Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: Emergent power shortages and power outages cause enormous economic losses. Special fees and rebates at times
Power shortage when the operating reserve is below standard are among the most popular mechanisms for driving demand
Operating reserve response, i.e., a spontaneous reduction in power usage, to help alleviate peak loads. However, a rebate which is
Demand response
uniform for all users usually under- or over-estimates companies’ incentive to respond, as product profitability
Power usage reduction
Scarcity price
and the cost of manufacturing vary from one company to another. As a result, the power provider either fails to
Game theory sufficiently reduce the load during a period of scarcity of electric power or unduly sacrifices profits to pay an
excessive rebate. This study presents a Stackelberg game-theoretic model that determines discriminative rebates
and scarcity pricing of electricity, thereby capturing incentives of different industrial users to respond rationally
and reduce their power usage. The model can be used to assess the current pricing and rebate structure and has
the potential to serve as the fee-and-rebate mechanism in an Emergency Demand Response Program. A case study
is used to simulate a game between the Taiwan Power Company and four classes of heterogeneous industrial
customers who are willing to cooperate with the power provider on implementation of an Emergency Demand
Response Program. The results show that the profits of the participants are higher than the profits of the non-
participants. Moreover, a high-value-added and non-interruptible manufacturing class should be paid higher
rebate.
* Corresponding author.
E-mail address: afu0330@gmail.com (H.-W. Hsu).
https://doi.org/10.1016/j.ijepes.2022.108831
Received 10 February 2022; Received in revised form 15 September 2022; Accepted 26 November 2022
Available online 13 December 2022
0142-0615/© 2022 Elsevier Ltd. All rights reserved.
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
feasible solutions if and only if the dual form has feasible solutions, in drive operating reserves back to a secure level and potentially save the
which case the optimal values of the primal and dual problems are equal. cost of opening extra power plants.
Thus the need for an equivalent formulation of the lower-level optimi
zation problems arises. Embedding the feasibility and strong duality 2. Literature review
conditions of the lower level optimization problem within the set of
constraints of the upper-level optimization problem eliminates the need 2.1. Emergency demand response programs (EDRPs)
for the bi-level structure and makes the non-cooperative Stackelberg
game a single optimization problem with nonlinear constraints. The authors of [3] presented an EDRP model and demonstrated
To implement this fee-and-rebate pricing model, we divide the EDRP appropriate tools for managing the local marginal prices (LMPs) of the
program into two phases. In the first phase, the power provider will ask power market more efficiently. Our model incorporates the cost of
all industrial users if they would like to participate in the calculation of restoring power to machinery at the end of a scarcity period (a period in
their reduction in electricity consumption during peak electricity con which machines were shut down to reduce electricity consumption)
sumption periods. In the second phase, the power provider will inform once regular electricity consumption has resumed—a cost that is
each industrial user of its results (rebate and reduction in electricity considerable for some industrial users.
consumption) and will ask users if they would like to participate in the The authors of [4] proposed an event-driven-based EDRP scheme to
EDRP. Thus during electricity scarcity periods, the power provider could choose the proper demand response (DR) participants and minimize the
2
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
total cost while holding a certain level of electric power in reserve. In our game solve an economic optimization problem subject to stochasticity in
approach, the demand reduction is estimated after calculating the users’ prices, weather-related variables, and the must-serve load. Also, their
costs, hence the overall structure is more rigorous. Another contribution study dealt with large-scale integration of renewable power sources,
is the choice of the proper DR participants. while our research focused on programs which can directly resolve the
The authors of [5] brought up the unit commitment (UC) problem of crisis in the electricity market in an emergency and is suitable for short-
thermal units in considering an EDRP, where the UC with EDRP problem term implementation. Compared programs in which there is the large-
was modeled as a nonlinear, non-convex problem and solved via the scale integration of renewable power sources, our model will not
simulated annealing method. They concluded that EDRP is a method cause fluctuations in the electricity market.
that can solve the power crisis. Our work further dealt with many details The authors of [14] presented a game-theoretic DR program which
of implementing an EDRP, such as how to set the size of the rebate, the comprises into two problems: the dynamic economic emission dispatch
scarcity price of electricity, and the power provider’s expected power problem and the price-based dynamic economic emission dispatch
acquisition from other suppliers, so that an EDRP can be implemented problem. For the leader, the objectives are to minimize fuel costs and
smoothly. emissions and determine the optimal incentive and load curtailment for
The authors of [6] proposed a two-stage stochastic game model and customers, while the follower’s model seeks to minimize emissions,
studied a risk-averse energy trading strategy within multi-energy maximize profits, and determine the optimal incentive and load
microgrids. In the context of multi-energy system, energy resources curtailment for customers. The authors of [15] constructed an
include electricity and DR capacity and all microgrid operators can sell incomplete-information game model between demand response aggre
energy resources to the energy market. Both uncertainty from the gen gators for selling energy previously stored, in order to predict and
eration capacity of the renewable energy and from the DR are taken into analyze the bid made for energy sale in the market. The authors of [16]
account in their formulation. To handle uncertainty, techniques of the studied the combined problem of power provider selection and DR
sample average approximation and the conditional value-at-risk mea management in a smart grid network consisting of multiple power
surement are employed. Their work characterizes the relationship be companies and multiple customers by adopting a reinforcement-
tween the price and the quantity of energy sold by microgrid operators, learning and game-theoretic technique. The authors of [17] consid
whereas our work aims to quantify the incentives to reduce load for ered bidirectional interaction between a power grid and a smart grid by
heterogeneous industrial users. using the Stackelberg game method to increase the net profit and reduce
The authors of [7] addressed the effect of the customer participation fluctuation in demand. The authors of [18] build up a framework for
level in an EDRP and the impact of irrational incentive values in distributed energy resources integration through a single-leader-multi-
implementing such a program for use in microgrid operation, and follower Stackelberg game model with the aggregator as the leader
pointed out that an unreasonable increase in the incentive increased and the system and distribution system operators as the followers.
