Plug-in Electric Vehicles Parking Lot Equilibria with Energy and Reserve Markets

Transcription

1 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems 1 Plug-in Elecric Vehicles Parking Lo Equilibria wih Energy and Reserve Markes Nilufar Neyesani, Member, IEEE, Maziar Y. Damavandi, Member, IEEE, Miadreza Shafie-khah, Member, IEEE, Anasasios G. Bakirzis, Fellow, IEEE, and João P. S. Caalão, Senior Member, IEEE Absrac This paper proposes a comprehensive model for he ineracions of he plug-in elecric vehicles (PEVs) involved paries. An aggregaor wih mixed resources is assumed o be he inerface beween he parking lo () and he upsream energy and reserve markes. On he oher hand, he ineracions of he PEV owners and he are also modeled as hey impose resricions o he s behavior. Therefore, a bilevel problem is consruced where in he upper-level he objecive of he aggregaor is o maximize is profi hrough is ineracions and in he lower-level he maximizes is own profi limied o he preferences of PEVs. The objecives of he upper and lower level are conradicory; hence, an equilibrium poin should be found o solve he problem. In his regard, he dualiy heorem is employed o conver he bilevel model o a mahemaical program wih equilibrium consrains (MPEC). The model is implemened on he IEEE 37-bus nework wih added disribued generaions (DGs). Various cases are horoughly invesigaed and conclusions are duly drawn. Index erms Aggregaor, energy and reserve markes, mahemaical programming wih equilibrium consrains (MPEC), parking lo (), plug-in elecric vehicle (PEV). NOMENCLATURE Capial leers denoe parameers and small ones denoe variables. Subscrips j, k Bus number l Power line m DG number Time inerval ω Scenario and scenario se Superscrips Agg Aggregaor Aux Auxiliary variable ar Arrived PEVs o he cha Charging mode D Demand dcha Discharging mode del Delegaed energy (probabiliy of reserve call) This work was suppored by FEDER funds (European Union) hrough COMPETE and by Poruguese funds hrough FCT, under Projecs FCOMP FEDER-2282 (Ref. PTDC/EEA-EEL/118519/21) and UID/CEC/521/213. Also, he research leading o hese resuls has received funding from he EU Sevenh Framework Programme FP7/ under gran agreemen no (projec SiNGULAR). N. Neyesani, M. Y. Damavandi, M. Shafie-khah and J. P. S. Caalão are wih INESC TEC and he Faculy of Engineering of he Universiy of Poro, Poro , Porugal, also wih C-MAST, Universiy of Beira Inerior, Covilhã 621-1, Porugal, and also wih INESC-ID, Insiuo Superior Técnico, Universiy of Lisbon, Lisbon 149-1, Porugal ( caalao@ubi.p). A. G. Bakirzis is wih he Arisole Universiy of Thessaloniki (AUTh), Thessaloniki, Greece ( bakiana@eng.auh.gr). dep Depared PEVs from he DG Disribued Generaion DSO Disribuion sysem operaor ED Energy delivery o demand EDG Energy purchase from DG EM Energy Marke Exra Exra paymen for V2G paricipaion EV Elecric vehicle fix Fixed SOC requiremen flex Flexible SOC requiremen G2V Grid o Vehicle IL Inerrupible load in Power injeced ino he sysem or he Incenive Incenive paymen o inerrupible loads Line Disribuion Lines LL Lower level problem Loss Power loss ou Oupu energy from Parking Lo Re Reserve RM Reserve Marke Sc Scenario Tariff Tariff from PEV owners enering TM Trade wih Marke T Trade wih Toal Toal amoun of demand ToU Time of Use UL Upper level problem V2G Vehicle o Grid Operaors, Maximum and minimum amoun of a variable Expeced value of a variable Idenificaion of variable on he seleced node Variables and Parameers C Capaciy of a (kw) Cd Cos of equipmen degradaion FOR Forced ouage rae (%) i, I Line curren (A) n, N Number of parked PEVs p, P Acive power (kw) q, Q Reacive Power (kvar) r, R Reserve (kw) R, X Resisance and reacance of a line (Ω) Si Binary variable indicaing he sie of resources soc, SOC Sae of Charge (kwh) Z Variable for linearizing he condiional erm α Percenage of inerrupible load from oal demand β Coefficien deermining he share of each PEV caegory from hourly vehicle deparure (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

2 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems θ Coefficien deermining he share of each PEV caegory from oal PEVs in he in each hour φ Coefficien deermining he minimum deparure SOC requiremen of each PEV caegory γ, Γ Charge/Discharge rae ρ Probabiliy η Efficiency π, Π Price λ Dual variable for equaliy consrains μ, μ Dual variable for lower and upper limis in inequaliy consrains I. INTRODUCTION A. Moivaion and Aim lecrificaion of ransporaion is an emerging rend in power sysem sudies, raffic planning and urban sudies. EPeneraion of elecric vehicles in everyday life has several aspecs ha should be deal wih. Deploymen of plug-in elecrical vehicles (PEVs) no only affecs he operaion of he power sysem, bu also imposes some necessary ineracions ha have no exised in he sysem before. These ineracions regard he echnical impacs of PEVs as well as economics, raffic and allocaion of PEVs and occur among all he paries ha are involved wih he PEVs. The paries could be he owners of he PEVs, he operaor of he charging saions, he disribuion sysem operaor (DSO), he urban planner, ec. Vas peneraion of PEVs in he sysem requires foreseeing he necessary infrasrucures. One of he recen soluions o provide he needed plaform for beer uilizaion of PEVs is he PEVs parking lo (). s provide a medium for he PEVs o charge heir baeries and an aggregaed version of PEVs o ac as sorage. The grid o vehicle (G2V) and vehicle o grid (V2G) modes gives he he poenial of being a resource in he sysem as well as he flexible load. Therefore, he will be a new pary in he ineracions of he PEVinvolved paries and will bring more conflics and challenges o he problem. On he oher hand, he raffic paern of he area where he is insalled and he behavior of he PEV owners ha use he considerably affec he operaion of he. The arrival and deparure paern of PEVs and heir say duraion influences he s behavior. Besides, heir charging requiremens impose oher resricions o he operaion. As a resul, modeling he PEVs behavior and heir obligaions effecs on he s behavior is necessary for sudying he ransacions of he and he marke. Confroning wih he above-menioned challenges, he aim of his paper is o invesigae he ineracions of he in he marke place affeced by he PEVs preferences in a mixed resource environmen. B. Lieraure Review The subjec of PEVs has been he focus of many recen sudies; however, he lieraure relaed o he subjec of his paper can be caegorized ino hree groups: The firs group which covers he inroducion of PEV aggregaor as a new eniy o he power sysem; he second group ha considers he marke paricipaion of PEVs hrough an aggregaor and he hird one ha invesigaes he ineracion of PEVs in a. As for he firs caegory, he preliminary impressions of agens for were brough by Kempon [1] indicaing ha he presence of an agen is necessary for he operaion of PEVs in he sysem; Lopes [2] encouraging he aggregaion of he PEVs in order o have a considerable effec on he sysem is ineviable; Guille [3] ha proposed he aggregaor as a criical eniy o enable he V2G operaion of EVs. A comprehensive survey on EV aggregaion can be found in [4]. The real-ime regulaion allocaion on EV aggregaors is presened in [5] wih welfare-maximizaion objecive. Jin e al. in [6] repored an opimized EV charging schedule hrough an aggregaor while considering he aggregaor s revenue and he EVs charging demand. In [7], he scheduling of EVs by aggregaors o ake par in V2G regulaion is sudied where he forecas of schedules based on he uncerainies of EVs is performed by muli-level aggregaors. Regarding he second caegory, a considerable number of available sudies have dedicaed he focus of heir sudy o he inegraion of EVs ino markeplace hrough aggregaors. Bessa e al. in [8] inroduce an EV aggregaion agen and propose an opimizaion approach for he agen o bid and paricipae in day-ahead and reserve markes. However, i considers individual EVs plugged o he grid from charging saions and he aggregaor conrols he EV charging for specific ime duraion based on he conrac beween each EV and he aggregaor. The auhors also invesigaed he model for hourahead marke in [9] as well as he manual reserve, no considering he V2G mode hough. In [1] a coordinaion approach beween EV aggregaor and sysem operaor is presened in boh elecriciy marke and ancillary services. The auhors in [11] developed a model for charging he EVs while he aggregaor rades wih energy and reserve markes. In [11], i is considered ha he charging of EVs is opimized wih he presence of elecric sorage. However, i does no consider he V2G mode of he EV operaion. Similarly, in [12] a bidding sraegy for he sochasic behavior of EV aggregaor is acquired o paricipae in energy and regulaion markes. Reference [12] also considers he EVs o be operaed in G2V mode only and he aggregaed EV poenial is deployed as regulaion up/down. Li e al. in [13] used an EV aggregaor model in heir locaional marginal pricing mehod o alleviae he congesion caused by EVs load. Alhough mos of he sudies have only considered he G2V mode of he EVs o paricipae in he elecriciy marke, here are some sudies ha consider he V2G mode. Soromme and El-Sharkawi in [14] and [15] developed a V2G algorihm for an EV aggregaor o paricipae in boh energy and ancillary service markes. The hird group, however, regards anoher poin of view in EVs paricipaion in power sysems. The inroducion of EV s o he sysem has changed he feaures of PEV peneraion sudies. Oher han he problem of allocaing s in he sysem [16], [17], he effecs caused by he procedure of charging/discharging in he have been he maer of ineres in he lieraure. The reason is ha, as firsly proposed by [18], he uilizaion of EVs V2G mode can be faciliaed by deploymen of s as an aggregaed source of PEVs. Furher sudies such as [19] - [21] addressed he V2G mode of. However, he sudy on he simulaneous charging/discharging of he is sill very limied. Some sudies such as [22] and [23] have sudied he managemen of he s ineracion, The significance of his work comparing o he abovemenioned sudies is ha he marke paricipaion of he (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

