin f the grid cnnectin pint. In the frmer case, EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 1
the user can revise his cntract with the distributin cmpany and cntracts higher pwer, but with higher assciated fixed csts. In the latter case, the limitatin is a result f a limited capacity f the internal (hme) netwrk r the external limitatin f the lw-vltage distributin netwrk and is definitive as such and requires an upgrade t the grid. The ecnmics f EV charging is related t the amunt f energy cnsumed and its price, which can change statically r dynamically. The user wishes t reduce his r her csts (with the use f favurable tariffs, planned charging,...) while enjying a high level f service. The cst f regular charging can be significant, especially cmpared t ther typical cnsumers in an exemplary husehld netwrk. The final custmer fr electricity has a direct business relatin with the electricity supplier and an indirect relatin with the electricity distributin cmpany. The mtivatin f the electricity supplier cmpany t influence EV charging lies in the ptimizatin f its prtfli f cnsumers. The distributin cmpany n the ther hand is mtivated t influence charging when the part f the grid where the EV charging is taking place is experiencing higher lads. 1.2 Public and semi-private parking areas (Charging Statin Operatr) Cmpared t private charging scenari, tw new entities are present in public and semi-private charging scenaris: Charging Statin Operatr (CSO) and Electrmbility Service Prvider (EMSP) [3]. The CSO can either wn the EVSE (electric vehicle supply equipment; hereafter charging statins) r make a deal with the wner f semi-private lcatins and manage nly the installatin and maintenance f charging statins. The EMSP ffers nly the charging services t EV users and des nt have t actually wn r maintain any charging statins. CSO's additinal interest can be t ffer system services t smart grid peratrs. The CSO is faced with different issues related t pwer management f charging statins which are publicly accessible. CSO is managing a prtfli f charging statins r clusters f charging statins that are cnsidered as end cnsumers f electricity. The CSO's strategy is t define a price plicy with flat rates r dynamic tariffs, ptimize the utilizatin f existing infrastructure and expand the charging infrastructure based n usage trends. On semi-private EVSE-equipped parking lts, the ttal nminal utput pwer f all installed charging statins may exceed the maximum cnnectin capacity. In additin t this cnstraint there might be ther cnsumers at the same grid cnnectin pint, fr example at a supermarket with a few charging statins fr visitrs r an emplyee EV parking lt. In these cases, the number f cars that are actually charging is usually lwer than the ttal number f cars parked at the charging statins at any given time. Increasing maximum cnnectin capacity is expensive and shuld be avided when pssible. Pwer management algrithms allw us t handle times f peak cnsumptin withut resrting t such measures. Withut smart charging infrastructure and remte management f pwer, it is practically impssible t ptimize the prtfli and stay cmpetitive in the lng run. The CSO can manage charging pwer n tw levels: Thrugh lcal ptimizatin f charging pwer in individual clusters f charging statins, Thrugh a central DSM system, which determines the charging pwer fr individual clusters f charging statins based n the demand f prtfli ptimizatin r system services. 2 Smart charging statin slutin The different levels f private and semi-private charging entail different prcesses with a number f business entities with smetimes cntradictry interests. The presented cncept f smart charging statin and algrithms can cver the needs f all invlved actrs. 2.1 Cmmunicatin is key Cmmunicatin links between individual systems are essential fr pwer management: Between the charging statin and the vehicle: IEC 61851 standard, the cming ISO 15118 standard, which will be essential fr precise pwer management; EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 2
