Electric Power Systems Research 174 (2019) 105847

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Stochastic robust optimization for smart grid considering various arbitrage T

opportunities
Mohammadali Saffaria, Mohammad Saeed Misaghiana, Mohsen Kiab, Alireza Heidaric,
⁎
Daming Zhangc, Payman Dehghaniand, Jamshid Aghaeie,
a
Department of Electrical Engineering, Shahid Beheshti University, Tehran, Iran
b
Department of Electrical Engineering, Faculty of Eng., Pardis Branch, Islamic Azad University, Pardis, Tehran, Iran
c
School of Electrical Engineering and Telecommunications, The University of New South Wales (UNSW), Sydney, Australia
d
Department of Electrical and Computer Engineering, George Washington University, Washington, DC, USA
e
Department of Electrical and Electronics Engineering, Shiraz University of Technology, Shiraz, Iran

A R T I C LE I N FO A B S T R A C T

Keywords: Because of power electronic advancements, nowadays Plug-in Electric Vehicles (PEVs) are capable to charge/
Robust/stochastic optimization discharge(absorb/inject) active(reactive) power. The storage capacity and the capability of bidirectional flowing
Microgrid active and reactive powers, lead PEVs to be contemplated as a viable option for energy arbitrage. By high
Plug-in electric vehicles penetration of PEVs, decentralized energy management of them, especially at peak hours causes serious pro-
Mix-integer non-linear programming
blems to the network operation and service quality. Therefore, it is important PEVs to be controlled as integrated
with microgrid (MG). Besides, PEVs uncertain behavior along with the prevalent uncertainties inherent to re-
newable resources and various pricing mechanism in electricity industry leaves the MG operators (MGOs) a
challenging decision making. So, it is vital to be applied an efficient strategy in order to deal with this challenges.
This paper proposes a centralized framework to co-optimize robust/stochastic optimization of MG with the PEVs
energy arbitrage in both active and reactive powers exchange. The problem is a mix-integer non-linear pro-
gramming (MINLP) problem, which is solved by GAMS software. The results of suggested model are investigated
on IEEE 18-bus and IEEE 33-bus test systems.

1. Introduction
Because of low emission and energy consumption, PEVs are con-
templated as efficient alternative to internal-combustion-engine.
Recently, there are drastically growth to use them as next generation of
vehicle. Therefore, governments and energy corporations have centered
their attempts to improve the development of PEVs. One of the bene-
ficial merits of PEVs is the energy arbitrage opportunity which moti-
vates their owners to minimize cost of energy. In Ref. [1] is investigated
the arbitrage strategy of PEVs owner using stochastic optimization to
estimate the potential profit from electricity price arbitrage of PEVs
under three scenarios with variant electricity tariff. Reference [2]
evaluates PEVs utilizing vehicle-to-grid (V2G) technology to behave as
a storage system, arbitraging in the energy market and providing an-
cillary services. Aforementioned references consider the arbitrage
strategy from PEVs viewpoint which are seeking to maximize their
profit and have not considered effects of PEVs energy arbitrage on
network. High penetration of PEVs without considering efficient control
strategy, brings challenges for their integration and energy manage-
ment and it can make side effects on MG e.g., overload of lines or
voltage drop [3]. In order to manage adverse effects of PEVs, it is es-
sential energy of PEVs to be controlled as integrated with MG. In Ref.
[4], a multistage droop-control mechanism has been suggested for PEVs
integrated islanded MGs. The droop characteristics of the distributed
generation (DG) units, load shedding of MG, and the charging and
discharging behavior of PEVs have been coordinated through the pro-
posed control mechanism in Ref. [4]. Reference [5] presents a hier-
archical stochastic control scheme for the coordination of PEVs char-
ging and wind power in a MG. In reference [6], a multi-objective
optimization problem is solved to schedule power sources in a typical
MG, while PEVs are viewed as a stochastic factor. Reference [7] pro-
poses two energy management strategies to effectively utilize V2G
potential of PEVs in managing energy imbalances in grid-connected
MGs. All this works have considered only charging/discharging of ac-
tive power by PEVs. In recent years, power electronic advancements
give PEVs the capability of charge/discharge active power and absorb/

⁎
Corresponding author.
E-mail address: aghaei@sutech.ac.ir (J. Aghaei).

