Electrical Power and Energy Systems 105 (2019) 237–248

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

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

Uncertainty-based electricity procurement by retailer using robust T

optimization approach in the presence of demand response exchange
⁎
Sayyad Nojavan , Ramin Nourollahi, Hamed Pashaei-Didani, Kazem Zare
Faculty of Electrical and Computer Engineering, University of Tabriz, P.O. Box: 51666-15813, Tabriz, Iran

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

Keywords: In this paper, robust optimization approach is proposed to handle market price uncertainty in which the upper
Demand response (DR) exchange deviation from forecasted value of pool price will be considered for risk analysis for a retailer. Objective of this
Electricity retailer paper is minimization of energy procurement cost for retailer from pool market, forward contracts and demand
Forward DR response programs (DRP). Therefore, three new designs of demand response (DR) programs have been proposed
Pool-order DR
in this study which retailer can use to procure their required energy. These new DR schemes consist of pool-order
Reward-based DR
Robust optimization approach
option, forward-DR and reward-based DR contracts. The robust optimization approach examines the retailer’s
performance at risk-averse and risk-neutral strategies, in which risk-neutral explains the normal performance of
retailer, and risk-averse explains the risky performance of retailer. The proposed robust scheduling of retailer is
modeled via MIP model which can be solved using CPLEX solver under GAMS software. The achieved results
show that the retailer cost in risk-neutral strategy is reduced due to use of new DR schemes. Also, in risk-averse
strategy, retailer cost reduction is more than the risk-neutral strategy use of new DR schemes.

1. Introduction programming framework based on risk-constrained is presented to

Before of the creating demand-side management (DSM) plans, re-
tailers can be only procured their energy through the pool market and
forward contracts [1]. But, due to the existing uncertainty in the pool
market price, this option is difficult to use in the day-ahead market by
retailers. In addition, after creating of demand side management (DSM),
retailers are able to supply some of their energy using of DR programs
[2]. The most important benefits of using DR are to flatten load curves
and reduce energy cost in peak hours [3]. Also, DRP is an efficient
option for risk and cost reduction for retailer under uncertainty con-
dition.
1.1. Literature review
In [4], a solution is presented to find optimal energy supply for
electricity retailers based on binary imperialist competitive algorithm
and binary particle swarm optimization. Also, pool market and bilateral
contracts are used in [5] to obtain an optimal strategy of electricity
retailers to procure their energy in electricity market. In [6], the pro-
blems of setting up contracts on the suppliers and end-user side to aim
of maximizing the profits of the retailer are reviewed in which the result
are presented at an acceptable level of risk. In [7], stochastic
choose forward contracts which retailer should sign in order to max-
imize their benefits. In order to consider the uncertainties in both
electricity prices and loads, Ref. [8] proposed a multistage stochastic
optimization approach, which permits the specification of conditional-
value-at-risk requirements to optimize hedging across intermediate
stages in the planning horizons. Also, in [9] a new stochastic approach
is considered to model uncertainty for retailer in order to maximize its
total expected rate of return. Researchers in [10], in order to find the
optimal sale price of electricity and determine the electricity procure-
ment policy of a retailer, a mixed-integer stochastic programming is
presented. In [11], constraint of financial risk associated with the
market price uncertainty is considered using of expected downside risk
in the mixed-integer stochastic optimization problem. In [12], to de-
termine the sale price of electricity and manage a portfolio of different
contracts in order to procure its demand, a decision-making framework
is proposed. In [13], from the retailer’s viewpoint, a model is provided
to set price changes in time-of-use tariffs in order to encourage custo-
mers to shift their loads. Ref [14] proposed a bi-level programming
approach in order to solve the medium-term decision-making problem
of retailer. In this ref, a retailer decides its level of involvement in the
futures market and in the pool as well as the selling price offered to its
potential clients with the goal of maximizing the expected profit at a

⁎
Corresponding author.
E-mail addresses: sayyad.nojavan@tabrizu.ac.ir (S. Nojavan), ramin.nourollahi96@ms.tabrizu.ac.ir (R. Nourollahi),
h.pashaei96@ms.tabrizu.ac.ir (H. Pashaei-Didani), kazem.zare@tabrizu.ac.ir (K. Zare).

https://doi.org/10.1016/j.ijepes.2018.08.041
Received 24 January 2018; Received in revised form 15 July 2018; Accepted 23 August 2018
0142-0615/ © 2018 Elsevier Ltd. All rights reserved.
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

Nomenclature NF number of forward contracts

NFB number of blocks in forward contracts
Set NFDR number of contract in forward DR
NJ number of steps in reward-base DR
t time period (week) Npo number of pool-order options

Parameters Variables

pen C (FDR) total cost of forward DR program ($)

f po (t ) penalty of not running pool-order DR in time period t
($/MWh) C (F ) total cost of forward contracts ($)
, MAX
PfDR ,b (t ) highest demand in block b of forward DR f in time period C (PO) total cost of pool-order options ($)
t (MWh) C (P ) total cost of power procurement from pool market ($)
P¯ jDR (t ) demand in jth step of reward-base DR in time period t C (RDR) total cost of reward-base DR
(MWh) P DR (t ) purchased power from reward-base DR in time period t
PfMAX, b (t ) highest demand in block b of forward contract in time (MW)
period t (MWh) P p (t ) purchased power from pool-order in time period t (MW)
MAX
Ppo (t ) highest demand in pool-order DR in time period t (MWh) P FDR (t ) purchased power from block b of forward DR f in time
P req (t ) value of purchased power by retailer in period t (MWh) period t (MW)
R¯ jDR (t ) highest value in jth step of reward-based DR in time period PfDR purchased power from block b of forward contract f in
, b (t )
t ($/MWh) time period t (MW)
λpo (t ) price of pool-order DR in period t ($/MWh) P p (t ) purchased power from the pool market in time period t
λ fDR
, b (t ) price of block b of forward DR f option in time period t (MW)
($/MWh) RDR (t ) value of reward in time period t ($/MWh)
λ fF, b (t ) price of the block b of forward contract f in time period t RjDR (t ) value of reward of step j in time period t ($/MWh)
($/MWh)
∼p vDR, j (t ) binary variable that shows which step is executed in time
λ (t ) forecasted pool market price ($/MWh)
period t
vpo (t ) binary variable which is 1 if pool-order is run in time
Numbers
period t
λ p (t ) actual pool market price ($/MWh)
NBDR number of blocks in forward DR

