Electric Power Systems Research 121 (2015) 368–378

Contents lists available at ScienceDirect

Electric Power Systems Research

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

Risk-based profit allocation to DERs integrated with a virtual power

plant using cooperative Game theory
Saeed Rahmani Dabbagh, Mohammad Kazem Sheikh-El-Eslami ∗
Faculty of Electrical and Computer Engineering, Tarbiat Modares University, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: Distributed energy resources (DERs) play a key role in the deregulated power systems with environmental
Received 24 August 2014 concerns. Their scales and the uncertainty pertaining to intermittent generation of renewable resources
Received in revised form 2 November 2014 are the major challenges of participating in wholesale electricity markets. The concept of virtual power
Accepted 24 November 2014
plant (VPP) makes their integration possible and also allows covering the risk due to uncertainties. It
Available online 12 December 2014
yields a surplus profit in comparison to profits made by uncoordinated DERs. In this paper, using a
novel stochastic programming approach, the participation of a VPP in the day-ahead market (DAM) and
Keywords:
the balancing (real-time) market (BM) is considered. The uncertainties involved in the electricity price,
Conditional value-at-risk
Distributed energy resources
generation of renewables, consumption of loads, and the losses allocation are taken into account. The
Nucleolus desired risk-aversion level of each independent DER owner is used to compute the conditional value-
Shapley value at-risk (CVaR) as a well-known risk measure. The role of each DER in covering the risk and making the
Virtual power plant total profit is evaluated. The Nucleolus and the Shapley value methods as the cooperative Game theory
Profit allocation approaches are implemented to allocate VPP’s profit to the DERs. The results of a numerical study are
presented and concluded.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction dispersed generation and consumption units can integrate into a

1.1. Competition and distributed energy resources
Nowadays, distributed energy resources (DERs) including dis-
persed generation (DG), storage facilities (SFs), and demand
response (DR) resources have a significant influence on the eco-
nomic development of power systems. Although DGs and small-
or medium-scale SFs seem highly successful in stimulating small
investments, they face strong challenges of economic activities.
Owners of these DERs will be limited to electricity price set-
ting mechanisms considered by the distribution system operators
(DSOs). Nodal pricing [1], locational marginal price (LMP) calcula-
tion [2], and contract pricing [3] are some of these mechanisms.
On the other hand, small- or medium-scale consumers which are
able to flexibly control their own loads are limited to DR programs
offered by DSOs, and hence they are not at the advantage of compet-
itive environment of wholesale electricity markets. Some of these
programs are evaluated in [4–7].
The concept of virtual power plant (VPP) is a practical way of
eliminating aforementioned limitations. According to this concept,
∗ Corresponding author. Tel.: +98 21 82884920; fax: +98 21 88220121.
E-mail address: aleslam110@gmail.com (M.K. Sheikh-El-Eslami).
single market agent to be large enough for participating in whole-
sale markets. So they can trade at the wholesale price similar to
large-scale producers and consumers. Centralized and distributed
dispatches of VPPs are analyzed in [8,9]. In this paper, VPP is
assumed to be centrally controlled. As smart grid infrastructure
develops, the VPP concept becomes more practical, even for small
DERs. Refs. [10,11] compare the concepts of VPP and micro-grid.
It is important to emphasize that, in comparison to micro-grids,
VPPs concept is much broader. This is so because it is not limited
to geographical location of DERs and the ownership of the grid.
However, participation in competitive markets poses the risk of
profit variability. For instance, there is no subsidy or fixed tariff
in competitive markets. DERs must compensate energy devia-
tions in respect of their scheduled generations or consumptions
in the day-ahead market (DAM). Ref. [12] presents the procedure
of including risk measures such as the shortfall probability, the
expected shortage, the value-at-risk (VaR), and the conditional
value-at-risk (CVaR) in the formulation of the stochastic program-
ming.
In some cases, an imbalance penalty must be paid [13–15], and
in a wider context, energy deviations shall be traded in competitive
markets, namely the balancing (real-time) market (BM). Trading in
the BM at a single price is the common practice in the US [16].
In European electricity markets, there is a dual pricing system for

http://dx.doi.org/10.1016/j.epsr.2014.11.025
0378-7796/© 2014 Elsevier B.V. All rights reserved.
 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378 369

