Risk management and participation planning of electric vehicles in smart grids for demand response

Energy 116 (2016) 836e850

Contents lists available at ScienceDirect

Energy journal homepage: www.elsevier.com/locate/energy

Risk management and participation planning of electric vehicles in smart grids for demand response Nasim Nezamoddini, Yong Wang* Department of Systems Science and Industrial Engineering, Binghamton University, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 May 2016 Received in revised form 12 September 2016 Accepted 1 October 2016

Demand response (DR) can serve as an effective tool to better balance the electricity demand and supply in the smart grid. It is defined as "the changes in electricity usage by end-use customers from their normal consumption patterns" in response to pricing and incentive payments. This paper focuses on new opportunities for DR with electric vehicles (EVs). EVs are potential distributed energy resources that support both the grid-to-vehicle and vehicle-to-grid modes. Their participation in the time-based (e.g., time-of-use) and incentive-based (e.g., regulation services) DR programs helps improve the stability and reduce the potential risks to the grid. Smart scheduling of EV charging and discharging activities also supports high penetration of renewables with volatile energy generation. This paper proposes a novel stochastic model from the Independent System Operator's perspective for risk management and participation planning of EVs in the smart grid for DR. The risk factors considered in this paper involve those caused by uncertainties in renewables (wind and solar), load patterns, parking patterns, and transmission lines' reliability. The effectiveness of the model in response to various settings such as the area type (residential, commercial, and industrial), the EV penetration level, and the risk level has been investigated. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Electric vehicle Grid-to-vehicle Vehicle-to-grid Smart grid Demand response

1. Introduction The overall sales of electric vehicles (EVs) have been steadily rising [1]. The worldwide sales of modern EVs have recently passed the one-million milestone as shown in Fig. 1. Multiple reasons have contributed to such an increasing trend in recent years. One reason is that significant savings can be achieved by driving EVs in place of traditional internal combustion engine vehicles. One example of a Nissan Leaf vs. a Toyota Camry is shown in Fig. 2[2]. In addition, EVs release almost no carbon dioxide (CO2) or air pollutants at the time of usage. As EV sales boom, the market share of EVs will also likely to increase in the future. However, the current power grids in many countries are not fully prepared for a high EV penetration. Therefore, unmanaged charging of EVs may cause problems such as system overload, power losses, and voltage fluctuations [3]. To deal with EVs' additional load and mitigate these potential issues, appropriate charging control is required. In literature, a wide variety of models were proposed for charging planning of EVs [4e7].

* Corresponding author. EB-T10, 4400 Vestal Pkwy E, Binghamton, NY 13902, USA. E-mail address: [email protected] (Y. Wang). http://dx.doi.org/10.1016/j.energy.2016.10.002 0360-5442/© 2016 Elsevier Ltd. All rights reserved.

Controlled charging during the valley hours (midnight to early morning) can partially reduce such issues and increase the utilization rate of existing infrastructure [8]. The charging service is commonly referred to as the grid-to-vehicle (G2V) service. Given the high battery capacity of EVs, they are also considered suitable distributed energy resources to discharge and feed power back to the grid when needed [9e11]. The discharging service is usually referred to as the vehicle-to-grid (V2G) service. The G2V and V2G services have the potential to enable the power grid to accommodate various EV penetration levels through the demand response (DR) programs [12,13], without significant system upgrades. The main idea of DR is to encourage electricity users to manage their demand during peak periods or when system's safety is at risk [14]. DR programs can be divided into two categories: time-based and incentive-based. Major time-based DR programs include time of use (TOU) [15], real time pricing [16], and critical peak pricing [17]. A common feature of such programs is the varying electricity price over time. The price will be higher during the on-peak periods and lower during the off-peak periods. The time-varying price intends to level the load, i.e., shift the load from the on-peak periods to the off-peak periods [18,19]. As a result, not only the generation costs of power grid decreases considerably, the

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

Abbreviations DR EV G2V ISO RBC SOC TOU V2G

Demand response Electric vehicle Grid-to-vehicle Independent system operator Remaining battery capacity State of charge Time of use Vehicle-to-grid

Indices and ranges a Aggregator index g Generator index j Wind power system index k Bus index l Line index n Solar power system index s Scenario index t Time index A Total number of aggregators G Total number of conventional generators J Total number of wind power systems K Total number of buses L Total number of lines N Total number of solar power systems S Total number of scenarios T Planning horizon Decision variables ba,t,s G2V reserve provided by aggregator a in time t of scenario s ck,t,s Renewable energy curtailment in bus k in time t of scenario s dþ Energy charged to aggregator a in time t of scenario s a;t;s d V2G reserve provided by aggregator a in time t of a;t;s scenario s fl,t,s Energy flow through line l in time t of scenario s pg,t,s Energy dispatched from generator g in time t of scenario s xV2G Required V2G reserve of aggregator a in time t a;t G2V xa;t Required G2V reserve of aggregator a in time t zk,t,s Unmet load of bus k in time t of scenario s qk,t,s Voltage angle at bus k in time t of scenario s Binary dummy variables 0 00 wt,s, wt;s , wt;s , qa,t,s Binary dummy variables EV owners pay less for their charging expenses. The EVs will be utilized only in the G2V mode when participating in the time-based DR programs. The TOU is recognized as the most efficient timebased tariff in reducing EVs' charging costs and emissions [20]. Incentive-based DR programs for EVs include frequency regulation and spinning reserve [21e24]. Such applications have not been widely implemented yet, but they have great foreseeable potentials [25,26]. Both programs are a part of ancillary services, which are designed to support the power grid's reliability and continuous flow of electricity so that supply will continually meet demand. The regulation service is a real-time service to balance load and power generation so that the frequency will be maintained within a specific range of the nominal frequency (e.g., 60 Hz). The frequency deviates from its nominal value when there

837

Dependent variables Agr RBCa;t;s

RBC of aggregator a in time t of scenario s

Agr SOCa;t;s

SOC of aggregator a in time t of scenario s

Parameters Bl Susceptance of line l þAgr

Ca CaAgr CkCur

Cost of V2G reserve provided by aggregator a Cost of G2V reserve capacity for aggregator a Cost of renewable energy curtailment at bus k

CkUL

Penalty cost for one unit of unmet load at bus k

CgGen CtCh CtDch CtDis Flmax

Generation cost of generator g Electricity price for charging EVs in time t Discharged energy cost of EVs in time t

M Pgmax Pgmin

Discount for providing G2V service at time t Maximum capacity of line l Incidence matrix coefficient (1, 0, or 1) at bus k of line l A very large number (big-M) Upper limit for power generation of generator g Lower limit for power generation of generator g