given risk level. In addition, authors in [15] introduced game theory as
an efficient method for the optimal operation of home MGs, which is
offers an advanced retail electricity market. As a good work, authors in
[16] in an integrated energy system, proposes the pricing and operation
strategy considering of DR for a MG retailer. In [17], information gap
decision theory is used to evaluation different strategies for a retailer
under pool price uncertainty, which this method can be used as a tool
for assessing the risk, levels, considering whether a retailer is risk-
taking or risk-averse regarding its midterm strategies. Ref. [18] paid to
explanation of demand response (DR) programs in deregulated elec-
tricity markets, which the definition and the classification of DR as well
as potential benefits and associated cost components are presented. Ref.
[19] analyzes the effect that the market structure can have on the
elasticity of the demand for electricity. It then describes how the con-
sumers' behavior can be modeled using a matrix of self- and cross-
elasticities. In [20], a new demand response, which called consumer
preference, based demand response model introduced in a game-theo-
retic framework. In addition, Ref. [21] introduced a clear reserve
market in the presence of uncertain responsive loads using of in-
formation gap decision theory (IGDT) concept.
Robust optimization approach is clearly in uncertainty modeling. In
Robust optimization approach is analyzed solution optimally under
forecasting errors in two risk-neutral strategy and risk-averse strategy
which risk-averse strategy modeled worst condition for uncertain
parameter. Nevertheless, IGDT is analyzed effects of various amounts of
deviation from optimal solution on the uncertain parameter.
In [22], responsive load economic model is presented that is a
model based on price elasticity and customer benefit function. In [23],
formulation and analysis of a new scheme of DR program targeting
retail customers who are equipped with smart meters yet still face a flat
rate. In [24], a two-way digital communication infrastructure proposed
for future, which will be used in the demand-side energy management
system between users. In [25], end-users performances are analyzed in
the restructured electricity market in order to deal with existing threats.
In order to find the optimal energy management scheduling scheme for
each end-users and utility company, a distributed real-time algorithm is
proposed in [26]. The consumer’s hourly behavior in response to hourly
changes in market prices introduced using of optimal analysis in [27].
In [28], the time-of-use poring for the electricity market is used. Also in
the mentioned reference with an illustrative example, the welfare
gains/losses are analyzed after an implementation of TOU pricing
scheme over the single pricing scheme. In [29–33], the technical con-
cepts of DR are discussed. For example, details on the control and
management of electrical loads like air conditioners, water heater, and
cooling systems is expressed. A new method for the exchange of DR has
been created in [34,35], which DR is a public good. In [36], a method
has been developed for the exchange of DR in which DR is trade directly
between the buyer and seller in a pool-base market. In addition, men-
tioned method has been improved in [37], which examine the economic
and technical perspectives of critical peak pricing plan as an active
demand response (DR) program. In [38,39], the formulation of three
known types of DR include load curtailment, load shifting and fuel
substitution is introduced in which consumers can decide participation
in mentioned DR programs. In order to evaluate the capacity of load
curtailment in industrial consumers, stochastic programming approach
has been used in [40].
In the retail market, retailer can use DR programs to reduce their
risk. For example, in [41], to control uncertainty in the pool market,
interruptible loads have been used. In [42], two interruptible loads
contracts, pay-in-advance and pay-as-you-go have been evaluated as a
retailer’s energy resource. In [43], self-production is introduced as a
source for reducing the risk of market price fluctuation. In [44], pre-
sents a multiperiod energy acquisition model for a distribution com-
pany (Disco) with distributed generation (DG) and interruptible load
(IL) in a day-ahead electricity market. Also, in [45], interruptible loads
are introduced as energy for distribution companies in the day-ahead
market. Finally, in [46], a robust optimization method is used to pool
price uncertainty modeling in order to obtain optimal bidding strategy

238
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

which is offered to the day-ahead market by retailer which pool-order
DR, forward-DR and reward-base DR programs aren't considered. In
addition, ref [47] presents new model for results on robust optimization
approach under correlated uncertainties.
In most of the article in the above paragraphs, only the concepts of
DR and its formulation are addressed. In the other word, in fewer ar-
ticles the financial concepts of DR have been taken into consideration.
In addition, in few article DR is exchanged as a commodity directly
between the buyer and seller of DR. Also, in fewer articles in above paid
to DR from retailer point of view.
According to reviewed papers in above, Table 1 can be presented in
order to clarify comparing different between reviewed papers.
1.2. Novelty and contributions
In this paper, a new method is introduced to model pool market
price uncertainty for a retailer in addition to previous articles, which is
as follows:
Direct DRP exchange between the buyer and seller, can be in-
troduced as a public benefit between the buyer and seller. For this
propose, three type of DR is introduced which retailer use to trade DR
with the consumers. These three options include pool-order option,
forward-DR option and reward-base DR. These schemes include long-
term and short-term contracts that the retailer can be used to procure
their needed energy. From view point of difference between this work
with previous works, proposed DR scheme allows a retailer to decide
how to buy DR from aggregators and consumers. The details of these
three contracts are as follows: pool-order has been developed using the
financial concepts described in [48,49] where retailers are used when
the pool market price is faced with small deviation from forecasted
value. Forward-DR is a secure contract for the future for retailers with a
specific volume and price. Finally, reward-based DR [50] is a real time
contract that the consumer receives reward for reducing the specified
load which the paid reward is increased by increasing loads reduction.
Finally, a robust optimization approach is proposed to consider the
pool market price uncertainty. This approach analyzes the retailer
performance in both the risk-neutral and risk-averse strategies.
The novelty of this article is briefly presented as follows:
1. Direct trade demand response between retailer and consumers.