operating costs. The authors of [8] presented a nonlinear model of an The Stackelberg game is a suitable model for describing the non-
incentive-based DR program based on the price elasticity of demand and cooperative competition between a leader (power provider) and multi
the benefit for the customer. Based on the results of these two studies, it ple followers (industrial users). However, Stackelberg games were not
is clear that an EDRP contributes to system reliability, but the balance employed in the literature to find the best scarcity price of electricity and
between setting a reasonable incentive and the willingness of users of the size of the rebates, or to forecast the contracted users’ reduction in
electric power to participate is still a major concern. electricity usage for the purpose of streamlining power distribution by a
The authors of [9] proposed a two-stage stochastic programming power provider under an EDRP.
model for minimizing the total operating cost and maximizing the DR
aggregator profit at the same time. The difference between their study 3. Formulation of a Bi-level optimization model
and ours is that they considered units’ ramp-up and ramp-down use of
wind power, while we considered the percent operating reserve and the To formulate the Stackelberg game, we consider a bi-level pro
size of the penalty which could ensure a stable power supply during gramming model that treats the power provider’s multi-objective
times of spikes in electricity usage. Their model considers the total cost maximization problem and the industrial users’ profit maximization
of DR scheduling but does not decide the price of electricity charged by problems as upper-level and lower-level problems, respectively. A flow
the power provider. diagram of building the bi-level optimization model is shown in Fig. 1.
Although the authors of [10] considered the electricity pricing and The left box contains the setup of the model: the power provider’s pa
incentive offering of EDRP with consumers preferences, they focus on rameters and the industrial user i’s parameters are summarized in the
the length of customer decision-making cycle which are short-range and beginning of this paper; the restrictions on coefficient in front of Rit are
long-range consumers. In our study, we focus on the impacts of different discussed formally in Section 3.2.2; to validate the model, we consider
industry categories on EDRP, and uses the Stackelberg game to find out 15 scenarios where four classes of the industrial users are the potential
the best decision between power companies and manufacturers. The participants of the EDRP. These 15 scenarios are formally introduced in
authors of [11] integrate renewable energy, incentives and price next section. In particular, when the number of participants increases,
mechanisms for modeling and analysis and consider on each other and the profits realized by the power provider must also increase.
supply side. However, this research focuses on market uncertainty and The central box of Fig. 1 illustrates the key techniques to build the bi-
reliability, and still does not fill in the interactive effects of power level optimization model. The selection of the weights for the three
companies and different industries on EDRP strategies. objectives of the power provider is discussed in Section 3.2.3. To embed
the lower-level optimization model within the upper-level optimization
2.2. Game theory in the electricity market model, we employ the typical Karush-Kuhn-Tucker condition, which
requires primal feasibility, dual feasibility, and equal primal and dual
The authors of [12] employed a Stackelberg game between utility values for the linear program. These techniques are shown in Section
companies and end-use customers to maximize the revenue of each 3.1.
utility company and the payoff of each user, and studied the impact of an The right box of Fig. 1 summarizes the major output of the bi-level
attacker who can manipulate the price information provided by the optimization model. The mechanism behind this bi-level optimization
utility companies. The authors of [13] posed a game-theoretic model program is a Stackelberg game, thus the optimal solution of the model is
that accounts for the Stackelberg relationship between retailers (leaders) theoretically a Nash equilibrium between the power provider and the
and consumers (followers) in a dynamic price environment. Unlike the industrial users.