3 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems in boh G2V andv2g mode has been considered while mos of he sudies only considered he G2V mode or he baery replacemen procedure (especially firs and second group). Moreover, he as he aggregaion of he PEVs is differen from a PEV aggregaor because in he all he PEVs locaion is fixed; hence, he vision of is marke paricipaion should be reaed differenly from an aggregaor. On he oher hand, as he has a limied capaciy due o resriced number of PEV saions, i may no be able o paricipae individually in he marke and should be examined in a mixed resource environmen. In addiion, he complex inerrelaion of he wih oher PEV-involved paries such as PEV owners, aggregaor, DSO, ec. is no addressed by he hird group sudies discussed in he lieraure review. Alhough hese sudies have somehow considered he preferences of PEV owners, he conradicory effecs of he vehicle owners behavior on s operaion have no been addressed. C. Conribuions Considering he issues discussed in he previous subsecions, i is necessary o derive a comprehensive model ha could address all he possibiliies and limiaions of he s operaion. The inenion of his paper is o presen he model for he ineracions of he wih he marke hrough an aggregaor while considering he resricions ha he preferences of PEV owners impose o is behavior. The increased level of flexibiliy due o he in he sysem is invesigaed hrough is inegraed operaion wih oher resources such as disribued generaion (DG) and demand response (DR). The impac of PEV owners preferences on he s operaion is addressed. In addiion, a novel and pracical framework o involve he PEVs preferences in he s operaion is proposed. The paper s main conribuions are: 1) To propose a model o impose he preferences of he PEVs who use he based on heir choice of G2V/V2G mode, ime of say and heir requiremen of SOC on deparure ime; 2) To model he ineracion of wih a mixed resource aggregaor based on a bilevel approach; 3) To invesigae he effecs of PEV preferences on equilibrium poin of and aggregaor ineracion. D. Paper Organizaion The res of he paper is organized as follows. Secion II describes he various ineracions of he componens in a disribuion sysem wih he and aggregaor. The mahemaical formulaion of he upper-level problem is described in secions III, while he lower-level problem is presened in secion IV. The uncerainy of he PEVs behavior in he is modeled in secion V. The numerical resuls are presened in secion VI. A comprehensive discussion on he role of PEVs preferences on he behavior is presened in secion VII. Finally, secion VIII concludes he paper. II. PROBLEM DESCRIPTION As comprehensively discussed in he lieraure review, numerous inerconnecions of PEVs should be managed hrough he new eniy of PEV aggregaor. Alhough a is an aggregaed form of PEV, resricions of is operaion confine he o compee independenly in he marke. However, he poenial of he as a resource in he sysem as well as is naure of being a flexible load canno be disregarded. In fac, he special role of he as a prosumer in he sysem can be bes employed along wih oher available resources in he sysem. Therefore, aggregaing s opporuniies wih oher resources such as DG and DR provides a suiable environmen for he aggregaor o achieve a higher level of flexibiliy. On he above premises, his paper proposes a model in which an aggregaor is he inerface of local resources wih he marke. The basic visual of such environmen is shown in Fig. 1. In his environmen, he paricipaes in he marke hrough an aggregaor which has o provide he required demand for he load reailer. Anoher resource (DG) is also presen in he sysem o sudy he variaions of price. The aggregaor combines all he resources in he local nework o maximize is profi when paricipaing in he upsream energy and reserve markes. However, each of he componens ha are aggregaed by he aggregaor has is own objecive and resricions ha may have conflic wih he objecive of he aggregaor. Therefore, a bilevel problem is encounered in his siuaion. In he upper-level (UL) problem, he objecive of he aggregaor is o maximize is own profi hrough is ineracion wih he upsream marke on one hand and he energy and reserve rade wih he, energy purchase from he DG and providing he required demand on he oher. On he lower-level (LL), he, he DG, and he load reailer are he componens who also wan o maximize heir profi. As a resul, an equilibrium poin should be found for he operaion of such sysem. The ineracions beween he wo levels of he model are described in follows. A. Aggregaor--PEV ineracions The provides he opporuniy for he PEV owners o charge heir baeries and ake par in he V2G mode if hey are willing o. The can ac more efficienly in he marke compared o charging saions because i enables he simulaneous G2V/V2G mode and i also benefis from he longer say of he PEVs in he. Consequenly, i can have he role of sorage as well as flexible load in he sysem. However, when operaing a, i is necessary o consider he preferences of he PEV owners. In some of he recen sudies on his subjec, such as [24]-[26] he behavior of PEVs has been considered peraining heir driven disance and sae of charge. However, he owners of hese vehicles may also have preferences oher han he limiaions of PEV. D RM ou, Re, EM in, DG Fig. 1. Ineracions of he componens in he environmen (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