between the charging statin and the user: the desired charging duratin, current state f charging, cnfiguratin f charging pwer ptimizatin, reprts; between charging statins in a cluster: the distributin f pwer amng the charging statins in the cluster; between the charging statin/cluster and the CSO (fr public charging infrastructure): infrmatin n the availability f charging statins and cnnected vehicles, pwer management based n the utput f the Demand Side Management (DSM) algrithm, nline diagnstics f the charging statins, reservatin f charging statins; between the charging statin and the distributin's smart grid system: pririty management f charging pwer. 2.2 Lcal pwer management algrithms The pwer management takes place n the level f individual charging statins r a cluster f charging statins. Pwer management can be perfrmed in several mdes, which can be enabled simultaneusly; hwever, nly the mde with the highest pririty at any given time is actually perfrmed. The supprted mdes are: fast charging with the maximum pwer f the charging statin (accrding t settings), manually planned charging n individual charging statin (e.g. delayed t adjust t ther cnsumers r t lwer tariffs), ptimal charging in a cluster f charging statins, where the limitatins are defined in the cnfiguratin f the cluster, ptimal charging in a cluster f charging statins based n the measured lad f installatin (and grid cnnectin pint). 2.2.1 Optimal charging in a cluster f charging statins Nt all EVs n EVSE-equipped parking lts will be charging all the time, fr example due t fllwing reasns: the EV will be parked fr lnger than the battery needs t be fully charged, the battery is almst full upn arrival, the user desn't plug in the charger fr sme reasn, etc. Nt all EVs that will be charging at any given mment will use all available pwer frm the cnnected charging statin. Fr example, the charging statin might allw three-phase charging at 32 A per phase, but the EV with a single-phase charger may nly draw 20 A n ne phase. If we add t this the fact that ther (nn-ev) cnsumers at the same grid cnnectin pint culd als be changing their cnsumptin, we see that the charging statins perate in a very dynamic envirnment and there is a lt f ptential t ptimize their charging behaviur. The IEC 61851 standard allws the EV t signal t the charging statin hw much current it is able t receive. Additinally, it allws the statin t signal t the EV hw much pwer it may draw at any ne mment. The idea behind cluster ptimizatin is that if we can make charging statins exchange this data amng themselves and at the same time gather sme mre data frm EV users, we can use it t ptimize the charging levels acrss the whle EVSE-equipped parking lt, which wuld allw the wner/peratr t use up mst f the pwer available n grid cnnectin pint withut paying extra when the demand wuld exceed the cntracted capacity. As a slutin we envisined a self-rganizing cluster f charging statins that cmmunicate with each ther ver the lcal netwrk cmbined with an installatin lad measurement device that measures pwer cnsumptin f all cnsumers at the grid cnnectin pint in real time. One f the statins has the rle f the»master«statin and hsts the mathematical algrithm that enables an ptimal distributin f pwer amng the statins in EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 3
the cluster. The mathematical algrithms can perate in tw regimes: pririty treatment f fair distributin f pwer amng the cnnected users (algrithm 1), pririty treatment f ptimal utilizatin f ttal available charging pwer (algrithm 2). 2.3 Upper level pwer management algrithms The algrithm fr the distributin f pwer within the cluster perates lcally. The CSO cntrls and manages its infrastructure thrugh the cntrl centre sftware. One f the mdules f the cntrl centre is DSM, which enables remte management f the charging pwer. [4] The DSM mdule perfms nline calculatins fr individual clusters, based n the current state f the infrastructure and demands (ptimizatin f prtfli, system services,...). These calculatins are distributed t the lcal level via cmmunicatin pathways. In respnse t any change, the DSM algrithm sends a new charging plan t the cluster. The cluster tries t cnfrm t the received charging plan, with the pririty treatment f lcal limitatins. 3 Charging cluster and its ptimizatin The cluster f charging statins is frmed simply by distributing a list f lcal IP addresses f the charging statins that are part f the cluster. The sftware in charging statins selects the statin that will act as the master. Alternatively, the rle f the master culd be given t a separate device in the netwrk, but this cnfiguratin wuld be less resilient as it wuld create a single pint f failure. In the event f master failure, the charging statins wuld nt be able t calculate pwer distributin by themselves. Our slutin was therefre t make each charging statin bth master and slave and have a master electin algrithm chse which rle will be perfrmed by the statin at any ne time. In the event that the charging statin which is currently the master stps wrking r requires maintenance, the master electin algrithm autmatically chses a new statin as the master. Even in a hypthetical situatin when the netwrk wuld split in tw and the resulting tw grups f statins wuld