https://doi.org/10.1016/j.epsr.2019.04.025
Received 27 September 2018; Received in revised form 26 February 2019; Accepted 23 April 2019
Available online 09 May 2019
0378-7796/ © 2019 Elsevier B.V. All rights reserved.
M. Saffari, et al.
inject reactive power simultaneously [8]. There are works that con-
template PEVs reactive power in MG operation. Reference [9] presents
a framework to manage energy in the smart MGs in the presence PEVs
charging facilities in controlling active and reactive powers. In Ref. [10]
is developed an energy management system (EMS) that is able to co-
ordinate voltage control devices, PVs, PEV aggregators, and dispatch-
able distributed generations (DDGs) in which active and reactive power
provision of PEVs along with voltage control devices lessen the plau-
sible violations. In refs. [9,10] the PEVs owners costs have not been
included in the objective function with the assumption that the MG
operator plans the daily charging/discharging scheduling of PEVs bat-
teries. This procedure has been executed during a mid-term decision
making of PEV owners and MGO. Accordingly, the both players ac-
complish their own cost-benefit analysis and negotiate on the agree-
ment terms and conditions. Therefore, this refs do not give any in-
formation about the cost of using PEVs active and reactive power.
Besides, they did not consider the uncertainty of PEVs in MG operation.
Owning to increasing use of renewable energy resources, transac-
tion in power market and also high penetration of PEVs, MG operation
is encountered with high percentage of uncertainties that must be
managed. There are papers have applied stochastic and robust opti-
mizations in order to deal with uncertainties in MG operation.
Reference [11] proposes a stochastic framework of MG operation with
consideration of uncertainty for participating in power market. Re-
ference [12] investigates optimal operation of MG in which uncertainty
of market price is captured by robust optimization. A scenario-based
robust energy management method accounting for the worst-case
amount of renewable generation and load is developed in Ref. [13].
MGO generally seeks the most optimal and cost-effective solution
for its MG operation, while satisfying the requisite network and security
constraints [16]. In MGs, the required energy is likely to be provided by
local sources and DGs. Transaction in power markets is another option
for MGO supply demand or even to benefit. Moreover, the capabilities
of PEVs to enhance power quality issues in the MGs can be activated by
means of the bidirectional converters. Accordingly, simultaneous
management of active power exchange and reactive power support of
the MG has been achieved for aggregators using the bidirectional
converters [8]. Integration of PEVs energy management to MG without
control strategy causes problems to MG operation and security [3].
This paper proposes a centralized strategy to integrate the optimi-
zation of smart MG with the energy arbitrage management of PEVs for
both the active and reactive powers. To do this, a stochastic/robust
optimization is considered to co-optimize MG operation and PEVs en-
ergy management. In this strategy, PEVs submit their preferences about
charging/discharging and absorbing/injecting to MGO. Then, MGO
generates a schedule for PEVs by optimizing the integrated MG op-
eration and PEVs energy management without violating preferences of
PEVs and the network constraints. In the PEVs model is considered “lost
opportunity cost” which refers to the state in which using reactive
power leads to losing the opportunity of using active power. The lost
revenue of the PEVs due to the reduced capacity of active power is
termed lost opportunity cost.
The major contributions of this work are summarized as follows:
• Proposing a centralized strategy in order to integration of PEVs
energy management with MG operation in which is co-optimized
the robust/stochastic optimization of MG and energy arbitrage of
PEVs. In this strategy, MGO seeks to minimize its cost with con-
sidering security constraints and is allowed to manage PEVs while
fulfils their expectations.
• Considering a model of PEVs in which they charge/discharge(ab-
sorb/inject) active(reactive) powers and submit active and reactive
bids. Furthermore, the lost opportunity cost of using PEVs reactive
power is considered.
• Investigating energy arbitrage of PEVs.
• Scrutinizing of the impact of risk management on cost(profit) using
2
Electric Power Systems Research 174 (2019) 105847
robust optimization.
The rest of the paper is organized as follows. Section 2 presents the
proposed model and mathematical formulations. Numerical results and
discussions are elaborated in Section 3. And finally comes the conclu-
sions in Section 4.
2. Modeling and problem formulation
2.1. Operational framework
In the proposed model, the uncertainties of the MG generating units,
PEVs arrival and departure times as well as that involved in both DA
and RT electricity market prices are captured and effectively taken into
account through a combination of stochastic and robust programming
techniques. As RT Market Price (RT MP) is unpredictable, its un-
certainty is handled via robust programming, while the uncertainty
associated with other parameters is contemplated via stochastic pro-
gramming. Robust programming assists MGO to effectively control the
risk level of its participation in RT market, where the risk-averse or risk-
conservative solutions and strategies can be approached based on the
MGO preference.
PEVs can play an important role in some MGs, most of which have
individual owners and behave stochastically. In this paper, it is assumed
that (a) there are some PEVs (parking lots) with random behavior in
their arrival and departure time; (b) PEVs can absorb/inject reactive
power that makes it a source for MGO to supply its reactive demands;
and (c) PEVs have individual owners that can have contractual trans-
actions with the MGO: paying for a charge and being paid for a dis-
charge. In addition, utilizing the capability of PEVs for absorbing/in-
jecting reactive power imposes some financial costs to MGOs since PEVs
have individual owners and, hence, the cost of using PEV reactive
power is driven by the owners. Accordingly, each PEV can bid for ac-
tive/reactive power. Under such scenarios, the MGO is responsible to
effectively utilize its units, PEVs, and the transactions in DA and RT
markets to optimize the cost and maximize the earned benefits. The
overall operational framework and the communication required be-
tween various agents is depicted in Fig. 1.
In this centralized energy management framework, first, required
data are provided by different agents and send to decision making en-
tity through communication links. Second, the stochastic/robust opti-
mization is applied by decision making entity, MGO, to find the optimal
planning. Third, optimal decisions are sent to agents. This decisions are
consist of (a) optimal scheduling of MG’s units; (b) optimal scheduling
of PEVs active and reactive power; and (c) optimal bids for DA and RT
markets.
In this strategy, the MG operation and PEVs energy management are
integrated which leads, effects of PEVs on operation and security of MG
to be controlled. Moreover, it gives the MGO the ability to manage PEVs
capacity in both active and reactive powers while considering PEVs
expectations. In addition, in PEVs bidding model is considered lost
opportunity cost. Considering lost opportunity cost causes win-win si-
tuation for both MGO and PEVs owners. Indeed, from one side, it is
given to MGO the opportunity to decide about optimal allocation of
PEVs capacity to active and reactive powers. For example, it is possible
MGO sacrifices the benefit of active power arbitrage to retain security
of network by allocating most or even all of PEVs capacity to injection
reactive power. From another side, because is given the lost opportunity
cost to PEVs owners, they don’t lose their benefit due to losing active
power arbitrage opportunity and are motivated to allow MGO to allo-
cate PEVs capacity to reactive power.
2.2. MG assets
An MG is typically composed of Dispatchable DGs (DDGs), WTs,
PVs, and ESSs. The total cost of providing power by DDGs are termed as
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Fig. 1. Overall centralized operational framework.

Cos t DDG and is indicated in (1) [14]. It is consist of cost of production of parking lots; (c) PEVs owners submit their active and reactive bids and
active power and inject/absorb reactive power. The DDGs output power is allowed MGO to manage PEVs capacity while fulfils PEVs expectation
is restricted by the minimum/maximum limits enforced in (2) [15]. The according to bids; and (d) PEVs have to be charged up to their expected
ESS degradation cost is contemplated and termed as Cos t ESS which is State-Of-Charge (SOC) at their departure time.
indicated in (3) [16]. ESSs constraints are presented in (4)–(5) [17]. Moreover, the underlying uncertainty in the PEV arrival and de-
The WT and PV models are represented in (6)–(7) [18]. parture times is captured by Normal probability distribution functions
[19–21]. Reasonably, each parking lots is modeled with an equivalent
Cos tiDDG 2 2
, kst = Ai (Pi, kst ) + Bi Pi, kst + Ci + A′i |Qi, kst | + B′i |Qi, kst | + C′i
PEV representing the behavior of all PEVs together. Since the charging/
(1) discharging power of several of PEVs is taken into account at the ag-
gregator level, this assumption seems to be adequate and plausible
Pimin ≤ Pi, kst ≤ Pimax
[9,19].
Qimin ≤ Qi, kst ≤ Qimax (2) Utilizing power electronic equipment makes PEVs capable of char-
ging/discharging (absorbing/injecting) active (reactive) power [22].
Cos teESS depreciation
, kst = ce
C
(Pe,kst D
+ Pe,kst ) (3) Fig. 2(a) illustrates the apparent power curve of a PEV, which is re-
C stricted to three curves.X1 and X2 represent the PEV maximum charge/
0 ≤ Pe,kst ≤ PeC,max
discharge and X3 represents the PEV maximum apparent power. Any
D
0 ≤ Pe,kst ≤ PeD,max (4) point in the PEV capability curve demonstrates its operating point that
composes of P(charge/discharge) and Q(absorb/inject): P and Q can
1 D
SOCe,kst=SOCe,ks,t−1 + ηeC Pe,kst
C
Δt − Pe,kst Δt increase/decrease in such a way that it does not intersect the curves X1,
ηeD
X2, and X3. Take point A(PM,Qb) as an example; it can be seen in
SOCemin ≤ SOCe,kst ≤ SOCemax (5) Fig. 2(b) that the PEV can rise its reactive power from Qb to QM, while
its active power (PM) is remained constant. Note that the PEV owners
⎧ 0 if vwt , st ≤ vciwind or vwt , st ≥ vcowind are paid the operational cost for injecting and/or absorbing reactive
⎪ (vw, st − vciwind ) power in region I and region III, respectively.
Pwt , st = Pwt , r . if vciwind ≤ vwt , st ≥ vrwind However, if additional reactive power is required at point M
⎨ (vrwind − vciwind )
⎪ (Q > QM), its active power (P) must decrease: when Q grows from QM
⎩ Pwt , r otherwise (6)
to QN, the PEV operating point changes to PN (PN < PM). Indeed, the
where, P wt , r
is the WT nominal power and and vwt , st , vciwind , vrwind
are vcowind operating point must be moving back along the curve to point N
the actual, cut-in, nominal, and cut-out speed of wind, respectively. (PN,QN). In other words, the PEV must dwindle its active power P to
where, η pv is the efficiency of solar panel conversion, Spv is the area of adhere to its restrictions at times that more reactive power is required.
solar panel, T st is the ambient temperature, and is the energy of sun The reduction in PEV revenue due to the shrunk range of P is called as
radiation. Based on scenarios of the wind speed(vwt , st ) and the energy of lost opportunity cost, meaning the payment for active power is much
sun radiation(), outputs power of wind turbine and PV units are cal- higher than that for reactive power [23]. Therefore, while PEVs are
culated. utilized in regions III and IV, PEV owners are paid the lost opportunity
cost in addition to the operational costs. Note that apart from the re-
2.3. PEV gions that PEV operates in, the availability cost is paid to its owner,
when its reactive power is utilized. Overall, the regions defined for the
In this paper, the presence of PEVs in MG is scrutinized as an ar- PEV reactive power statuses are as follows.
bitrage opportunity for MGO. The following assumptions are con-
sidered here: (a) PEVs are not MG assets and they have individual • Regions I and II: Availability and operational costs are paid to PEV
owners; (b) PEVs are reachable only at times that are parked in the owners based on their bidding values.

