Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
1
Abstract—Active consumers who engage in energy consump- a∗i Optimal action of i th active consumer that maxi-
tion, production and provide ancillary services in a dynamic mizes the perceived payoff
and interactive manner will be an integral part of the future P̂ Perceived price of electricity
grid. This paper models and analyzes interaction between active
consumers and aggregators with a specific focus on consumer A Energy purchased by the aggregator in day-ahead
actions in response to real-time electricity pricing and their market
impacts on grid voltage. Firstly, the paper introduces a unique ρda Price of electricity paid by the aggregator in day-
prospect theory based consumer behavior model that captures ahead market
wide range of consumers each with their individual preferences. ρrt Price of electricity paid by the aggregator in real-
Secondly, the interaction between consumers and aggregators is
modeled as a Stackelberg game. However, unlike existing game time market
theoretic efforts that assume rational behavior of consumers, the v Vector of distribution grid voltage
prospect theory based models systematically incorporate realistic S Voltage sensitivity matrix of the distribution system
consumer behavior including irrationality. Another unique aspect a Vector of best response actions of all active con-
of the formulation is the inclusion of the physical grid constraints sumers
(in terms of voltage violations) within the analysis. In contrast to
prior approaches that limit themselves to economic aspects, the â Vector of previous actions of all active consumers
proposed techno-economic perspective provides an understanding 1 Vector of all 1’s
of the impact of large scale penetration of active consumers on the θ Coefficient of grid awareness for the aggregator
physical grid. The proposed model is analyzed using simulations
on the IEEE 69 bus test system and the impact of irrationality
on both consumer/aggregator payoffs and voltage violations is I. Introduction
quantified.
A smart power grid leverages advances in sensing, commu-
Keywords: Active Consumer, Game Theory, Power Distri- nication, computing, and control to increase grid reliability,
bution, Prospect Theory, Sensitivity, Stackelberg Game resiliency, efficiency, and flexibility [1]. Integration of these
technologies also allows higher deployment of large and small-
Nomenclature scale renewable energy resources. The most significant impact
of these changes is being felt at the consumer level. The
ai Action of active consumer connected to node i new generation of appliances within the home are becoming
i Node and active consumer index smarter with load management and communication capabil-
ρ Price of electricity ities. As a result of these advances, future consumers will
P Probability distribution of electricity price be very different from those today. It is expected that active
ω(·) Probability weighting function consumers who engage in managing energy use, production,
ω(p) Subjective probability storage and providing other ancillary services in a dynamic
p Objective probability and interactive manner will be an integral part of the future
α Coefficient of rationality grid. Therefore, understanding, modeling and quantifying the
Ri Renewable generation of i th active consumer impact of such active consumers under the umbrella of cyber
βi Comfort coefficient of i th active consumer physical social and economic systems is an important and
v(i) Distribution grid voltage at node i much needed research thrust. There will be a continuing need
âi Previous action of i th active consumer to integrate cyber-physical system (CPS) advances into the
S(i, i) ith diagonal element of the voltage sensitivity ma- operation, control, and design of these systems.
trix The biggest change the consumers will see in the future
γi Coefficient of grid awareness for i th active con- is more engagement and ability to actively participate in the
sumer Transactive Energy Market (TEM) [2]. As envisioned, large
ui (ai , ρ) Payoff function of i th active consumer commercial, industrial customers, and individual homeowners
ΠiPT Perceived payoff of i th active consumer will be able to participate directly in markets set up by
the utility or by third party aggregators. For an individual
K. Jhala, B. Natarajan and A. Pahwa are with the Department of Elec- customer, it will be impossible to keep track of various
trical and Computer Engineering, Kansas State University, Manhattan, KS-
66506, USA, (e-mail: kumarsinh@k-state.edu; bala@k-state.edu; pahwa@k- decisions that will be required in real-time as well as in
state.edu) the immediate future in a highly dynamic transactive energy
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
2
market. Therefore, each home is expected to be equipped maximize some utility function (e.g., energy costs [9]) under
with a home energy management system [3]–[6] or a complex some constraints (levels of comfort/convenience [10]). One
computer, which will use real-time electricity price, real-time approach is to implement direct load control where a utility or
data on loads/generation, and customer preferences to schedule an aggregator can remotely control certain loads in a house-
different devices in homes. These efforts primarily focus on hold based on an a priori agreement [11]. User privacy is the
cost saving for the consumer as discussed thoroughly in the primary barrier for large scale implementation of such direct
next subsection. However, it is important to note that at the load control methods. Alternately, smart pricing (e.g., critical
utility level, new dynamics may emerge due to participation peak pricing (CPP), TOU pricing and real time pricing (RTP))
of new active customers in the market place that could create can be used to encourage consumers to individually manage
conditions detrimental to reliability and resiliency of the their loads [12]. A plethora of deterministic centralized and
system. Utilities will have to implement appropriate controls distributed optimization [13], model predictive control [14],
to keep the system stable and within bounds. High penetration reinforcement learning [15] as well as game theoretic methods
of solar PV can also cause technical issues as recently reported [16] have been proposed to attack these problems. Recently,
for Hawaii [7], where the utilities are refusing interconnection there have been some efforts to systematically model uncer-
of additional rooftop PV systems due to operating issues. To tainties in this framework and implement stochastic versions
address some of these challenges it is important to integrate of optimization [17], and dynamic programming [18] methods.