Table 1
Reviewed paper compare.
Ref. Considered DRP Uncertain parameter Uncertainty modeling Time period Market player

[1] – 1. Market Price IGDT Short term Retailer

[2] TOU 1. Market Price Scenario Short term Retailer
[3] TOU 1. Market Price Scenario Short term Retailer
2. Demand
3. Outage RERs
[5] – 1. Market Price Scenario 1. Mid-term Retailer
2. Demand 2. Short-term
[6] – 1. Demand Scenario Long-term Retailer
[7] – 1. Market Price Scenario Long-term Retailer
[8] – 1. Market Price Scenario Long-term Retailer
[10] – 1. Market Price Scenario Mid-term Retailer
2. Demand
[11] – 1. Market Price Scenario Mid-term Retailer
2. Demand
[12] – 1. Market Price Scenario Long-term Retailer
2. Demand
[13] TOU 1. Market Price Scenario Mid-term Retailer
2. Demand
[14] – 1. Market Price Scenario Mid-term Retailer
2. Demand
3. Rival-retailer prices
[17] TOU Market Price IGDT Mid-term Retailer
[20] 1. RTP, 2. TOU – – Short-term Consumers
[22] 1. RTP, 2. TOU, 3. CPP – – Short-term Consumers
[23] 1. RTP, 2. TOU – – Short-term Consumers
3. CPP, 4. PLP
[24] 1. RTP, 2. TOU 1. Market Price Scenario Short-term Utility
3. CPP 2. Demand company and its Customers/users
[27] RTP 1. Market Price Robust Short-term Consumers
[28] TOU – – Short-term Retailer
[29] TOU – – Short-term Consumers
[32] Physical-based DR – – Short-term Consumers
[33] DR of Water Load – – Short-term Consumers
[34] Pool-Based DR – – Short-term DR buyers and sellers
[35] DR exchange (DRX) – – Short-term DR buyers and DR sellers
[37] CPP – – Short-term Energy service provider
[38] CPP – – Short-term Utilities
[39] Other DRP 1. Demand Scenario Mid-term Consumers
[40] – 1. Demand Scenario Short-term Consumers
[41] – 1. Market Price Scenario Mid-term Retailer
[42] – 1. Market Price Scenario Long-term Large consumer
[43] – 1. Market Price Scenario Short-term Retailer
[44] 1. RTP, 2. TOU – – Short-term Retailer
[49] 1. Reward-based 1. Demand Scenario Short-term Retailer
[50] – – – Short-term Ancillary service
[51] – – – Short-term Electricity markets
This paper 1. Pool-order DR 1. Pool market price Robust optimization approach Medium-term Electricity retailer
2. Forward DR
3. Reward-based DR

239
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

2. Pool-order DR, forward DR and reward-base DR are proposed as

new DRP scheme.

1.3. Paper organization

The remainder of this paper includes the following Sections: Section

2 introduces the optimal retailer performance, objective function and
mathematical formulation of the new DR schemes. The introduction
and formulation of robust optimization approach is proposed in Section
3 and it is applied to the problem of optimal retailer performance in
same section. The input data, the detailed results and discussion are
presented in Section 4. Finally, the conclusion of paper is presented in
Section 5.

2. Problem formulation

Retailer can procure their energy in electricity market from different

ways such as pool market or bilateral contracts. After creating of de-
mand response (DR) program based on demand side management
(DSM), the retailer is able to use DR as a proper option to procure their
energy use of consumers demand reduction. Demand response program
can be provide several benefits for retailer such as reducing of power
procurement cost in the peak times and risk. For example, Use of DRP
makes to reduce power purchase from pool markets that have un-
certainty. Therefore, risk and cost reduction will be increased in the
pool market price uncertainty condition. In addition, risky retailers can
be use of DR to reduce their damage in the case of an error in the
forecast. In this paper, three types of DR programs are proposed for use
of retailer. Retailer can use these three schemes to buy DR from con-
sumers. The structure of new DR scheme and how trade between the
retailer and consumers are shown in Fig. 1. These DR programs are
signed in the form of long-term and short-term contracts between the
consumers and the retailer. Here, aggregators are only allowed to sell
DR toy retailer but in principle, they can sell their DR in different
markets such as pool market or ancillary services [51,52].
2.1. Objective function and power balance constraint
In this paper, the objective is reducing the retailer cost that is pre-
sented in Eq. (1) as the objective function. The objective function in-
volves the purchase costs of new DR scheme (pool-order option, for-
ward-DR, and reward-based DR) as well as the costs of purchasing from
the pool market and bilateral contracts.
Min C (p , λ ) = C (P ) + C (F ) + C (PO) + C (FDR) + C (RDR) (1)
The power balance Equation is provided in (2) in which in it re-
quired energy of retailer is equal to the total power purchased from the
market (pool market and bilateral contracts) and new DR schemes
Fig. 2. Structure of pool-order option.
(pool-order option, forward-DR contracts and reward-base DR).
P erq (t ) = P p (t ) + P F (t ) + P total (t ) + P FDR (t ) + P DR (t ) (2)
2.2. New DR schemes
The retailers procure part of their power using of introduced three
new DR options. These new DR schemes consist of pool-order option,
forward-DR and reward-base DR.
2.2.1. Pool-order option
The pool-order option flowchart can be seen in Fig. 2. This option is
used by retailer at the time of few increasing in the pool market prices.
Retailer after signing this contract does not coercion to implement it. In
other words, the retailer decides when implementing the contract on
the execution or not execution of the contract.
Actually, the implementation of the contract depends on the pool
market price at the time of execution of the contract. If energy purchase
price from pool-order option is less than the energy purchase price from
the pool market, retailer act to the pool-order option contract.
Nevertheless, if energy purchased price from pool-order option is
higher than the energy purchased price from the pool market, the re-
tailer will be dissuaded from acting to the pool-order contract. Then by
paying amount of penalties to consumers, it will buy its energy from the
pool market. The total cost formulation of the pool-order option defined

Fig. 1. New DR scheme for energy supply of retailer.

240
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

as follows: curve can be selected. Total cost of reward-base DR is defined as fol-