Stackelberg game employed in our model, both types of players in their A unique characteristic of our model is a constraint on the minimum
3
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
( ( ))
⎧ ∑ Qt +qt +Gt ⎫
⎪
⎪ (Uit − n V
it i )Y1,it +U Y
it 6,it + U ′ +u t − Y8,it + ⎪
⎪
⎪ i t
1+PORt ⎪
∑⎪ ⎨ ′
i ∈I,
⎪
⎬
minimize ( ) (12)
⎪
t∈T ⎪ ∑ ⎪
⎪
⎪
⎪ ⎪
⎪
⎩ Qt +qt − Ui′ t − ut Y9,it ⎭
′
i ∈I,
Industrial user i’s optimization problem for a scarcity period is Y3,it + Y4,it ≥ 0, i ∈ I, t = 0 (13)
( ) ( )
∑{ ( ) ( + ) } +
Xi,|T+1| −
+ Xi,|T+1| Wi Pi
Pi Xit + Xit− Wi Y1,it + − λit − St + Y2,it − Y3,i,t− 1 + Y3,it + Y6,it − Y7,it + Y8,t − Y9,t
maximize Rit λit + St − − − Vi
t∈T
Vi Vi Vi
Pi
(2) ≥ λit + St − , i ∈ I, t ∈ T (14)
Vi
subject to the following constraints: − Y3,i,t− 1 + Y5,it ≥ 0, i ∈ I, t = |T| + 1 (15)
Uit − Rit ( ) ( )
nit ≤ , i ∈ I, t ∈ T (3) Wi − Wi
Vi Y2,it − Y3,i,t− 1 + Y7,it ≥ , i ∈ I, t ∈ T (16)
Vi Vi
( ) ( + )
Pi Xit + Xit− Wi ( ) ( )
Rit − St ≤ Rit (λit ) − , i ∈ I, t ∈ T (4) Wi
Y2,it + Y3,i,t− 1 + Y7,it ≥
− Wi
, i ∈ I, t ∈ T (17)
Vi Vi Vi Vi
( )
+
Rit − Ri,t+1 = Xi,t+1 −
− Xi,t+1 , i ∈ I, t = 0, ⋯, |T| (5) − Y3,i,t− 1 ≥ − VWi i , i ∈ I, t = |T| + 1 (18).
( )
Rit = 0, i ∈ I, t = 0, |T| + 1 (6) Y3,i,t− 1 ≥ − VWi i , i ∈ I, t = |T| + 1 (19)
4
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
{ (
∑ ∑ ∑ (USD). Constraint (27) requires the rebates received by industrial user i
maximize St Ui′ t − Ri′ t at time t = 0 and at time|T| + 1 to be equal to 0. Constraint (28) requires
the product of the restored quantity of electricity and the diminished
t∈T ′ ′
i ∈I, i ∈I,
) }
∑ quantity of electricity from time unit t = 1 to time unit t + 1 to be zero.
+ ut − Bt Qt − Ct qt − πλi′ t Ri′ t − Gt Π (21)
′
This constraint is sufficient for condition (1) in the lower-level optimi
zation problem to hold. Constraint (29) requires sum of the industrial
i ∈I,
5
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
on its electricity consumption). operating reserve is expected to be at or above PORt. Then another
If not all users reduce their electricity consumption, their total scarcity period could occur later, when the percent operating reserve is
electricity use will exceed the capacity of the power provider, that is, the again expected to be below PORt.
percent operating reserve will be less than PORt, so the power provider The participants in an EDRP are the industrial users who sign an
has to provide incentives that will reduce the total electricity con EDRP contract with the power provider. The power provider informs the
∑ ∑
sumption, i∈I, Uit − i∈I, Rit , to a level that will achieve the percent contracted industrial users before the start of each scarcity period. The
operating reserve set by constraint (9): contracted users then plan their use of electricity on their own. If the
(∑ ∑ ) percent operating reserve is at least PORt at the end of the scarcity
(Qt + qt + Gt ) − i∈I, Uit − i∈I, Rit + ut period, the price of electricity reverts to the public price that applies to
(∑ ∑ ) ≤ PORt all users. Otherwise, the power provider may consider continuing the
i∈I, Uit − i∈I, Rit + ut
scarcity period until the percent operating reserve is at least PORt.
The objective of maximizing profits will drive users to reduce their
consumption as much as possible, hence the power provider sets an
upper bound on the percent operating reserve. This serves as another 4.2. Two-phase procedure of the load-reduction EDRP
upper limit on reduction in power consumption for all industrial users
combined. In this game, the power provider divides the establishment of an the
Building the multi-objective optimization function: selection of EDRP into two phases. In the first phase, the power provider will ask all
weights: users if they would like to participate in the calculation of the reduction
To build the multi-objective optimization problem, there are three in their electricity consumption during peak electricity consumption
different sizes of weights (Π, π, and 1) that should be related as follows: periods, the purpose of which is to reduce the probability of scarcity in
the availability of power and/or outright failure of the system. In the
1≪π≪Π second phase, the power provider publishes the scarcity price of elec
where the symbol ≪ means “much less than,” and the weights need tricity, together with the sizes of the rebates per unit reduction in
to be tuned to ensure the priorities of the objectives. electricity usage and the amount of the reduction in electricity usage
required of each industrial user. Users will consider whether to partic
1. To meet the secure percent operating reserve (weight Π): The highest ipate in the EDRP program based on the results of the calculations. This
priority is given to avoiding an emergent scarcity. Whether Π is large study assumes that the fees and rebates are published in near-real time,
enough is determined by the solution of the model. The secure that is, close to the time when the demand-responsive strategy is acti
operating reserve should be met unless it is impossible to do so, in vated, whereas the standard level of consumption is stated in the con
which case the gap between PORt and the actual operating reserve tract between the power provider and the individual industrial user,
should be minimized. which is signed before any period of scarcity occurs. The power provider
2. To reduce the cost of rebate payments (weight π): If the first objective has an option to buy electricity from other suppliers at a price deter
(achieving PORt) is optimized, the second goal is to minimize the mined by those suppliers.