3
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Fig. 2. PEV features (a) PEV capability curve (b) PEV Reactive Power Bidding (PRB) curve.

• Regions III and IV: Availability, operational, and lost opportunity C

0 ≤ Pυ,kst ≤ βυ,kst PυC,max
costs are paid to PEV owners based on their bidding values. D
0 ≤ Pυ,kst ≤ (1 − βυ,kst ) PυD,max
C D
PEV Reactive power Bidding (PRB) curve is given in Fig. 2(b). Based P υ,kst = Pυ,kst + Pυ,kst (11)
on the reactive power Q of a PEV (horizontal axis), the integrated area
is assessed as the cost that must be paid to PEV owners. ξ υavalability , (P υ,kst ) 2+ (Qυ,kst ) 2 ≤ (Sυnominal )2 (12)
ξ υoperation and ξ υopportunity are biddings of PEVs for reactive power Q in
1 D
different operation regions, where ξ υavalability stands for availability price Ev, kst = Ev, ks, t − 1 + ηυC Pυ,kst
C
Δt− Pυ,kst Δt
ηυD (13)
of PEV v for absorbing/injecting reactive power; ξ υoperation represents the
operational price for using PEV v in all regions; and ξ υopportunity represents
the opportunity price of PEV v in regions III and IV. In general, the PEV Eυmin ≤ Eυ,kst ≤ Eυmax
cost is formulated in (8), where PRB function is the first term in which Eυ,kst = συEXP Eυmax ∀ t=t υdeparture (14)
αr, υ, r ∈ {I , II , III , IV } are binary variables indicating the operational
regions of PEV v . First line indicates the cost of PEVs availability for
injecting/absorbing reactive power, second line represents the cost of 2.4. Transactions of DA and RT markets
PEVs injecting/absorbing reactive power and lost opportunity cost is
demonstrated in third line. The second term in (8) indicates cost of MGO’s bidding in DA and RT electricity markets is constrained by
RT
PEVs active power charging/discharging. QI,υ , QII,υ , QIII , υ and QIV,υ in (15), where PktDA and Pkst can be positive/negative depending on the
Fig. 2 represent the operation regions (Q b′ to Q M′), (Qb to QM ) , buying/selling action in the markets. Overall, the cost and revenue of
(QM to QN ) , and (Q M′ to Q N ′) , respectively. Eqs. (9) and (10) demon- transactions is presented as Cost DA market in (16) and Cost RT market in (17)
strate the region of operation and reactive power which is injected/ for participating in DA and RT electricity markets, respectively, where
absorbed by PEV in that region, respectively. Constraints for PEV the positive/negative values represent cost/revenue. The deviation
charging/discharging (absorbing/injecting) active (reactive) power are from the RT MP is maximized in the second term in (17) to achieve the
denoted in (11)–(12). Energy balance of PEV storage is enforced in (13). maximum robustness accounting for the decision uncertainties and the
The limits on battery capacity and the expected energy of PEVs at their risk level adopted by the MGO with the corresponding constraints en-
departure time are accounted for in (14). forced in (18)–(19). Constraint (18) defines a range for RT MP fluc-
RT
tuations(Uncertainty Set) where δkst represents deviation from λkst . Γks
PEV
Cos tk,s in (19) is an integer control parameter for the MG risk level in RT
Nt Nυ market. If Γks = 0 , the uncertainty of the RT market price can be ne-
= ∑ ∑ ⎡⎢ξυavalability α0,υ,kst + ∑ ξ υoperation αr,υ,kst Qυ,kst
2
+ glected; nevertheless, if Γks = |Lks |, the uncertainty of the RT market
t=1 υ=1 ⎣ r ∈ {I , II , III , IV } price would be entirely accounted for leading to the most conservative
solution eventually [24].
2 ⎤
∑ ξ υopportunity αr′,υ,kst Qυ,kst ⎥
r ′∈ {III , IV } ⎦ |PktDA| ≤ P DA,max
RT
Nt Nυ |Pkst | ≤ P RT ,max (15)
+ ∑ ∑ [=γυD Pυ,kst
D
− γυC Pυ,kst
C
]
t= υ=1 (8) Nt
Cos tkDAmarket = ∑ λktDA PktDA
∑ αr,υ,kst ≤ α 0,υ,kst t=1 (16)
r ∈ {I , II , III , IV } (9)
RT RT RT
λkst ∈ [λkst − δkst , λkst + δkst ] , δkst ≥ 0 (18)
∑ Qr,υ,kst = Qυ,kst
r ∈ {I , II , III , IV } (10) Γks ∈ [0, |Lks |] , Lks = {(kst ) | δks ≥ 0} (19)

4
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

2.5. Objective function

The objective function is formulated in (20).