Npo lows:
pen
C (PO) = ∑ ∑ [Ppo (t ). λpo (po). νpo (t ). d (t ) + (1−νpo (t )). f po (t )] NJ
⎡ ⎤
t ∈ T po = 1 C (RDR) = ∑ ⎢∑ P¯ jDR (t ). RjDR (t )⎥
(3) t∈T ⎣ j=1 ⎦ (13)
Max
0 ⩽ Ppo (t ) ⩽ Ppo (t ) ∀ po, t (4)
2.3. Wholesale market suppliers
Npo
total
Ppo (t ) = ∑ Ppo (t ). νpo (t ) ∀t
In addition to DR programs, the retailer can use market options such
po = 1 (5)
as forward contract and pool market to procure their energy. Modeling
Eq. (3) shows the retailer cost for the execution and not execution of of these options is provided in bellow.
pool-order contract. The first part of Eq. (3) shows the cost of contract
execution, and the second part shows the cost of not executed contracts 2.3.1. Pool market
(penalties). Power bound of pool-order option contract is shown in Eq. Real-time Pool market is considered as important energy source for
(4). The total demand reduced in executed pool-order contracts is retailers which retailer can buy their needed energy at any moment. In
shown in Eq. (5). this paper, the market price is uncertain parameter, which robust op-
timization approach is proposed to model this uncertainty in order to
2.2.2. Forward DR obtain robust scheduling of retailer in the uncertain environment.
Forward contracts is an agreement that is contracted for future Eq. (14) is used to calculate the cost of purchasing energy from the
periods which is created by combining forward contracts with DR. pool market.
Forward-DR contracts are considered to be safe contracts for retailers
with a defined volume and price [41]. Forward-DR contracts are de-
C (P ) = ∑ P p (t ). λ p (t )
t∈T (14)
signed in different blocks, which are offered by consumers’ aggregator.
Because, in this paper, the DR is traded directly between the retailer
and aggregator, so the price of the forward-DR contracts is determined 2.3.2. Forward contract
by negotiation between the retailer and aggregators. The total cost Forward contract is a bilateral agreement, which will be created
formulation of the forward-DR defined as follows: between retailers and aggregators for future periods. These contracts
NFDR NBDR
have specific price and volume that will increase as a stepwise trend in
C (FDR) = ∑ ∑ ∑ PfDR DR blocks.
, b (t ). λ f , b (t )
t∈T f =1 b=1 (6) Eq. (15) is used to calculate the cost of purchasing energy from the
all blocks of forward contracts.
Eq. (6) shows the total cost of all executed blocks of forward-DR
NF NFB
contracts in all periods. Other forward-DR constraints are as follows:
C (F ) = ∑ ∑ ∑ PfF,b (t ). λfF,b (t )
0 ⩽ PfDR DR, Max
, b (t ) ⩽ P f , b (t ) ∀ f , b, t (7) t∈T f =1 b=1 (15)

NFDR NBDR 0 ⩽ PfF, b (t ) ⩽ PfMAX

, b (t ) ∀ f , b, t (16)
P FDR (t ) = ∑ ∑ PfDR
, b (t ) ∀t
NF NFB
f =1 b=1 (8)
P F (t ) = ∑ ∑ PfF,b (t ) ∀t
Eq. (7) shows the power band of each block from the forward-DR f =1 b=1 (17)
contracts. The total demand reduced by forward-DR contracts can be
Similar to other contracts, power amount for each blocks of forward
calculated as Eq. (8).
contract is shown in Eq. (16). Also, total energy of forward contracts is
shown in Eq. (17).
2.2.3. Reward-based DR
Retailer pays reward to consumers in the reward-base DR for each
3. Robust optimization based scheduling of retailer
reduction in their demand. By increasing consumers demand reduction,
the retailer paid reward is increased. Therefore, reward-base DR de-
Many methods are used in literature to analyze uncertainty over the
pends on consumer behavior in amount of load reduction. The reward-
base DR curve is shown in Fig. 3.
Reward-base DR is modeled as follows:
NJ
P DR (t ) = ∑ P¯ jDR (t ). νDR,j (t ) ∀t
j=1 (9)
NJ
RDR (t ) = ∑ RjDR (t ) ∀t
j=1 (10)

R¯ jDR DR ¯ DR
− 1 (t ). νDR, j (t ) ⩽ Rj (t ) ⩽ Rj (t ). νDR, j (t ) ∀ j, t (11)
NJ
∑ νDR,j (t ) = 1 ∀t
j=1 (12)
Eq. (9) shows the total demand reduced by consumers in reward-
base DR. The total reward paid by retailer is defined in Eq. (10). Also,
the amount of reward in each step of reward-base DR is determined by
Eq. (11). Eq. (12) states that only one-step from the reward-base DR Fig. 3. Reward-base DR curve.

241
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

years. The most commonly used methods are stochastic probabilistic ωt ⩾ x t ∀ t = 1, ...,n (28)
methods, possibilistic methods and hybrid probabilistic-possibilistic.
The above formulation (22)–(26) is derived from the duality theory
The method mentioned above require a lot of information about un-
[55] and specially detailed liner robust description are given in [55,56].
certain parameters, which for this reason, they are more used than the
In the solved base problem that includes Eqs. (1)–(17), the uncertain
other mentioned methods.
parameter is the only pool market price. In order to model of this un-
In order to modeling of uncertainty parameters, a new method that
certain parameter, the robust optimization approach is used that the
called robust optimization approach has just been provided. The robust
formulation is as follows:
optimization method investigates the effect of an uncertain parameter
on the optimal result, which target is to reduce the sensitivity of the Min C (p , λ ) = [Ec (P ) + C (F ) + C (PO) + C (FDR) + C (RDR)] + β . Γ0
optimal result to the uncertain parameter. This approach has been used + ∑ ζt
in recent years as a substitute for stochastic programming to addressing t∈T (29)
data uncertainty in mathematical programming mode.
The worst case is used to deal with uncertainty in the robust opti- Eqs.(2)−(17) (30)
mization approach while scenario production is used in stochastic
β + ζt ⩾ (λtmax −λtmin ). ωt ∀ t (31)
programming. The probability of the uncertain parameter must be de-
termined in the stochastic programming which is a big challenge in the ωt ⩾ P p (t ) ∀ t = 1, ...,T (32)
stochastic programming. Therefore, robust optimization approach has
the following advantages [53]: where β and ζt are two other variables the problem which is to con-
sidering variation of λt , and ωt is an auxiliary variable for the formation
(1) For a specific problem, the computation is less and the result is of linear expressions. Finally Γ0 has a value in the range of [0, 32] if
more effective than the stochastic programming. (λtmax −λtmin ) ⩾ 0 , while Γ0 = 0 if (λtmax −λtmin ) = 0 .
(2) Because of the considering worst conditions, its result is very reli-
able. 4. Case study
(3) There is no need to probabilistic distribution function.
The retailer performance under the uncertainty of the pool market
Robust optimization approach is a risk management method that price is examined in continue of paper. Used method in this paper,
has a low computing volume those other methods. firstly the optimal results are obtained for forecasted price. Then, robust
The standard MILP formulation of proposed model based on Eqs. optimization approach has been used to analyzing performance of re-
(1)–(17) is defined as follows [54]: tailer under the different risk conditions relative to pool price deviation
n from forecasted value. This problem are modeled as mixed-integer
Min ∑ ct . xt programming (MIP) which can be solved using CPLEX solver [57]
t=1 (18) under the GAMS optimization software on Intel(R) Core(TM) i7-7500U
CPU @ 2.70 GHz (4 CPUs), ∼2.9 GHz, RAM 8 GB system in the 30 s of
s.t
time.
n
∑ ait . x t ⩽ bi ∀ i = 1, ....m
4.1. Data
j=1 (19)