Nomenclature
D
Ptω power traded in the day-ahead market, positive val-
Sets ues for selling and negative values for purchasing
(N)DL set of (non-)dispatchable loads (MW)
(N)DP set of (non-)dispatchable producers Putω net power injected by unit u (MW)
(N)DU set of (non-)dispatchable units + −
Putω , Putω power generated and consumed by SFs, respec-
SF set of storage facilities tively (MW)
P
utω forecasted power injection of NDUs (MW), positive
Indices values for generation and negative values for con-
b index of net power injection blocks of DERs sumption
f index of combinations of the units, running from 1 sutω binary variable denoting on/off status of DPs, 1 for
to 2U−1 on and 0 for off
h index of combinations of the units except one of s , zd
zutω utω binary variables denoting respectively start-up and
them, running from 1 to 2U−1 shut-down decisions for DPs
t index of time periods  ubtω binary variable for selection of block b of DUs power
u index of units (DERs) integrated into a VPP, running injections. It is equal to 1 if block b is selected and 0
from 1 to U otherwise
v index of units expect one of them, running from 1 ω auxiliary variable for calculating the CVaR ($)
to U − 1  value-at-risk ($)
ω, ω index of scenarios ˘ f , ˘ hu profit of combination f, and combination h separat-
ing out unit u, respectively ($)
Constants
˘u , ˘u allocated profit to unit u and uncoordinated
cum marginal operation cost of DPs/SFs and marginal
expected risk-neutral profit of unit u, respectively
curtailment cost of DLs ($/MWh)
f
($)
cu , cus , cud fixed cost, start-up cost, and shut-down cost of DPs, ˘ tω profit realized in time period t and scenario ω ($)
respectively ($) B+
tω B−
, tω imbalance prices for positive and negative energy
dt duration of period t (h) deviations, respectively ($/MWh)
EuS storage capacity of SFs (MWh) D
tω day-ahead market price ($/MWh)
kfu binary coefficient equal to 1 if unit u is available in
combination f and 0 otherwise
khv binary coefficient equal to 1 if remaining unit v is
available in combination h and 0 otherwise the BM [17], i.e. the positive deviation (the generation surplus or
ku binary coefficient equal to kfu or khv , pertaining to consumption deficit) shall be sold to the BM+ and the negative
Nucleolus- or Shapley value-based methods deviation (the generation deficit or consumption surplus) shall be
Pub upper limit of net power injection associated with bought from the BM− [18–20]. The dual pricing system is consid-
block b (MW) ered in this paper.
Put upper limit of net power injection for unit u due to
the generation capacity of DPs/SFs (constant dur- 1.2. Risk-hedging tools
ing the time), hourly non-curtailable demand of DLs,
and transmission/distribution limitations (MW) The general mechanism for imbalance prices is described in
Pu technical minimum of DPs and rated demand of DLs [12,20]. It is shown that, in competitive environments, the DAM
(MW) price is hourly equal to or higher than the BM+ price, and equal to
Ptc previously contracted power (MW) or lower than the BM−. The owners of the non-dispatchable sources
Ru , Ru respectively, up and down ramp rate limits for DPs must cope with the generation intermittency. Decision-making of
(MW/h) large-scale wind power plants (WPPs) under uncertainty are pre-
˛, ˛u confidence levels of VPP and the owner of unit u, sented in [20,21]. These studies show that, as the risk aversion of
respectively non-dispatchable producers (NDPs) increases, the sale in the DAM
ˇ, ˇu weighting factor used to materialize the tradeoff decreases and in the BM increases at the expense of reducing the
between the expected profit and the CVaR, respec- expected profit.
tively pertaining to VPP and unit u A joint configuration of a WPP and a pumped-hydro-storage
u whole efficiency factor of SFs (per unit) plant (PHSP) is modeled in [13–15,18,19], and compared with an
ω probability of occurrence of scenario ω uncoordinated operation. These studies show that the joint oper-
Tu , Tu minimum up and down time limits for DPs, respec- ation results in a profit higher than the summation of profits
tively (h) uncoordinatedly obtained. In this paper, we call this added value
“surplus profit”.
Variables The price volatility in competitive electricity markets is the
Cutω operation cost of DUs ($) source of uncertainty from viewpoint of dispatchable units (DUs),
B+ B−
Etω , Etω respectively, positive/negative energy deviation such as conventional power plants (CPPs), SFs, and dispatchable
sold to/purchased in the balancing market (MWh) loads (DLs). Decision-making of large producers under uncertainty
S
Eutω energy stored by SF u at the end of time period t is discussed in [22,23], and energy procurement problem for large
(MWh) consumers is addressed in [24,25]. These researches show that the
Lutω loss allocated to unit u (MW) energy trades through bilateral contracts and the futures market
L
ubtω forecasted loss allocation pertaining to block b of (FM) grow as the risk aversion becomes more significant. It often
DUs (MW) leads to a reduction in the expected profit of producers and the
expected cost of consumers.
370 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378
1.3. Offering/bidding model for a VPP
It has been shown that non-dispatchable renewable energy
resources such as a photovoltaic power plant (PVP) can be com-
bined with a DP (e.g. a quick response CPP), or with a SF (e.g. a
battery energy storage system (BSS) or a PHSP) to lower the risk of
intermittent generation. Since VPP may have both generation and
demand units (DERs), it may offer (to sell) or bid (to purchase) corre-
sponding to the desired amount of power totally injected to the grid
in a time period. Optimal offering problem of a VPP for participating
in the DAM and the BM is presented in [26–28]. Optimal mid-term
dispatch problem of a VPP is addressed in [29]. For taking uncer-
tainties into account, Refs. [26,29] use stochastic mixed-integer
programming (MIP) approach, and Refs. [27,28] use point-estimate
method. Ref. [30] presents optimal bidding strategy of a VPP for par-
ticipating in a joint market of energy and spinning reserve service.
It takes no uncertainty into consideration and solves the proposed
non-linear model via a genetic algorithm.
Refs. [27,30] assume that VPP operates the distribution system,
and so loss management is included in the proposed models. It is
important to point out that, in general, VPPs are not equivalent to
micro-grids, and so they have no control on all of DERs existing
on the grid. Also they do not act as a DSO which is responsible for
allocating electrical losses in deregulated distribution systems.
1.4. Game theory-based approaches
Non-cooperative and cooperative Game theories are widely
used in power industry, e.g. for network cost allocation [31,32], per-
formance evaluation of thermal power plants [33], allocation of unit
start-up costs [34], estimating the pricing of transmission [35,36],
LMP calculation [2], reactive control [37], making bid strategies
[38], loss allocation [39], and profit allocation of independent power
plants (IPPs) [40]. Since DERs integrated into a VPP play a coopera-
tive Game in competitive electricity markets to earn the maximum
economic profit, the procedure for using cooperative Game theory
is presented in this paper to impartially share VPP’s profit between
the DERs. For this reason, the Nucleolus-based and the Shapley
value-based methods which are two common ways to solve the
problem of cooperative Games are implemented.
1.5. Contributions and paper layout
Since the owners of the DERs can be independent, a method
based on the desired risk-aversion of each DER owner is proposed
in this paper to determine the total risk-aversion degree of a VPP.
Using stochastic programming approach, short-term market oper-
ations of a VPP under uncertainty are presented in detail.
To assess the impact of each DER on making the surplus profit, all
possible states of DER combinations are taken into account via the
concepts of the Nucleolus and the Shapley value. This study indi-
cates that all of the possible coalitions of the following categories
of the DERs can always yield a surplus profit:
• NDUs (non-dispatchable units) including NDPs (e.g. WPP, PVP)
and NDLs (non-dispatchable loads).
• DUs including DPs (e.g. CPP), SFs (e.g. PHSP, BSS), and DLs (e.g.
flexible DR).
Note that NDLs are considered similar to other NDUs in this
paper. Also, for the reason explained in Section 1.3, allocated losses
are considered in the profit allocation problem in a way similar to
other uncertain parameters.
The rest of this paper is organized as follows. Section 2 pro-
vides a general view of the scope of this study. Market operations
of a VPP are analyzed in detail in Section 3. Section 4 introduces
the profit allocation mechanisms. An illustrative numerical study
is presented in Section 5. Lastly, Section 6 concludes the paper.
2. The scope of this study
2.1. Uncertainties
A typical VPP consisting of several DERs participates in whole-
sale electricity markets to turn the maximum profit. If the energy
consumed by VPP is in the majority, the profit will be negative.
In this case, the problem will be equivalent to the cost minimi-
zation. Whereas the DAM is cleared the day before in which the
power delivery takes place, VPP needs to optimize offering/bidding
curves, and so it needs to forecast all uncertain parameters. In this
paper, the following uncertainties are considered: The prices of
the DAM and the BM (both sale and purchase), the generation of
each NDP, the consumption of each NDL, and the loss allocation
of each DU. The uncertainty of powers injected by units not inte-
grated into VPP are considered in a way similar to powers injected
by NDUs. After characterizing these parameters as random vari-
ables, scenario-generation and scenario-reduction techniques [41]
should be respectively implemented.
2.2. Surplus profit
The risk of decision-making under uncertainty pertaining to
each DER can be covered by integration into a VPP. This integra-
tion yields a surplus profit, i.e., if there is no uncertainty, there will
be no risk of decision-making, and so there will be no surplus profit.
For example, when the electricity price is fixed, the offering/bidding
provided by DUs has no risk. Likewise, when the mismatch between
the traded and delivered electricity results no penalty, or the prices
in the BM and the DAM are the same, there is no risk to NDUs. But,
such circumstances are not acceptable in competitive electricity
markets.
On the short-term horizon, irrespective of previously contracted
powers and also irrespective of being dispatchable, the only mea-
sure that a producer/consumer can take to hedge against profit
variability is to trade less/more energy in the DAM in the hope
of trading its generation surplus/consumption deficit in the BM
at a good price. Consequently, as the risk aversion increases, the
negative deviation all over the market horizon decreases at the
expense of increasing the positive deviation. The following illus-
trative example clarifies how the surplus profit is made by the
integration.
Suppose that the generation of a NDP during an hour is higher
than the previously scheduled amount. In the case of an individual
NDP, this generation surplus must be sold in the BM+ at a price
lower than or equal to the DAM price [12]. But, in the case of the
integration into a VPP:
• It can be delivered instead of the generation deficit of another
NDP which needs to cover the deficit in the BM− at a price higher
than or equal to the DAM price.
• It can be consumed by a NDL which has a consumption surplus.
• It can be substituted for the generation of a DP whose marginal
generation cost is higher than the difference between the prices
of the DAM and the BM+.
• It can be stored by a PHSP in order to cover a subsequent energy
deviation.
Thus, the integrated operation of DERs under a central control
reduces the energy sold/bought at a lower/higher price which is
taken place during the day-of transactions. This integration con-
stitutes an efficacious dynamic risk-hedging tool for DERs, and as a
 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378 371