Ps

Probability of occurrence of scenario s

Rdn g

Ramp down limit of generator g

Rup g Agr SOCa;0 a(l) b(l)

Ramp up limit of generator g Initial SOC of aggregator a Bus that line l starts Bus that line l ends User defined risk factor Charge efficiency of EVs for aggregator a Discharge efficiency of EVs for aggregator a

Hl,k

g hþAgr a hAgr a

Random parameters SOCa Desired leaving state of charge in percentage for aggregator a SOCaþ Expected joining state of charge in percentage for aggregator a dl;t;s Whether or not line l fails in time t of scenario s lk,t,s Load at bus k in time t in scenario s pþ Joining EVs' capacity of aggregator a in time t of a;t;s scenario s p Leaving EVs' capacity of aggregator a in time t of a;t;s scenario s 4j,t,s Generation of wind system j in time t of scenario s jn,t,s Generation of solar system n in time t of scenario s

is a mismatch between load and electricity supply. EVs can be utilized in both the G2V and the V2G modes when participating in the incentive-based DR programs. The regulation down service is implemented in the G2V mode when power generation exceeds the load, and the regulation up service is implemented in the V2G mode when the power generation is insufficient for the load [27,28]. Several studies examined the strengths, weaknesses, opportunities, and threats of using EVs for frequency regulation. The participation of EV owners is motivated by the additional revenue for the bidirectional energy flow [29]. Various techniques were proposed for cost benefit analysis of applying EVs in the V2G mode [30e32]. It was shown that the benefits justify the battery degradation and replacement expenses [33]. Han et al. [34] investigated

838

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

Fig. 1. Cumulative EV sales worldwide from 2009 to 2015.

Fig. 2. Annual Fuel Savings: Nissan Leaf vs. Toyota Camry (US$/10,000 miles).

the management of EVs charging rate and sequence. Dallinger et al. [35] studied the required capacity for bidding regulation services considering the stochastic behavior of the EVs [35]. Ansari et al. [36] applied fuzzy optimization to model uncertainties involved in bidding ancillary services. Zeng et al. showed that the factors such as battery sizes and charging rates affect the resultant profit of participation in the regulation market [37]. Jargstorf and Wickert discussed the financial benefits of providing regulation using EVs on German market [38] and conclude that the high infrastructure cost may still be a barrier. The benefits in terms of costs and system reliability for EVs participating in spinning reserve were also studied in Refs. [39,40]. A great advantage of using EVs for spinning reserve is that they are able to respond to the request rapidly. This capability is especially useful to ensure grid stability with the increased use of intermittent renewable generation such as wind and solar [41,42]. It was shown that bidirectional connection of EVs to the power system with wind power generation reduces the grid operation costs significantly [43,44]. Similar results for well-being assessment of smart grids with solar and wind power generation are presented in Ref. [45]. It was also shown that the provided reserve by EVs can compensate the forecasting errors of power generation by the volatile renewable energy sources [46]. Another photovoltaic-based energy management mechanism using DC link voltage sensing is presented in Ref. [47]. Other than renewable power availability, the charging and discharging can be affected by electric pricing as presented in Ref. [48]. The proposed charging control mechanisms in Refs. [47,48] are only concerned with charging control of single EVs. System level day-ahead reserve planning and coordination of a fleet of EVs are not incorporated.

In summary, the analysis of previous literature shows that the research on EV integration in the power grid has attracted much attention. Previous studies have created a solid knowledge base for further research. However, in most of the existing literature, EVs' integration is analyzed from the perspective of the EV service aggregators. In practice, it is imperative to consider the interactions between aggregators and the Independent System Operator (ISO) simultaneously while each of them tries to optimize their own benefits. In the existing literature, the simultaneous consideration of both time-based and incentive based DR programs in one model is rare. In addition, uncertainty may rise from any components in the smart grid with EV and renewable integration. Previous consideration of such uncertainties are scattered, each focusing on one or two separated aspects. There lacks a systems approach to comprehensively consider the stochastic factors that may greatly affect the efficiency and effectiveness of the established models. Motivated by this status quo, we plan to contribute to the knowledge base in the following aspects in this paper: (i) We propose a new stochastic model for the EV participation in both time-based and incentive-based DR programs by considering the interactions between the ISO and the aggregators. For the time-based DR, we focus on the TOU program; for the incentive-based DR, we focus on the regulation services. (ii) The risk factors considered in this paper involve those caused by uncertainties in renewables (wind and solar), load patterns, parking patterns, and transmission lines' reliability. (iii) The effectiveness of the model in response to various settings such as the area type (residential, commercial, and industrial), the EV penetration level, and the risk level has been investigated. (iv) The model presented in this paper also addresses part of the concerns summarized in Ref. [49] regarding control and planning of EV parking lots. We highlight the need for a smart planning mechanism that not only optimize EVs charging but also supports grid-stabilizing demand response and ancillary services.

2. Problem description The overview of the system discussed in this paper is shown in Fig. 3. The ISO plays a central role in the system. It operates the market by collecting and issuing information among various players in the market such as the power generation units, the load centers, and the EV aggregators. The ISO tries to minimize the system operation costs while maintaining the balance between supply and demand at all times. The power system consists of distributed energy generation units such as conventional power generators as well as renewable wind and solar energy systems. Due to the limited capacity of each EV battery, the isolated contribution of an individual EV to the grid is negligible. In existing literature, it was shown that planning the participation of smallsized consumers in wholesale electricity market is inefficient [24]. However, controlled charging and discharging activities of a large number of EVs through an aggregator service can be an effective way to achieve the desired DR goals. Therefore, the intermediate entity like the aggregator is required to participate in biddings and coordinate EVs' charging and discharging activities. The aggregators support both the G2V and the V2G modes. To enable aggregators to properly schedule EVs', the EV owners report their battery capacity and trip plans with expected arrival and departure time. To participate in the load leveling and frequency regulation services, the aggregators inform ISO about their

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

839

Market

Commercial

Demand forecast

Industrial

Power generation plan

Conventional generators

Curtailments plan Independent system operator (ISO)

Wind power

Residential EVs availability

Solar power

Renewables forecasts

Reserve plan

Power generation units

Load centers

Prices

Aggregators Charge/discharge arge schedule

Trip plan EV owners

Fig. 3. Overview of the system examined in this paper.

available capacity and forecasted SOC. In this research, the aggregator's SOC is expressed in terms of its total EV capacity. The example in Fig. 4 shows the state change of SOC and remaining battery capacity (RBC) for an aggregator at start and end of time t. On the left side, the EVs are discharged in the V2G mode; the aggregator's total SOC is decreased and the RBC is increased. On the right side, the EVs are charged in the G2V mode; the aggregator's total SOC is increased and the RBC is decreased. Aggregators support both the time-based and the incentivebased DR programs. In this research, TOU is selected as the timebased program to provide load leveling service. If the EVs participate in this program, they should pay different prices for off-peak, mid-peak, and on-peak periods [50,51]. Aggregators also participate in the incentive-based DR program to provide V2G and G2V needed for services like regulation. These services incur two different types of costs for the ISO: the reserve capacity payment and the energy payment [29] to the aggregators. The capacity payment is associated with the maximum capacity that each aggregator can provide during the contract period. The energy payment is related to the cost of actual dispatched energy in the V2G or G2V mode. A certain level of EV capacity will be needed to serve as the reserve to quickly respond and mitigate the supply surplus or deficiency for the regulation service. The required reserve levels should be determined properly. The decisions related to EVs charging/discharging and power generation plans from

various resources should also be carefully determined. The decisions should be made by considering possible future scenarios with various risk factors. The risk is expressed in terms of the probability of unbalance between load and supply.