xt ⩾ 0 ∀ t = 1, ...,n (20) Thirty-two periods are considered in this article, which is included
32 weeks of the one year. In fact, eight months from a year that includes
x t ∈ {0, 1} for some t = 1, ...,n (21) 32 weeks. Each of these periods is calculated from the average of peak
where known coefficient ct is considered for the objective function, times of one week. These periods are composed of 12 weeks of
which is an unknown value, and a known bound. The value considered January–March, 17 weeks of June–September and 3 weeks of
for c, are in range [ct −dt, ct+dt] which is d is a deviation from the December. Also, daily peak hours consist of, from 11 am to 9 pm in
nominal value of ct. In addition to formulating the robust mixed-integer summer days, while those of winter days are from 6 am–10 am to
liner programming, it is necessary to define an integer control para- 4 pm–10 pm. It is noteworthy that these peak times are calculated only
meter. Γ0 (GAMA) represents integer control parameter, which is a real for Monday to Friday of each week. According to [58,59], load data in
value in the interval of [0, |T0|]. The parameter Γ0 controls the trade-off this paper come from the Queensland daily curve in 2012 [60]. Input
between the probability of the variation of the variable parameter and data of required demand and forecasted market price for retailer in
its effects on the objective function of the base problem. If Γ0 = 0 re- Figs. 4 and 5 are shown respectively.
flects the risk-neutral strategy and happens for other values the risk- Pool-order option is a cost-efficient option at the time of slight in-
averse strategy. For Γ0 = |T0|, the worst possible deviations have been crease in the pool market price. Four types of pool-order contracts have
occurred and the most conservative strategy has been chosen. been proposed that have a specific price and volume of load reduction.
The robust model of the main problem (18)–(21) is defined as Maximum amount of load reduction in each contract is equal to 50 MW.
follow: Also, if the contract is not-executed, the retailer must pay 15% of the
contract price as penalty. Input data of pool-order option is shown in
n n
Fig. 6. Aggregator has proposed this curve to the retailer in order to
Min ∑ ct . x t + β . Γ0 + ∑ ζt
t=1 t=1 (22) decide for use of the pool-order option.
Forward-DR contract consist of six blocks, each block having a
Eqs. (19)−(21) (23) certain amount of load reduction in which each contract encompass
one-month period. Therefore, eight contracts are set for forward-DR
β + ζt ⩾ dt . ωt t ∈ T0 (24)
contracts to cover the all periods. Maximum load reduction for each
β⩾0 (25) block of forward-DR contract is equal to 75 MW. Table 2, presented the
forward-DR input data. According to this table, number of selected
ζt ⩾ 0 ∀ t = 1, ...,n (26) block is depends on pool market price increase at each period.
In reward-base DR, a reward is paid to consumers for each amount
ωt ⩾ 0 ∀ t = 1, ...,n (27) of reduced load. In this study, 14 steps are considered as a reward

242
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

Pool-order price ($/MWh)

Demand (MWh)

Period (weeks)
Period (week)
Fig. 4. Demand data for the retailer.
Fig. 6. Pool-order option prices.

which is paid to consumers for each reduction of load. Input data of

reward-base DR is shown in Fig. 7. According to this (Fig. 7), proposed Table 2
reward per MWh by retailer to consumer is increased in the high Forward DR prices ($/MWh).
amounts of demands. B1 B2 B3 B4 B5 B6
The forward contract that is signed at the beginning of the period
FDR1 35 37 39 41 43 45
and covers the whole period includes of three contracts (F1-F3). F1
FDR2 32 34 36 38 40 42
encompass the first twelve week of the time horizons (twelve periods). FDR3 29 31 33 35 37 39
F2 also encompass the second 17 weeks of time horizons and finally, F3 FDR4 33 35 37 39 41 43
has been encompassing 3 weeks of December. Also, forward contract FDR5 45 47 49 51 53 55
FDR6 51 53 55 57 59 61
like the forward-DR consist of six blocks which each block contains of
FDR7 56 58 60 62 64 66
specific power and price. In this study, the forward contract price for FDR8 69 71 73 75 77 79
each quarter is determined from data of Queensland in 2012 [60].
Maximum power for each block of forward contract is equal to 450 MW.
Table 3 presents the forward contract input data. According to this according to Fig. 8 and the result obtained in the risk-neutral strategy
table, number of selected block is depends on pool market price in- (Γ0 = 0 ), the retailer cost is 4.68 M$ and 4.87 M$ for the with and
crease at each period. without DR modes, respectively. These results show that the use of DR
in the risk-neutral strategy (Γ0 = 0 ), the retailer cost reduced to the 4%
(reduced cost is equal to 0.19 M$).
4.2. Results Also, according to Fig. 8, and by solving robust function (29) and
taking into account constraints (30)–(32), obtained result for the risk-
The retailer cost is obtained by solving function (1) and taking into averse strategy (Γ0 = 32 ) represent the worst possible condition in the
account constraints (2)–(17). The value obtained in this case is the pool market price. The retailer cost in the worst condition pool market
retailer’s expense in the risk-neutral strategy. Also, in order to illustrate price are equal to 4.9 M$ and 5.16 M$ in with and without DR modes,
the effect of the proposed new DR scheme, this problem has been solved respectively. These results show that the use of DR in the risk-averse
once by with considering DR and once without considering DR.

Fig. 5. Forecasted pool market price.

243
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

order option are shown in Figs. 9 and 10, respectively.