4 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems 4 Fig. 2. The sequence of ineracions from PEVs o Marke In his sudy, i is assumed ha he PEVs who ener he resric he s behavior in he markeplace wih heir choice on heir paricipaion as G2V or V2G. The reason is o consider he owners concern on heir baery degradaion in V2G mode or he probabiliy of he sudden deparure. In such cases, he owners may no wan o paricipae in he V2G mode o be sure of adequae charge in heir baeries for heir nex ravel. Noe ha by V2G mode, we mean ha PEVs are willing o ake boh G2V and V2G mode. For he sake of breviy, only he V2G erm is used for his ype of PEVs. Moreover, i is assumed ha all PEVs specify a minimum amoun of SOC of heir baeries a he ime of deparure from he ; however, some of he PEVs need a fixed amoun of deparure SOC while ohers agree o have a flexible deparure SOC and he only limi for hem is heir minimum SOC. The reason for considering fixed deparure SOC is o ake ino accoun he possible conracs of PEV owners wih oher PEV-aggregaors which oblige hem o keep a specific porion of heir capaciy empy. Therefore, four differen caegories of he PEVs ener he : G2V mode wih fixed deparure SOC; G2V mode wih flexible deparure SOC; Boh G2V, V2G mode wih fixed deparure SOC; Boh G2V, V2G mode wih flexible deparure SOC. Each of hese caegories and heir requiremens resric he in uilizing he oal available capaciy in he. Figure 2 shows various ineracions ha occur from marke o PEVs hrough he aggregaor and. As shown, wo main physical and financial ineracions exis. The objecives of he and he aggregaor due o is ineracions wih are based on financial ransacions shown in Fig. 2. In each inerface (aggregaor or ) differen prices are applied o he ransacions and are illusraed wih differen line ypes. As a resul, an equilibrium poin should be found beween all he objecives B. Mixed Resource Environmen As a main feaure of he forhcoming power sysems and for enabling he aggregaor o have access o more resources, DGs are also considered in he sysem. I is assumed ha he DGs offer heir price and quaniy o he aggregaor, bu hey should reach an equilibrium poin in heir rade. Hence, he price ha he aggregaor buys he power from DG is he decision variable for he UL and he amoun of power ha DG should sell o he aggregaor is he decision variable for he LL. All he end-users in he sysem are served by a load-reailer which purchases he required amoun from he aggregaor. On he oher hand, he load-reailer can play wih is capabiliy in providing he DR opion. The DR opion here is supposed o be he inerrupible load (IL) which is a definie percen of he oal demand. Therefore, he reailer has he opporuniy o reduce is oal demand by IL when he aggregaor increases he demand price. On he oher hand, i should consider paying an incenive o he inerruped loads. C. Approach for solving he problem The problem discussed in his paper is a bilevel problem wih iner-relaed objecives. In his model, he UL problem is he aggregaor s decision making and he LL problem is he decision making of local resources. As also employed in [27], he decision making conflic beween wo levels of players is modeled as a bilevel problem and convered o a mahemaical program wih equilibrium consrains (MPEC). This non-linear bilevel problem is convered o a single level mixed-ineger linear programming (MILP) by implemening he dualiy heorem. The procedure is as follows and is based on [28] and [29]: Formulae he LL problem as a linear and convex problem. Consider he decision vecor of he UL problem as an inpu parameer for he LL problem. Implemen he dualiy heorem and replace he LL problem wih is Karush-Kuhn-Tucker (KKT) opimaliy condiions. Apply srong dualiy o he LL problem and linearize he non-linear erms of he UL objecive funcion. In his paper, he UL and LL are presened wih heir mahemaical models in Secions III and IV, respecively. In order o implemen he dualiy heorem, all he consrains of he LL problem are succeeded by he respecive dual variables separaed by a colon. They are classified ino equaliy and inequaliy consrains wih he respecive dual variables represened by λ and µ, respecively. Finally, he Lagrangian equaion for he LL problem is developed (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

5 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems III. UPPER LEVEL MATHEMATICAL MODEL In he UL, he aggregaor manages is ineracions wih he upsream energy and reserve markes and is resriced by he objecives of is componens as well as he loss ariff of he disribuion nework. I is assumed ha he aggregaor pays he loss ariff o he disribuion sysem operaor in response o he energy purchased from he disribuion nework. Therefore, he objecive of he UL problem will be as (1): TM T Agg profi profi Max profi Max EDG ED (1) profi profi Each of he componens of he objecive funcion is explained below. The aggregaor rades energy and reserve wih he upsream marke based on he marke prices, which are reaed in he problem as known parameers (2). The aggregaor paricipaes in he reserve marke hrough offering he s SOC in he marke. Therefore, i is reimbursed for being ready o deliver reserve (Π ) and if by he probabiliy of reserve call (ρ ) i is summoned o provide he reserve, i will be paid by energy price (Π ). Oherwise, if he aggregaor fails o deliver he amoun of reserve due o FOR, i is subjeced o a penaly based on he hourly energy price. The amoun of FOR is dependen o he nework and he LL resources failure rae. profi p r TM Agg EM Agg RM r Agg del EM Agg del Agg EM r FOR The profi of he aggregaor from is ineracion wih he is caused by he revenue from selling power o for charging is vehicles minus he coss of purchasing energy and reserve from he. The ineracs wih he aggregaor wih he equilibrium prices of energy and reserve (π,, π,, π, ). Noe ha in his sudy various uncerainy scenarios are considered for arrival, deparure and duraion of say in he. As a resul, he amoun of s inpu/oupu power will be differen for each scenario. However, as he s inernal ineracions does no affec he aggregaor s decision making i rades wih he aggregaor wih he expeced values (i.e. p,, p,, r ). T in, in, ou, ou, Re, profi pˆ ˆ ˆ p r (3) del ou, del ou, rˆ rˆ FOR I is assumed ha here can be muliple numbers of DGs in he nework and sell heir power o he aggregaor wih equilibrium price of DG ( π ) as in (4). EDG DG DG profi pm, (4) m The demand is delivered o he end-users wih he hourly equilibrium demand price (π ). I is also assumed ha he aggregaor has o pay for he nework loss (5). ED 2 D D Loss profi p R j,k i j, k, (5) k j The assumpions and consrains of he above objecive are as follows. I is assumed ha he only reserve provider in he sysem is he. Hence, he oal reserve ha he aggregaor (2) can presen in he marke is equal o he expeced amoun of reserve ha he can provide (6). Agg r rˆ (6) The expeced value for he s reserve, inpu and oupu power is he summaion of heir amoun in each scenario muliplied by he probabiliy of each scenario. These are shown in (7)-(9) for he reserve, inpu and oupu power, respecively. r ˆ r (7), in, in, pˆ p, ou, ou, pˆ p, (8) (9) The oal power of he aggregaor is equal o he amoun of demand in each node, he inpu power of he o he node on which i is insalled minus he oupu power of he on ha node and he oupu power of he DG (1). In order o idenify he node on which he or DG is insalled he binary variable (Si) is defined as in (11) o (13). p p p p p (1) Agg D in, ou, DG j, j, j, j, j in, in, p ˆ j, Si j p (11) p Si p (12) ou, ou, ˆ j, j The load flow equaions are presened in (13) (17). I is assumed he power injeced from he upsream nework (p,, ) or delivered o i is affeced by he efficiency of he connecor ransformer. In order o calculae he share of IL on each node, he assumpion of spread share of IL on all loads is used. As a resul, he share of he demand afer IL (p ) from oal demand (p, ) is muliplied by he load of each node (p, ). Besides, he power facor of IL is considered equal o he power facor of he whole sysem; hence, he same approach can be used for he reacive power. The approach o perform he load flow of he sysem is based on [3], [31] and is linearized in he problem as explained in [17]. p p p p R i 2 DSO,ou DSO,in Trans j, Line Line j, j Trans k, j, j,k, j l l j,k j,k, p P D D, Toal P p p p D in, ou, DG j, j, j, j, p q q q q X i Q D DSO,in DSO,ou Line Line D j, j, k, j, j,k, j,k j,k,, j, l l P 5 (13) 2 (14) D Toal Line Line Line Line v 2 R p p X q q Z i v 2 j, j,k j, k, k, j, j,k j, k, k, j, j, k j, k, k, j Line Line j, j, k, j,k, j,k, (15) v i p q (16) V v V, I j, k i I j, k (17) j, j j, k, Considering he objecive and consrains of he UL problem, he decision vecor of he UL for he bilevel model will be as (18). UL Agg Agg in, ou, Re, DG D DV p, r,,,,, (18) (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