result, turns a surplus profit. The resulted profit should be allocated coefficients are determined by kfu or khv depending on the imple-
between the DERs. Obviously, the profit allocation method should mented Game theory-based method. It should be emphasize that,
fairly take the role of each DER in risk reduction into account. only the offering curve pertaining to the integration of all DERs
is actually submitted, and others will be only used to solving the
3. Market operations of a VPP profit allocation problem. After clearing of the markets, the sched-
uled power of VPP is determined by the market operator (MO).
3.1. Optimal offering/bidding Considering the submitted and the accepted offers, and the cleared
prices in the DAM, the virtual offering curves pertaining to each
The virtual offering curves pertaining to each combination of combination can be used to assign the virtual scheduled powers.
available DERs can be obtained by the following two-stage stochas-
tic programming which is run by VPP. 3.1.3. Power injection constraints
sutω Pu ku ≤ Putω ≤ sutω Put ku , ∀u, ∀t, ∀ω (9)
3.1.1. Objective function
Putω = P
utω ku , ∀u ∈ NDU, ∀t, ∀ω (10)
Setting confidence level ˛, to a constant value for the CVaR cal-
+
culation, the tradeoff between the expected value of the profit and Putω dt S
≤ Eutω ≤ EuS , ∀u ∈ SF, ∀t, ∀ω (11)
the CVaR can be materialized through desired weighting factor ˇ,
− +
in the objective function as
S
Eutω S
= Eut−1ω + (u Putω − Putω )dt , ∀u ∈ SF, ∀t ∀ω (12)
  Putω = + −
− Putω ∀u ∈ SF, ∀t, ∀ω
 1 
Putω , (13)
Maximize (1 − ˇ) ω ˘tω + ˇ − ω ω (1) ±
1−˛ Putω ≥ 0, ∀u ∈ SF, ∀t, ∀ω (14)
ω t ω
Constraints (9) define the output limits regarding the rated val-
By varying the confidence level ˛, as a desired risk-aversion
ues of DERs or transmission/distribution limitations imposed by
parameter, there is no need to use the weighting factor ˇ, and
the DSO, e.g. the minimum power output and capacity of a CPP,
the CVaR can be expressed as the objective function [12]. It can
the pumping and turbine capacities of a PHSP, and the maximum
be formulated as
and hourly non-curtailable demands of a DL. Constraints (10) state
1  that the outputs of the NDUs are not dispatchable. They are equal to
Maximize  − ω ω (2)
1−˛ the power generation/consumption scenarios. The available energy
ω
stored by SFs at the end of each time period is defined in (11) and
Briefly, values of ˛ or ˇ close to 0 result in solutions involving low (12) regarding to their storage capacity and whole efficiency. The
risk-aversion, whereas values of ˛ or ˇ close to 1 yield risk-averse net power injection of SFs is defined by (13). Constraints (14) ensure
solutions. that both storage and retrieval powers are non-negative.

3.1.2. Trading constraints 3.1.4. Constraints of DUs


˘tω = D D
tω Ptω dt B+ B+
+ tω Etω B− B−
− tω Etω − Cutω ku , ∀t, ∀ω (3) −Ru ≤ Putω − Put−1ω ≤ Ru , ∀u ∈ DP, ∀t, ∀ω (15)
u d
−zutω s
≤ sutω − sut−1ω ≤ zutω , ∀u ∈ DP, ∀t, ∀ω (16)

− ˘tω ≤ ω , ∀ω, ω ≥ 0, ∀ω (4) 
u −1
s
sut+iω dt+i ≥ zutω u , ∀u ∈ DP, ∀t, ∀ω (17)
t
 i=0

Putω dt = (PtC + Ptω

D B+
)dt + Etω B−
− Etω , ∀t, ∀ω (5) 
u −1

u
d
(1 − sut+iω )dt+i ≥ zutω u , ∀u ∈ DP, ∀t, ∀ω (18)
i=0
D
Ptω D
< Ptω  , ∀t, ∀ω,
D
if tω D
< tω  (6) ⎧
⎪ f
⎨ cu Putω dt + cu sutω + cu zutω + cu zutω ,
m s s d d
∀u ∈ DP, ∀t, ∀ω
 , ∀t, ∀ω,
D D D D
Ptω = Ptω if tω = tω  (7) + −
Cutω = Cum (Putω + Putω )dt , ∀u ∈ SF, ∀t, ∀ω (19)
⎪
⎩
B±
Etω ≥ 0, ∀t, ∀ω (8) Cum (Putω − P
∀u ∈ DL, ∀t, ∀ω
utω )dt ,

The profit of VPP participating in the DAM and the BM is s

described in (3). Since a typical VPP is a price-taker agency in elec-
tricity markets due to its scale, the prices are random variables.
The powers offered in the DAM are here-and-now decision vari-
ables, and the energy deviations traded in the BM are wait-and-see
decision variables. Constraints (4) are required for the linear for-
mulation of the CVaR. Energy balance equations (5) denote that the
energy generated by (N)DPs and SFs, plus the energy totally pur-
chased has to be equal to the summation of the energy consumed by
(N)DLs and SFs, and the energy totally sold. Constraints (6) ensure
that offering curves submitted to the DAM are monotonously non-
decreasing. Equations (7) are non-anticipativity constraints for the
DAM. Constraints (8) state that both sold and purchased energy
deviations are non-negative. The sign ± is used instead of either +
and − to save on the space.
Binary coefficients ku , e.g. in (3), are used to represent the pres-
ence or absence of unit (DER) u in each re-run of the model for
each possible combination of DERs. As explained in Section 4, these
sutω , zutω d
, zutω , ku ∈ {0, 1}, ∀u, ∀t, ∀ω (20)
Constraints (15) enforce the ramp rate limits of DPs. The
required correlation between on/off status and start-up/shut-down
decisions are enforced by (16), while the minimum up and down
time constraints are defined in (17) and (18), respectively. Oper-
ation costs of DPs and SFs, and curtailment costs of DLs are
considered in (19). The operation costs of NDPs can be expressed
as cum Putω dt if they are not negligible. The binary coefficients are
introduced in (20).
3.2. Risk-aversion of DER owners
The control center of a VPP should know in advance the level
of risk it is willing to assume. Since the owners of the DERs may
be independent, their risk-aversion levels may be different. These
levels must be taken into account to obtain the offering curves (pro-
vided to the DAM) and the energy deviations (traded in the BM) of
372 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378

each combination. Using the CVaR as the risk measure, the desired
confidence level of each DER owner ˛u , should be considered to
obtain the total confidence level ˛, as an input parameter for the
model. Considering (2) as the objective function, each confidence
level of DERs can be weighted by the expected absolute value of
their uncoordinated profits. It can be formulated as