According to Fig. 9, purchased demand from pool-order option in
both risk-neutral and risk-averse strategies is different, which is more in
risk-averse strategy than the risk-neutral strategy. Also, this Fig in-
Load reduction (MWh)

dicated that pool-order DR option is a safe and efficient option for re-
tailer in the uncertainty condition. In other word in the small increase
of pool market price, effects of pool-order option is significant. In ad-
dition Fig. 9 is shown that read line is related to without DRP case
results, which in this case DRP is not considered. Therefore, power
exchange and procurement cost of pool order DR is zero.
According to Fig. 10, the demand purchase cost from pool-order
options in the risk-neutral strategy (Γ0 = 0 ) is less than the risk-averse
strategy (Γ0 = 32 ). The maximum cost is obtain after from Γ0 = 13which
is equal to 0.095 M$.
Purchased demand from forward-DR contracts and traded demand
cost from forward-DR contracts are depicted in Figs. 11 and 12, re-
Offered reward ($ /MWh) spectively. According to Fig. 11, purchased demand from forward-DR
contracts in risk-averse strategy is more than the risk-neutral strategy,
Fig. 7. Paid reward for each reduce demand. which is because of the safety and suitability on forward-DR contracts
in high market prices.
Table 3 According to Fig. 12, the demand purchase cost from forward-DR
Forward contract prices ($/MWh). contracts in the risk-neutral strategy (Γ0 = 0 ) is less than the risk-averse
B1 B2 B3 B4 B5 B6
strategy. The maximum cost is obtain for after from Γ0 = 14 which is
equal to 1.3 M$.
F1 40 45 50 55 60 65 Purchased demand scheduling in the reward-base DR and traded
F2 38 42 46 50 54 58 demand cost from reward-base DR are illustrated in Figs. 13 and 14,
F3 39 44 49 54 59 64
respectively.
According to Fig. 13, reduced demand in reward-base DR in risk-
averse strategy is more than the risk-neutral strategy which is due to
rising energy price in the pool market and the desire to buy DR from the
consumers. Also, according to Fig. 13, it can be shown that at the low
market price periods, retailer does not want to pay reward to consumers
in order to reduce load.
According to Fig. 14, the reduced demand cost in reward-base DR
contracts in the risk-neutral strategy (Γ0 = 0 ) is less than the risk-averse
strategy (Γ0 = 32). In other word, an effect of reward-based DR in the
worst condition of the pool market price is more compared to the
normal condition of pool market price. The maximum cost is occurred
for after from (Γ0 = 14) , which is equal to 0.22 M$.
Purchased energy from pool market and traded power cost from
pool market are illustrated in Figs. 15 and 16, respectively.
According to Fig. 15, purchased power from pool market in risk-
neutral strategy is more than the risk-averse strategy which is due to
rising energy price in the pool market in risk-averse strategy. Also, it
can be seen that in both risk-neutral and risk-averse strategies pur-
chased energy from the pool market in the case with DR is less than the

Fig. 8. Robust operation cost of retailer.

strategy (Γ0 = 32 ) the retailer cost reduced to the 5% (reduced cost is

equal to 0.26 M$).
In addition, according to Fig. 8, it can be shown that the energy
procurement cost of retailer in the robust case as risk-averse strategy is
increased in comparison with non-robust case as risk-neutral strategy.
This increased cost is due to model pool market price uncertainty which
it can be exchanged 20% from forecasted value. In other words, the
risk-averse strategy of retailer can be used in the worst condition
(maximum pool market price).
According to the presented result in above and with respect to
Fig. 9, it can be seen that the effect of the new DR schemes on the worst
saturation (Γ0 = 32) is more than the risk-neutral strategy (Γ0 = 0 ). In
other word, by increasing the pool market price deviation from the base
value, effects of the proposed new DR scheme on the total cost of re-
tailer is greater.
Purchased demand scheduling and traded demand cost from pool- Fig. 9. Traded power from the pool-order DR.

244
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

Fig. 13. Traded power from the reward-base DR.

Fig. 10. Cost of pool-order DR.

Fig. 14. Cost of reward-base DR.

Fig. 11. Traded power from the forward-DR.

Fig. 12. Cost of forward-DR contract.

case without DR. Therefore, retailer in the during of worst condition is

Fig. 15. Traded power from the pool market.
rely on the new DR scheme which proposed in this paper.
As show in Fig. 16, the cost of purchased power from pool market in
the risk-neutral strategy (Γ0 = 0 ) is more than the risk-averse strategy,

245
S. Nojavan et al. Electrical Power and Energy Systems 105 (2019) 237–248

Fig. 16. Cost of pool market.

Fig. 19. Sensitivity of total cost under forward contract price increasing.

Fig. 20. Sensitivity of total cost under pool order price increasing.
Fig. 17. Traded power from the forward contract.

which is due to the increase pool market price in the risk-averse

strategy. So that the retailer cost in risk-averse strategy compared to
risk-neutral strategy is reduced 57%. Also, using of proposed new DR
schemes, pool market cost in risk-neutral strategy 15% and in risk-
averse strategy 30% has been reduced. Also, retailer in the worst con-
dition compared to normal condition, can be more use of the benefits of
the introduced new DR scheme in this paper which illustrated in
Fig. 16.
Energy purchased scheduling and traded power costs from forward
contracts are shown in Figs. 17 and 18, respectively. According to
Fig. 17, purchased power from forward contracts in risk-averse strategy
is more than the risk-neutral strategy which is due to increase energy
price in the pool market in risk-averse strategy. This market price in-
crease will be desire to buy energy from the forward contracts and DR
options. Also, it can be seen that in both risk-neutral and risk-averse
strategies, purchased energy from the forward contracts in the case with
DR is less than the without DR. Therefore, use of DR makes to reduce
contracted power (forward contract) by retailer.
As show in Fig. 18, the cost of purchased power from forward
contracts in the risk-neutral strategy (Γ0 = 0 ) is lower than the risk-
Fig. 18. Cost of forward contract.