6 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems IV. LOWER LEVEL MATHEMATICAL MODEL The objecive on he LL problem consiss of he objecives of he players on he LL based on heir conribuion in he equilibrium price. These objecives are for he rades of wih he aggregaor, he ineracions of he wih PEV owners, he rade of DG wih he aggregaor and he opporuniy of IL on behalf of he reailer. In his regard, he objecive of he LL will be as (19). Max profi Max profi profi LL Agg PEV profi profi DG Agg D Agg (19) A. -Aggregaor ineracion The profi gained by he owner hrough is ineracion wih he aggregaor is shown in (2). In his level, he vehicles ha paricipae in boh G2V and V2G mode are separaed from hose who are only operaed in G2V mode. I is obvious ha he oupu power of he is only due o he opporuniy of V2G. On he oher hand, he reserve presened o he marke is from he opporuniy of V2G, hence is reaed wih he same price of oupu power whenever i is called. If he fails o deliver he required reserve amoun (FOR ), i will be charged wih he oupu energy rae. I can be observed ha in (2), here are common erms wih he UL objecive (3) ha make he equilibrium poin wih UL. Agg ou, V 2 G ou, in, V 2 G in, G 2V in, profi p, p, p, (2) r r r FOR Re, del ou, del ou,,,, The ineracion of he wih he PEV owners ha use he is modeled wih deails in (21). The financial ransacion of wih each group of vehicles should be differen and proporional o he opporuniy hey bring because hey lead o differen levels of profi for he. The naming and clusering of hese caegories are shown in Table I. Moreover, in each hour and in each scenario, he share of each caegory should be deermined. On he oher hand, i should be specified ha he amoun of depared SOC belong o which caegory. The share of each caegory in he deparure SOC is needed for precisely calculaing he hourly revenue and coss of. For his purpose, wo coefficiens are defined o impose he preferences of he PEV owners o he objecive of he. The coefficien β is defined o deermine he share of each caegory from deparing vehicles. Anoher coefficien φ is defined o deermine he preference of each caegory for he minimum required SOC a heir deparure. Besides, he coefficien θ deermines he amoun of PEVs in G2V or V2G mode in each hour. In his sudy, he PEVs ha agree o ake par in he V2G mode are paid an incenive amoun for being ready (as reserve or energy). This amoun is calculaed hrough he muliplicaion of heir available capaciy by he incenive price (Π ). However, when acual energy is purchased from V2G PEVs, hey are paid by V2G price (Π ) as well as he degradaion cos. Moreover, all he PEVs ha ener he have o pay he usage ariff based on he oal hours ha hey have sayed in he muliplied by he ariff (Π ). Also in Table I i is shown ha differen G2V price are considered for differen caegories. The reason is ha he owner encourages he PEVs o paricipae in flexible modes by selling he energy wih lower prices o hem (Π <Π <Π ). As shown in (23) and (24) he s SOC in each hour is separaed for G2V and V2G modes. I is assumed ha he sars wih an iniial amoun of SOC a =1 and he arrival and deparure SOC as well as he power raded wih grid form he hourly SOC of he. The faciliies in he resric he charging/discharging of due o heir efficiencies (η,, η, ). TABLE I PEV OWNERS CLUSTERING Fixed deparure SOC Flexible deparure SOC requiremen requiremen Mode Price of Price Price of Price of Naming Naming G2V of V2G G2V V2G G2V fix1 Π - flex1 Π - V2G fix2 Π Π flex2 Π Π, Π soc soc SOC, G2 V, G2 V, G2V,, 1 1, 1 soc soc p ar, G2 V dep, G2 V in, G2 V cha,,,, soc soc SOC soc, V 2 G, V 2 G, V 2 G ar, V 2G,, 1 1, 1, soc p p dep, V 2 G in, V 2 G cha, ou, V 2 G dcha,,,, : :, G2V, 6 (21) (22), V 2G, The hourly deparure SOC of he is equal o he minimum requiremen of PEVs wih fixed deparure SOC and hose who accep o have flexible deparure SOC. This is applicable o boh G2V and V2G modes as (24) and (25), respecively. dep, G 2V soc C soc : (23) dep, G2V fix1 fix1 dep, dep, flex1,,,,, soc C soc : dep, V 2G fix 2 fix2 dep, dep, flex2,,,,,, (24) dep, V 2G, Alhough some PEVs agree o have a flexible amoun of deparure SOC, he deparure SOC is sill limied o heir minimum preference and he maximum possible SOC due o he limiaion of heir capaciy as in (26) and (27). profi C SOC soc SOC 1 1,, 1, 2 1, 1, 1, 2 2,,,,,, fix2 fix 2 dep, dep, fix 2, Sc G 2V 3 fix 2 dep, fix 2 fix2 dep, Exra Cd C,, C, SOC,,,,, C, dep, flex 2 dep, flex2, Sc G 2V 3 dep, flex2, Sc dep, flex 2 V 2G soc, SOC, Cd dep, flex 2 dep, flex 2, Sc SOC, soc, dep, flex 2 soc, SOC, soc, SOC flex 2 dep, flex2 flex 2 dep, Exra ou, V 2 G in, V 2G del Tariff del V 2G, C,,, C, p, p, r, Cd N, r,, 2, dep flex Sc, PEV fix fix dep dep fix Sc G V dep flex dep flex Sc G V (25) (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

7 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems 7 flex1 flex1 dep, dep, flex1 EV flex1 dep,,, C, soc, soc, C, :, (26) dep, flex1 dep, flex1,, flex2 flex2 dep, dep, flex2 EV flex2 dep,,, C, soc, soc, C, :, (27) dep, flex 2 dep, flex2,, The SOC of in G2V mode should no pass he maximum available capaciy of G2V vehicles in he muliplied by he maximum possible SOC of each EV (28). For he V2G vehicles, as he has he conrol o discharge he PEVs baeries a minimum limi also should be bounded by he SOC of in each hour (29). Due o variable levels of s capaciy resuling from PEVs arrival/deparure, he hourly SOC of he is considered in kwh insead of he raio of he oal capaciy., G2V,,, EV soc C soc : EV 1, 2 1 V G,,,,, C soc soc C soc : EV (28), G 2V,, (29) P L, V 2 G P L, V 2 G,, The faciliies in he s saions have a charging/discharging rae (γ ) ha limis he maximum amoun of inpu/oupu power of he (3)-(32). p N : in, G2V,,, p 1 N : in, V 2G,,, p r 1 N : ou, 2,,,,, (3) in, G2 V in, G2V,,, (31) in, V 2 G in, V 2G,,, (32) in, V 2G1 in, V 2G1,, The maximum amoun ha can offer in he marke (including energy and reserve) should no pass he limi of available SOC from V2G vehicles and he minimum SOC ha can remain in he PEVs baeries (33)., 2,, 2 EV p r soc soc 1 C : ou V G dcha V G,,,,,, (33) in, V 2G 2 in, V 2G 2,, The reserve and energy oupu of he are defined as posiive variables (34), (35). r, : ou, V 2G p, : (34) Re,, (35) ou,, For he purpose of linearizaion in (22), a variable (Z, ) is defined o compare he depared SOC in each hour wih he scenario paern (36)-(39). dep, flex 2 dep, flex 2, Sc G2V 3 soc, SOC, Cd dep, flex2 dep, flex 2, Sc soc, SOC, dep, flex2, Sc dep, flex 2 V 2G SOC, soc, dep, flex2 dep, flex 2, Sc soc, SOC, dep, flex 2 dep, flex 2, Sc G2V 3 V 2G soc, SOC, Cd 2 V 2G G2V 3 V 2G Z, Cd 2, dep, flex 2 dep, flex 2, Sc soc, SOC, Z :,, dep, flex 2, Sc dep, flex 2 SOC, soc, Z :,, (36) Min Z (37) Aux, 1 (38) Aux, 2 (39) B. DG-Aggregaor ineracion In (4) he profi gained by he DG owner from selling energy o he aggregaor is shown m, m m, (4) DG Agg DG DG DG DG profi p A p m where A is he marginal cos for m h DG. All he DGs should be limied o heir maximum generaing power (41). DG DG p P m : m,, m, (41) DG m, DG C. Demand-Aggregaor ineracion The loads in he sysem are supplied by a load reailer who purchases he required amoun of energy from he aggregaor wih equilibrium price (π ) and sell i o he load wih he ime of use ariff (Π ). The users who paricipae as IL are also paid an incenive (Π ). D Agg D ToU D IL Incenive profi p p (42) I is assumed ha he amoun of demand ha is purchased from he aggregaor (p ) is afer he implemenaion of IL (43). Moreover, he demand afer he IL implemenaion should be limied o he maximum oal demand ( p, ) and he minimum of no inerrupible load (44). IL D, oal D IL p p p : (43) D D oal D D oal D, oal D, oal 1 P p P :, (44),, Considering all he equaions presened for he LL problem he decision vecor for lower level will be as (45). LL in, ou, dep, DG D DV p,, p,, r,, soc,, pm,, p (45) As previously menioned, he problem is formulaed o conver he bilevel problem ino an MPEC. For his purpose, firsly he Lagrangian of he LL problem is developed. The variables in his equaion are he decision variable vecors in LL problem. The componens of he Lagrangian are he LL objecive (19), equaliy consrains (22-25, 43) and inequaliy consrains (26-39, 41, 44). For linearizaion of he non-linear erms in he UL, he srong dualiy heorem is employed which saes when a problem is convex, he primal and dual objecive funcions are equal a he opimum. V. PEV SCENARIO GENERATION Considering he real daa from he surveys and he say duraion classificaion, he scenarios for he arrival of PEVs in he is generaed using he approach in [17] where a lognormal disribuion funcion is considered. Then he deparure scenarios are derived from he arrival scenario and say duraion. However, due o he fixed number of saions in he, he scenarios generaed for arrival/deparure may resul in PEVs number in he more han he s capaciy. To preven his, a procedure is implemened on he scenario generaion as shown in Fig. 3. The scenario of PEV numbers in he is generaed from he summaion of he remainder PEVs in he from he previous hour and he arrived PEVs in each hour minus he depared PEV. Whenever he PEV numbers exceed he s saions, he number of excess PEVs is reduced from he arrival scenario on ha hour (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