3. Proposed model The proposed model for management of EVs participation in DR programs is based on the basic DC OPF model. Eqs. (1)e(5) represent the basic OPF formulation that minimizes the costs associated with different generating units and shed load considering technical constraints of the power grid [52].

min

K X

CgGen pg þ CkUL zk

(1)

k¼1 L X

Hl;k fl þ pg þ zk ¼ lk

ck

(2)

l¼1

  fl ¼ Bl qaðlÞ  qbðlÞ

cl

Fig. 4. An aggregator's state change in the V2G (left) and G2V (right) modes.

(3)

840

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

Flmax  fl  Flmax

cl

(4)

constraints shown in Eqs. (3)e(5). The flow constraint is not considered for failed lines (dl,t,s ¼ 1). Constraint (13) ensures that the generators operate within the ramp limits.

Pgmin  pg  Pgmax

cg

(5)

  M dl;t;s  fl;t;s  Bl qaðlÞ;t;s  qbðlÞ;t;s  M dl;t;s

Eqs. (2) and (3) show the power balance in each bus and power flow in each line respectively. Lines' thermal flow limit and generation capacity of generators are represented in Eqs. (4) and (5). The above OPF model is modified to incorporate dynamic decisions in our proposed model. The main goal is to manage the required V2G and G2V reserve levels considering the expected future costs of the system. The modified model for managing EVs participation is presented in Eqs. (6)e(25). The first part of the objective function shown in Eq. (6) is related to the reserved EVs capacity payment for contracting the V2G and G2V services. The second part is the expected operation costs of the ISO for actual dispatched energy payments for regulation services. The last part is about the expected generation costs of the conventional generators, shed load costs, and renewable energy curtailment costs. The curtailment costs are included in the model because when extra energy is produced by renewables, the ISO has to pay others to take such additional energy [53]. T X A  X

min

Agr G2V CaþAgr xV2G xa;t a;t þ Ca

þ

S X

"

Ps

s¼1

þ

S X

" Ps

s¼1

T X A  X

CtDis ba;t;s

þ

CtDch d a;t;s



(6)

CgGen pg;t;s þ CkUL zk;t;s þ CkCur ck;t;s



#

t¼1 k¼1

Eq. (7) is similar to Eq. (2) and shows the power balance in each bus. The total energy flow that enters a bus (generated electricity from conventional generators and renewable energy sources, and discharged energy from aggregators) should be equal to the total energy flow that leaves the bus (base load, energy charged to aggregators, and renewable energy curtailment). L X

ck; t; s

(7)

The SOC for each aggregator is presented in (8). It changes based on the charge and discharge status as well as the efficiency of the batteries. The aggregators should first provide the charge to the EVs that will be leaving soon. The SOC of leaving and joining EVs also affect the total SOC of aggregator. Constraint (9) shows the change in the aggregators' RBC. The time that the EV arrives at the parking location as well as the charging and discharging patterns will influence the RBC. The RBC is also affected by the SOC of joining and leaving EVs' SOC.

¼

Agr SOCa;t1;s

þ

hþAgr dþ a a;t;s

þ SOCaþ pþ a;t;s



1

d a;t;s hAgr a



SOCa p a;t;s

(8) Agr

Agr

þAgr þ da;t;s

þ

(12)

up

(13)

Rdn g  pg;t;s  pg;t1;s  Rg

1  da;t;s hAgr a

      SOCaþ pþ a;t;s  1  SOCa pa;t;s

cg; t; s

Constraint (14) reflects the risk perspective of the decision maker based on the Value at Risk concept [54]. According to this constraint, the probability of any mismatch (reflected by unmet load and energy curtailment) between energy supply and load should be less than or equal to a specified risk tolerance limit (g). Constraints (15) and (16) guarantee that the aggregators operate only in one of the two modes, V2G or G2V, at each instant of time. K X

zk;t;s þ

K X

! ck;t;s  0

g

cs; t

 þ 1 ca; t; s

ca; t; s

  0  d a;t;s  M 1  qa;t;s

(15)

ca; t; s

(16)

Constraint (17) shows that the discharge is limited by available energy considering discharging efficiency. Constraint (18) guarantees that the charge does not exceed the available empty capacity of the aggregator.

1

hAgr a

Des  þ þ d a;t;s  SOCa;t1;s  SOCa pa;t;s þ SOCa pa;t;s Agr

Constraints (10) through (12) represent the similar technical

ca; t; s (17)

   1  SOCa p a;t;s



ca; t; s

(18)

Constraints (19) through (22) are the bounding constraints. The unmet demand and energy curtailment are bounded by the actual demand and available renewable energy in (19) and (20), respectively. The required reserve is determined in the aggregators' contracts, and their support for the G2V and V2G services is bounded by their capacities. Constraint (21) demonstrates this capacity limit for G2V. Similarly, aggregators' discharged energy is limited by the maximum V2G reserve defined in their contract (22).