physical grid constraints and economic aspects resulting in
techno-economic analysis of the underlying grid. There is a plethora of research in the area of smart buildings
With the anticipated growth in active consumers, modeling energy management systems. Authors in [19] define a building
consumer behavior to changing price of electricity is a crucial management system that controls energy flow inside a smart
aspect of this work. Although the most common reaction for building with renewable energy sources and energy storage
any consumer is to take actions which will save money, there using a heuristic optimization process. [20] reviews some of
are consumers who give more importance to other factors such the technical opportunities provided by internet of things in
as environment and comfort. Frugal consumers are typically smart building management area. Authors in [21] build a build-
more concerned about saving money at the cost of comfort ing energy management system including a grid-connected
and environment. Green consumers give higher importance to PV system and a storage system using the mixed-integer
the environment relative to money and comfort. On the other linear programming framework for the purpose of optimizing
hand, comfort maybe the most important aspect for affluent scheduling of building elements in order to achieve a pre-
consumers. Additionally, there may be consumers who do not specified objective. In [22], an intelligent residential energy
care and thus they will not make any adjustments to their management system to reduce electricity bills for prosumers
power usage pattern based on pricing. Many prior efforts of smart residential buildings is proposed, and its benefits are
attempt to model consumer actions in response to pricing demonstrated through a case study.
within the framework of game theory.
Traditional game theory assumes that players in the game Many of the prior efforts in the area of DR and DG
are rational and uninfluenced by real-life perceptions. Most of management tend to take a user/home centric approach. That
game theoretic work in smart grid are based on assumption that is, the goal is to minimize energy costs for a home owner by
consumers make decisions according to their expected utility. scheduling and reducing load and managing generation and
However, real-life decision making of consumers is influenced storage [23]–[26]. While this is an essential step in under-
by their perceived subjectivity which cannot be explained by standing optimal actions for a consumer, it is important to note
expected utility theory (EUT). A Nobel Prize winning theory, that the consumer is not acting in isolation and the operational
Prospect Theory [8] explains the fact that people usually over- cost and stability of the grid is dependent on the cumulative
weigh low probability bad outcomes and under-weigh high actions of multiple consumers. This aspect was brought out
probability favorite outcomes. This work uses the Prospect in [16], where a demand side management scheme based on
Theory to capture subjective behavior of active consumers for collaborative game theoretic approach is proposed. The efforts
uncertain price of electricity. A unique aspect of this work within the control theory community that focuses on modeling
is the modeling of a wide range of consumers and their consumer behavior also rely on deterministic frameworks [27].
behavior/response to pricing using the Prospect Theory. A data driven approach to model consumers demand response
behavior was presented in [28], where authors hypothesize a
A. Related work long term steady behavior and short term dynamic response
Over the past decade, there have been numerous efforts that behavior. Since, the analysis was data based, the generality of
have attempted to model, manage and optimize participation of this model is unclear. There have been many game theoretic
consumers in the power distribution system. These efforts can efforts to address some of the challenges in smart grid. All
be broadly classified into two classes: (1) demand response these efforts use classical EUT assuming rational behavior
(DR) programs focusing only on load management; (2) de- of players [29], [30], which is not pragmatic. There are few
mand response and distributed generation (DG) management. efforts that use the Prospect Theory to address some of the
The first dimension of the work relates to residential demand challenges in power systems. Most of these efforts are limited
response, which typically leverages price and generation fore- to investigating the energy interactions between micro-grids
casts to either shift or reduce load consumption in order to [31], [32].