246
S. Nojavan et al.
Fig. 21. Sensitivity of total cost under forward-DR price increasing.
averse strategy (Γ0 = 32), which is due to the increase pool market price
in the risk-averse strategy. So that the retailer cost in risk-averse
strategy compared to risk-neutral strategy is increased 20%. Also, using
of proposed new DR schemes, forward contract cost in risk-neutral
strategy and in risk-averse strategy has been reduced 64% and 61%,
respectively.
4.3. Sensitivity analyze
In this subsection, sensitivity analyze is conducted on the input
parameter such as pool-order price, forward-DR price and forward
contract price and summary of sensitivity are presented as follow; this
analyze is applied under 10% and 20% increase on the these prices and
obtained results has been illustrated in Figs. 19–21. According to
Fig. 19 that illustrated the sensitivity of total cost under forward con-
tract price increasing can be shown that the effect of forward contract
price increasing on the total cost is greater than the other input para-
meters which are shown in Figs. 20 and 21. Also, according to Fig. 20
which illustrated sensitivity of total cost under pool order price in-
creasing, effects of pool-order price on the total cast is very low com-
pared to forward contract price and forward DR price effects.
4.4. Discussion
According to work donned in this paper, it can be concluded that
robust optimization approach can be used as a risk assessment method
in order to considering various condition in the uncertain parameter
and effects of these condition on the retailer strategies. In addition,
indicated that proposed DRP schemes is more efficient on retailer risk in
the uncertain condition. On other word, position effects of these
schemes are more in the worst condition of pool market price un-
certainty.
5. Conclusion
In this paper, retailer strategies were analyzed to meet their energy
needs. Actually, in this article the retailer performance was compared at
two levels of risks (risk-neutral and risk-averse strategies). The used
tool in this study for this purpose is robust optimization approach. This
approach examines the uncertainty of pool market price, which is the
only uncertain factor in this study. Also, in this paper new DR schemes
are presented and its impact on retailer cost is examined. According to
the obtained results, it has be shown that in the risk-neutral strategy,
247
Electrical Power and Energy Systems 105 (2019) 237–248
because of the use of new DR schemes, the retailer cost reduced from
4.87 M$ to 4.68 M$ which represent a 4% reduce in retailer cost. Also,
in risk-averse strategy, new DR schemes reduced the retailer cost from
5.16 M$ to 4.9 M$ which represent a 5% reduce in retailer cost. So, it
can be concluded that cost reduction in the risk-averse strategy is more
than the risk-neutral strategy, which this more cost reduction in risk-
averse strategy is due to the increased pool market price in risk-averse
strategy and positive effect of new DRP scheme on the retailer cost. In
addition, the results show that the retailer tends to be able to procure
their energy from forward contracts and DR program in the risk-averse
strategy. According to the obtained result, it can be seen that the most
purchasing power is by retailer from forward contract which is due to
the security this contract. Finally, robust optimization approach results
showed that in the risk-averse strategy purchased power from pool
market is reduced 54%.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.ijepes.2018.08.041.
References
[1] Nojavan S, Zare K, Mohammadi-Ivatloo B. Risk-based framework for supplying
electricity from renewable generation-owning retailers to price-sensitive customers
using information gap decision theory. Int J Electr Power Energy Syst
2017;31(93):156–70.
[2] Nojavan S, Zare K, Mohammadi-Ivatloo B. Optimal stochastic energy management
of retailer based on selling price determination under smart grid environment in the
presence of demand response program. Appl Energy 2017;1(187):449–64.
[3] Nojavan S, Zare K, Mohammadi-Ivatloo B. Selling price determination by electricity
retailer in the smart grid under demand side management in the presence of the
electrolyser and fuel cell as hydrogen storage system. Int J Hydrogen Energy
2017;42(5):3294–308.
[4] Nojavan S, Mehdinejad M, Zare K, Mohammadi-Ivatloo B. Energy procurement
management for electricity retailer using new hybrid approach based on combined
BICA–BPSO. Int J Electr Power Energy Syst 2015;31(73):411–9.
[5] Nazari M, Foroud AA. Optimal strategy planning for a retailer considering medium
and short-term decisions. Int J Electr Power Energy Syst 2013;45(1):107–16.
[6] Gabriel SA, Conejo AJ, Plazas MA, Balakrishnan S. Optimal price and quantity
determination for retail electric power contracts. IEEE Trans Power Syst
2006;21(1):180–7.
[7] Carrion M, Conejo AJ, Arroyo JM. Forward contracting and selling price determi-
nation for a retailer. IEEE Trans Power Syst 2007;22(4):2105–14.
[8] Kettunen J, Salo A, Bunn DW. Optimization of electricity retailer's contract portfolio
subject to risk preferences. IEEE Trans Power Syst 2010;25(1):117–28.
[9] Charwand M, Gitizadeh M, Siano P. A new active portfolio risk management for an
electricity retailer based on a drawdown risk preference. Energy.
2017;1(118):387–98.
[10] Hatami AR, Seifi H, Sheikh-El-Eslami MK. Optimal selling price and energy pro-
curement strategies for a retailer in an electricity market. Electric Power Systems
Research. 2009;79(1):246–54.j.
[11] Ahmadi A, Charwand M, Aghaei J. Risk-constrained optimal strategy for retailer
forward contract portfolio. Int J Electr Power Energy Syst 2013;31(53):704–13.
[12] Hatami A, Seifi H, Sheikh-El-Eslami MK. A stochastic-based decision-making fra-
mework for an electricity retailer: time-of-use pricing and electricity portfolio op-
timization. IEEE Trans Power Syst 2011;26(4):1808–16.
[13] Garcia-Bertrand R. Sale prices setting tool for retailers. IEEE Trans Smart Grid
2013;4(4):2028–35.
[14] Carrion M, Arroyo JM, Conejo AJ. A bilevel stochastic programming approach for
retailer futures market trading. IEEE Trans Power Syst 2009;24(3):1446–56.
[15] Marzband M, Javadi M, Pourmousavi SA, Lightbody G. An advanced retail elec-
tricity market for active distribution systems and home microgrid interoperability
based on game theory. Electr Power Syst Res 2018;30(157):187–99.
[16] Jin M, Feng W, Marnay C, Spanos C. Microgrid to enable optimal distributed energy
retail and end-user demand response. Appl Energy 2018;15(210):1321–35.
[17] Charwand M, Moshavash Z. Midterm decision-making framework for an electricity
retailer based on information gap decision theory. Int J Electr Power Energy Syst