˛ |˘u |ku
u u
˛=  
(21)
u
|˘u |ku

For the calculation of ˘u the net power injection scenarios of

NDUs and the constraints of DUs should be taken into account.
For this reason, the optimal offering model should be re-run for
U risk-neutral uncoordinated units. So the confidence level corre-
sponding to each combination can be obtained by (21). But, when
the weighting factor ˇ, is used as the risk-aversion parameter in the
objective function formulated in (1), ˛ is set to a constant value and
to obtain the weighting factor corresponding to each combination,
parameters ˛ and ˛u can be respectively replaced with ˇ and ˇu in
(21).

3.3. Loss allocation scenarios

Ordinarily, DERs integrated into a VPP are geographically dis-

persed in a distribution grid and VPP as a single market agent trades
energy with the upstream grid through one or more buses. The net
power injections (power generations or consumptions) at the buses
of DERs have a dynamic impact on power losses in the distribution
grid. In competitive environment of distribution industry, the DSO
is responsible for allocating losses among all of the units including
the DERs of a VPP. Some of loss allocation methods are presented in
[42,43] for transmission or meshed distribution grids, and in [44,45]
for radial distribution grids with DGs. Paying attention to impact of
allocated losses on the profit share of each DER, they should be con- Fig. 1. Market operations of a VPP.
sidered in the profit allocation problem. Besides the site of a DER,
the quantity and the polarity of the allocated loss in a time period
depends on not only the power injection of that DER but also the 3.4. Optimal real-time dispatching
power injection of other units which may not integrated with the
same VPP. For this reason, the scenarios of power injections of effec- During the day-of actual energy exchanges, VPP dispatches the
tive units on the grid (both integrated and not integrated with the DERs to optimally respond to scenarios occurring in real-time, such
VPP) can be taken into consideration to generate the loss alloca- as generation of NDPs and consumption of NDLs. Thus, a model
tion scenarios for the day-ahead. Using a loss allocation technique similar to (3)–(20) and (22)–(26) can be run for maximizing ˘ tω ,
but tωD and B± as random variables, and P D as a decision vari-
to provide step-wise curves of forecasted losses in terms of DUs tω tω
outputs, the net power injections of each DU can be optimized by able shall be respectively substituted by tD , tB± and PtD as input
taking loss allocation scenarios into account. For this reason, the fol- data. A multi-stage stochastic programming can be used to imply
lowing MIP approach can be incorporated in the model described perfect information on intermittent parameters in a few hours-
in (1)–(20) to select the provided power injection blocks; ahead. In addition to actual optimal dispatching, the same model
is virtually re-run for each combination of DERs so that the virtual
Pub−1 ubtω ku ≤ Putω ≤ ubtω Pub ku , ∀u ∈ DU, ∀b, ∀t, ∀ω (22) profit earned by each combination can be obtained. The flowchart
 of market operations of a VPP in the day-before, the day-of and the
Lutω = ubtω L
ubtω ku , ∀u ∈ DU, ∀t, ∀ω (23) day-after participation in the DAM is illustrated in Fig. 1.
b

 4. Profit allocation
ubtω = 1, ∀u ∈ DU, ∀t, ∀ω (24)
b 4.1. The basis of cooperative Games

ubtω ∈ {0, 1}, ∀u ∈ DU, ∀b, ∀t, ∀ω (25) To recognize the role of each DER in covering the risks and earn-
ing the total profit, cooperative Game approach can be used. The
However, the losses allocated to DERs should be considered in Nucleolus-based and the Shapley value-based methods are imple-
their profit shares. As an example, if an independent operator clears mented in this paper. All of the possible combinations of DERs are
them at the DAM prices, they can be written as below which should taken into consideration to form the basis of these methods. For
be added to ˘ tω in (3). this reason, virtual profits pertaining to all of the combinations of
 DERs should be obtained. Using state 1 and 0 to respectively repre-
D
− tω Lutω (26) sent the presence and absence of each DER in a combination, there
u are 2U combinations for a VPP consisted of U units (DERs), minus
 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378 373

one empty (Ø) combination (totally 2U−1 ). Having cleared the BM 1 DER
Rated
1 Power
Type
and allocated losses, the total profit corresponding to the actual #
(MW)
combination (in the presence of all available DERs) and the virtual 1 WPP 24.8

Integrated
2 PVP 6.1

into a VPP
curtailed combinations, is realized. 2
3 NDL 6.2
It is important to emphasize that when the number of DERs is 4 CPP 17.4
large, some techniques should be used to lower the computational 5 PHSP 24 / 18
6 DL 9.5
burden, e.g. each re-run required for each combination can be inde- 14 7 WPP 9.6
pendently computed in parallel with others in separate machines. 3 3 8 PVP 3.2
Also it is necessary to use scenario-reduction techniques [41]. 9 CPP 11.34
10 DL 8.7
4 13 15 Remaining NDLs 194.2
4.2. Nucleolus concept 12
132:33 kV
16
Basically, every cooperative Game has one and only one nucleo-
lus, and the nucleolus is in the core unless the core is empty. The aim 18 4
23
20 6 30
of using the nucleolus concept is to fairly allocate the total profit 19
which is jointly earned by all transactions of DERs. Defining binary 17 21
132:33 kV 24
coefficient kfu , that indicates the presence (1) and the absence (0) of 5 7 10
8
unit u in combination f, the occurred profits can be organized as ˘ f

132:33 kV
6 10 27 29
for 2U − 1 combinations. So the Nucleolus-based profit allocation 9 11
22
25
method can be expressed as 28

Minimize ε

subject to kfu ˘u ≥ ˘f − ε
u 7 5
4.3. Shapley value concept
The Shapley value is another approach to solve the problem of
cooperation Games. Separating out each DER from a VPP, the profits
corresponding to the combinations of the remaining DERs (2U−1
combinations) can be calculated one time with the separated DER
and another time without that (totally 2 × 2U−1 = 2U re-runs).
Defining binary coefficient khv , which indicates remaining unit v
in combination h, the difference between the resulted profits after
and before joining the separated unit u can be organized as ˘ hu for
2U combinations. It is important to note that, binary coefficients kfu
can be used to order khu , and vice versa. This is also true for ˘ f and
˘ hu . Thus, the Shapley value-based profit allocation method can
be written as
(27) 26
(28) 2
8
Fig. 2. DERs located in IEEE 30-bus test grid.
5.2. Random variables
The forecasted price scenarios in the DAM and the BM (for the
optimal offering problem) and the prices actually happened during
the day-of power delivery (related to the optimal real-time dis-
patching problem) are depicted in Fig. 3. All scenarios are assumed
to be equi-probable. Likewise, Fig. 4(a)–(c) depicts the power gen-
eration/consumption scenarios of WPP, PVP and NDL. It is assumed
that, DL whose power consumption scenarios are shown in Fig. 4(d),
has the hourly non-curtailable load, and the hourly marginal cost of
curtailment. These values are also depicted in Fig. 4(d). The power