8 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems 8 n n n n ar dep,,,, 1 n, saions n n n surp saions,, n n n ar, new ar surp,, max n n n n ar, new dep, new,,, 1 Fig. 3. Flowchar of generaing scenario for PEVs number in. Now, he arrival scenarios need o be changed which consequenly cause he change in he say duraion. Considering he discree disribuion of say duraion paern, he new arrival scenario and say duraion scenario is formed. Based on he new arrival and say duraion scenario, he new deparure scenario is generaed. Once again he number of PEVs in is calculaed. The procedure is performed unil he s number scenario does no exceed he oal s saion number (Fig. 3). VI. NUMERICAL RESULTS The proposed model is implemened on a sandard disribuion nework wih, DG, and DR program. The IEEE 37-bus nework [32] as shown in Fig. 4 is seleced for he sudy. The locaion of he resources and heir capaciies are based on previous sudies of DG inegraion in IEEE 37-bus nework as in [33] and [34]. In his sudy, a wih 25 saions in a commercial area is considered. Figures 5 and 6 depic he arrival and deparure scenarios employed in his sudy. The oal SOC of PEVs in he is shown in Fig. 7. The mean values for he scenarios are derived from repors and surveys on European driving paern presened in [35] and [36] and he household ravel survey in [37]. The daa presened in [38] are employed o acquire he expeced say duraion of PEVs as shown in Fig. 8. As a he is assumed o be in a commercial cener, he PEVs ha ener he may say from 1 o 12 hours in he. The values for coefficien φ which deermines he minimum deparure SOC requiremen of each PEV caegory is shown in Table II. The values o deermine he share of each caegory from he oal depared PEVs are presened by coefficien β in Table III. Fig. 4. IEEE 37-bus nework under sudy wih added resources. Number of PEV Arrival Fig. 5. Expeced value of PEV arrival o and is scenarios. Number of PEV Deparure Expeced PEV arrival Arrival number scenario Fig. 6. Expeced value of PEV deparure from he and is scenarios. Toal SOC of PEVs in Expeced PEV deparure Deparure number scenario Fig. 7. Expeced value of PEV SOC in he and is scenarios. Expeced PEVs SOC in Scenario of PEVs SOC in (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

9 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems Fig. 8. Toal number of PEVs in he in each hour based on heir expeced say duraion. Mode G2V G2V+V2G Mode G2V G2V+V2G Number of Vehicles 1 hour 2 hours 3 hours 4 hours 5 hours 6 hours 7 hours 8 hours 9 hours 1 hours 11 hours 12 hours TABLE II VALUES OF φ FOR DIFFERENT PEV CATEGORIES Deparure SOC Duraion of say (hours) Requiremen Fix Flex Fix Flex TABLE III VALUES OF β FOR DIFFERENT PEV CATEGORIES Deparure SOC Duraion of say (hours) Requiremen Fix Flex Fix Flex The prices for energy and reserve marke are from he Spanish elecriciy marke [39] and are adaped o he disribuion level based on [4]. In [4] i is menioned ha a surplus should be added o he upsream energy prices when i is implemened o lower volage levels. This surplus is divided beween he aggregaor and he LL componens of he problem. In his sudy he surplus is considered as 5 cens. Therefore, 3 cens are added o he upsream marke energy price and hen implemened o he energy rades of he aggregaor wih he upsream nework. The remaining 2 cens is added o he LL resources ransacion price. Noe ha he rades beween he aggregaor and LL resources ake place based on he equilibrium prices In his sudy, all he saions in he are he same and are quick charging saions wih a charging rae of 11 kw per hour as in [17]. Oher specificaions of he PEVs and ariffs are based on [41]. In his paper, wo case sudies are invesigaed o evaluae he proposed model. On he firs approach, he Pay as Bid pricing model is invesigaed o examine he individual ineracion of LL resource wih he aggregaor. In he second approach, he cross effec of he resources in heir marke paricipaion is invesigaed hrough uniform pricing. The problem is modeled as an MILP problem and implemened in GAMS using CEX12 solver. A. Case I: Pay as Bid In his case, as each resource receives paymen based on is bidding, he aggregaor ineracs wih each componen individually. Therefore, he aggregaion approach causes a leader/follower framework. In order o bind he profi of he leader, a price cap is pu on he maximum rade price beween he aggregaor and each of he LL componens. The cap is 1 cens per kwh. This case is sudied as a base case o show he differen behavior of he aggregaor and he resources when one equilibrium poin is found comparing o individual ineracion. In Fig. 9 he prices of he upsream EM (i.e., he amoun paid by he aggregaor o he upsream marke), he price (i.e., he price paid by he o he aggregaor for energy purchase), and he DG price are shown. The price reaches he price cap for he whole 24 hours. The variaions of prices in his case comply wih he energy ineracion balance of he sysem in Fig. 1. I can be seen ha he behavior of he aggregaor is relaively jusifiable o he EM price variaions. For example, during hours 2-7 A.M. when he upsream energy price has he lowes amoun, none of he resources in he LL is acivaed. For he remaining hours, only DG1 is commied o supply energy. Therefore, he s power exchange is only for inpu power as in Fig. 11. Moreover, wih his price cap, i is no profiable for he load reailer o acivae he IL and hus i provides he oal demand from he EM. The reserve marke price and he aggregaor- reserve price are shown in Fig. 12. I shows ha during he high commuing hours he reserve price reaches is peak amoun. The can make a profi hrough is paricipaion in he upsream reserve marke. Being he only reserve source of he aggregaor, all he possible SOC of is presened in he reserve marke. Accordingly, he price of he reserve paid o he can be a moivaing facor o change he s behavior. I indicaes ha considerable higher paymen o he in order o mainain is SOC for paricipaing in reserve marke is profiable in his case. The SOC of he for various PEV caegories in he is shown in Fig. 13. Energy Price (Cen/kWh) EM Price DG Price Price Fig. 9. Energy prices for aggregaor,, and DG in Case I (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

10 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems 1 DSO Energy Balance (kwh) Fig. 1. Energy balance of sysem in Case I. Power Exchange (kwh) Pin_G2V P Fig. 11. s power exchange in Case I. Resrve Price (Cen/kWh) P1 Agg P2 V2G DG3 DG2 DG1 Toal oaldemand Demand new afer IL Pou_V2G P RM Price REq Price Fig. 12. Upsream reserve marke and LL reserve equilibrium prices in Case I. Deparure SOC (kwh) SOC_dep_flex2 SOC SOC_dep_flex1 SOC SOC_dep min SOC, SOC_dep_fix2 SOC SOC_dep_fix1 SOC SOC max SOC dep max, Fig. 13. s sae of charge for various caegories of PEVs in in Case I. The mos challenging resource in his model is. In his case as shown in Fig. 9, he price for s energy rade is a consan value for he whole 24 hours. This price is he equilibrium price derived from he behavior of he and he aggregaor considering oher resources available o he aggregaor. In oher words, if he changes is behavior, he price will also change. However, boh he s behavior and he price are propelled o he equilibrium price as in his price he opimum profi is obained. During he early hours of he day (hours 1-9) he sars o charge he PEVs in he because he energy price is low. The can make profi from selling energy o he PEVs, however he preferences of PEVs on requiring a fixed amoun of deparure SOC limis he charging behavior of he. Meanwhile, he aggregaor wans o increase is profi from selling energy o he ; as a resul, i will encourage he o charge is PEVs by increasing he price of reserve a hours 1 and 14 (see Fig. 12). The price of reserve is increased by he aggregaor so ha he will be moivaed for charging; however, he preferences of he PEVs limi he maximum charging of. In fac, noing Fig. 13, i is shown ha he PEVs are charged almos he same as heir minimum requiremen of deparure SOC. The reason is ha from hour 16, he PEVs deparure from he increases. As a resul, in order o mee he PEV s preferences he charging of is limied. For he reserve provision, excep where he reserve price faces a spike a hour 15, in oher hours he price is almos equal o he marginal price of for providing reserve. B. Case II: Uniform Pricing In his case, all he resources on he LL rade wih he aggregaor wih a uniform price which is he equilibrium price. As a resul, he LL resources can have more flexibiliy on heir ransacions wih he aggregaor comparing o Case I. As can be seen in Fig. 14, he LL energy equilibrium (EEq) price has significan differences from he EM price and he pay as bid case. Moreover, Fig. 15 shows he conribuion of all resources in he sysem. In conrary o Case I, in his case all he resources (i.e., DG, s V2G mode, and IL) ake par in he schedule. The reason is ha one equilibrium price concerning all he consrains and objecives of various componens is calculaed and hence more flexibiliy for he aggregaor o compromise beween he various objecives is provided. From anoher poin of view, he reserve price in his case in Fig. 16 is higher han he firs case and in some hours he aggregaor is persuaded o increase he reserve price up o he upsream reserve marke price. As a resul, he LL resources will be encouraged o paricipae more effecively in he marke. In Fig. 16, i is shown ha in he reserve price experience a spike from hour 19 o 23. A hours 19 o 23 he aggregaor increases he reserve price o encourage he o charge is PEVs. In fac, he equilibrium price is a compromise beween he lowes amoun of EM price and RM prices (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