total cost incurred by the power company in the way of rebates paid Given the time periods of electricity scarcity based on a forecast
to users in order to achieve the needed demand responses. If there is made by the power provider, the decisions of the power provider on the
no upper limit on the sizes of the rebate payments in the objective pricing of electricity and the decisions of the contracted users on their
function, user i’s optimality conditions that are imposed as con usage of power are made (and are changeable) at each time unit during
straints can technically drive up the value of λit Rit , which is the profit the period of electricity scarcity. The overall profit that’s to be maxi
of user i but also a part of the costs incurred by the power provider. mized (the profit of the power provider and the profits of the contracted
Whether the weight π has been properly selected can be observed users) is the sum of the profits measured during that period.
from the solution of Rit . We now present the details of one particular load-reduction EDRP.
3. To maximize the profit of the power provider (weight 1): The final Given a set of potential cooperative users, the steps taken by the power
objective is to consider the profit of the power provider. By setting provider before the time periods in T are as follows:
the penalties (for failing to satisfy the previous two objectives) high Phase 1.
enough, this objective will help find the most profitable solution
(among all of those that can maximally deal with power scarcity) • Step 0: Retrieve the values of ut, Tt, Mt, Ct, Bt, Fit, nit, Vi, Wi, and Pi for
that provides an incentive of the power provider to execute such an all industrial users i ∈ I and all time periods t ∈ T for which the
EDRP. reserve is expected to be below PORt.
• Step 1: Solve the fee-and-rebate pricing model. The solution obtained
4. Execution and implementation of the proposed emergency consists of R*it , λ*it , S*t , Q*t and q*t for all industrial users i ∈ I and all
demand response program prospective time periods t ∈ T.
• Step 2: Remove from the set I all users i for which R*it = 0, and return
4.1. Time to execute the emergency demand response program to Step 1. If there are no such users, proceed to Step 3.
6
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
responds to the public price S*t and to its individual rebate offer λ*it by returning to normal means that the price is reset at the standard elec
adjusting its demand. tricity price and that no rebate will be provided to any user for reducing
• Step 5: Att = 1, ⋯, |T| + 1, scarcity period will end. The power their demand.
provider then either starts the next scarcity period or returns to
normal. 5. Results and discussions with case study
In Step 0, the power provider communicates with the potential Along with global warming, the temperature in Taiwan in summer is
cooperative users through a designated association. Information on user increasing significantly. As a result, electricity consumption reaches a
i (Fit, nit, Vi, Wi, and Pi) comprises its historical statistics and some pre new peak load of power usage each year. According to Taiwan Power
dictions regarding its consumption of electricity. Other internal oper Company (TPC) statistics, the number of days on which the percent
ating parameters (ut, Tt, Mt, Ct, and Bt) are derived from the power operating reserve was less than the secure operating reserve set at 6 %
provider’s regular statistics. In Step 1, the fee-and-rebate pricing model increased from 9 in 2014 to 104 in 2017.
can be solved by any existing nonlinear optimization solver. A case study The load reduction EDRP is one of several methods of receiving a
that shows the resulting values of the decision variables R*it , λ*it , S*t , Q*t monetary reward from TPC that became effective in July 2018 [19].
Industrial users with consumption of at least 100 kW of electricity
and q*t is given in Section 4. It will be seen in the results of that case study
(standard quantity of electricity consumed) are eligible to sign an EDRP
that there exists a scenario in which R*it = 0 is possible for some user i ∈ I.
contract with TPC. In the contract, each user with electricity consump
This situation implies that user i is indifferent (in terms of profits) be
tion that does not exceed 5,000 kW agrees to a reduction in consumption
tween participating in the EDRP and not participating in it, and that the
that is at least 20 % of its standard electricity consumption during every
power provider can achieve an operating reserve of PORt from the de
period of electricity scarcity, and TPC is required to officially inform
mand responses of the other users, and at a cost no higher than if the
demand response of user i were included. Thus we eliminate user i from users of the start and end times of each scarcity period at least T
̂ minutes
the set I in Step 3 and solve the nonlinear optimization model. Step 4 prior to starting. If a company’s actual reduction in electricity use is R ≥
describes the actions taken by the power provider and the users during R,
̂ an implementation rate of δ% is calculated as R/ R ̂ × 100 %.
the scarcity period. Step 5 describes the two possible situations at the Depending on the value of T,
̂ a base monthly rebate is calculated as
end of the scarcity period. For the second situation mentioned in Step 5, R×P̂T × δ%, where the values of P̂T (in USD/kW-month) that
̂
7
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
Table 2
Parameters for four classes of aggregate manufacturers.
Source Parameters t¼1 t¼ |T|-1 t¼ |T|
8
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
Table 4
Result for the 15 scenarios with a fixed scarcity price.