Obj
Nk Ns
⎛
=min ∑ πk ⎜ Cos tkDAmarket + ∑ πs ( Cos tksDDG +CostksESS +CostksPEV
k=1 ⎝ s=1

⎞
+CostksRTmarket ) ⎟
⎠ (20)
The following AC power flow constraints are enforced in (21)–(23).
Ni Nυ
∑ Pn,i,kst + Pn,wt,kst + Pn,pv,kst + Pn,kt
DA RT
+ Pn,kst + ∑ P υ,kst
i=1 υ=1
Ne
+ ∑ (Pn,e,kst
D C
− Pn,e,kst ) − DnP, t
e=1
Nm
= ∑ Vn,kst Vm,kst Ynm. cos (θn,kst − θm,kst − φnm)
m=1 (21)
Ni Nυ Nm
∑ Qn,i,kst + ∑ Qn,υ,kst − Dn,tQ = ∑ Vn,kst Vm,kst Ynm.
i=1 υ=1 m=1 Fig. 4. Modified 18-bus IEEE test system.
sin(θn,kst − θm,kst − φnm) (22)
2 2 max 2
Pnm , kst + Qnm, kst ≤ (Snm )

Vmin
n ≤ Vn,kst ≤ Vmax
n (23)
Proposed model is a combined stochastic/robust optimization
method in order to optimize MG operation considering PEVs bids with

Fig. 3. Framework of the proposed model.

5
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Table 1
Parameters of ESS.

SOCemin (kWh) SOCemax (kWh) PeC,max (kW) PeD,max (kW) η

0 1500 350 350 0.9

Table 2
Parameters of parking lots.
DDG Pmin(kW) Pmax(kW) Qmin(kVAr) Qmax(kVAr)

1 500 5000 0 450 Fig. 5. Five wind power output scenario.

2 300 3000 0 450
3 300 3000 0 580
4 100 2000 0 300
5 100 2000 0 220
6 100 2000 0 220

the aim of cost minimization. Fig. 3 represents whole framework of the

proposed model optimization process.

3. Case studies1

A modified 18-bus IEEE test system [14] is employed as the case

study to demonstrate the effectiveness of the suggested framework. To Fig. 6. Five PV power output scenario.
transfer power between grid and PEVs, there must be a suitable com-
munication infrastructure in the MG, which can be found in a smart
grid. Fig. 4 indicates modified 18-bus IEEE test system. The MINLP
optimization problem is formulated in GAMS environment [25] and is
solved using DICOPT solver.
Parameters of ESS, DDGs and parking lots are shown in Tables 1–3
[26] respectively.
In order to generate a meaningful set of possible scenarios, Latin
Hypercube Sampling (LHS) [11] technique is applied following by the
Kantorovich distance [27] method as an effective mean for scenario
reduction considering possible correlations. In this paper is modeled
uncertainty in arrival and departure time of PEV with Normal Dis-
tribution Function(NDF). It has been considered parking lots in dif-
ferent buses of network. Each parking lots has different NDF. Based on
NDF, scenarios of arrival and departure of PEVs in each parking lots are
determined by PEV aggregator that indicate PEV’s uncertainty beha-
vior. The 600 kW wind turbine model is adopted from [28]. Based on
the wind speed forecast result, wind speed scenarios are generated and
the corresponding wind generation power outputs scenarios are calcu-
lated based on Eq. (6). Fig. 5 shows five selected scenario for wind
power outputs. The 1550 kW solar farm is considered which consist of
five 310 kW PV model from [29]. Scenarios of solar irradiance and
temperature are generated and the corresponding PV output power is
calculated based on Eq. (7). Fig. 6 illustrates five selected scenarios for
PV output. Uncertainty in solar irradiance, temperature and wind speed
are modeled by NDF. Wind speed, solar irradiance and air temperature Fig. 7. Market Prices and MG demands. (a) DA, RT, and MEV prices; (b) Active
data are from [30]. and reactive power of MG.
The DA/RT electricity market prices, and their Mean Value (MEV)
for one selected scenario are illustrated in Fig. 7(a). Active/Reactive demand in the studied MG is also depicted in Fig. 7(b). In this paper, the
consumers are considered as price taker loads and inelastic to avoid the
Table 3 unnecessary intricacies. So, the active and reactive loads are con-
Parameters of parking lots. templated as non-deferrable and non-interruptible demands that must
Parking lots Number of Aggregate Aggregate maximum always be served.
PEVs capacity(kWh) charging/discharging power
(kW)
3.1. Analysis impact of components on MG operation cost
1 50 1640 330
2 40 1440 290 In order to evaluate the effect of each component on operation cost,
3 60 1700 310 five distinct cases are contemplated. Case1 is the normal operation of
4 40 1400 310
MG and it is considered as the base case scenario. Case2 is the scenario
5 50 1600 330
6 60 1700 300 representing a MG operation in an island mode, where no participation
in DA and RT markets is considered. Case3 is the case without ESS; in

6
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Table 4 In case1, 20% of total reactive demand is supplied by PVEs at 24 h

Expected costs in different cases. and the rest of it is served by DDGs. As it is seen in Fig. 8(a), at peak
Expected cost ($) Cases load hours, PEVs contribution to inject reactive power increases. This is
because at peak hours energy price is high, so it is financially efficient
Case1 Case2 Case3 Case4 Case5 more fraction of reactive demand to be supplied by PEVs, however it is
possible MGO to be forced to pay lost opportunity cost to PEVs owners.
Active power by DDGs 21653 33619 21695 24196 23655
Reactive power by DDGs 5623 5245 5684 7528 7831 Fig. 6(b) and (c), illustrate active load and demand. Fig. 8(b) in-
Transaction in RT market −410 0 −437 −419 −437 dicates contribution of DDGs and renewable energy resources to supply
Transaction in DA market 7101 0 7230 5281 5424 demand, also, in Fig. 8(c) contribution of EES, PEVs, DA and RT market
Active power by PEVs −1278 −1342 −1282 −1172 0 is represented.
Reactive power by PEVs 1460 1795 1962 0 0
In case2, MG is operated in island mode. According to Table 4,
Active power by ESS 29 31 0 29 29
Total expected cost ($) 34,746 39348 34852 35443 36502 Case2 has most effect on MG operation cost in comparison with other
Cost increment (%) 0 11.6 0.3 1.96 4.81 cases. In this case, DDGs active power generation goes up because they
are viable alternatives for market transactions. DDGs active power
generation increases 24% in comparison with Case1.
MG revenue from PEVs active power increase. This is because in
grid-connected mode, MGO has this opportunity to accept PEVs active
bids and sells them with higher price in markets to get benefit. To do
this MG must discharge PEVs active power and will pay them according
to their expectations. But in island mode, MG does not have this op-
portunity, so discharging PEVs active power dwindles. In case2, amount
of charging/discharging active power by PEVs diminishes about 24% in
comparison with Case1. This situation provides this opportunity for
MGO to allocate more PEVs capacity to inject reactive power to supply
reactive demand. PEVs contribution in reactive power injection in-
creases from 20% in Case1 to 22.5% in Case2. It is seen in Table 1 that
DDGs reactive power cost reduces in Case2 in comparison with case1
because of more PEVs contribution to supply reactive demand. Fig. 9
represents contribution of all components in optimal operation of MG in
Case2.
In case3, ESS is neglected, according to Table 4, absent of ESS causes
MG operation cost goes up 0.3% in comparison to case1.