11 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems 11 Energy Price (Cen/kWh) EM Price EEQ Price Fig. 14. Energy Marke and Energy Equilibrium prices in Case II. DSO Energy Balance (kwh) P Agg 2P V2G DG3 DG2 DG1 Toal oal new Demand Demand afer IL Fig. 15. Energy balance of sysem in Case II. Reserve Price (Cen/kWh) RM Price REQ Price Fig. 16. Reserve Marke and Reserve Equilibrium prices in Case II. 1) DGs behavior During hours 2 o 6, he EM price is in is lowes amoun; however, during hose hours DG1 is commied for he energy generaion bu wo of he DGs canno compee and are no operaing. A hour 7, DG2 is commied and afer ha all DGs are paricipaing in he energy producion of he sysem. This happens because of he EEq price increase on ha hour. The reason is ha from hour 7 he arrival o he is increasing; as a resul, he will be able o charge he baeries of arriving PEVs, increase is poenial of reserve provision and consequenly increase is own profi. 2) Load Reailer s behavior The aggregaor s decision making on operaing is resources impose significan changes on he EEq price during 24 hours. However, hese changes are no emping enough for he load reailer o acivae is IL unil hours when he end users demand is on is peak amoun. As a resul, he load reailer will use he IL o reduce is coss. 3) s behavior During early hours of he day, he is encouraged o charge he PEVs due o low energy prices. Afer hour 6 up o 9, alhough he EEq price is increased, i is sill mainained in low amoun; herefore, he keeps charging he baeries. In oher words, in hese hours he aggregaor holds he EEq price relaively low so ha he coninues on is behavior of charging. In hour 1, he price of he EM increases and consequenly he aggregaor increases he price o make benefi from selling energy o load reailer and. Referring o Fig. 8, i is observed ha from hour 1 mos of he PEVs ha ener he are hose who need o say in he for a shor say. As a resul, he EEq price is reduced and he energy rade is reduced (Fig. 14 and 16). Consequenly, from hours 1 o 15, he changes is sraegy. Alhough he EEq price is reduced a hour 11 comparing o hour 1, he is no moivaed o increase is charging. During his period, he will charge mosly he PEVs ha only ake par as G2V mode. The reason is ha he price reducion is up o he G2V2 price considering he efficiency of he saion charger. In oher words, he Fix2 conracs are he mos preferred conracs for boh and he aggregaor, because he aggregaor benefis from selling energy o and he benefis from selling o he PEVs. Afer ha in hours 16 and 17, he EEq price is decreased. The reason is no only due o he price reducion in he EM, bu also due o he fac ha from hour 16, he number of PEV deparure increases. Hence, he needs o charge he baeries, especially he flexible ones, o increase is own profi. As a resul, he aggregaor decreases he EEq price so ha he is encouraged o charge he PEVs which are abou o depar he. This increase in he SOC can be seen in Fig. 17. Unlike Case I, in his case he V2G power is injeced ino he grid (see Fig. 18). The reason is ha in hours 16 and 17 he charges he PEVs bu from hour 19, i has o discharge he baeries because i ges near o he ending hours and he PEVs leaves he. As a resul, in order o mee he requiremens imposed o he by PEVs caegories, i will injec he excess power o he grid. Consequenly, he price of energy spikes in hour 19 and remains high afer ha, boh due o his reason and he fac ha he demand peak is also during hose hours. Alhough he sraegy of from hour 19 o 24 is o discharge mos of he energy sored in PEV baeries, he aggregaor will equilibrae he siuaion by increasing he reserve up o he upsream reserve marke (see Fig. 16). The oal deparure SOC of for Flex2 conracs in his case is shown in Fig. 17. As shown, he difference of SOC in he wih he minimum requiremen of PEVs deparure SOC is higher in hours which is due o higher reserve price encouraged by he aggregaor. For oher hours he ends o keep he PEVs on heir minimum requiremen. Figure 19 shows he oal s SOC and capaciy and he reserve provision of he (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

12 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems Deparure SOC (kwh) SOC SOC_dep_flex2 flex min SOC, SOC max SOC dep max, Fig. 17. The behavior of in charging Flex2 conracs in Case II. Power Exchange (kwh) Pin_G2V P Fig. 18. s power exchange in Case II. Pou_V2G P In his regard, in Figs. 2 and 21 he profis of he aggregaor and for he variaion of G2V2 and G2V3 prices in case II are shown, respecively. For he aggregaor in Fig. 2, he oal profi is reduced consanly wih he reducion of he G2V prices. The reason is ha as he prices decrease, he endency of he o charge is PEVs will decrease and consequenly he aggregaor s profi will decrease. However, i can be seen ha when he G2V2 price goes less han 11 cens, he aggregaor s profi decreases drasically. I is due o he fac ha G2V2 price is for hose PEVs ha only paricipae in G2V mode, bu hey agree o have flexible deparure SOC (i.e., Flex1 conrac). In his case, he s choice of profi is only hrough charging hese PEVs and no encouragemen for charging he PEVs o ake benefi from hem in he reserve marke does no exis. As a resul, he reduces is flexibiliy and considerably affecs he aggregaor s profi. In Fig. 21, i is observed ha he oal profi of he can have significan changes wih he changes in G2V2 and G2V3 prices. These prices are he incenives ha he deermines for is rade wih hose PEVs ha agree o have flexible deparure SOC requiremens. I should be noed ha hese ariffs can considerably affec he role of he as a flexible load or as a resource. In oher words, hese wo prices can change he marginal price of he and change is behavior in conac wih he aggregaor. 12 SOC and Capaciy G2V SOC Pou_V2G P Capaciy V2G SOC Reserve Provision SOC Fig. 19. Comparison of s capaciy and SOC divided by G2V and V2G PEVs in Case II. Fig. 2. Aggregaor profi in Case II for various G2V2 and G2V3 prices. VII. THE ROLE OF PEV PREFERENCES ON AGGREGATOR EQUILIBRIUM In his paper, he as he main concern of he sudy changes is behavior based on is rade wih he PEV owners and he aggregaor. As a resul, he ariffs ha are implemened o he PEVs can significanly change he sraegy of he in he marke. The variaion of he behavior also leads o differen levels of profi gain for he and aggregaor. In his sudy, is a complicaed resource in he sysem which can ac as a flexible demand and as a resource as well. Therefore, he aggregaor can benefi he mos from he s poenial o ac as he flexible load. However, he aggregaor needs o manage he marke wisely o encourage he o show more flexibiliy. Fig. 21. profi in Case II for various G2V2 and G2V3 prices (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