2014;31(63):185–95.
[18] Albadi MH, El-Saadany EF. A summary of demand response in electricity markets.
Electr Power Syst Res 2008;78(11):1989–96.
[19] Kirschen DS, Strbac G, Cumperayot P, de Paiva Mendes D. Factoring the elasticity of
demand in electricity prices. IEEE Trans Power Syst 2000;15(2):612–7.
[20] Pal S, Thakur S, Kumar R, Panigrahi BK. A strategical game theoretic based demand
response model for residential consumers in a fair environment. Int J Electr Power
Energy Syst 2018;30(97):201–10.
[21] Ghahary K, Abdollahi A, Rashidinejad M, Alizadeh MI. Optimal reserve market
clearing considering uncertain demand response using information gap decision
theory. Int J Electr Power Energy Syst 2018;31(101):213–22.
S. Nojavan et al.
[22] Aalami HA, Moghaddam MP, Yousefi GR. Modeling and prioritizing demand re-
sponse programs in power markets. Electr Power Syst Res 2010;80(4):426–35.
[23] Zhong H, Xie L, Xia Q. Coupon incentive-based demand response: theory and case
study. IEEE Trans Power Syst 2013;28(2):1266–76.
[24] Mohsenian-Rad AH, Wong VW, Jatskevich J, Schober R, Leon-Garcia A.
Autonomous demand-side management based on game-theoretic energy consump-
tion scheduling for the future smart grid. IEEE Trans Smart Grid 2010;1(3):320–31.
[25] Sharifi R, Fathi SH, Vahidinasab V. A review on demand-side tools in electricity
market. Renew Sustain Energy Rev 2017;1(72):565–72.
[26] Hu M, Xiao JW, Cui SC, Wang YW. Distributed real-time demand response for en-
ergy management scheduling in smart grid. Int J Electr Power Energy Syst
2018;31(99):233–45.
[27] Conejo AJ, Morales JM, Baringo L. Real-time demand response model. IEEE Trans
Smart Grid 2010;1(3):236–42.
[28] Celebi E, Fuller JD. Time-of-use pricing in electricity markets under different
market structures. IEEE Trans Power Syst 2012;27(3):1170–81.
[29] Chu CM, Jong TL. A novel direct air-conditioning load control method. IEEE Trans
Power Syst 2008;23(3):1356–63.
[30] Thomas AG, Jahangiri P, Wu D, Cai C, Zhao H, Aliprantis DC, et al. Intelligent
residential air-conditioning system with smart-grid functionality. IEEE Trans Smart
Grid 2012;3(4):2240–51.
[31] Lu N, Chassin DP. A state-queueing model of thermostatically controlled appliances.
IEEE Trans Power Syst 2004;19(3):1666–73.
[32] Shao S, Pipattanasomporn M, Rahman S. Development of physical-based demand
response-enabled residential load models. IEEE Trans Power Syst
2013;28(2):607–14.
[33] Heffner GC, Goldman CA, Moezzi MM. Innovative approaches to verifying demand
response of water heater load control. IEEE Trans Power Delivery
2006;21(1):388–97.
[34] Nguyen DT, Negnevitsky M, De Groot M. Pool-based demand response ex-
change—concept and modeling. IEEE Trans Power Syst 2011;26(3):1677–85.
[35] Nguyen DT, Negnevitsky M, de Groot M. Walrasian market clearing for demand
response exchange. IEEE Trans Power Syst 2012;27(1):535–44.
[36] Sezgen O, Goldman CA, Krishnarao P. Option value of electricity demand response.
Energy. 2007;32(2):108–19.
[37] Joo JY, Ahn SH, Yoon YT, Choi JW. Option valuation applied to implementing
demand response via critical peak pricing. In: Power Engineering Society General
Meeting, 2007. IEEE 2007 Jun 24. p. 1–7 [IEEE].
[38] Tyagi R, Black JW, Petersen J. Optimal scheduling of demand response events using
options valuation methods. In: Power and Energy Society General Meeting, 2011
IEEE 2011 July 24. p. 1–5 [IEEE].
[39] Oh SC, D'Arcy JB, Arinez JF, Biller SR, Hildreth AJ. Assessment of energy demand
response options in smart grid utilizing the stochastic programming approach.
InPower and Energy Society General Meeting, 2011 IEEE 2011 July 24. p. 1–5
[IEEE].
[40] Hatami AR, Seifi H, Sheikh-El-Eslami MK. Hedging risks with interruptible load
programs for a load serving entity. Decis Support Syst 2009 Dec 31;48(1):150–7.
248
Electrical Power and Energy Systems 105 (2019) 237–248
[41] Baldick R, Kolos S, Tompaidis S. Interruptible electricity contracts from an elec-
tricity retailer's point of view: valuation and optimal interruption. Oper Res 2006
Aug;54(4):627–42.
[42] Carrion M, Philpott AB, Conejo AJ, Arroyo JM. A stochastic programming approach
to electric energy procurement for large consumers. IEEE Trans Power Syst
2007;22(2):744–54.
[43] Algarni AA, Bhattacharya K. A generic operations framework for discos in retail
electricity markets. IEEE Trans Power Syst 2009;24(1):356–67.
[44] Li H, Li Y, Li Z. A multiperiod energy acquisition model for a distribution company
with distributed generation and interruptible load. IEEE Trans Power Syst
2007;22(2):588–96.
[45] Horowitz I, Woo CK. Designing Pareto-superior demand-response rate options.
Energy. 2006;31(6):1040–51.
[46] Hull JC, Basu S. Options, futures, and other derivatives. Pearson Education India;
2016.
[47] Yuan Y, Li Z, Huang B. Robust optimization under correlated uncertainty:
Formulations and computational study. Comput Chem Eng 2016 Feb;2(85):58–71.
[48] Kirschen DS, Strbac G. Fundamentals of Power System Economics. John Wiley &
Sons; 2004.
[49] Mahmoudi N, Saha TK, Eghbal M. Developing a scenario-based demand response
for short-term decisions of electricity retailers. In: Power and Energy Society
General Meeting (PES), 2013 IEEE 2013 July 21. p. 1–5 [IEEE].
[50] Chen Y, Li J. Comparison of security constrained economic dispatch formulations to
incorporate reliability standards on demand response resources into Midwest ISO
co-optimized energy and ancillary service market. Electr Power Syst Res
2011;81(9):1786–95.
[51] Su CL, Kirschen D. Quantifying the effect of demand response on electricity mar-
kets. IEEE Trans Power Syst 2009;24(3):1199–207.
[52] Lima RM, Novais AQ, Conejo AJ. Weekly self-scheduling, forward contracting, and
pool involvement for an electricity producer. An adaptive robust optimization ap-
proach. Eur J Oper Res. 2015;240(2):457–75.
[53] Bertsimas D, Sim M. The price of robustness. Oper Res 2004;52(1):35–53.
[54] Bertsimas D, Sim M. Robust discrete optimization and network flows. Math
Program 2003;98(1):49–71.
[55] Dominguez R, Baringo L, Conejo AJ. Optimal offering strategy for a concentrating
solar power plant. Appl Energy 2012;31(98):316–25.
[56] Bhattacharya K. Competitive framework for procurement of interruptible load
services. IEEE Trans Power Syst 2003;18(2):889–97.
[57] The GAMS Software Website; 2014 < http://www.gams.com/dd/docs/solvers/
cplex.pdf > .
[58] Lee TF, Cho MY, Hsiao YC, Chao PJ, Fang FM. Optimization and implementation of
a load control scheduler using relaxed dynamic programming for large air condi-
tioner loads. IEEE Trans Power Syst 2008;23(2):691–702.
[59] AEMO Price & Demand Data [Online]. Available: < http://www.aemo.com.au/
data/pricedemand.html > .
[60] AER, State of the Energy Market, Melbourn, Australia, 2011.