0  zk;t;s  lk;t;s

0  ba;t;s  xG2V a;t V2G 0  d a;t;s  xa;t

(9)

(14)

k¼1

ck; t; s

0  ck;t;s  4j;t;s þ jn;t;s

ca; t; s

RBCa;t;s ¼ RBCa;t1;s  ha

cg; t; s

Agr þ þ hþAgr dþ a a;t;s  RBCa;t1;s þ 1  SOCa pa;t;s

l¼1

Agr SOCa;t;s

Pgmin  pg;t;s  Pgmax



Hl;k fl;t;s þ pg;t;s þ zk;t;s þ d a;t;s þ 4j;t;s þ jn;t;s

¼ lk;t;s þ dþ a;t;s þ ck;t;s

(11)

0  dþ a;t;s  Mqa;t;s

#

t¼1 a¼1 T X K  X

cl; t; s

k¼1

t¼1 a¼1

(10)

Flmax  fl;t;s  Flmax

pr



cl; t; s

(19) ck; t; s

(20)

ca; t; s

(21)

ca; t; s

(22)

Constraint (23) defines the G2V reserve. The max term guarantees that G2V reserve is needed when the energy generation is higher than the system's load. In that case, the EVs are charged and their G2V reserve is calculated based on their participation for

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

consuming extra energy. Constraint (24) shows that the G2V service provided by each aggregator cannot exceed its charging at that time. The ranges for reserve variables are shown in (25). A X

0 ba;t;s

¼ min@

a¼1

0

A X a¼1

K X



max@0;

dþ a;t;s ;

J X

4j;t;s þ

jn;t;s

n¼1

j¼1

11

N X

lk;t;s AA ca; t; s

(23)

k¼1

ba;t;s  dþ a;t;s

ct; s; a

G2V  0; xV2G xa;t a;t  0

(24)

ca; t

(25)

The model established above is a mixed integer nonlinear programing problem. This model can be transformed to a mixed integer linear programming problem. To linearize Constraint (14), we replace it with Constraints (26) and (27). The binary dummy variable ws,t is set to one if mismatch between energy supply and demand exist and zero otherwise. To linearize Constraint (23), we introduce Constraints (28) through (31) to cover all possible cases. K X

K X

zk;t;s þ

k¼1 S X

ck;t;s  Mwt;s

ct; s

(26)

k¼1

Ps wt;s  g

ct; s

(27)

s¼1 J N K  0  X X X 4j;t;s þ jn;t;s  lk;t;s M wt;s  n¼1

j¼1



0

 M 1  wt;s J X

4j;t;s þ

N X

jn;t;s 

n¼1

j¼1



K X

k¼1

c t; s

lk;t;s 

(28)

A X

  00 dþ a;t;s  M 1  wt;s

c t; s

a¼1

k¼1

(29) A X

0 ba;t;s

@

a¼1

J X

4j;t;s þ

N X

jn;t;s 

n¼1

j¼1 0

 Mwt;s

K X

1



lk;t;s A  M 1  wt;s 00



k¼1

c t; s (30)

A X a¼1

ba;t;s 

A X

00

0

dþ a;t;s  Mwt;s  Mwt;s

c t; s

(31)

841

4.1. The base case The system configuration is presented in Fig. 5. It includes three conventional generators with the same ramp limit of 200 kW per hour. The maximum generation capacity is 700 kW for each generator. No minimum generation limit is set for conventional generators. The transmission capacity of each line is 500 kW with line susceptance equal to 10 p.u. It is assumed that the three load centers have the same base load as shown in Table 1. The load profile is scaled down so it has smaller magnitudes but the same shape as the data in Ref. [19]. EV charging will impose an additional load for the system, and the additional load is not included in the base load. The parking pattern in Ref. [36] is used to estimate the daily numbers of joining and leaving EVs (Fig. 6). Each parking area has a maximum capacity of 200 vehicles and is managed by one aggregator. It is assumed that each EV has a battery capacity of 24 kWh with 99% charging and discharging efficiencies. It is expected that 35% of parked vehicles are electric vehicles (i.e., the penetration level of EVs is 35%). The EVs join the parking with 30% battery charged and prefer to leave with 90% charged battery [56]. The wind and solar power generations (Fig. 7) are scaled down based on California ISO's wind and solar power generation data [57]. The costs related to renewable energy curtailment and lost load are set to $1.5/kWh and $5/kWh, respectively [23]. The cost of emergency electricity generation by the conventional generator is set to $0.20/kWh. Aggregators are paid $0.02/kWh for their available capacity for V2G and G2V [29]. The aggregators receive 100% discount if they charge when energy curtailment is needed. The ISO also pays aggregators for the actual dispatched energy for regulation up or other purposes. The payments for this service differs and mostly depends on market electricity price. For the base case, the electricity price is considered $0.01/kWh. The model is developed in Matlab and solved using the CPLEX solver in both the deterministic and the stochastic forms. The results for the deterministic case are analyzed in this subsection, and the results for the stochastic case are analyzed in the subsequent subsections. The deterministic case serves as the base case. For the deterministic case, the risk constraint is relaxed, and the model is solved after setting random parameters to their expected values. The results for load leveling in the deterministic case are shown in Fig. 8. It shows that participation of EVs in load leveling helps better utilize available renewable energy and shift the charging load to the off-peak periods. The results for the V2G and G2V services in the deterministic case are shown in Fig. 9. It shows that the EVs provides G2V reserve in hours with higher renewable energy generation and insufficient base load. The EVs are discharged in peak periods to provide the V2G service. Conventional generators are used to generate the required electricity in periods that the sum of the renewables and the dispatched energy by EVs is not sufficient to meet the base load. 4.2. Charging policy

a¼1

4. Case studies and effects of various model parameters To demonstrate the effectiveness of the proposed model, we solve it for a one-day plan using the modified IEEE six-bus microgrid[55]. A wide range of experiments are conducted in this section to derive insights about effects of various model parameter on the optimal solutions.

In this subsection, the effect of different EV charging schedules on the power system operation costs is demonstrated. The base case deterministic model is run under three different charging policies. In the first policy, it is assumed that EVs are not participating in DR programs and they are charged once they arrive at parking. The second policy represents that EVs participate in the time-based DR program. It enables the aggregators to plan their charging time to minimize the electricity bills and flatten the total load of the grid. When aggregators participate in the time-based DR, the ISO only responds to their charging patterns by minimizing its own operation costs. Table 2 presents the TOU pricing

842

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

Fig. 5. Six-bus test system configuration.

Table 1 Load information for the test system. 1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

Load (kW)

100

110

120

90

120

140

150

180

250

325

320

325

260

280

250

250

190

195

200

230

280

250

240

190

Parking utilization (%)

Clock

100% 80% 60% 40% 20% 0% 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (hour)

Fig. 6. Typical parking patterns.

incentive-based program also generates positive income for aggregators. Fig. 10 shows that unplanned charging forces the conventional power systems to generate more electricity during the peak periods when the system experiences higher loads. The incentive-based program almost completely eliminates the need for conventional energy generation by appropriately utilizing

400 350 300 250 200 150 100 50 0

Power geeneration (kW)

data used for managed charging [15]. The total charging costs of aggregators participating in TOU is calculated using PT PA Ch þ t¼1 a¼1 Ct da;t;s . In the third policy, it is assumed that EVs participate in the incentive-based DR program. They receive incentives for their participation in the V2G and G2V modes. The total PT PA energy cost of the aggregators in this policy is t¼1 a¼1 Reg þAgr Agr Dch d V 2G  C ðCtCh dþ xa;t xG2V a a;t;s  Ca a;t Þ. Negaa;t;s  Ct ba;t;s  Ct tive values represent that aggregators not only do not have to pay any cost, but also make revenue through received incentives. The results of these charging strategies are shown in Table 3. Conventional power generation and charging patterns for the three policies are presented in Figs. 10 and 11. Participation in the DR programs reduce the costs for both the aggregators and the ISO. Compared to the time-based DR program, the incentive-based program entails more cost savings for ISO. This is mainly because EVs' utilization for managed G2V and V2G reduces costs associated with lost load and energy curtailment. Participation in the

Solar Wind

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

Fig. 7. Typical wind and solar power generation patterns.