nt be able t cmmunicate with each ther, the algrithm wuld frm tw separate clusters with the knwledge that there might be anther cluster that is unreachable and wuld regulate pwer cnsumptin accrdingly. The master statin is respnsible fr gathering data frm ther statins, fr example: is there an EV currently charging n the statin, when did the EV start charging, what is the minimum and maximum current with which the EV can charge, des the EV have a single-phase r a three-phase charger, what is the battery status, when wuld the user like t have the EV charged and ready t g. Figure 1: CSO-cntrlled pwer management If there is enugh pwer t charge all the cars at the full capacity f their chargers then all the master des is cnfirm charging plans. It is when the pwer is limited when the interesting things happen. EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 4
In such case the abve infrmatin is used t priritize amng EVs. We nrmally favur thse that need t be charged faster and/r thse that need mre energy, but the exact criteria is nt imprtant in this discussin as it can be defined by the wner/peratr f charging statins. As we have mentined befre, we have tw algrithms fr pwer management within the cluster. Algrithm 1 assumes that users are the happiest when the EV is fully charged by the time they specified and we adjust the algrithm s that there are nly a few cases in which the EV is nt yet fully charged at that time. Algrithm 2 n the ther hand maximizes the pwer cnsumptin and enables thse users with mre pwerful chargers t hit the rad sner. Our slutin can perate in either mde and can switch between them at any time. 3.1 Algrithm 1: Maximizing the number f EVs that are fully charged by the appinted time After the master statin priritizes the cars accrding t the selected criteria, it calculates the s called "fair current" fr all vehicles - this is the current that the EV wuld be assigned if its charger wuld have n restrictins. At this current, the car that wuld need t be full by 1 PM wuld always be full befre the car that needs t be full by 2 PM. But since EV manufacturers d set sme restrictins, bth in minimum and maximum current that can be drawn by the EVs, it can happen that ne EV's fair current is lwer than what the EV manufacturer allws while anther EV's fair current is mre than its charger can draw. In such cases the master needs t adjust the current and redistribute the remaining current amng ther vehicles. This wuld be a relatively straightfrward prblem t slve in a wrld where nly singlephase r three-phase chargers wuld exist, but mst charging areas will need t deal with a mix f bth in the same cluster. Our gal is that single-phase and three-phase chargers get an equal amunt f pwer if all ther variables like charging time, amunt f energy needed, minimum and maximum pwer, etc., are equal. Therefre the current we calculate fr three-phase chargers is generally 3 times lwer than the current fr single-phase chargers. T get the fair current fr three-phase chargers, we calculate it n all three phases and take the lwest value. Usually we will nt be able t assign fair current t the chargers because f minimum and maximum current limitatins f thse chargers. Each EV is assigned at least the minimum current that is allwed by the EV manufacturer (i.e. n EV is left waiting withut charging). There are als cars that get a "fair current" that is higher than what they can use, s we try t redistribute that current in a way that thse cars that need t be charged sner (r need mre energy) will get mre. T that end, we calculate a pririty factr fr each EV and distribute the remaining current accrding t it. Since all ur calculatins are dne per-phase, we can crrectly distribute pwer in an envirnment where there are ther single-phase cnsumers and where we might nt have the same amunt f current available n all three phases. 3.2 Algrithm 2: Maximizing cluster pwer In this mde the gal is t use up as much f the available pwer as pssible, while staying within limitatins f the lcal pwer netwrk. Fr this reasn the algrithm favurs three-phase vehicles (if any) as they cnsume pwer n all three phases. First the maximum current f all three-phase chargers cmbined is calculated (M3), as well as the minimum current f all single-phase chargers (m1). If the cmbined current (M3 + m1) is greater than the available current, the maximum current f three-phase chargers is adjusted accrdingly. On the ther hand, if the cmbined current is lwer than the available current, mre current is given t single-phase chargers. The calculated target current is distributed amng single-phase and three-phase chargers in separate calculatins. The relative pririty f charging is calculated in the same way as in algrithm 1. 3.3 Algrithm cmparisn The easiest way t understand the effects f bth pwer distributin algrithms is t take a lk at sme examples. In this sectin we present cases in which algrithms give very different results, hwever there are als many cases in which results frm bth algrithms are similar. Fr example, if EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 5