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
3
B. Contributions
This work proposes a Prospect Theory based model of active
consumers making subjective actions under uncertain price
of electricity and analyzes the impact of their actions on the
distribution grid voltage. Major contributions of this paper are
listed below.
• This work models interaction between an aggregator and
multiple active consumers connected to a power distribu-
tion system as a Stackelberg leader-follower game, where
the aggregator acts as a leader and the active consumers
are followers.
• A general model that efficiently captures consumers
with varying preferences for electricity consumption is
considered. For the first time, the subjective behavior
of consumers in response to variations in the price of
electricity is modeled using the Prospect Theory.
• Behavior of the aggregator is modeled by considering
both technical and economic aspects of grid operation.
Fig. 1: System Model
The economic goal of maximizing profit is integrated with
the technical goal of maintaining the grid voltage within
permissible limits in the game formulation.
decides the price of electricity that maximizes the profit while
• Actions of active consumers at the Nash equilibrium and
keeping voltages of distribution system within limits.
its impact on the distribution grid voltage is analyzed
In a Transactive Energy Market, each active consumer can
using simulations on the modified IEEE 69 bus test
interact with the grid by changing its energy consumption
system.
or by providing real power support from the local renewable
• The difference between expected utility theory to the
generation. In essence, each consumer chooses an action a i ,
Prospect Theory formulations are highlighted. Specifi-
which is power drawn from the grid or power supplied back
cally the impact of lack of rationality on the physical
to the grid. The consumer’s choice of action a i is based
grid voltage and monetary payoffs for both the consumer
on its perceived payoff that is determined by the price of
and the aggregator is quantified.
electricity. It should be noted that the payoff could depend
The paper is organized as follows. Section II presents a on more than a single metric. For example, the payoff for
system model with an overview of the Prospect Theory and the an active consumer could depend on (1) cost of electricity;
Stackelberg game. A model of active consumers with different (2) comfort achieved by consuming energy, and (3) impact
preferences for electricity consumption is derived in section of their action on the distribution grid voltage. Aggregation
III. Behavior of the aggregator and its best response action is of payoff metrics in the payoff vector to decide the optimal
derived in section IV. Nash equilibrium of the proposed game action would depend on the class to which the consumer
and effect of active consumers on the grid voltage is observed belongs. Goal of a comfort seeking consumer is to maximize
using simulation on the modified IEEE 69 bus test system in the comfort achieved by consuming energy. Some consumers
section V. Conclusions are presented in section VI. may want to minimize the cost of electricity consumption.
Active consumers who consider the physical grid voltage while
II. System Model making their decision want to keep the distribution grid voltage
The future power distribution systems will involve multiple at the node of connection close to the rated value. While
active consumers interacting with a third party aggregator as at this point of time, a grid-conscious consumers may seem
shown in Fig. 1. This paper considers a scenario where mul- utopian, it is interesting to model and consider the impact
tiple active consumers that are part of the power distribution of such customers. One may expect such consumer types
network have a contractual agreement with a third party aggre- in community owned microgrids where the social aspect of
gator. The aggregator buys electricity in a day-ahead wholesale maintaining grid constraints play an important role. In this
market from the independent system operator (ISO) and sells work, it is assumed that a consumer connected to node i
it to consumers. Any additional energy required is purchased of the distribution system knows i th diagonal element of the
in the real-time market. Based on the day-ahead agreement sensitivity matrix, which is used to compute change in voltage
that the aggregator makes with ISO, the aggregator decides at node i due to change in action of i th consumer. Additionally,
the electricity pricing to influence electricity consumption of electricity prices would be communicated to consumers by the
active consumers. In addition to all financial transactions, the aggregator over the cyber infrastructure. Price of electricity,
aggregator is also responsible for maintaining the grid voltage which is decided by the aggregator, can take values from
within allowable limits. In case of a voltage violation, the a set ρ, where ρ = {ρ1 , ρ2 , · · · , ρn }. However exact price of
aggregator pays a penalty to the electric utility company that electricity is unknown to consumers. The aggregator decides
owns the physical grid infrastructure. Therefore, the aggregator the probability distribution of electricity prices. Let P be vector
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
4
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
5
behavior. A time dependent model of active consumer behavior the ith node in the distribution system is also assumed to
along with correlated load scheduling will be pursued as part know ith diagonal element of the sensitivity matrix, which
of the future work which will capture shift in demands due to can be used to compute change in voltage at node i due to
price of electricity. change in power at node i. An analytical method to compute
sensitivity matrix is proposed in [36], [37], using which an
III. Active Consumer Modeling active consumer can compute i th diagonal element of the
sensitivity matrix (S(i, i)). Goal of grid aware active consumers
Modeling different consumer types, capturing their prefer-
is to keep the voltage at point of connection close to the
ences and reactions to changes in electricity price and renew-
rated value by adjusting electricity consumption by minimizing
able generation is challenging. Consumers may belong to dif-
γi (v(i) + S(i, i)(â i − ai ) − 1)2 .