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

1200

Base load

1000

Charge load Renewables

800

Power (kW)

843

600 400 200

0 1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

Fig. 8. Results of the EV participation for load leveling in the base case.

900

Total reserve (kW)

700

G2V V2G

500 300 100 -100

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

-300 -500

Fig. 9. Results of the EV participation for providing reserve in the base case.

Table 2 TOU pricing. Clock

1

2

3

4

5

6

7

8

9

10

11

12

Price ($/kWh) Clock Price ($/kWh)

0.05 13 0.19

0.05 14 0.19

0.05 15 0.19

0.05 16 0.12

0.05 17 0.12

0.05 18 0.12

0.05 19 0.19

0.12 20 0.19

0.12 21 0.19

0.12 22 0.12

0.19 23 0.12

0.19 24 0.05

Table 3 Results for the three charging policies. DR charging policy

ISO reserve cost ($)

ISO operation cost ($)

Aggregators' energy payment ($)

Generation (kWh)

Curtailment (kWh)

Peak load (kW)

No participation Time-based Incentive-based

0 0 134

8288 3577 684

253 127 206

4903 2132 74

4871 2100 0

1725 975 1068

1000

No participation Time-based Incentive-based

Power (kW)

800 600 400 200 0 1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour) Fig. 10. Conventional power generation for the three different charging policies.

renewable sources. The main goal of demand response programs is peak load reduction. As the results show, participation in the timebased and incentive-based programs leads to 48% and 51% peak load reduction, respectively. Fig. 11 displays that aggregators' participation in DR programs will motivate them to avoid charging during on-peak hours and the charging activities will mostly happen during off-peak periods.

4.3. Risk perspective and stochastic solutions In this subsection, the model is solved in the stochastic form with a predefined risk level of 0.01, which means the probability of a mismatch between load and supply should be less than 1%. It is assumed that load, renewable energy generation, EV owners' behavior, joining and leaving SOC of aggregators, and line failures are uncertain. To minimize the computation time of solving the

844

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

1000

No participation Time-based Incentive-based

Charge (kW)

800 600 400 200 0 1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

Fig. 11. EV charging activities for the three different charging policies.

Table 4 Reserve (kW) for the G2V service for the deterministic and the stochastic models. Clock

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

Deterministic Stochastic

647 693

791 846

748 804

798 852

642 693

439 483

257 294

91 95

0 0

0 0

0 0

0 0

82 82

0 0

46 47

0 0

94 95

0 0

0 0

0 0

0 0

0 0

0 0

0 0

Table 5 Reserve (kW) for the V2G service for the deterministic and the stochastic models. Clock

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

Deterministic Stochastic

0 0

0 0

0 0

0 0

0 0

0 0

0 0

0 0

89 103

266 316

125 140

114 129

0 0

18 43

0 0

1 30

0 5

56 64

219 231

314 333

352 372

201 261

200 264

104 105

stochastic model, the scenario reduction method presented in Ref. [58] is used to build the scenario tree with 10 scenarios. Tables 4 and 5 show the reserve plan results of the G2V (which is for regulation down) and V2G (which is for regulation up) services, respectively, for the deterministic and stochastic model. The results show that higher levels of reserves are required (thus higher costs are involved) by considering the effect of uncertain parameters and the risk level. The stochastic model is also solved for various risk thresholds including 0, 0.001, 0.01, 0.1, and 1. The results in Fig. 12 show that a higher risk threshold tolerates a higher probability of mismatch between supply and load and thus entails less reserve requirements.

4.4. Parking availability Parking availability is another influencing factor for determining the reserve capacities. Three different types of parking patterns are tested for the incentive-based DR participation. The same test system presented in Fig. 5 is used and it is assumed that parking lots in Buses 4, 5, and 6 are located in residential, commercial, and industrial areas, respectively. Each parking lot has a total capacity of 160 vehicles. The parking patterns for these three areas are shown in Fig. 13. Figs. 14e16 present the results for the participation of

these aggregators in the G2V and V2G services. The aggregator in the residential and industrial areas are more willing to provide G2V because of their capacity availability after midnight. During day time, especially in the peak periods, there is less need for curtailing renewable energy generation. The parking lots in the commercial area have more available capacities during day time and the corresponding aggregator can participate in the G2V service if needed. However, as Fig. 15 shows, this area is not a good choice for providing the V2G service. This is because the frequent leaving and joining patterns in this area greatly affect its SOC and RBC. Fig. 17 shows the charging patterns for all three areas. The results are shown as the proportion of SOC to the total available capacity of parked EVs in each parking area. For example, the EVs parked in the residential and industrial areas from 3:00 a.m. to 6:00 a.m. are kept fully charged because of their earlier participation in the G2V regulation. Participation of the aggregators placed in the residential and industrial areas for providing V2G reduce the charged capacity during peak periods. Appropriate charging planning of vehicles in the commercial area also prevents full charging during the peak periods.

4.5. Load patterns The effects of load patterns in residential, commercial, and

5600 5200 5000 4800 4600

4400 4200 0

0.001

0.01 0.1 Risk tolerance

1

Regulation up (kW)

Regulation down (kW)

5400

2500 2400 2300 2200 2100 2000 1900 1800 0

Fig. 12. Effects of risk tolerance on reserve plans.

0.001

0.01 0.1 Risk tolerance

1

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850 100%

100%

100%

80%

80%

80%

60%

60%

60%

40%

40%

40%

20%

20%

20%

0%

0% 1 3 5 7 9 11 13 15 17 19 21 23 Time (hour)

1 3 5 7 9 11 13 15 17 19 21 23 Time (hour)

845

0% 1 3 5 7 9 11 13 15 17 19 21 23 Time (hour)

Fig. 13. Parking patterns in the residential (left), commercial (middle), and industrial (right) areas.

Total reserve (kW)

500 G2V V2G

300 100 1

-100

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour) -300 Fig. 14. Participation of the aggregator located in the residential area.

Total reserve (kW)

200 G2V V2G

100

0 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

-100 Fig. 15. Participation of the aggregator located in the commercial area.