Scenario St λ1 λ2 λ3 λ4 Qt qt R1 R2 R3 R4 POR Gt
A 0.1000 0.8129 0.1000 0.5076 0.1000 11,165 27,300 0.000 6.120 0.000 966.180 6% 0
B 0.1000 0.2298 0.1000 0.2122 – 11,165 27,300 0.000 972.310 0.000 – 6% 0
C 0.1000 0.2533 0.1000 – 0.1000 11,165 27,300 0.000 482.700 – 489.610 6% 0
D 0.1000 0.2289 – 0.1789 0.1000 11,165 27,300 0.000 – 0.000 972.310 6% 0
E 0.1000 – 0.1000 0.2138 0.1000 11,165 27,300 – 61.480 0.000 910.830 6% 0
F 0.1000 0.2120 0.1000 – – 11,165 27,300 0.000 972.310 – – 6% 0
G 0.1000 0.2188 – 0.2125 – 11,165 27,300 281.620 – 690.690 – 6% 0
H 0.1000 0.2331 – – 0.1000 11,165 27,300 0.000 – – 972.310 6% 0
I 0.1000 – 0.1000 0.2101 – 11,165 27,300 – 972.310 0.000 – 6% 0
J 0.1000 – 0.1000 – 0.1000 11,165 27,300 – 485.990 – 486.320 6% 0
K 0.1000 – – 0.1371 0.1000 11,165 27,300 – – 0.000 972.310 6% 0
L 0.1000 0.2196 – – – 11,165 27,300 972.310 – – – 6% 0
M 0.1000 – 0.1000 – – 11,165 27,300 – 972.310 – – 6% 0
N 0.1000 – – 0.2125 – 11,165 27,300 – – 691.980 – 5% 297
O 0.1000 – – – 0.1000 11,165 27,300 – – – 972.310 6% 0
participated in the EDRP. In scenarios L–O, a single aggregate industrial rebates, the mission of reducing electricity consumption is concentrated
user signed the EDRP contract. Based on Intel(R) core (TM) i5-4590, on low-value-added users (user 4 and user 2). The results in Table 4
3.30 GHz CPU, 16 GB RAM, OS Windows X operation environment, show that TPC gives priority to providing rewards for reducing elec
the run-time (sec) associated with 15 scenarios are between 0.0149 s and tricity consumption according to the following order: user 4 ≥ user 2 ≥
0,0160 s with average of 0.0155. user 3 ≥ user 1. In a rational outcome, users 4 and 2 would be more
willing than users 3 and 1 to participate in the second phase of such an
5.3. The EDRP with a fixed price of electricity EDRP.
We first focused on the incentive-based demand response that would 5.4. Profits by TPC and EDRP participants in the 15 scenarios with a
promote reducing electricity consumption spontaneously and achieve a fixed price of electricity
6 % operating reserve during peak periods of electricity consumption. In
this simplified program, TPC would grant rebates to the users who sign The data given in Table 5 show the profits of users that do not
the demand response contract without making adjustments in the participate in the EDRP (User_n), the profits of users that do participate
scarcity price of electricity. The results of this type of EDRP for the 15 in the EDRP (User_ER), and TPC’s profit from executing the EDRP. The
scenarios are shown in Table 4, where the data are expressed as the highest profit is realized by user 4 under different scenarios. Also, the
results for the average of |T| hours of the scarcity period for every user. maximum value of TPC’s profit is 2196.643 K (USD), in scenario A,
The scarcity price published by the power provider is 0.1 USD/kWh, which would always be deployed instead of any of the scenarios with a
which was always set to the original price of electricity. single participant. With a fixed scarcity price, TPC’s profit from
TPC will give different rebates according to the product value added executing the EDRP rises by an amount that ranges from 0 % to 5.57 % of
and the setup costs. Larger rebates would be needed to entice high- its minimum profit, 2080.68 K (USD). In the short term, the participa
value-added users to participate in the second phase of an EDRP and tion of user 4 allows TPC to meet the goal of mitigating power shortages
reduce their electricity consumption. According to Table 4, the order of and stabilizing the power supply during peak electricity consumption
the four users in terms of their average rebate is as follows: user 4 ≤ user periods.
2 ≤ user 3 ≤ user 1 (that is, user 4′ s rebate is less than or equal to user 3′ s
rebate, and so on). This phenomenon is referred to as the users’ hierarchy 5.5. Reasons for switching from a fixed scarcity price to a variable
for rebates. scarcity price
TPC minimizes the gap between the actual operating reserve and a 6
% operating reserve during peak periods of electricity consumption, At a fixed price of electricity, the users’ hierarchy for rebates is sig
hence the instantaneous peak load of the system must be reduced by nificant in all scenarios. According to that hierarchy, the values of the Rit
972.310 kW during peak periods in order to fulfill the cooperative users’ calculated by TPC are concentrated on the user that is lowest in the
minimum production levels. According to the users’ hierarchy for hierarchy. In the long term, if the users do not wish to participate in the
Table 5
Profits (thousand USD) of TPC and EDRP participants in the 15 scenarios.