Fig. 8. Contribution of components in Case1. (a) Reactive, (b) and (c) Active.

Case4, PEVs reactive power is ignored and there is no PEV considered in

Case5. The cost/revenue of each MG component, total expected cost in
each case study, and the cost increment in comparison with the base
case scenario is tabulated in Table 4.
Positive/Negative values stand for the cost/revenue. Table 4 de-
monstrates that concurrent existence of DA and RT markets together
with the presence of ESS and PEVs lead to 11.6%, 0.3%, 1.96%, and
4.81% cost savings, respectively due to their contributions in making
arbitrage profits.
Fig. 8 represents contribution of all components in MG optimal
operation for Case1. In Fig. 8(a) reactive power demand and generation
is indicated. Reactive demand can be supplied by DDGs and PEVs.
Note, PPEVs, QPEVs, PDDGs and QDDGs indicate net effect of all DDGs
and parking lots in figures.
Fig. 9. Contribution of components in Case2. (a) Reactive, (b) and (c) Active.

7
M. Saffari, et al.
Fig. 10. Contribution of components in Case4. (a) Reactive, (b) and (c) Active.
Fig. 10 represents contribution of components in optimal operation
of MG in Case4. In this case, PEVs reactive power is neglected. It causes
cost of DDGs reactive power increases, because DDGs must supply all of
reactive demand by themselves. Owning to all of PEVs capacity can be
allocated to charge/discharge of active power, there is this opportunity
for MGO to increase amount of PEVs charging/discharging to get profit.
In case4, amount of charging/discharging active power by PEVs in-
creases 19% in comparison with Case1.
At last, case5 is without PEVs. In this case MG operation cost in-
creases 4.81%.
In order to investigate contribution of ESS, PEVs and markets in
detail, they are analyzed in next sections separately.
3.2. Markets contribution and impact of robust optimization on submitted
bids to RT and DA markets
With MG participation in DA and RT electricity markets, MGO
submits DA and RT bids as depicted in Fig. 11 for one selected scenario.
Take t = 2 as an example, where the MEV is at its lowest value re-
presenting an optimal scenario for MGO to buy electricity from markets
at this specific hour. Observe that, for Γ = 0 , MGO buys 1000 kW from
RT and 2500 kW from DA markets. For Γ = 6, Γ = 10 , and Γ = 24 , the
RT (DA) biding values change to 646.2 (3010.43), 439.86 (3158.44),
and 0 (3500) kW demonstrating a significant reduction (increment) in
RT (DA) biddings. In some hours, it would be optimal for MGO to buy
power from one market and sell it in the other. For instance, at hours
3–5 that the RT MP is higher than the DA MP, the MGO bids for buying
power from DA market and bids for selling it in RT market, it is called
arbitrage opportunity between DA and RT markets. On the contrary at
hours 13–14, the RT MP is lower than the DA MP, creating an
8
Electric Power Systems Research 174 (2019) 105847
opportunity for the MGO to buy a lower-price power form the RT
market and sell it with higher price in DA market. In this studied sce-
nario, there is no optimal opportunity found for MGO to sell power in
both markets simultaneously. Using the robust control parameter en-
sures an appropriate risk level for the MGO to participate in the RT
market. As the MGO considers DA and RT markets simultaneously,
imposing such limitations on RT bids indirectly affects the DA bids of
MG as well. Indeed, according to the desired robustness level selected,
the MGO will have to set a trade-off between the DA and RT bids. For
instance, at t = 21, the RT MP is higher than the DA MP. Observe that,
for Γ = 0 , MGO bids 3500 kW for buying power from DA market and
1000 kW to sell in RT market. In fact, MGO buys 3500 kW from the
cheaper market, of which 2500 kW is utilized for its own consumption
and 1000 kW would be allocated for selling with higher price in RT
market. Since the risk controller is not taken into account here, MGO
submits high-risk bids to the RT market in order to maximize its profit.
On the other hand, for Γ = 24 , when the risk factors are applied, the
MGO only bids 2500 kW for buying power from DA market for its own
consumption, while it does not bid for selling in RT market due to its
conservative preference imposed by the selected robustness level.
To recapitulate, MGO faces with arbitrage opportunities in the
presence of DA and RT markets and it can transact between DA and RT
markets to earn additional revenue.
Fig. 12 illustrates all components contribution to supply active and
reactive demand for Γ = 24 . In comparison with Fig. 8 which indicates
components contribution for Γ = 0 , it is seen that Γ mainly influences
on submitted bids to RT and DA markets. For instance, at t = 24 for
Γ = 0 , MG submits 1177 kW and 1000 kW purchase bids from DA and
RT markets respectively. For Γ = 24 these bids become 2117 kw and
0 kW and also, the contribution of other components will be un-
changed. In fact, in order to have riskless behavior, instead of suppling
a fraction of demand from RT market, MG only buy power from DA
market. It happens for most hours, but there are some hours that Γ
influences on other components contribution as well. For example, at
t = 20 for Γ = 0 , submitted bids are 3500 kW and −1000 kW for DA
and RT markets respectively. For Γ = 24 , they change to 2793 kW and
0 kW. Moreover, DDGs active power generation alters from 6495 kW
for Γ = 0 to 6247 kW for Γ = 24 and also PEVs charging from 295 kW
to 250 kw. Indeed, at t = 20 for Γ = 0 in order to submit 1000 kW sale
bid to RT market, 707 kW is bought from DA market, 247 kW is sup-
plied by DDGs and 46 kW is supplied by decreasing the amount of PEVs
charging. For Γ = 24 which MGO takes riskless behavior in order to
participate in RT market, its bid for RT market become 0 kW. In fact,
MGO conservative decision making, causes it to avoid submitting bid to
RT market.
3.3. ESS contributions
The ESS is one of the cost-effective units that belongs to MG and its
operational costs are restricted merely to its degradation cost. Hence, in
the case of high demand, it would be optimal for MGO to use ESS for
supplying a fraction of demands rather than increasing the output of its
DDGs. Further, it would be optimal to charge ESS in low price hours
and discharge them in high price hours to gain benefit. Therefore, the
arbitrage strategy can be applied from two perspectives. Firstly, sup-
plying a fraction of demand in peak-load hours by ESS in lieu of de-
ploying costly DDGs with the aim of decreasing operational costs.
Secondly, charging ESS in low-price hours and discharging it in high
price-hours. In Fig. 13, the charging and discharging of ESS in 24-hour
is depicted. Noted, the positive/negative values stand for charging/
discharging of ESS. In hours 1–7 that the market price is low and also
the demand is not high, the ESS is charged. However, in hours 14–15,
when the market price is the highest, the ESS is discharged. Noted, in
hours 19–21 that the market price is relatively high and the demand is
elevated, ESS is discharged.
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Fig. 11. MG bids in DA and RT markets. (a) Γ = 0 ; (b) Γ = 6 ; (c) Γ = 10 ; (d) Γ = 24 .