ferent socioeconomic classes with different value or preference
In this work, type of consumers are differentiated based on
for electricity consumption. In the Transactive Energy Market,
their comfort requirement, price responsiveness and awareness
the active consumers decide their energy consumption based
towards grid voltage. The proposed framework is general
on price of electricity that they receive from the aggregator.
enough to accommodate size and elasticity of different types
Actions of consumers are based on their perceived payoff.
of consumers. So, other categories/types of consumers (i.e.
The payoff function could be composed of more than a single
residential, commercial and industrial) can be seamlessly in-
objective. For example, in this work, the payoff function of
tegrated into the model. Since actions of such consumers will
an active consumer incorporates three metrics corresponding
be primarily dictated by their priorities/preferences, we use
to economics, comfort and environmental costs. The payoff
the comfort, price, and grid awareness categories to highlight
function for i th active consumer can be written as:
the efficiency of the proposed approach. For example, among
ui (ai , ρ) = βi log(ai + Ri + 1) − ρai − γi (v(i) + S(i, i)(â i − ai ) − 1)2 residential consumers we may have some who prefer to keep
(3) cost low while some may choose to pay higher cost to maintain
where, ai is action of active consumer that represents energy their comfort levels. Similarly among industrial consumers,
purchased from the grid or sold to the grid. R i is the amount grid stability might be more important.
of renewable generation available to the i th consumer, which
Theorem 1. The payoff function of active consumer u i (ai , ρ)
is known to the aggregator, and β i is the comfort coefficient to
given by equation (3) is a concave function of a i .
represent comfort seeking level of a particular active consumer.
Each active consumer gains some utility for consuming elec- Proof. The payoff function for an active consumer is sum of
tricity as per the energy gain function, which can be written as following three terms: 1) β i log(ai + Ri + 1), 2) −ρai , and 3)
βi log(ai +Ri +1) [35]. The energy gain function of a consumer −γi (v(i) + S(i, i)(â i − ai ) − 1)2 . First term βi log(ai + Ri + 1) is
is an increasing function of a i with diminishing returns. For concave function of a i for βi > 0 and ai > −Ri . Second term
instance, an active consumer consuming less energy benefits is an affine function of a i , and third term −γ i (v(i) + S(i, i)(â i −
much more from receiving additional unit of energy than the ai ) − 1)2 is also concave function of a i . As sum of concave
active consumer consuming higher energy. ρ is the price of functions is concave, payoff function for an active consumer
electricity that a consumer pays to the aggregator. The active shown in equation (3) is concave function of a i .
consumer have to pay ρa i amount of money to the aggregator.
The electricity price, which is decided by the aggregator
v(i) is the distribution grid voltage at node i, and â i indicates
is unknown to the active consumers. The active consumers
previous action of the i th active consumer when voltage at node
observe the probability distribution of uncertain electricity
i of the distribution grid was v(i). S(i, i) is i th diagonal element
price and decide their action. The active consumers decides
of the voltage sensitivity matrix, and γ i is coefficient of grid
the optimal strategy based on the perceived payoff. Equation
awareness that indicates level of the i th consumer’s concerns
(4) represent expected payoff of an active consumer according
for the grid voltage while taking action. Positive values of a i
to the EUT, whereas equation (5) shows perceived payoff of
indicates that the active consumer is buying energy from the
an active consumer according to the Prospect Theory.