Scenario User1_n User2_n User3_n User4_n User1_ER User2_ER User3_ER User4_ER Company
9
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
second phase of the EDRP, the probability of having electricity rationing interruptible manufacturing class could be lower than that paid to a non-
during peak periods will increase, and the reliability will become too interruptible manufacturing class. Similarly, compared to a low-value-
low to achieve a stable power supply. Ultimately, TPC would need to added manufacturing class, a high-value-added manufacturing class
invest resources in the acquisition of more generating units. should be paid slightly higher rebates.
Taking advantage of the demand response to a price that varies with Comparison of a fixed scarcity price of electricity to a variable
the time of use, the users’ hierarchy for rebates switches from being scarcity price shows that TPC is more willing to allocate different re
significant to being nonsignificant. In a nonsignificant hierarchy, the ductions in electricity consumption to different users with a variable
sizes of the reductions in electricity consumption calculated by TPC are scarcity price, which means that there are more scenarios in which the
more evenly distributed among the users. In the long term, the reliability equilibrium distribution can be achieved, as shown in Fig. 4. TPC guides
of electricity with a variable scarcity price should outperform that with a users to change their electricity consumption behavior and also in
fixed price of electricity. creases their willingness to participate in the second phase of the EDRP.
This makes it possible to steadily reduce the risk of power rationing
during peak electricity consumption periods in order to achieve the goal
5.6. Results of EDRP with a variable scarcity price of electricity and
of a 6 % operating reserve. In the long term, more users will be willing to
variable sizes of rebate payments in the 15 scenarios
cooperate with demand responses that suppress peak loads and achieve
a stable market equilibrium.
The results for the 15 scenarios with a variable price and variable
sizes of rebate payments (and for the average of |T| hours in the scarcity
period for every user) are shown in Table 6, where it is observed that the 5.7. Profits realized by TPC and EDRP participants with a variable
maximum scarcity price published by TPC is 0.1618 USD/kWh, in sce scarcity price
nario A, which is which would always be deployed instead of the orig
inal price of electricity, 0.1 USD/kWh. With a variable scarcity price, the We have restricted TPC’s minimum required profit by stating that, in
increase in the scarcity price from executing the EDRP rises by an every scenario, the profit realized by the power provider must be at least
amount that ranges from 0 % to 61.8 % of the minimum scarcity price, as high as the power provider’s profits realized for all subsets of users in
0.1 USD/kWh. The scarcity price is closely related to the users’ hierar that scenario. This can be done by setting the value of the power pro
chy for rebates. Because of constraint (24), the scarcity price will be vider’s minimum profit serially: First, we obtained TPC’s profits in
determined by the user that is lowest in the users’ hierarchy for rebates; scenarios with one user: L, M, N, and O. For scenarios with two users
otherwise, the scarcity price would increase in order to satisfy the (scenario F for example, which consists of users 1 and 2), the minimum
objective of maximizing TPC’s profit in the model. required profit for TPC was set to the maximum of TPC’s profits for the
In terms of the rebate payments with a variable scarcity price, the one-element subsets of users that contain only users 1 and 2, that is,
order is always user 4 ≤ user 2 ≤ user 3 ≤ user 1. There are differences in scenarios L and M. For scenarios with three users (scenario B for
the significance of the hierarchy for the 15 scenarios, as shown in Fig. 3. example, which consists of users 1, 2, and 3), the minimum required
For instance, the range of the rebates in scenarios B, F, G, H, L, M, N, O is profit for TPC was set to the maximum of TPC’s profits for the one- and
larger, relatively speaking, than the range of the rebates in scenarios A, two-element subsets of users that contain only users 1, 2, and 3, that is,
C, D, E, I, J, K. TPC has set up constraint (24) to prevent the phenomenon scenarios F, G, I, L, M, and N. For the scenario with all four users, the
of buying low and selling high. Thus the order of the users in the users’ minimum required profit for TPC set to the maximum of TPC’s profits for
hierarchy for rebates will be the same as the order of the users in terms of all of the other 14 scenarios. The greater the number of EDRP partici
the scarcity price. pants, the higher TPC’s profit. The maximum value of TPC’s profit is
In terms of the reduction in electricity consumption, the order of the 4380 K (USD), in scenario A, which would always be deployed instead of
users is determined by the sizes of the rebates. There is actually little any of the scenarios with a single participant. With a variable scarcity
difference from one scenario to another. As shown in Table 6, a percent price, TPC’s profit from executing the EDRP rises by an amount that
operating reserve of 6 % was achieved in all but one of the 15 scenarios, ranges from 0 % to 99.4 % of the minimum profit, 2196.643 K (USD) as
and it can be seen that this model has a significant effect on mitigating shown in Table 7.
power shortages. Only scenario N, which consists of just user 3, did not The profits realized by the four users in all 15 scenarios with a var
achieve that level, because the original power consumption of user 3 was iable scarcity price are shown in Fig. 5. For each user, profits of four
small. Therefore, even if user 3 consumed no power, TPC would not be different types are shown there: the absolute values of the profits of the
able to achieve the 6 % level. It is easier for an interruptible participants (User_ER_abs), the absolute values of the profits of the non-
manufacturing class to change its electricity usage habits than for a non- participants (User_n_abs), the positive profits of the participants
interruptible manufacturing class to do so. Thus the rebate paid to an (User_ER), and the positive profits of the non-participants (User_n).