Fig. 13. Charge/discharge of ESS.

3.4. PEVs contributions

PEVs are similar to ESS with some additional restrictions. Therefore,

MGO can also exploit PEVs to gain additional benefits; however, this is
totally dependent on the availability of PEVs in parking lots and the
electricity market status at those hours. Fig. 14(a) and Fig. 14(b) il-
lustrate the charge/discharge (absorb/inject) of active (reactive) power
for the 1st and 5th PEV parking lots when they are available. For in-
stance, PEV5(5th parking lots) are discharged by MGO at hours 15 and
16, when the MP is high and it is charged at hours 17–18 and 21–23
when the MP is relatively low (the terms low and high for MP is defined
for PEVs5 based on its availability in parking lots.
As stated earlier (see Section 2.3), QM is defined as the point where
any increase in reactive power leads to a decrease in active power ca-
pacity, hence, resulting in a lost opportunity cost. QM is illustrated in
red dashed-line in Fig. 6. According to Fig. 7(b), reactive demand is
high at hours 17–23 and, as one can see in Fig. 14(a) and Fig. 14(b),
reactive power provided by the PEVs is higher than QM at these hours.
Indeed, it is optimal for the MGO to utilize the PEV reactive power and
Fig. 12. Contribution of components for Γ = 24 . (a) Reactive, (b) and (c) pay them the lost opportunity cost for supplying parts of the reactive
Active. demand, rather than serving it through DDGs. Therefore, arbitrage
opportunity between PEV active and reactive power makes it possible
for the MGO to set a trade-off between the active and reactive power so
as to maximize its benefits. Such benefits can be well realized in
Fig. 14(c) and Fig. 14(d), through which the utilized PEV capacity and

9
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Fig. 14. Charge/discharge and absorb/inject of active and reactive power of PEVs: (a) PEV1; (b) PEV5. Contribution of P2 and Q2 in S2: (c) PEV1; (d) PEV5.

also their active and reactive power contributions at each hour are il-
lustrated. For simplicity, only P2, Q2, and S2 are demonstrated. It is
observed that at most hours, the full capacity of PEVs are utilized. Take
PEVs1(1th parking lots) as an example. They are at a parking lot at the
time, when the MP is low at first and then reaches the highest level.
Hence, it would be optimal for the MGO to exploit the active power of
PEVs1 to charge it in low-price hours and discharge it in high-price
hours rather than utilizing its reactive power. It is also illustrated in
Fig. 14(a) that active power makes up the highest percentage of the
total capacity of PEVs1. On the contrary, PEVs5 are in the parking lot at
the reactive demand peak time and it is observed in Fig. 14(d) that most
of the PEVs5 capacity is allocated to reactive power. It can be seen that
after hour 16 when the reactive demand reaches its peak time interval,
the contribution of injected reactive power by PEVs5 is higher than its
active power and this continues to the point where all its capacity is
devoted to the reactive power at hours 19–20. Afterward and as the
reactive power demand decreases, the contribution of reactive power in Fig. 15. Voltage: (a) bus 6; (b) bus 8 (c) bus17.
PEVs5 drops gradually and the proportion of its active power improves
steadily. Consequently, the combined scheduling of active and reactive
Table 5
powers capabilities of PEVs is a technical requirement that could bring
Impact of various risk-driven policies on the MG costs.
further potentials for the network operation. Also, brings about addi-
tional flexibility in MGO’s decisions. Γ 0 3 6 9 12 15 24

Total Expected Cost 34,746 35058 35291 35474 35554 35558 35558
3.5. Voltage analysis ($)

Fig. 15 indicates voltage magnitude of buses for two cases at 24 h. In
CaseI PEVs can submit active and reactive bids, CaseII represents state
in which PEVs only submit active bids and reactive power bids are
neglected. Fig. 15(a), (b) and (c) illustrate the grid voltage profile of
buses 6, 8 and 17 for two designated cases. It is seen that in these cases,
the proposed framework is able to maintain the MG’s voltage profile
within the permissible ranges, voltage magnitude of buses must be
between 0.98 p.u to 1.02 p.u.
At buses 6 and 8 reactive power injection by PEVs improve voltage
profile, while at bus 17 it is worsened. These different effects are due to
objective function of problem is cost minimization, indeed reactive
power is provided by PEVs with the aim of minimization cost of reactive
power supplement subjected to retain the MG’s voltage profile within
the permissible ranges.
3.6. Impact of risk attitude
Unpredictable behavior in market prices imposes a significant risk
in making decisions on participating in electricity markets. In order to
tackle the risk level of the MGO’s decisions to participate in RT markets,
robust programming has been employed in this paper. Table 5 deline-
ates the impact of robust control parameter on the total expected cost of
the MG. One can see, in Table 5, that as the robust control parameter

10
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Table 6
Impact of distinct QM scenarios on the MG costs.

QM Limitation −10% −5% 0 −5% +10%

Total expected cost ($) 34935 34842 34,746 34670 34580

Table 7
Impact of DA and RT market constraints.
DA restriction(kW) Total expected cost($)

2500 34942
3000 34830
3500 34,746
4000 34725
4500 34724
RT restriction(kW) Total expected cost($)
500 35171
750 34956
1000 34,746
1250 34565
Fig. 16. Expectation of RT bids. 1500 34386

increases, the total expected cost of the MG will rise. Indeed, by in-
creasing the control parameter, MGO sacrifices its economic benefits to
maintain its security. As a result, risk-driven policies significantly affect
the arbitrage opportunity concerning participation in electricity mar-
kets. For instance, in Γ = 0 , where the MGO is optimistic, the total
expected cost of the MG is $34,746. However, when MGO is pessimistic
(Γ = 24 ), the costs increases by 2.34%. The reason mainly lies in the
fact that MGO’s transactions in RT market is restricted which in turn
prevents taking full advantage of the possible arbitrage opportunities.
The expected bided power in the RT market is demonstrated in
Fig. 16 by parameter. As mentioned before, the bidding power in the RT
market would be diminished by growing and this procedure will be
continued until the bidding value in the RT market becomes approxi-
mately zero. In this case, the most robust and conservative solution
would be achieved, in which the bidding risk in the RT market reaches
the least value and the operational cost peaks at the highest value.
3.7. Impact of set points on lost opportunity cost of PEVs
We defined QM as the point where any increase in reactive power of
PEVs results in a decline in the active power. Hence, the lost opportunity
cost due to such active power drops is studied in this section. In order to
evaluate the impact of QM, Table 6 presents the total expected cost of

Fig. 17. Modified 33-bus IEEE test system.