13 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems VIII. CONCLUSIONS In his paper a comprehensive bilevel model o derive he equilibrium price of energy and reserve rade of has been proposed considering he preferences of he PEV owners. I is obvious ha a criical influence is pu on he manipulaion of he elecric vehicles in fuure sysems by heir owners. The behavior of he PEV users can significanly change he process of he sysem operaor. On he oher hand, in such environmen wih various componens and complicaed ineracions, an organized iner-relaion should be defined so ha all he involved paries in his sysem could assure heir own profi. In his regard, his paper inended o propose a model for such siuaion. The main characerisic of his model is ha in he upper level he price was specified while he lower level deermined he quaniy. This was accuraely compaible wih he realiy of operaion. In fac, he main role of as a flexible load is o add he poenial of possible load increase or decrease. Considering he resuls obained from his sudy, several influenial conclusions can be deduced which are lised below: I is shown ha in he uniform pricing model where he LL componens can have more flexibiliy he model is more effecive and he equilibrium poin is found in a more suiable way for all paries profi. On he oher hand, i is shown ha in an environmen wih mixed resources, he model can provide he soluion o compromise beween all he poenials in he sysem. I is shown ha he equilibrium price is affeced by various facors ha may change he behavior of he players in he model. When he behavior is changed, he equilibrium price is going o be changed; however, he bilevel model is designed in a way ha he opimum soluion is found in his compromising siuaion. Alhough he can be considered as a resource in he sysem, he compromise beween he compeiiveness of oher resources in he sysem such as DGs and he expenses of V2G vehicles will lead o less endency owards V2G mode operaion. However, i was deduced from he sudy ha he can provide various opporuniies for he aggregaor in erms of flexibiliy and increase he oal profi. The aggregaor can decrease he equilibrium price o increase is own share of he local marke, riggering he load flexibiliy poenial of he (increasing he quaniy in lower level) which causes higher profi for boh aggregaor and. I is deduced ha he reserve price also had a criical role wih which he aggregaor conrols he inpu energy o he and encourages he for purchasing more energy. From anoher poin of view, oher local resources have proved ha hey influence he problem. Wih higher levels of local resources peneraion in he sysem, he equilibrium price can go as low as he marginal price of hese resources, which affecs he charging saus of he as well. In final words, his sudy proposed a model for he combinaion of fuure sysem componens wih high peneraion local resources considering he wo-fold role of he. The inheren naure of he bi-level ineracion of he wih he grid is coordinaed wih he ineviable effec of he PEV owners preferences. The oucomes of his model can be useful for fuure ariff deerminaion or incenive calculaions for furher deploymen of s in he sysem. The relaion of muliple s, as he commue beween wo s in one zone can affec he oal SOC and capaciy of each and make i dependen o he oher s behavior in charging is vehicles, will be a subjec of fuure work. REFERENCES [1] W. Kempon, J. Tomic, S. Leendre, A. Brooks, and T. Lipman, Vehicle o grid power: Baery, hybrid, and fuel cell vehicles as resources for disribued elecric power in California, Univ. California Davis Ins. for Transporaion Sudies, Rep. ECD-ITS-RR-1-3, 21. [2] J. A. P. Lopes, F. J. Soares, and P. M. R. Almeida, Inegraion of elecric vehicles in he elecric power sysem, Proc. IEEE, vol. 99, no. 1, pp , Jan [3] C. Guille and G. Gross, A concepual framework for he vehicle-o-grid (V2G) implemenaion, Energy Policy, vol. 37, pp , 29. [4] R. J. Bessa and M. A. Maos, Economic and echnical managemen of an aggregaion agen for elecric vehicles: a lieraure survey, Euro Trans. Elec. Power, vol. 22, pp , 212. [5] S. Sun, M. Dong, and B. Liang, Real-Time Welfare-Maximizing Regulaion Allocaion in Aggregaor-EVs Sysems, in Proc. IEEE INFOCOM WKSHPS 213, pp [6] C. Jin, J. Tang, and P. Ghosh, Opimizing Elecric Vehicle Charging: A Cusomer s Perspecive, IEEE Trans. Vehicular Technology, vol. 62, No. 7, pp , Sep [7] J. Lin, K.C. Leung, and V. O.K. Li, Opimal Scheduling Wih Vehicleo-Grid Regulaion Service, IEEE Inerne of Things Journal, vol. 1, No. 6, pp , Dec [8] R. J. Bessa, M. A. Maos, F. J. Soares, and J.A.P. Lopes, Opimized Bidding of a EV Aggregaion Agen in he Elecriciy Marke, IEEE Trans. Smar Grid, vol. 3, No. 1, pp , Mar [9] R.J. Bessa and M.A. Maos, Opimizaion Models for EV Aggregaor Paricipaion in a Manual Reserve Marke, IEEE Trans. Power Sys., vol. 28, No. 3, pp , Aug [1] M.A. Orega-Vazquez, F. Bouffard, and V. Silva, Elecric Vehicle Aggregaor/Sysem Operaor Coordinaion for Charging Scheduling and Services Procuremen, IEEE Trans. Power Sys., vol. 28, No. 2, pp , May 213. [11] C. Jin, J. Tang, and P. Chosh, Opimizing Elecric Vehicle Charging Wih Energy Sorage in he Elecriciy Marke, IEEE Trans. Smar Grid, vol. 4, No. 1, pp , Mar [12] S. I. Vagropoulos and A. G. Bakirzis, Opimal Bidding Sraegy for Elecric Vehicle Aggregaors in Elecriciy Markes, IEEE Trans. Power Sys., vol. 28, No. 4, pp , Nov [13] R. Li, Q. Wu, and S.S. Oren. Disribuion Locaional Marginal Pricing for Opimal Elecric Vehicle Charging Managemen, IEEE Trans. Power Sys., vol. 29, No. 1, pp , Jan [14] E. Soromme and M. A. El-Sharkawi, Opimal Scheduling of Vehicleo-Grid Energy and Ancillary Services, IEEE Trans. Smar Grid, vol. 3, No. 1, pp , Mar [15] E. Soromme, M. A. El-Sharkawi, Opimal Combined Bidding of Vehicle-o-Grid Ancillary Services, IEEE Trans. Smar Grid, vol. 3, No. 1, pp. 7-79, Mar [16] M. Moradijoz, M. P. Moghaddam, M.R. Haghifam, and E. Alishahi, A muli-objecive opimizaion problem for allocaing parking los in a disribuion nework, Elec. Power Energy Sys., vol. 46, pp , 213. [17] N. Neyesani, M. Yazdani Damavandi, M.R. Shafie-khah, J. Conreras, and J.P.S. Caalao, Allocaion of Plug-In Vehicles Parking Los in Disribuion Sysems Considering Nework-Consrained Objecives, IEEE Trans. Power Sys., vol. 3, no. 5, pp , Sep [18] W. Kempon and S. E. Leendre, Elecric vehicles as a new power source for elecric uiliies, Transporaion Research Par D. Transpor and Environmen, vol. 2, pp , Sep [19] U.C. Chukwu and S.M. Mahajan, V2G Parking Lo Wih PV Roofop for Capaciy Enhancemen of a Disribuion Sysem, IEEE Trans. Susain. Energy, vol. 5, no. 1, Jan [2] M. Honarmand, A. Zakariazadeh, and S. Jadid, Opimal scheduling of elecric vehicles in an inelligen parking lo considering vehicle-o-grid concep and baery condiion, Energy, vol. 65, pp , 214. [21] U.C. Chukwu and S.M. Mahajan, Real-Time Managemen of Power Sysems wih V2G Faciliy for Smar-Grid Applicaions, IEEE Trans. Susain. Energy, vol. 5, no. 2, Apr [22] M. F. Shaaban, M. Ismail, E. F. El-Saadany, and W. Zhuang, Real- Time PEV Charging/Discharging Coordinaion in Smar Disribuion Sysems, IEEE Trans. Smar Grid, vol. 5, pp , Jul (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.