Total reserve (kW)

500 G2V V2G

300 100 1

-100

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

-300 Fig. 16. Participation of the aggregator located in the industrial area.

Charged capacity (%)

100%

Residential Commercial Industrial

50%

0% 1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

Fig. 17. Charged capacity percentages in three different areas.

industrial areas on reserve plans are investigated in this subsection. Previous studies have shown that the load patterns differ greatly

for these three areas [59]. Fig. 18 shows example load patterns scaled down based on the data provided by Ref. [59]. The peak loads

846

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

Load (kw)

450

Residential Commercial Industrial

350 250 150 50 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

Fig. 18. Load information for the residential, commercial, and industrial areas.

are 350 kW for all three areas. To study the influence of load patterns on the reserve services provided by EVs, three different cases defined in Table 6 are examined in this subsection. Figs. 19e21 show that depending on the load and parking patterns, the EVs participation differs in each case. The residential areas with a total reserve of 6131 (kWh) are the best choices for utilizing EVs for regulation reserve (Fig. 19). The EVs of residential areas usually are parked in off-peak times during the morning. They can be charged during these periods and used for V2G during onpeak periods. The parking availability is highly volatile in the

commercial areas during peak times and they are not reliable sources for V2G. However, the results show that they can be utilized for G2V in the early morning and at night after 21:00. During the morning, noon, and afternoon hours in the industrial areas, EVs usually are parked until the end of the work shift. Therefore, they are good choices to provide G2V considering the constant load profile of industrial areas. They can also provide limited V2G reserve. Another set of experiments is implemented to investigate the time-based and incentive-based DR programs' influence on the system operation cost for the residential, commercial, and industrial areas. The results are compared with unplanned charging without DR participation (Fig. 22). The time-based DR program is more efficient in reducing the peak load while the incentive-based DR program has more capability in balancing load and supply in the power grid. The cost results summary in Table 7 shows that compared to the time-based program, the incentive-based program entails more cost savings in all three areas. Participation of EVs in these DR programs helps aggregators save their electricity payments and in some cases results in extra income.

Table 6 Three different cases. Case

Test system modification in Fig. 5

Parking pattern in Fig. 13

Load pattern in Fig. 18

1 2 3

All load areas being residential areas All load areas being commercial areas All load areas being industrial areas

Residential parking pattern Commercial parking pattern Industrial parking pattern

Residential load Commercial load Industrial load

900

-100

V2G

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour) -600 Fig. 19. Reserve participation of the aggregators in residential areas.

Total reserve (kW)

500 G2V

300

V2G

100 -100

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour) -300 Fig. 20. Reserve participation of the aggregators in commercial areas.

600

Total reserve (kW)

Total reserve (kW)

G2V 400

G2V

400

V2G

200 0 -200

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

-400 Fig. 21. Reserve participation of the aggregators in industrial areas.

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

Generation reduction

Aggregator cost reduction

ISO cost reduction

Generation reduction

Residential Commercial Industrial

60% 50% 40% 30% 20% 10% 0%

Curtailment reduction

120% 100% 80% 60% 40% 20% 0%

Aggregator cost reduction

ISO cost reduction

Peak load reduction

847

Residential Commercial Industrial Curtailment reduction

Peak load reduction

Fig. 22. Results summary for time-base (left) and incentive-based (right) programs for three different areas.

Table 7 Costs and income summary for comparing two DR programs for three different areas. Areas

DR type

ISO reserve plan cost ($)

ISO operation cost ($)

Aggregators' energy payment ($)

Aggregators' income ($)

Residential

No DR participation Time-based DR Incentive-based DR No DR participation Time-based DR Incentive-based DR No DR participation Time-based DR Incentive-based DR

0 0 123 0 0 23 0 0 70

9872 7417 4391 7645 6144 6144 4513 2584 1032

188 94 403 199 173 206 127 63 239

0 0 736 0 0 135 0 0 418

Commercial

Industrial

4.6. Green energy utilization The problem is also solved for different combinations of EV penetration and renewable energy capacities. The results for renewable energy curtailment, conventional electricity generation, and the related reserve are tested for three cases explained in

Table 6. Figs. 23e25 summarize these results for residential, commercial, and industrial areas. The axis for “renewables utilization” represents the scaling factor for both wind and solar energy compared to their capacities in the base case. Thus, number “1” in this axis represents the same level of renewable energy as used in the base case, while “2” means doubled renewable energy

Fig. 23. Results summary for residential area.

848

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

Fig. 24. Results summary for commercial area.

Fig. 25. Results summary for industrial area.

utilization. The renewable energy curtailment is expressed as a percentage of the total available renewable energy generation. For example, if the total renewable energy generation is 1000 kWh and 250 kWh is curtailed by the ISO, the curtailment percentage will be 25%. The conventional power generation is expressed as a percentage of the total load (including EVs' charging) met by the conventional generator. For example, if the total load (the base load plus the charge load) is 1000 kW and 750 kW is met by the conventional generators, then the generation percentage is 75%. The main results based on the observations presented in Figs. 23e25 are summarized as follows:

(i) Fig. 23(a) reveals that increasing EVs penetration reduces the renewable energy curtailment in residential areas. These energy curtailments are achieved by dedicating higher G2V reserve as shown in Fig. 23(c). (ii) Fig. 23(b) shows that in settings with low renewable utilization (0.5), increasing EVs penetration results in a rising need for conventional generation. In settings with medium renewable utilization (1), however, increasing EVs penetration will reduce the conventional generation initially. After a certain level (50%), the available renewable energy is not sufficient to meet the required energy to charge EVs. Then the system may need to utilize more conventional

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850

generation. It was shown that utilizing more renewable energy in residential areas completely eliminate the need for conventional energy generation. (iii) Fig. 23(d) shows that in the presence of renewables, EVs' participation in V2G increases with increasing penetration rate initially. However, after a certain level (e.g., 50% in default renewables utilization) their participation in V2G may decrease so they can use the stored energy for their own needs. (iv) The similar patterns for commercial areas' curtailment and conventional generation are observed in Fig. 24(a) and (b). The only difference in commercial areas is that their energy need mostly occurs during periods with lower renewables availability. Therefore, they require more utilization of conventional generators. Due to volatile EV capacity availability in commercial areas, aggregators prefer not to participate in V2G unless in higher rates of renewables. However, they are actively utilized for G2V reserve in periods with higher renewable generations. The results for G2V and V2G in Fig. 24(c) and (d) are consistent with this analysis. (v) Fig. 25(a), (b), and (c) have similar patterns as the counterparts in Fig. 23. The only difference lies in the magnitudes. Fig. 25(d) illustrate that EVs also participate in V2G reserve. However, since the base load is higher in industrial areas, less energy can be stored in the G2V mode and discharged later in the V2G mode.