 

 v khv ! U − 1 − v khv !
injection scenarios for NDUs integrated with the VPP and effec-
˘u = ˘hu (29) tive units not integrated with the VPP were generated. Then, Z-bus
u U!
method [41] was implemented to obtain the loss allocation sce-
narios. Fig. 5 shows the scenarios pertaining to PVP and PHSP for
5. Numerical study
example.
5.1. Description
100 (a)
For a comprehensive numerical study, all types of DERs Actually occurred
80 Forecast ed scenarios
described in Section 1.4 shall be taken into consideration. For this
reason, DERs located in IEEE 30-bus test grid [46] are assumed as 60
depicted in Fig. 2. The integrated units into a VPP are delineated. It
40
is assumed that VPP exchanges the energy with upstream network
via bus 2 in order to participate in wholesale markets. The rated 20
capacities of the DERs are indicated in Fig. 2. The turbine and pump 0
capacities of PHSP are 16 and 12 MW. The equivalent energy capac- 5 10 15 20
150
ity and the total efficiency of the PHSP are set to 110 MWh and 68%. (b)
Actually occurred, BM+
The operation costs of NDPs and PHSP are ignored. Actually occurred, BM-
In relation to CPP, the technical minimum generation is 1.2 MW, 100 Forecast ed scenarios, BM+
both up and down ramp rates are 3 MW/h, and both minimum up Forecast ed scenarios, BM-
and down times are 2 h. Its fixed and marginal generation costs
50
are $4.8 and 34.2 $/MWh, and its start-up and shut-down costs are
$19.3 and $12.9, respectively. It is considered that, at the beginning
of the day under study, the upper basin of PHSP is empty, and CPP 0
5 10 15 20
is off. It is also assumed that the loads have a power factor ranging
t ime (hour)
from 0.95 lead to 0.9 lag, regulated by their automatic capacitor
banks. Fig. 3. Price ($/MWh) scenarios in the DAM (a) and the BM (b).
374 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378

Table 1
(a)
20 Actually occurred Total energy traded in the markets (MWh).
Forecast ed scenarios
15 DER # DAM BM+ BM−

ˇ=0 ˇ→1 ˇ=0 ˇ→1 ˇ=0 ˇ→1

10
1 203.2 64.5 74.5 171.2 42.2 0.2
5 2 24.4 22.2 2.2 3.4 1.2 0.1
3 112.2 116 7.7 9.1 3.1 0.7
4 245.9 18 45 95.5 10.1 0.8
0
5 10 15 20 5 177.5 97.4 0 0 0 0
6 152 164 0.8 3.9 5 0.3
6 VPP 163.1 62.3 35.2 84.3 4.2 3.7
(b)
Actually occurred
Forecast ed scenarios
4

5.3. Results and discussion

2
For a confidence level ˛ equal to 0.9, the risk-neutral and risk-
0
averse behavior of each DER owner is considered by two values of
5 10 15 20 the weighting factor; ˇ = 0 and ˇ → 1 (0.8, 0.65, 0.9, 0.7, 0.85 and
0.72 for units 1–6 respectively). Due to the limited space only the
6 (c) results corresponding to these values of ˇ for all DERs are indicated.
The described case study is solved using CPLEX [47] under GAMS
5 [48]. After that DAM prices and scheduled powers are cleared by the
MO, the accepted power trades are calculated for each combination
4 as previously stated in Section 3.1.2.
The resulted schedules for the VPP and for the uncoordi-
3 Actually occurred
Forecast ed scenarios nated DERs are shown in Fig. 6, and the corresponding net
2 energy deviations occurred in the day-of power delivery are pro-
5 10 15 20 vided in Fig. 7. In Fig. 6, positive/negative values correspond to
accepted sales/purchases in the DAM for uncoordinated produc-
(d) ers/consumers and SF which acts as either producer or consumer.
8
Accepted trading for VPP is also illustrated and compared with the
6 summation of the resulted schedules for the uncoordinated DERs.
As can be seen, the uncoordinated summation curve goes to nega-
4 Non -curtailable (MW) tive values when the summation of accepted consumptions in VPP
Marginal cost of curtailment (10$ /MW h)
is larger than the summation of accepted generations.
2 Actually occurred (MW )
Forcasted sce narios (MW )
Positive/negative values in Fig. 7 correspond to posi-
0 tive/negative energy deviations. Integrating into a VPP, the
5 10 15 20 offering curve and as a result the accepted curve varies in a
t ime (hour) way that the energy deviations reduce. As can be observed in
Figs. 6 and 7, the VPP curve has lower values than the uncoordi-
Fig. 4. Power generation/consumption (MW) scenarios of WPP (a), PVP (b), and NDL
(c), and the characteristics of DL (d).
nated summation curve for both positive and negative deviations.
This is true for both values of ˇ. These figures indicate that due
to the integration, total energy deviation reduces, irrespective of
0.05 (a) risk-aversion level. It corroborates that the integration acts as a
Actually occurred
risk-hedging instrument for DERs. Expressing as a whole, total
Forecast ed scenarios
energy traded by VPP and the uncoordinated DERs in the markets
(DAM and BM) are summarized in Table 1. As pointed out in
0 Section 2.2 and as can be seen in Table 1 and Figs. 6–7, irrespective
of being dispatchable, when a producer/consumer becomes more
risk-averse, there is a decrease/increase in the energy traded in the
-0.05 DAM, and there is also a decrease in the negative deviation at the
5 10 15 20 expense of increasing the positive deviation. Table 1 also shows
that VPP behaves as though it is a producer. This is so because
(b) considering the type and scale of the DERs, and the realized
0.1 scenarios, the total production of VPP is dominant in comparison
to its total consumption.
0 The virtual values of the CVaR, the total profit and the sur-
plus profit corresponding to the possible combinations are listed
-0.1 in Table 2 for different values of ˇ. The obtained profits help us to
Actually occurred
Forecast ed scenarios analyze the role of each DER in covering the risk and obtaining the
-0.2 surplus profit. They are considered as input data in the profit allo-
5 10 15 20 cation problem. As can be inferred from Table 2, the profit of every
t ime (hour) coalition (according to all combinations except 1, 2, 4, 8, 16 and 32)
is higher than the summation of the profits of the uncoordinated
Fig. 5. Loss (MW) allocation scenarios of PVP (a) and PHSP (b).
DERs (according to combinations 1, 2, 4, 8, 16 and 32) being present
Table 2
The results of the possible combinations.

f Available DERs # CVaR ($) Profit ($) Surplus profit ($) f Available DERs # CVaR ($) Profit ($) Surplus profit ($)