Table 6
Results for the 15 scenarios with a variable scarcity price.
Scenario St λ1 λ2 λ3 λ4 Qt qt R1 R2 R3 R4 POR Gt
A 0.1618 0.1645 0.1618 0.1618 0.1618 11,165 27,300 0.000 326.780 319.010 326.530 6% 0
B 0.1613 0.1637 0.1613 0.1629 – 11,165 27,300 0.000 706.050 266.260 – 6% 0
C 0.1617 0.1618 0.1617 – 0.1617 11,165 27,300 0.000 483.840 – 488.470 6% 0
D 0.1609 0.1649 – 0.1609 0.1609 11,165 27,300 94.570 – 439.020 438.720 6% 0
E 0.1608 – 0.1608 0.1608 0.1608 11,165 27,300 – 361.180 300.920 310.210 6% 0
F 0.1598 0.1638 0.1598 – – 11,165 27,300 0.000 972.310 – – 6% 0
G 0.1599 0.1651 – 0.1599 – 11,165 27,300 282.900 – 689.410 – 6% 0
H 0.1598 0.1676 – – 0.1598 11,165 27,300 0.000 – – 972.310 6% 0
I 0.1598 – 0.1598 0.1598 – 11,165 27,300 – 485.930 486.380 – 6% 0
J 0.1000 – 0.1000 – 0.1000 11,165 27,300 – 484.900 – 487.410 6% 0
K 0.1598 – – 0.1598 0.1598 11,165 27,300 – – 596.270 376.040 6% 0
L 0.1598 0.1598 – – – 11,165 27,300 972.310 – – – 6% 0
M 0.1000 – 0.1000 – – 11,165 27,300 – 972.310 – – 6% 0
N 0.1563 – – 0.1563 – 11,165 27,300 – – 691.980 – 5% 297
O 0.1000 – – – 0.1000 11,165 27,300 – – – 972.310 6% 0
10
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
Fig. 3. Average rebate for users in time unit t for fixed and variable scarcity prices.
Users can achieve the estimated profit if the amount of their reduction in power company to participate in emergency demand response program;
electricity consumption meets the amount that was calculated and and finds the best possible benefits and cooperation methods for the
offered by TPC. The profits of the participants (User_ER, User_ER_abs) industries at the same time. The results show that the proposed model
are higher than the profits of the non-participants (User_n, User_n_abs). can find a Nash equilibrium point for such a situation which can also be
expressed as the policy cost of adopting the emergency demand response
6. Conclusions program as a system resiliency measure.
A case study in Taiwan was undertaken for purposes of model vali
This study developed a game-theoretic model for a power provider dation and policy simulation, as well as general possible recommenda
and industrial customers to negotiate power usage during peak periods tions for EDRP, which will provide a clearer understanding of the
with proper rewards based on the users’ hierarchy for rebates in an partnership for utilities implementing EDRP around the world. The re
emergency demand response program. The model can assist a power sults show that the profits of the participants in the emergency demand
provider in making its decision on the scarcity price of electricity, and response program are higher than those of the non-participants. For the
heterogeneous industrial users in making decisions on their power different classes of manufacturers considered, the one which is lowest in
usage, to achieve a win–win situation. In practice, this research provides the users’ hierarchy for rebates consists of the low-value-added inter
a reasonable portfolio from different industries and rebate prices for ruptible manufacturers, and the one which is highest in the users’
11
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
Fig. 4. Reduction in electricity consumption by users for fixed and variable prices.
Table 7
Profits (thousand USD) of TPC and EDRP participants in the 15 scenarios.
Scenario User1_n User2_n User3_n User4_n User1_ER User2_ER User3_ER User4_ER Company
12
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
hierarchy for rebates consists of the high-value-added non-interruptible Declaration of Competing Interest
manufacturers. The amounts of the rebates are based on the users’ hi
erarchy for rebates and the reductions in electricity consumption on the The authors declare that they have no known competing financial
part of the manufacturers. However, observing equilibrium in the interests or personal relationships that could have appeared to influence
required electricity reductions on the part of different manufacturers the work reported in this paper.
could result in participation of a set of manufacturers which is different
from the set with the highest order in the users’ hierarchy for rebates. A Data availability
percent operating reserve of 6 % was achieved in all but one of the 15
scenarios studied, and that significantly reduced the risk of a power Data will be made available on request.
shortage.
This study has some limitations. The first is the assumption that in Acknowledgements
dustrial users have no way of knowing the decisions made by other users
when making their own decision. The second is that the current model The authors would like to thank the Ministry of Science and Tech
does not capture the situation for users who do not participate in the nology of the Republic of China, Taiwan for financially supporting this
second phase of the emergency demand response program to reduce research under Contract no. MOST 109-2628-E-007 -002 -MY3 and
their electricity consumption as the scarcity price of electricity changes. MOST 109-2221-E-033 -039 -MY3.