11
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Table 9
Expected costs in different cases.
Expected Cost ($) Cases

Case1 Case2 Case3 Case4 Case5

Active power by DDGs 5003 16142 5039 5141 4882

Reactive power by DDGs 3005 2976 3003 4626 4599
Transaction in RT market −437 0 −437 −437 −437
Transaction in DA market 7041 0 7158 6945 6168
Active power by PEVs −1017 −1035 −1053 −1026 0
Reactive power by PEVs 203 208 204 0 0
Fig. 18. MG demands. Active power by ESS 25 27 25 22 25
Total expected cost ($) 13823 18318 13939 15271 15237
Cost increment (%) 0 24 0.8 9 9
the MG with different QM scenarios. It is observed that as QM increases,
MG costs decreases since the PEVs cross the lost opportunity cost region
with higher Q values. On the other side, lower QM results in higher costs
for MGO, since PEVs reach the lost opportunity cost region with lower
Q values.

3.8. Impact of DA and RT market restrictions

For illustrating the effect of markets limitations, Table 7 demon-

strates different upper limits for DA and RT markets. DA and RT market
constraints pertinent to base case are shown with bold letters. As can be
seen, by increasing (decreasing) the upper limits, the total expected cost
of MG drops (grows). For instance, by limiting the RT constraint to
1500 kW, MG undergoes 360$ depletion in its costs in comparison with
the base case. On the contrary, by restricting it to 500 kW, there would
be 425$ rise in MG costs.

4. Case studies 2

In this section, a modified IEEE 33-bus test system is employed as

the second case study to illustrate the effectiveness of the proposed
model. Fig.17 indicates schematic of case study system and its com-
ponents. Fig. 18 represents active and reactive demand at 24 h and
Tables 7 and 8 demonstrate parameters of DDGs and parking lots, re-
spectively. Other data such markets prices, parameters of ESS, PEVs,
wind turbine and PV are the same IEEE 18-bus case study (Table 8).

4.1. Analysis impact of components on operation MG cost

In order to investigate effect of components on MG operation cost,

five different cases is considered. Case1 is MG normal operation. In
case2, both RT and DA markets are neglected. In case3, PEVs can ab-
sorb/inject reactive power and only be charged. Case4 indicates state of
operation in which PEVs reactive power is neglected and Case5 is
without PEVs presence. Table 9 demonstrates contribution of compo-
Fig. 19. Contribution of components in Case1. (a) Reactive, (b) and (c) Active.
nents in MG operation cost in different cases.
Fig. 19 shows contribution of components to supply active and re-
active demand in case1. It is seen in Fig. 20(a) that reactive demand is demands by its internal components. As it is seen in Table 9, cost of
supplied by DDGs and PEVs. In this case, 15% of reactive demand is DDGs active power generation increases in case2 owning to DGGs must
served by PEVs. compensate the power that was bought from markets. In this case DDGs
In cace2 which MG is operated in island mode, the opportunity to active power generation goes up 36% in comparison with case1. Be-
use market transaction is not available. So, MG must supply all of its sides, amount of PEVs charging/discharging dwindles. This is because
markets are viable option which causes MGO accept PEVs active power
Table 8 bids and sells purchased power in market to get profit. In this case,
Parameters of parking lots. amount of PEVs charging/discharging decreases 10% in comparison to
DDG Pmin(kW) Pmax(kW) Qmin(kVAr) Qmax(kVAr)
case1. In case3 because MGO cannot discharge PEVs power, MG op-
eration cost increases in comparison to case1 because MGO loses PEVs
1 50 2000 0 450 arbitrage opportunity. In case4 which PEVs reactive power is ignored,
2 50 2000 0 450 cost of DDGs reactive power injection goes up because all of reactive
3 50 1500 0 580
demand must be served by DDGs. ignoring PEVs reactive power gives
4 50 1500 0 300
5 50 1000 0 220 this opportunity to MGO to allocate all of PEVs capacity to charge/
6 50 1000 0 220 discharge. In this case, amount of PEVs charging/discharging goes up

12
M. Saffari, et al. Electric Power Systems Research 174 (2019) 105847

Fig. 20. MG bids in DA and RT markets. (a) Γ = 0 ; (b) Γ = 6 ; (c) Γ = 10 ; (d) Γ = 24 .

18% in comparison to case1. At last, Absence of PEVs in case6 makes

MG operation cost increases 9%.

4.2. Markets contribution and impact of robust optimization on submitted

bids to RT and DA markets

In this section impact of robust control parameter(Γ ) on MG sub-

mitted bids to RT and DA markets is investigated. By increasing the
amount of Γ , MGO decision making in order to submit bids to RT
market becomes more conservative. As it is seen in Fig. 20, increasing Γ
causes that MGO submitted bids to RT market decreases. Indeed, this
procedure indicates trade-off between revenue and risk. According to
Table 10, as the amount of Γ goes up(riskless decision making), the MG
operation cost increases.

4.3. Voltage analysis

Fig. 21 demonstrates voltage magnitude of buses for two different

Fig. 21. Voltage: (a) bus8; (b) bus18 (c) bus28.
cases. In caseI, PEVs can inject/absorb reactive power, while it is ne-
glected in caseII. It is seen in both cases voltage of buses remains in
permissible range. this, it is considered a stochastic/robust optimization as efficient tool to
deal with uncertainties in renewable energy resources, DA and RT
markets prices and arrival and departure times of PEVs. Moreover, a
5. Conclusion
model of PEVs in which they charge/discharge(absorb/inject) active
(reactive) power is contemplated. In this strategy, PEVs active and re-
This paper proposes a centralized energy management framework to
active bids are submitted to decision making by aggregators. Besides, in
integrate operation of MG with PEVs arbitrage energy management in
PEVs bids is considered lost opportunity cost. This cost is paid by MGO
order to control adverse effects of PEVs high penetration in MG. To do
because of using PEVs reactive power so that decreasing of active power
exchange capacity for PEVs owners to be compensated. It causes deci-
Table 10
sion making in MG operation to become more flexible and also guar-
Impact of various risk-driven policies on the MG Costs.
antees PEVs benefits. At last, considering robust optimization in this
Γ 0 6 10 24 strategy leads MGO manages level of risk-taking behavior in decision
making when participates in RT markets.
Total expected cost ($) 13823 14295 14554 14554