S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378
6 5 4 3 2 1 ˇ=0 ˇ→1 ˇ=0 ˇ→1 ˇ=0 ˇ→1 6 5 4 3 2 1 ˇ=0 ˇ→1 ˇ=0 ˇ→1 ˇ=0 ˇ→1
√
32 −8449 −8186 −6656 −6612 0 0
√ √ √
1 899 2409 8269 7778 0 0 33 −6648 −4412 1881 1344 268 177
√ √ √
2 597 725 919 914 0 0 34 −7432 −6994 −5727 −5695 10 3
√ √ √ √ √
3 1550 3349 9210 8700 22 7 35 −5678 −3376 2823 2313 290 233
√ √ √
4 −5811 −5445 −4259 −4261 0 0 36 −14,234 −13,566 −10,833 −10,823 82 49
√ √ √ √ √
5 −4579 −2024 4102 3553 91 35 37 −12,622 −8917 −2308 −2870 338 225
√ √ √ √ √
6 −4899 −4332 −3307 −3344 34 2 38 −13,231 −12,367 −9897 −9919 99 39
√ √ √ √ √ √ √
7 −3604 −991 5031 4533 102 100 39 −11,657 −7872 −1358 −1912 369 268
√ √ √
8 4 19 1667 837 0 0 40 −5057 −4919 −4855 −4936 133 839
√ √ √ √ √
9 1023 3942 10,426 10,208 489 1593 41 −4038 −1743 4019 3701 738 1698
√ √ √ √ √
10 616 745 2600 1758 14 7 42 −4459 −4181 −3914 −4005 156 856
√ √ √ √ √ √ √
11 1643 4767 11,379 11,152 524 1622 43 −3401 −863 4967 4658 767 1740
√ √ √ √ √
12 −3415 −3179 −2555 −2767 37 657 44 −9360 −8803 −9051 −9069 197 968
√ √ √ √ √ √ √
13 −2405 293 6251 5933 574 1578 45 −7640 −5859 −185 −578 793 1679
√ √ √ √ √ √ √
14 −2840 −2432 −1593 −1846 81 663 46 −8470 −7818 −8107 −8115 222 1007
√ √ √ √ √ √ √ √ √
15 −1794 1170 7202 6888 606 1619 47 −6856 −4862 761 368 820 1711
√ √ √
16 87 143 1107 911 0 0 48 −6516 −6383 −5525 −5496 24 205
√ √ √ √ √
17 1057 3215 9655 9065 279 375 49 −5012 −3274 3179 2594 459 516
√ √ √ √ √
18 708 877 2033 1835 8 9 50 −5623 −5290 −4601 −4570 29 217
√ √ √ √ √ √ √
19 1712 4075 10,631 10,057 336 453 51 −4164 −2287 4120 3573 481 581
√ √ √ √ √
20 −4077 −3859 −3134 −3167 19 182 52 −12,095 −11,573 −9705 −9698 104 263
√ √ √ √ √ √ √
21 −2801 −1003 5488 4812 371 383 53 −10,599 −7763 −1020 −1600 519 583
√ √ √ √ √ √ √
22 −3272 −2829 −2189 −2219 45 216 54 −11,090 −10,419 −8769 −8772 121 275
√ √ √ √ √ √ √ √ √
23 −1969 −18 6374 5783 339 440 55 −9628 −6711 −68 −658 553 611
√ √ √ √ √
24 154 162 2982 1873 209 125 56 −4916 −4761 −3554 −4063 328 802
√ √ √ √ √ √ √
25 1299 4312 11,722 11,477 679 1951 57 −3789 −1294 5267 4985 879 2071
√ √ √ √ √ √ √
26 754 895 3926 2732 233 70 58 −4302 −4020 −2594 −3185 370 765
√ √ √ √ √ √ √ √ √
27 1935 5141 12,680 12440 718 1999 59 −3137 −411 6234 5933 928 2104
√ √ √ √ √ √ √
28 −3292 −3023 −1251 −1963 234 550 60 −8472 −8165 −7765 −7737 377 1387
√ √ √ √ √ √ √ √ √
29 −2108 746 7537 7236 753 1971 61 −7218 −5020 1088 720 960 2066
√ √ √ √ √ √ √ √ √
30 −2667 −2270 −292 −1072 275 527 62 −7308 −7218 −6791 −6786 432 1425
√ √ √ √ √ √ √ √ √ √ √
31 −1498 1607 8490 8196 788 2016 63 −6598 −4098 2034 1710 987 2142

375
376 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378

30 β=0

20

10

-10

-20 Unit 1 Unit 2 Unit 3 Unit 4 Unit 5 Unit 6 Sum VP P

2 4 6 8 10 12 14 16 18 20 22 24

30 β→1

20

10

-10

-20 Unit 1 Unit 2 Unit 3 Unit 4 Unit 5 Unit 6 Sum VP P

2 4 6 8 10 12 14 16 18 20 22 24
t ime (hour)

Fig. 6. Accepted power trades (MW) in the DAM for ˇ = 0 and ˇ → 1.

in the coalition. So the integration of two or more types of DERs can
result in a surplus profit.
It can be observed that, for high ˇ values, the CVaR obtained for
all combinations rises at the expense of reducing the probability of
higher profits. As expected, increasing the CVaR of VPP (shown in
combination 63 in Table 2 where all DERs are available), its profit
reduces and the surplus profit increases. This is so because the risk-
hedging role of the integration is more salient as DERs become more
risk-averse. It is important to emphasize that, a higher risk-aversion
of VPP or other virtual combinations does not necessarily lead to a
lower profit and a higher surplus profit, because these profits are
not the expected values resulted from solving the optimal offering
problem, as previously illustrated in Fig. 1.
The difference between the profit of VPP and the profit resulted
from separating out each DER from VPP (combinations 31, 47, 55,
59, 61 and 62) is indicated in Table 3. For example, for ˇ = 0, the
profit of VPP is $2034 and profit without DER 1 is −$6791 (in
Table 2), so the difference is equal to 2034 − (−6791) = 8825. As

β=0
20

15

10

-5
Unit 1 Unit 2 Unit 3 Unit 4 Unit 6 Sum VP P
-10
2 4 6 8 10 12 14 16 18 20 22 24

β→1
20

15

10

-5
Unit 1 Unit 2 Unit 3 Unit 4 Unit 6 Sum VP P
-10
2 4 6 8 10 12 14 16 18 20 22 24
t ime (hour)

Fig. 7. Occurred net energy deviations (MWh) for ˇ = 0 and ˇ → 1.

 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378 377

Table 3
The allocated profits.