14 This aricle has been acceped for publicaion in a fuure issue of his journal, bu has no been fully edied. Conen may change prior o final publicaion. Ciaion informaion: DOI 1.119/TPWRS , IEEE Transacions on Power Sysems 14 [23] S. Rezaee, E. Farjah, and B. Khorramdel, Probabilisic analysis of plugin elecric vehicles impac on elecrical grid hrough homes and parking los, IEEE Trans. Susain. Energy, vol. 4, pp , Oc [24] S. Shao, M. Pipaanasomporn, and S. Rahman, Grid Inegraion of Elecric Vehicles and Demand Response Wih Cusomer Choice, IEE Trans. Smar Grid, vol. 3, no. 1, pp , 212. [25] R.J. Bessa and M.A. Maos, Global agains divided opimizaion for he paricipaion of an EV aggregaor in he day-ahead elecriciy marke. Par I: Theory, Elec. Pow. Sys. Res., vol. 95, pp , 213. [26] X. Xi and R. Sioshansi, Using Price-Based Signals o Conrol Plug-in Elecric Vehicle Flee Charging, IEE Trans. Smar Grid, vol. 5, no. 3, pp , 214. [27] M. Gonzalez Vaya and G. Anderson, Opimal Bidding Sraegy of a Plug-In Elecric Vehicle Aggregaor in Day-Ahead Elecriciy Markes Under Uncerainy, IEEE Trans. Power Sys., DOI: 1.119/TPWRS [28] S. J. Kazempour, A. J. Conejo, and C. Ruiz, Sraegic Generaion Invesmen Using a Complemenariy Approach, IEEE Trans. Power Sys., vol. 26, pp , May 211. [29] S. J. Kazempour, A. J. Conejo, and C. Ruiz, Generaion Invesmen Equilibria wih Sraegic Producers Par I: Formulaion, IEEE Trans. Power Sys., vol. 28, no. 3, pp , Aug [3] A. C. Rueda-Medina, J. F. Franco, M. J. Rider, A. Padilha-Felrin, and R. Romero, A mixed-ineger linear programming approach for opimal ype, size and allocaion of disribued generaion in radial disribuion sysems, Elec. Power Sys. Res., vol. 97, pp , Apr [31] J. F. Franco, M. J. Rider, M. Lavorao, and R. Romero, A mixed ineger LP model for he reconfiguraion of radial elecric disribuion sysems considering disribued generaion, Elec. Power Sys. Res., vol. 97, pp. 51 6, Apr [32] W. H. Kersing, Radial disribuion es feeders, in Proc. IEEE Power Eng. Soc. Winer Meeing, 21, vol. 2, pp [33] Z. Liu, F. Wen, and G. Ledwich, Opimal Siing and Sizing of Disribued Generaors in Disribuion Sysems Considering Uncerainies, IEEE Trans. Power Del., vol. 26, no. 4, pp , Oc [34] Y. G. Hegazy, M.M. Ohman, W. El-Khaam, and A. Y. Abdelaziz, Opimal sizing and siing of disribued generaors using Big Bang Big Crunch mehod, in Proc. Inernaional Universiies Power Engineering Conference (UPEC), pp. 1-6, Sep [35] G. Pasaoglu, D. Fiorello, L. Zani, A. Marino, A. Zubaryeva, and C. Thiel, Projecions for Elecric Vehicle Load Profiles in Europe Based on Travel Survey Daa, European Comm. Rep. EUR EN, 213, [Online]. Available: hps://seis.ec.europa.eu/publicaions/jrc-seisrepors. [36] G. Pasaoglu, D. Fiorello, A. Marino, G. Scarcella, A. Alemanno, A. Zubaryeva, and C. Thiel, Driving and parking paerns of European car drivers - a mobiliy survey, European Comm. Rep. EUR EN, 212, [Online]. Available: hps://seis.ec.europa.eu/publicaions/jrcseis-repors. [37] R. van Haaren, Assessmen of elecric cars range requiremens and usage paerns based on driving behavior recorded in he Naional Household Travel Survey of 29. Sudy of he Solar Journey USA. Earh and Environmenal Eng. Dep., Columbia Univ., Fu Foundaion School of Eng. and Applied Science, New York, December 211. [38] J. Smar, W. Powell, S. Schey, Exended Range Elecric Vehicle Driving and Charging Behavior Observed Early in he EV Projec, SAE Technical Rep , doi: / [39] Red Elécrica de España, Maximum average hourly power demand and daily energy, [Online]. Available: hp:// [4] ERSE, Porugal Annual Repor o he European Commission, Naional reporing for Council of European Energy Regulaors (CEER), [Online]. Available: [41] W. Kempon and J. Tomic, Vehicle-o-grid power fundamenals: Calculaing capaciy and ne revenue, Power Sources, vol. 144, pp , Jun. 25. BIOGRAPHIES Nilufar Neyesani (S 9) received he M.Sc. degrees in elecrical engineering from Iran Universiy of Science and Technology, Tehran, Iran, in 21. She is currenly pursuing her Ph.D. sudies in he laboraory of Susainable Energy Sysems wih Professor João P. S. Caalão, Universiy of Beira Inerior (UBI), Covilha, Porugal. Her research ineress include elecric vehicles, smar grids, power sysem planning, power sysem opimizaion, muli-energy sysems and energy hub. Maziar Yazdani Damavandi (S 8) received he M.Sc. degree in elecrical engineering from he Tarbia Modares Universiy, Tehran, Iran, in 21. He is currenly pursuing his Ph.D. sudies in he Energy Managemen laboraory in he Tarbia Modares Universiy (TMU) and working as a researcher in SiNGULAR projec in Universiy of Beira Inerior (UBI), Covilha, Porugal. His research ineress include opimizaion models in power sysem sudies and muli-energy sysem modeling. Miadreza Shafie-khah (S 8, M 13) received he M.Sc. and Ph.D. degrees in elecrical engineering from Tarbia Modares Universiy, Tehran, Iran, in 28 and 212, respecively. He is a posdocoral fellow in he laboraory of Susainable Energy Sysems wih Professor João P. S. Caalão, Universiy of Beira Inerior (UBI), Covilha, Porugal. He is working on he EU-funded FP7 projec SiNGULAR. His research ineress include power marke simulaion, marke power monioring, power sysem opimizaion, operaion of elecriciy markes, price forecasing and smar grids. Anasasios G. Bakirzis (S 77-M 79-SM 95-F 15) received he Dipl. Eng. Degree from he Deparmen of Elecrical Engineering, Naional Technical Universiy, Ahens, Greece, in 1979 and he M.S.E.E. and Ph.D. degrees from Georgia Insiue of Technology, Alana, in 1981 and 1984, respecively. Since 1986 he has been wih he Elecrical Engineering Deparmen, Arisole Universiy of Thessaloniki, Greece, where he is currenly Professor. His research ineress are in power sysem operaion, planning and economics. João P. S. Caalão (M 4-SM 12) received he M.Sc. degree from he Insiuo Superior Técnico (IST), Lisbon, Porugal, in 23, and he Ph.D. degree and Habiliaion for Full Professor ("Agregação") from he Universiy of Beira Inerior (UBI), Covilha, Porugal, in 27 and 213, respecively. Currenly, he is a Professor a he Faculy of Engineering of he Universiy of Poro (FEUP), Poro, Porugal, and Researcher a INESC TEC, INESC- ID/IST-UL, and C-MAST/UBI. He was he Primary Coordinaor of he EUfunded FP7 projec SiNGULAR ("Smar and Susainable Insular Elecriciy Grids Under Large-Scale Renewable Inegraion"), a 5.2-million-euro projec involving 11 indusry parners. He has auhored or coauhored more han 48 publicaions, including 16 journal papers, 281 conference proceedings papers, 23 book chapers, and 14 echnical repors, wih an h-index of 28 and over 345 ciaions (according o Google Scholar), having supervised more han 45 pos-docs, Ph.D. and M.Sc. sudens. He is he Edior of he books eniled Elecric Power Sysems: Advanced Forecasing Techniques and Opimal Generaion Scheduling and Smar and Susainable Power Sysems: Operaions, Planning and Economics of Insular Elecriciy Grids (Boca Raon, FL, USA: CRC Press, 212 and 215, respecively). His research ineress include power sysem operaions and planning, hydro and hermal scheduling, wind and price forecasing, disribued renewable generaion, demand response and smar grids. Prof. Caalão is an Edior of he IEEE TRANSACTIONS ON SMART GRID, an Edior of he IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, and an Associae Edior of he IET Renewable Power Generaion. He was he Gues Edior-in-Chief for he Special Secion on "Real-Time Demand Response" of he IEEE TRANSACTIONS ON SMART GRID, published in December 212, and he Gues Edior-in-Chief for he Special Secion on "Reserve and Flexibiliy for Handling Variabiliy and Uncerainy of Renewable Generaion" of he IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, published in April 216. He was he recipien of he 211 Scienific Meri Award UBI-FE/Sanander Universiies and he 212 Scienific Award UTL/Sanander Toa. Also, he has won 4 Bes Paper Awards a IEEE Conferences (c) 216 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp:// for more informaion.