5. Conclusions and future work This paper presents a novel model for planning participation of EVs in the DR programs and their scheduling in the smart grid. The system is subject to risks of uncertain factors such as loads, renewable energy generation, lines reliability, and EV owners' behaviors. EVs participation in both the time-based (i.e., TOU) and incentive based (i.e., regulation) DR programs has been investigated. The proposed stochastic model enables the decision maker to specify the risk level and balance its costs and benefits according to this factor. The presented model optimizes the required reserve levels of the ISO while considering benefits of aggregators. The ISO minimizes the operation costs by optimally scheduling conventional generation and renewable energy curtailment. Simultaneously, the aggregators try to minimize their electricity payments by optimally scheduling EVs charging and discharging to receive the maximum discounts and income from the DR participation. A wide range of experiments on deterministic and stochastic cases are implemented to show the effects of parameters such as charging policies, parking availability, risk levels, renewable energy penetration, and load patterns. The results show that utilizing services such as the time-based DR programs are efficient in reducing the electricity costs for the ISO and the aggregators. It is also shown that the EVs' participation in the incentive-based DR programs are highly affected by the parking patterns. The available EVs' capacity differs in various areas such as residential, commercial, and industrial sections, and not all of them are good choices for both G2V and V2G services. The required reserve also depends on the decision makers' risk perspective. The presented research has the potential to be extended in different directions. For example, the mobility of EVs can be considered to investigate the influence of transportation patterns on capacity availability of the aggregators. Besides, the model can be modified for distributed charging cases in which EV owners have more freedom and can individually decide about their charging and discharging schedules. Such research topics will be investigated in the future work.

849

References [1] Lutsey N. Global milestone: the first million electric vehicles. 2015. Retrieved March 15, 2016, http://www.theicct.org/blogs/staff/global-milestone-firstmillion-electric-vehicles. [2] Wilson L. The EV wedge: how electric vehicle fuel savings vary by country (and car)?. 2015. Retrieved March 15, 2016, http://shrinkthatfootprint.com/ electric-vehicle-fuel-savings. [3] Yang Z, Li K, Foley A. Computational scheduling methods for integrating plugin electric vehicles with power systems: a review. Renew Sustain Energy Rev 2015;51:396e416.
voa A, [4] Cardoso G, Stadler M, Bozchalui MC, Sharma R, Marnay C, Barbosa-Po et al. Optimal investment and scheduling of distributed energy resources with uncertainty in electric vehicle driving schedules. Energy 2014;64:17e30. [5] Luo Y, Zhu T, Wan S, Zhang S, Li K. Optimal charging scheduling for large-scale EV (electric vehicle) deployment based on the interaction of the smart-grid and intelligent-transport systems. Energy 2016;97:359e68. [6] Luo Z, Hu Z, Song Y, Xu Z, Lu H. Optimal coordination of plug-in electric vehicles in power grids with cost-benefit analysisdpart I: enabling techniques. Power Syst IEEE Trans 2013;28(4):3546e55. [7] Alonso M, Amaris H, Germain JG, Galan JM. Optimal charging scheduling of electric vehicles in smart grids by Heuristic algorithms. Energies 2014;7(4): 2449e75. [8] Mets K, D'hulst R, Develder C. Comparison of intelligent charging algorithms for electric vehicles to reduce peak load and demand variability in a distribution grid. Commun Netw J 2012;14(6):672e81. [9] Musio M, Lombardi P, Damiano A. Vehicles to grid (V2G) concept applied to a virtual power plant structure. In: Proceedings of XIX international conference on electrical machines - ICEM 2010, Rome, Italy; 2010. [10] White C, Zhang K. Using vehicle-to-grid technology for frequency regulation and peak-load reduction. J Power Sources 2011;196(8):3972e80. [11] Honarmand M, Zakariazadeh A, Jadid S. Optimal scheduling of electric vehicles in an intelligent parking lot considering vehicle-to-grid concept and battery condition. Energy 2014;65:572e9. [12] Shao S, Pipattanasomporn M, Rahman S. Grid integration of electric vehicles and demand response with customer choice. IEEE Trans Smart Grid 2012;3(1):543e50. [13] Pipattanasomporn M, Kuzlu M, Rahman S. An algorithm for intelligent home energy management and demand response analysis. IEEE Trans Smart Grid 2012;3(4):2166e73. [14] Kathan D, Aldina R, Lee MP, Medearis L, Sporborg P, Tita M, et al. Assessment of demand response and advanced metering. 2012 [Federal Energy Regulatory Commission Report]. [15] Cao Y, Tang S, Li C, Zhang P, Tan Y, Zhang Z, et al. An optimized EV charging model considering TOU price and SOC curve. IEEE Trans Smart Grid 2012;3(1): 388e93. [16] Sarker MR, Ortega-Vazquez MA, Kirschen DS. Optimal coordination and scheduling of demand response via monetary incentives. IEEE Trans Smart Grid 2015;6(3):1341e52. [17] Fan Z. A distributed demand response algorithm and its application to PHEV charging in smart grids. IEEE Trans Smart Grid 2012;3(3):1280e90. [18] Zhao J, Kucuksari S, Mazhari E, Son YJ. Integrated analysis of high-penetration PV and PHEV with energy storage and demand response. Appl Energy 2013;112:35e51. [19] Zakariazadeh A, Jadid S, Siano P. Integrated operation of electric vehicles and renewable generation in a smart distribution system. Energy Convers Manag 2015;89:99e110. [20] Sioshansi R. OR Forum-Modeling the impacts of electricity tariffs on plug-in hybrid electric vehicle charging, costs, and emissions. Oper Res 2012;60(3): 506e16. [21] Habib S, Kamran M, Rashid U. Impact analysis of vehicle-to-grid technology and charging strategies of electric vehicles on distribution networksea review. J Power Sources 2015;277:205e14. [22] Yilmaz M, Krein PT. Review of the impact of vehicle-to-grid technologies on distribution systems and utility interfaces. IEEE Trans Power Electron 2013;28(12):5673e89. [23] Khazali A, Kalantar M. A stochasticeprobabilistic energy and reserve market clearing scheme for smart power systems with plug-in electrical vehicles. Energy Convers Manag 2015;105:1046e58. [24] Sortomme E, El-Sharkawi MA. Optimal combined bidding of vehicle-to-grid ancillary services. IEEE Trans Smart Grid 2012;3(1):70e9. [25] Damiano A, Gatto G, Marongiu I, Porru M, Serpi A. Vehicle-to-Grid technology: state-of-the-art and future scenarios. J Energy Power Eng 2014;8(1):152. [26] Sortomme E, El-Sharkawi MA. Optimal scheduling of vehicle-to-grid energy and ancillary services. IEEE Trans Smart Grid 2012;3(1):351e9. [27] Kempton W, Udo V, Huber K, Komara K, Letendre S, Baker S, et al. A test of vehicle-to-grid (V2G) for energy storage and frequency regulation in the PJM system. Results Ind Univ Res Partnersh 2008;32. [28] Khalid MW, Al-Awami AT, Sortomme E. Stochastic-programming-based bidding strategy for V2G services. In: Innovative smart grid technologies Europe (ISGT EUROPE), 2013 4th IEEE/PES; 2013. p. 1e5. [29] Kempton W, Tomi
c J. Vehicle-to-grid power fundamentals: calculating capacity and net revenue. J Power Sources 2005;144(1):268e79. [30] Bai X, Qiao W, Wei H, Huang F, Chen Y. Bidirectional coordinating dispatch of