DER # Uncoordinated profit ($) a ($) Using Nucleolus-based method Using Shapley value-based method

ˇ=0 ˇ→1 ˇ=0 ˇ→1 Profit ($) Surplus profit ($) Profit ($) Surplus profit ($)

ˇ=0 ˇ→1 ˇ=0 ˇ→1 ˇ=0 ˇ→1 ˇ=0 ˇ→1

1 8269 7778 8825 8496 8734 8482 464 703 8632 8372 363 594
2 919 914 946 991 946 966 27 52 941 943 23 28
3 −4259 −4261 −4200 −4223 −4220 −4241 39 20 −4208 −4151 51 109
4 1667 837 2102 2368 1876 1922 209 1085 1947 1802 280 966
5 1107 911 1273 1342 1155 1074 48 162 1239 1154 132 243
6 −6656 −6612 −6457 −6486 −6457 −6492 199 120 −6518 −6410 138 202

1047 −432 2489 2488 2034 1710 987 2142 2034 1710 987 2142
a
The difference between the profit of VPP and the profit of VPP without a DER.

can be seen, the summation of these differences is higher than the
profit of VPP and so they cannot be directly used to share the profit.
The results of the profit allocation using both the Nucleolus-
based and the Shapley value-based methods are presented in
Table 3. As can be observed, irrespective of profit allocation method
and the risk-aversion degree, the shares of the DERs in the total
profit are higher than their uncoordinated profits, i.e., their surplus
profits are positive. It corroborates the fact that, the surplus profit
of each player is inevitably non-negative in every rational cooper-
ative Game. Also the summation of profits allocated to the DERs
is equal to the total profit of VPP (according to combination 63 in
Table 2). It is also true for allocated surplus profits and the total
surplus profit of VPP. It indicates that the obtained surplus profit is
completely allocated.
Due to integration into a VPP, total profit of DERs rises from
$1047 to $2034 for ˇ = 0 (an increase of 94.3%) and from −$432
to $1710 for ˇ → 1 (an increase of 495.8%). The surplus profit rises
from $987 for ˇ = 0 to $2142 for ˇ → 1 (an increase of 117%). It can
be inferred that, as DER owners in a VPP become more risk-averse,
a reduction in the total profit and an increase in the surplus profit
are expected. If all of DER owners become fully risk-averse (ˇ → 1
in comparison with ˇ = 0), total profit of uncoordinated DERs expe-
riences a reduction of 141.3% (from $1047 for ˇ = 0 to −$432 for
ˇ → 1), but total profit of integrated DERs reduces only 15.9% (from
$2034 for ˇ = 0 to $1710 for ˇ → 1). This means that, increasing
the risk-aversion level, the reduction experienced by the expected
total profit of integrated DERs is less in comparison to uncoordi-
nated DERs. Table 3 also shows that the profits of all DERs allocated
by the Shapley value-based method, increase as the parameter ˇ
grows. This is not necessarily true for the profits allocated by the
Nucleolus-based method. Also the difference between the shares
in the total profit calculated by two different Game theory-based
methods increases as the value of ˇ increases.
6. Conclusions
After clearing markets by the MO and allocating losses by the
independent DSO, the profit turned by a VPP should be allocated to
the DERs. Since the risk-aversion level has a significant impact on
this profit, the time horizon of the profit allocation problem begins
with building the offering curves. For this reason, the optimal
offering model of VPP under uncertainty should be run to obtain
the scheduled powers, and then the optimal real-time dispatching
model should be used to calculate energy deviations occurred in the
day-of power delivery. In this paper, considering the desired risk-
aversions of independent DER owners and the scenarios of allocated
losses, and also integrating NDLs into VPP, a novel approach is pro-
posed for the optimal offering/bidding problem of a VPP.
To find out the role of each DER in covering the risks and turning
the total profit, a cooperative Game theory-based approach can be
implemented. For this purpose, the models of offering and real-
time dispatching shall be re-run for all possible combinations of
the DERs.
Irrespective of being dispatchable, a dispersed risk-averse pro-
ducer/consumer prefers to trade less/more energy in the DAM to
have less negative energy deviation all over the market horizon at
the expense of more positive deviation. The numerical results indi-
cate that the integration of two or more types of DERs leads to less
energy deviations too. It corroborates that such an integration acts
as a risk-hedging mechanism which results in a surplus profit.
In the context of a VPP trading in the DAM and BM, as the
owners of the DERs become more risk-averse, a decrease in the
total profit and an increase in the surplus profit (due to more salient
risk-hedging role of the integration) are expected. Increasing the
risk-aversion level, total profit of integrated DERs experiences less
reduction in comparison to uncoordinated DERs.
References
[1] P.M. Sotkiewicz, J.M. Vignolo, Nodal pricing for distribution networks: effi-
cient pricing for efficient enhancing DG, IEEE Trans. Power Syst. 21 (2) (2006)
1013–1014.
[2] K. Shaloudegi, N. Madinehi, S.H. Hosseinian, H. Askarian Abyaneh, A novel pol-
icy for locational marginal price calculation in distribution systems based on
loss reduction allocation using Game theory, IEEE Trans. Power Syst. 27 (2)
(2012) 811–820.
[3] J.M. López-Lezama, A. Padilha-Feltrin, J. Contreras, J.I. Muñoz, Optimal contract
pricing of distributed generation in distribution networks, IEEE Trans. Power
Syst. 26 (1) (2011) 128–136.
[4] M. Parsa Moghaddam, A. Abdollahi, M. Rashidinejad, Flexible demand response
programs modeling in competitive electricity markets, Appl. Energy 88 (9)
(2011) 3257–3269.
[5] H.A. Aalami, M. Parsa Moghaddam, G.R. Yousefi, Modeling and prioritizing
demand response programs in power markets, Electr. Power Syst. Res. 80
(2010) 426–435.
[6] J.M. Lujano-Rojas, C. Monteiro, R. Dufo-López, J.L. Bernal-Agustín, Optimum res-
idential load management strategy for real time pricing (RTP) demand response
programs, Energy Policy 45 (2012) 671–679.
[7] P. Teimourzadeh Baboli, M. Parsa Moghaddam, M. Eghbal, Present status and
future trends in enabling demand response programs, in: IEEE PES General
Meeting, San Diego, CA, 2011.
[8] S. Sučić, T. Dragičević, T. Capuder, M. Delimar, Economic dispatch of virtual
power plants in an event-driven service-oriented framework using standards-
based communications, Electr. Power Syst. Res. 81 (2011) 2108–2119.
[9] H. Yang, D. Yi, J. Zhao, Z.Y. Dong, Distributed optimal dispatch of virtual