Finally, the data for each user should be included as parameters of this
model. Among the possibilities for future study in this area is incorpo
References
ration of the decisions and costs of building new power plants into the
current game-theoretic model to gain further insight, and the uncer [1] Aalami HA, Moghadam MP, Yousefi GR. Modeling and prioritizing demand
tainty from the integration should be considered into an emergency response programs in power markets. Electr Pow Syst Res 2010;80:426–35.
demand response program. [2] Aalami, H.A., Yousefi, G.R., Moghadam, M.P. Demand response model considering
EDRP and TOU programs. Transmission and Distribution Conference and Exposition,
April 21-24, Chicago, IL, USA: 2008.
CRediT authorship contribution statement [3] Aazami R, Aflaki K, Haghifam MR. A demand response based solution for LMP
management in power markets. Int J Electr Power Energy Syst 2011;33(5):
1125–32.
Yu-Ching Lee: Conceptualization, Methodology, Validation, Soft [4] Wang T, Rahimi Pordanjani I, Xu W. An event-driven demand response scheme for
ware, Supervision, Writing – review & editing. Hsin-Wei Hsu: power system security enhancement. IEEE Trans Smart Grid 2011;2(1):23–9.
Conceptualization, Methodology, Validation, Investigation, Resources, [5] Rahmani-Andebili, M., Abdollahi, A., Parsa Moghaddam, M. An investigation of
implementing Emergency Demand Response Program (EDRP) in unit commitment
Supervision, Writing – review & editing, Project administration. Yen- problem. Power and Energy Society General Meeting, July 24-29, Detroit, MI, USA:
Chang Chang: Software, Formal analysis, Data curation, Writing – 2011.
original draft, Visualization. Ying-Lien Chen: Software, Formal anal [6] Li C, Xu Y, Yu X, Ryan C, Huang T. Risk-averse energy trading in multienergy
microgrids: A two-stage stochastic game approach. IEEE Trans Ind Inf 2017;13(5):
ysis, Data curation, Writing – original draft, Visualization, Validation,
2620–30.
Resources, Supervision, Project administration. Ming-Chuan Chiu: [7] Imani HM, Niknejad P, Barzegaran MR. The impact of customers’ participation
Validation, Resources, Supervision, Project administration. level and various incentive values on implementing emergency demand response
program in microgrid operation. International Journal of Electrical Power & Energy
Systems 2018;96:114–25.
[8] Aalami HA, Pashaei-Didani H, Nojavan S. Deriving nonlinear models for incentive-
based demand response programs. Int J Electr Power Energy Syst 2019;106:
223–31.
13
Y.-C. Lee et al. International Journal of Electrical Power and Energy Systems 147 (2023) 108831
[9] Talari S, Shafie-khah M, Wang F, Aghaei J, Catalão JPS. Optimal scheduling of [15] Motalleb M, Ghorbani R. Non-cooperative game-theoretic model of demand
demand response in pre-emptive markets based on stochastic bilevel programming response aggregator competition for selling stored energy in storage devices. Appl
method. IEEE Trans Ind Electron 2019;66(2):453–1464. Energy 2017;202:581–96.
[10] Dadkh, A., Bayati, N., Shafie-khah, M., Vandevelde, L., P.S.Catalãod, J. Optimal [16] Apostolopoulos PA, Tsiropoulou EE, Papavassiliou S. Demand response
price-based and emergency demand response programs considering consumers management in smart grid networks: A two-stage game-theoretic learning based
preferences. International Journal of Electrical Power & Energy Systems, 138 (2022), approach. Mobile Networks and Applications 2018:1–14.
Article 107890. [17] Tang R, Wang S, Lia H. Game theory based interactive demand side management
[11] Pourramezan A, Samadi M. A novel approach for incorporating incentive-based responding to dynamic pricing in price-based demand response of smart grids. Appl
and price-based demand response programs in long-term generation investment Energy 2019;250:118–30.
planning. Int J Electr Power Energy Syst 2022;142(B):108315. [18] Shen Z, Liu M, Xu L, Lu W. An accelerated Stackelberg game approach for
[12] Maharjan S, Zhu Q, Zhang T, Gjessing S, Başar T. Dependable demand response distributed energy resource aggregator participating in energy and reserve markets
management in the smart grid: A Stackelberg game approach. IEEE Trans Smart considering security check. Int J Electr Power Energy Syst 2022;142(B):108376.
Grid 2013;4(1):120–32. [19] Taiwan Power Company (TPC). Load Reduction Program Report: 2018. https://
[13] Zugno M, Morales JM, Pinson P, Madsen H. A bilevel model for electricity retailers’ www.taipower.com.tw/tc/page.aspx?mid=135cid=46cchk=d7bd7816-24e8-
participation in a demand response market environment. Energy Econ 2013;36: 4fc4-9ec2-43e7246637bb (Accessed 29 December 2021).
182–97. [20] Bureau of Energy: Official Website 2022. https://www.moeaboe.gov.tw/ECW/
[14] Nwulu N, Xia XH. Implementing a model predictive control strategy on the english/home/English.aspx (Accessed 28Augest 2022).
dynamic economic emission dispatch problem with game theory based demand
response programs. Energy 2015;91:404–19.
14