850

[31]

[32]

[33]

[34] [35]

[36]

[37]

[38] [39] [40]

[41]

[42]

[43]

[44]

N. Nezamoddini, Y. Wang / Energy 116 (2016) 836e850 large-scale V2G in a future smart grid using complementarity optimization. Int J Electr Power & Energy Syst 2015;68:269e77. Liang H, Choi BJ, Zhuang W, Shen X. Optimizing the energy delivery via V2G systems based on stochastic inventory theory. IEEE Trans Smart Grid 2013;4(4):2230e43. Singh M, Kumar P, Kar I. A multi charging station for electric vehicles and its utilization for load management and the grid support. IEEE Trans Smart Grid 2013;4(2):1026e37. Bishop JD, Axon CJ, Bonilla D, Tran M, Banister D, McCulloch MD. Evaluating the impact of V2G services on the degradation of batteries in PHEV and EV. Appl Energy 2013;111:206e18. Han S, Han S, Sezaki K. Development of an optimal vehicle-to-grid aggregator for frequency regulation. IEEE Trans Smart Grid 2010;1(1):65e72. Dallinger D, Krampe D, Wietschel M. Vehicle-to-grid regulation reserves based on a dynamic simulation of mobility behavior. IEEE Trans Smart Grid 2011;2(2):302e13. Ansari M, Al-Awami AT, Sortomme E, Abidoeric MA. Coordinated bidding of ancillary services for vehicle-to-grid using fuzzy optimization. Smart Grid IEEE Trans 2015;6(1):261e70. Zeng W, Gibeau J, Chow MY. Economic benefits of plug-in electric vehicles using V2G for grid performance-based regulation service. In: Industrial electronics society, IECON 2015-41st annual conference of the IEEE; 2015. p. 004322e7. Jargstorf J, Wickert M. Offer of secondary reserve with a pool of electric vehicles on the German market. Energy Policy 2013;62:185e95. Pavi
c I, Capuder T, Kuzle I. Value of flexible electric vehicles in providing spinning reserve services. Appl Energy 2015;157:60e74. Rahmani-andebili M. Spinning reserve supply with presence of electric vehicles aggregator considering compromise between cost and reliability. Gener Transm Distrib IET 2013;7(12):1442e52. Liu L, Kong F, Liu X, Peng Y, Wang Q. A review on electric vehicles interacting with renewable energy in smart grid. Renew Sustain Energy Rev 2015;51: 648e61. Honarmand M, Zakariazadeh A, Jadid S. Integrated scheduling of renewable generation and electric vehicles parking lot in a smart microgrid. Energy Convers Manag 2014;86:745e55. Khodayar ME, Wu L, Shahidehpour M. Hourly coordination of electric vehicle operation and volatile wind power generation in SCUC. IEEE Trans Smart Grid 2012;3(3):1271e9. Huang Q, Jia QS, Qiu Z, Guan X, Deconinck G. Matching EV charging load with uncertain wind: a simulation-based policy improvement approach. Smart Grid IEEE Trans 2015;6(3):1425e33.

[45] Hashemi-Dezaki H, Hamzeh M, Askarian-Abyaneh H, Haeri-Khiavi H. Risk management of smart grids based on managed charging of PHEVs and vehicle-to-grid strategy using Monte Carlo simulation. Energy Convers Manag 2015;100:262e76. [46] Zhang N, Hu Z, Han X, Zhang J, Zhou Y. A fuzzy chance-constrained program for unit commitment problem considering demand response, electric vehicle and wind power. Int J Electr Power & Energy Syst 2015;65:201e9. [47] Goli P, Shireen W. Control and management of PV integrated charging facilities for PEVs. In: Plug-in electric vehicles in smart grids. Singapore: Springer; 2015. p. 23e53. €rges D, Liu S. Energy management in smart grids with electric vehicles [48] Go based on pricing. IFAC Proc Vol 2013;46(27):182e9. [49] Su W, Wang J, Hu Z. Planning, control, and management strategies for parking lots for PEVs. In: Plug-in electric vehicles in smart grids. Singapore: Springer; 2015. p. 61e98. [50] Wang Y, Li L. Time-of-use based electricity demand response for sustainable manufacturing systems. Energy 2013;63:233e44. [51] Wang Y, Li L. Time-of-use electricity pricing for industrial customers: a survey of US utilities. Appl Energy 2015;149:89e103. [52] Sun J, Tesfatsion L. DC optimal power flow formulation and solution using QuadProgJ. In: Proceedings of the IEEE power and energy society general meeting, tampa, Florida; 2007. [53] Howarth D, Monsen B. Renewable energy faces daytime curtailment in California. 2015. Available at: http://nawindpower.com/online/issues/NAW1412/ FEAT_04_Renewable-Energy-Faces-Daytime-Curtailment-In-California.html. [54] Chen W, Sim M, Sun J, Teo C. From CVaR to uncertainty set: implications in joint chance-constrained optimization. Oper Res 2010;58(2):470e85. [55] Wood AJ, Wollenberg BF. Power generation, operation, and control. John Wiley & Sons; 2012. [56] Zhang K, Xu L, Ouyang M, Wang H, Lu L, Li J, et al. Optimal decentralized valley-filling charging strategy for electric vehicles. Energy Convers Manag 2014;78:537e50. [57] CAISO renewable watch data: Retrieved April 20 2016. http://www.caiso.com/ market/Pages/ReportsBulletins/DailyRenewablesWatch.aspx. € we-Kuska N, Heitsch H, Ro €misch W. Scenario reduction and scenario tree [58] Gro construction for power management problems. In: Proceedings of IEEE power tech conference, Bologna, Italy; 2003. [59] Hruska J. Europe's solar power deployments sailed through feared eclipse with flying colors. 2015. Retrieved March 12 2016, http://www.extremetech. com/extreme/201890-europes-solar-power-deployments-sailed-throughfeared-eclipse-with-flying-colors.