power plant via limited communication, IEEE Trans. Power Syst. 28 (3) (2013)
3511–3512.
[10] P. Asmus, Microgrids, virtual power plants and our distributed energy future,
Electr. J. (2010) 72–82.
[11] E. Mashhour, S.M. Moghaddas-Tafreshi, A review on operation of micro grids
and virtual power plants in the power markets, in: IEEE 2nd Conference on
Adaptive Science and Technology, Accra, 2009.
[12] A.J. Conejo, M. Carrión, J.M. Morales, Decision Making Under Uncertainty in
Electricity Markets, Springer, 2010.
[13] A. Karimi Varkani, A. Daraeepour, H. Monsef, A new self-scheduling strategy
for integrated operation of wind and pumped-storage power plants in power
markets, Appl. Energy 88 (2011) 5002–5012.
[14] N. Zhang, Ch. Kang, D.S. Kirschen, Q. Xia, W. Xi, J. Huang, Q. Zhang, Plan-
ning pumped storage capacity for wind power integration, IEEE Trans. Sustain.
Energy 4 (2) (2013) 393–401.
378 S.R. Dabbagh, M.K. Sheikh-El-Eslami / Electric Power Systems Research 121 (2015) 368–378
[15] H. Ding, Z. Hu, Y. Song, Stochastic optimization of the daily operation of wind
farm and pumped-hydro-storage plant, Renew. Energy 48 (2012) 571–578.
[16] New York ISO, 2014, http://www.nyiso.com
[17] European Energy Exchange, 2014, http://www.eex.com
[18] I. Gerami Moghaddam, M. Nick, F. Fallahi, M. Sanei, S. Mortazavi, Risk-
averse profit-based optimal operation strategy of a combined wind
farm–cascade hydro system in an electricity market, Renew. Energy 55 (2013)
252–259.
[19] J. García-González, R.M. Ruiz de la Muela, L.M. Santos, A.M. González, Stochastic
joint optimization of wind generation and pumped-storage units in an electric-
ity market, IEEE Trans. Power Syst. 23 (2) (2008) 460–468.
[20] J.M. Morales, A.J. Conejo, J. Pérez-Ruiz, Short-term trading for a wind power
producer, IEEE Trans. Power Syst. 25 (1) (2010) 554–564.
[21] P. Pinson, C. Chevallier, G.N. Kariniotakis, Trading wind generation from short-
term probabilistic forecasts of wind power, IEEE Trans. Power Syst. 22 (3) (2007)
1148–1156.
[22] M. Shahidehpour, T. Li, Z. Li, Risk-constrained bidding strategy with stochastic
unit commitment, IEEE Trans. Power Syst. 22 (1) (2007) 449–458.
[23] D. Das, B.F. Wollenberg, Risk assessment of generators bidding in day-ahead
market, IEEE Trans. Power Syst. 20 (1) (2005) 416–424.
[24] M. Carrión, A.B. Philpott, A.J. Conejo, J.M. Arroyo, A stochastic programming
approach to electric energy procurement for large consumers, IEEE Trans.
Power Syst. 22 (2) (2007) 744–754.
[25] A.J. Conejo, M. Carrión, Risk-constrained electricity procurement for a large
consumer, IEE Proc. Gener. Transm. Distrib. 153 (4) (2006) 407–413.
[26] H. Pandžić, J.M. Morales, A.J. Conejo, I. Kuzle, Offering model for a virtual power
plant based on stochastic programming, Appl. Energy 105 (2013) 282–292.
[27] M. Peik-Herfeh, H. Seifi, M.K. Sheikh-El-Eslami, Decision making of a virtual
power plant under uncertainties for bidding in a day-ahead market using point
estimate method, Electr. Power Energy Syst. 44 (2013) 88–98.
[28] M. Peik-Herfeh, H. Seifi, M.K. Sheikh-El-Eslami, Two-stage approach for optimal
dispatch of distributed energy resources in distribution networks considering
virtual power plant concept, Int. Trans. Electr. Energy Syst. 24 (1) (2012) 43–63.
[29] H. Pandžić, I. Kuzle, T. Capuder, Virtual power plant mid-term dispatch opti-
mization, Appl. Energy 101 (2013) 134–141.
[30] E. Mashhour, S.M. Moghaddas-Tafreshi, Bidding strategy of virtual power plant
for participating in energy and spinning reserve markets—Part I: Problem for-
mulation, IEEE Trans. Power Syst. 26 (2) (2011) 949–956.
[31] G. Erli, K. Takahasi, L. Chen, I. Kurihaa, Transmission expansion cost allocation
based on cooperative Game theory for congestion relief, Electr. Power Energy
Syst. 27 (2005) 61–67.
[32] R. Bhakar, V.S. Sriram, N.P. Padhy, H.O. Gupta, Probabilistic Game approaches
for network cost allocation, IEEE Trans. Power Syst. 25 (1) (2010) 51–58.
[33] M. Jahangoshai Rezaee, A. Moini, A. Makui, Operational and non-operational
performance evaluation of thermal power plants in Iran: a Game theory
approach, Energy 38 (2012) 96–103.
[34] Z. Hu, L. Chen, D. Gan, D. Chattopadhyay, Allocation of unit start-up costs using
cooperative Game theory, IEEE Trans. Power Syst. 21 (2) (2006) 653–662.
[35] M. Junqueira, L. Carlos da Costa, L.A. Barroso, G.C. Oliveira, L.M. Thomé, M.V.
Pereira, An Aumann–Shapley approach to allocate transmission service cost
among network users in electricity markets, IEEE Trans. Power Syst. 22 (4)
(2007) 1532–1546.
[36] M.S.S. Rao, S.A. Soman, P. Chitkara, R.K. Gajbhiye, N. Hemachandra, B.L.
Menezes, Min–max fair power flow tracing for transmission system usage cost
allocation: a large system perspective, IEEE Trans. Power Syst. 25 (3) (2010)
1457–1468.
[37] E.S. Hoji, A. Padilha-Feltrin, J. Contreras, Reactive control for transmission over-
load relief based on sensitivity analysis and cooperative Game theory, IEEE
Trans. Power Syst. 27 (3) (2012) 1192–1203.
[38] D. Kang, B.H. Kim, D. Hur, Supplier bidding strategy based on non-cooperative
Game theory concepts in single auction power pools, Electr. Power Syst. Res.
77 (2007) 630–636.
[39] D. Songhuai, Z. Xinghua, M. Lu, X. Hui, A novel Nucleolus-based loss allocation
method in bilateral electricity markets, IEEE Trans. Power Syst. 21 (1) (2006)
28–33.
[40] N.X. Jia, R. Yokoyama, Profit allocation of independent power producers based
on cooperative Game theory, Electr. Power Energy Syst. 25 (2003) 633–641.
[41] J.M. Morales, S. Pineda, A.J. Conejo, M. Carrión, Scenario reduction for futures
market trading in electricity markets, IEEE Trans. Power Syst. 24 (2) (2009)
878–888.
[42] A.J. Conejo, F.D. Galiana, I. Kockar, Z-bus loss allocation, IEEE Trans. Power Syst.
16 (1) (2001) 105–110.
[43] P.M. Costa, M.A. Matos, Loss allocation in distribution networks with embedded
generation, IEEE Trans. Power Syst. 19 (1) (2004) 384–389.
[44] E. Carpaneto, G. Chicco, J.S. Akilimali, Characterization of the loss allocation
techniques for radial systems with distributed generation, Electr. Power Syst.
Res. 78 (2008) 1396–1406.
[45] Z. Ghofrani-Jahromi, Z. Mahmoodzadeh, M. Ehsan, Distribution loss allocation
for radial systems including DGs, IEEE Trans. Power Deliv. 29 (1) (2014) 72–80.
[46] http://www.ee.washington.edu/research/pstca/pf30/pg tca30bus.htm
[47] IBM ILOG CPLEX 12.6, 2013, http://www.cplex.com
[48] The GAMS Development Co., 2013, http://www.gams.com