13
M. Saffari, et al.
References
[1] D. Shang, G. Sun, Electricity-price arbitrage with plug-in hybrid electric vehicle:
gain or loss? Energy Policy 95 (8) (2016) 402–410.
[2] H. Zhang, Z. Hu, Z. Xu, Y. Song, Evaluation of achievable vehicle-to-grid capacity
using aggregate PEV model, IEEE Trans. Power Syst. 32 (2017) 784–794.
[3] Q. Gong, S. Midlam-Mohler, V. Marano, G. Rizzoni, Distribution of PEV charging
resources to balance transformer life and customer satisfaction, 2012 IEEE
International Electric Vehicle Conference, (2012), pp. 1–7.
[4] M.M.A. Abdelaziz, M.F. Shaaban, H.E. Farag, E.F. El-Saadany, A multistage cen-
tralized control Scheme for islanded microgrids with PEVs, IEEE Trans. Sustainable
Energy 5 (2014) 927–937.
[5] P. Kou, D. Liang, L. Gao, F. Gao, Stochastic coordination of plug-in electric vehicles
and wind turbines in microgrid: a model predictive control approach, IEEE Trans.
Smart Grid 7 (2016) 1537–1551.
[6] B. Khorramdel, H. Khorramdel, J. Aghaei, A. Heidari, V.G. Agelidis, Voltage se-
curity considerations in optimal operation of BEVs/PHEVs integrated microgrids,
IEEE Trans. Smart Grid 6 (2015) 1575–1587.
[7] H.S.V.S.K. Nunna, S. Battula, S. Doolla, D. Srinivasan, Energy management in smart
distribution systems with vehicle-to-grid integrated microgrids, IEEE Trans. Smart
Grid (2018) 1-1.
[8] M. Kesler, M.C. Kisacikoglu, L.M. Tolbert, Vehicle-to-grid reactive power operation
using plug-in electric vehicle bidirectional offboard charger, IEEE Trans. Ind.
Electronics 61 (2014) 6778–6784.
[9] T. Shekari, S. Golshannavaz, F. Aminifar, Techno-economic collaboration of PEV
fleets in energy management of microgrids, IEEE Trans. Power Syst. 32 (2017)
3833–3841.
[10] A. Hamidi, D. Nazarpour, S. Golshannavaz, Multiobjective scheduling of microgrids
to harvest Higher photovoltaic energy, IEEE Trans. Industrial Inform. 14 (2018)
47–57.
[11] L. Shi, Y. Luo, G.Y. Tu, Bidding strategy of microgrid with consideration of un-
certainty for participating in power market, Int. J. Electrical Power Energy Sys. 59
(7) (2014) 1–13.
[12] A. Hussain, V. Bui, H. Kim, Robust optimal operation of AC/DC hybrid microgrids
under Market price uncertainties, IEEE Access 6 (2018) 2654–2667.
[13] Y. Xiang, J. Liu, Y. Liu, Robust energy management of microgrid with uncertain
renewable generation and load, IEEE Trans. Smart Grid 7 (2016) 1034–1043.
[14] S.Y. Derakhshandeh, A.S. Masoum, S. Deilami, M.A.S. Masoum, M.E.H. Golshan,
Coordination of generation scheduling with PEVs charging in industrial microgrids,
IEEE Trans. Power Syst 28 (2013) 3451–3461.
[15] S.Y. Derakhshandeh, M.E.H. Golshan, M.A.S. Masoum, Profit-based unit commit-
ment with security constraints and fair allocation of cost saving in industrial
14
Electric Power Systems Research 174 (2019) 105847
microgrids, IET Sci., Measur Technol. 7 (2013) 315–325.
[16] G. Liu, Y. Xu, K. Tomsovic, Bidding strategy for microgrid in day-ahead Market
based on hybrid Stochastic/Robust optimization, IEEE Trans. Smart Grid 7 (2016)
227–237.
[17] M. Kia, M. Setayesh Nazar, M.S. Sepasian, A. Heidari, J.P.S. Catalão, New frame-
work for optimal scheduling of combined heat and power with electric and thermal
storage systems considering industrial customers inter-zonal power exchanges,
Energy 138 (11) (2017) 1006–1015.
[18] D.T. Nguyen, L.B. Le, Optimal bidding strategy for microgrids considering renew-
able energy and building thermal dynamics, IEEE Trans. Smart Grid 5 (2014)
1608–1620.
[19] Y. Xiang, J. Liu, Y. Liu, Optimal active distribution system management considering
aggregated plug-in electric vehicles, Electr. Power Syst. Res. 131 (2) (2016)
105–115.
[20] F. Rassaei, W. Soh, K. Chua, A statistical modelling and analysis of residential
electric vehicles’ charging demand in smart grids, 2015 IEEE Power & Energy
Society Innovative Smart Grid Technologies Conference (ISGT), (2015), pp. 1–5.
[21] S.L. Andersson, A.K. Elofsson, M.D. Galus, L. Göransson, S. Karlsson, F. Johnsson,
et al., Plug-in hybrid electric vehicles as regulating power providers: case studies of
Sweden and Germany, Energy Policy 38 (6) (2010) 2751–2762.
[22] N. Mehboob, M. Restrepo, C.A. Ca, Claudio A. Canizares, C. Rosenberg,
M. Kazerani, Smart operation of electric vehicles with Four-quadrant chargers
considering uncertainties, IEEE Trans. Smart Grid 10 (3) (2019) 2999–3009.
[23] A. Rabiee, H.F. Farahani, M. Khalili, J. Aghaei, K.M. Muttaqi, Integration of plug-in
electric vehicles into microgrids as energy and reactive power providers in Market
environment, IEEE Trans. Industrial Inform. 12 (2016) 1312–1320.
[24] M. Shabanzadeh, M.-K. Sheikh-El-Eslami, M.-R. Haghifam, The design of a risk-
hedging tool for virtual power plants via robust optimization approach, Applied
Energy 155 (10) (2015) 766–777.
[25] GAMS Development Corporation, General Algebraic Modeling System (GAMS)
Release 24.2.1, Washington, DC, USA (2013).
[26] S. Shafiee, M. Fotuhi-Firuzabad, M. Rastegar, Investigating the impacts of plug-in
hybrid electric vehicles on power distribution systems, IEEE Trans. Smart Grid 4
(2013) 1351–1360.
[27] F.S. Gazijahani, J. Salehi, Optimal Bi-level model for stochastic risk-based planning
of microgrids under uncertainty, IEEE Trans. Industrial Inform. PP (2017) 1-1.
[28] Wind Turbine Model. Available: https://en.wind-turbine-models.com/turbines/
123-an-bonus-600-41.
[29] PV model. Available: https://www.solarelectricsupply.com/solar-panels/rec/rec-
310w-rec310tp2m-twinpeak-2-mono-solar-panel-low-price.
[30] (2018). Rotating Shadowband Radiometer. Available: https://midcdmz.nrel.gov/
ornl_rsr/.