all EVs have three-phase r single-phase charging, bth algrithms will return identical results. 3.3.1 Example 1 This example shws the allcated current fr three EVs, tw f them with single-phase chargers and ne with a three-phase charger. We will assume that they all arrived at the same time and that each needs 8 kwh t fill their batteries; hwever, the desired charging time and charger pwer are different fr each EV. Let us say the cluster has 32 A available n each phase fr EV charging, s it is impssible t charge all three EVs at their full capacity. The table belw shws the effect f algrithms 1 and 2 n charging time. (DC = duratin f charging) Table 1: Example 1 Car 1 2 3 Charger Threephase Singlephase (L1) Singlephase (L2) Max current 32 20 16 (A) Desired DC 2:00 2:00 3:00 Ideal DC* 0:22 1:44 2:10 Algrithm 1 0:45 1:45 2:40 DC Algrithm 2 DC 0:28 2:22 2:45 *Ideal cnditins are thse in which the EV can be charged at its charger's full capacity. Figure 3: Algrithm 1 - percentage f available current used Algrithm 2 maximized the cluster pwer which resulted in EV 1 drawing 26 A n all three phases fr its entire charging time (28 min), while EVs 2 and 3 each drew 6 A n phases 1 and 2, respectively. The maximum cnsumptin by phase was therefre 32 A / 32 A / 26 A. But this came at the expense f EV user 2, since the EV had t charge fr 22 minutes lnger than desired. Figure 4: Algrithm 2 - cnsumed current by EV Algrithm 1, which maximizes the number f EVs charged n time, managed t charge all three EVs in the alltted time, but since EV 2 uses nly phase 1 and EV 3 uses nly phase 2, phase 3 was used nly at 12 A thrughut the time f charging f EV 1. Maximum cnsumptin by phase was 32 A / 28 A / 12 A. Figure 5: Algrithm 2 - percentage f available current used Figure 2: Algrithm 1 - cnsumed current by EV 3.3.2 Example 2 In this example we have three EVs again, ne with a single-phase charger and tw with three-phase EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 6
chargers. Each needs 10 kwh and the cluster has 32 A available n each phase thrughut the duratin f charging. There is enugh time and pwer available t charge all f their batteries using either f the tw algrithms. Table 2: Example 2 Car 1 2 3 Charger Threephase Threephase Singlephase (L1) Max current 20 20 32 (A) Desired DC 2:00 2:00 4:00 Ideal DC* 0:43 0:43 1:21 Algrithm 1 1:48 1:48 2:15 DC Algrithm 2 DC 1:00 1:00 2:15 Algrithm 2 maximized the cluster pwer which in this example resulted in a much better verall perfrmance f the cluster. Figure 8: Algrithm 2 - cnsumed current by EV Algrithm 1, which maximizes number f cars charged n time (but des nt guarantee the fastest charging pssible), needed 48 minutes lnger t charge the three-phase EVs than algrithm 2, while the single-phase EV was charged at the same time in bth instances. Figure 9: Algrithm 2 - percentage f available current used The cluster master is able t calculate effects f these tw algrithms in real time and switch between them as necessary t ensure high quality f service. Alternatively, the wner/peratr f charging statins can select the algrithm t be used. Figure 6: Algrithm 1 - cnsumed current by EV Figure 7: Algrithm 1 - percentage f available current used References [1] Andreja Žitnik (2012), Data Transfer in Electric Vehicle Charging System. EngD thesis. [2] Prject MERGE, http://www.ev-merge.eu/, accessed n 2013-2-10 [3] Jure Ratej, Brut Mehle, Miha Kcbek, Glbal Service Prvider fr Electric Vehicle Raming, EVS27, 2013 [4] Brut Mehle, Andreja Žitnik, Jure Ratej, Plnjenje električnih vzil in prilagajanje djema ptrebam SODO (EV charging pwer management in the scpe f DSO smart grid system), Slvenian CIGRE-CIRED cnference 2013 EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 7
Authrs Andreja Žitnik is a Bachelr f Engineering in Cmputer Science and Infrmatin Technlgy frm the University f Ljubljana and is currently the Head f the applicatin develpment department at Etrel d... Brut Mehle is a Bachelr f Science in Electrical Engineering frm the University f Ljubljana and is currently the chief technical fficer f Etrel d... verseeing the develpment f charging hardware and sftware slutins. EVS27 Internatinal Battery, Hybrid and Fuel Cell Electric Vehicle Sympsium 8