grid and negative values of a i indicate that the active consumer
is selling energy back to the grid. a i > −Ri , which implies that ΠiEUT = ui (ai , ρ j )P(ρ = ρ j ) (4)
the active consumer cannot sell more energy than produced j
by the renewable generation. Different consumers may give
different weightage to different metrics. Aggregation of these ΠiPT = ui (ai , ρ j )ω(P(ρ = ρ j )) (5)
metrics into a single payoff function to decide the optimal j
action would depend on the class to which the active consumer Goal of the active consumers is to maximize their perceived
belongs. For example, a highly comfort seeking consumer may payoff based on their subjective probabilities, which is given
have higher values of β i compared to a low comfort seeking by equation (5). Theorem 2 shows the best response action of
consumer. an active consumer.
This work assumes that consumers can measure the dis-
tribution grid voltage at their node of connection. However, Theorem 2. The best response action of an active consumer
information about voltage at nodes other than node i is not that maximizes its perceived payoff is given by equation (6).
available to an active consumer not connected at node i −2z
a∗i = arg max ΠiPT = , (6)
of the distribution grid. An active consumer connected to ai y − y2 − 4xz
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
6
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
7
to the independent system operator (ISO) for buying bulk Consumer Type β γ
electricity in the day-ahead market. Last term in equation (18), A - Highly comfort-seeking and grid aware 0.85 1
B - Comfort-seeking and grid aware 0.75 1
ρrt (aT 1 − A), indicates the cost of additional energy purchased C - Comfort-seeking and grid unaware 0.75 0
from the real-time market. Finally, v + S( â − a) − 1 is the D - Low comfort-seeking and grid aware 0.40 1
metric that indicates deviation of the distribution grid voltage
from the rated voltage. The aggregator decides the probability
distribution of electricity prices that maximizes (18). observe the aggregator’s mixed strategy and decide their opti-
Theorem 3. For 0.11 ≤ P ≤ 0.91, payoff function of an mal response based on their perceived probability distribution.
aggregator is a concave function of P . This leader follower game results in a Stackelberg equilibrium
where all the active consumers and the aggregator are playing
Proof. Payoff function of the aggregator is given by equation their optimal strategy. Change in the Stackelberg equilibrium
(18). From lemma 1, it can be proven that the optimal action with change in consumer type and rationality and their impact
of an active consumer a i is a convex decreasing function of P. on physical grid voltage is investigated in the next section.
As sum of convex functions is convex, a Ti 1 is also a convex
function of P for 0.11 ≤ P ≤ 0.91. V. Simulation and Results
The first term of equation (18), P T ρaT 1 is a concave
To investigate the Stackelberg equilibrium of the proposed
function of P as it is product of a linearly increasing function
system, we use a modified IEEE 69 bus system as shown in
of P (PT ρ) and a convex decreasing function of P (a T 1). The
Fig. 3. A scenario is considered where four type of active con-
second term of equation (18), −ρ da A is a constant. As convex
sumers exist in the distribution grid. Type-A active consumers
function of a convex function is convex, −θv + S( â − a) − 1
are highly comfort-seeking and consider grid voltage while
is a concave function of P. From lemma 1, it can be shown
making decisions. Type-B active consumers are relatively
that −ρrt (aT 1 − A) is a concave function of P as ρ rt and A are
low comfort-seeking and are also grid aware. Type-C active
constant. Summation of concave functions is concave leading
consumers have the same level of comfort requirement as type-
to Theorem 3.
B active consumers but they do not consider physical grid
voltage while making decisions, and type-D active consumers
The optimal mixed strategy of the aggregator corresponds are the least comfort-seeking with the lowest value of β.
to (19) Table I summarizes the type of active consumers and their
P∗ = arg max ua , (19) value of β and γ. In order to maintain fairness of weights,
P values of β and γ are normalized using weight fairness model
which can be found numerically. Based on estimated best proposed in [38]. Fig. 3 shows location of different kind of
response of the active consumers, the aggregator decides active consumers in the distribution grid. There are no active
the optimal probability distribution of electricity prices that consumers connected to bus 1 to 3. Note that the modeling
maximize the aggregator’s objective function, which is the and analysis approach is general and is applicable to other
Stackelberg equilibrium strategy of the aggregator. In a highly scenarios as well. For simplicity in interpretation, renewable
dynamic transactive energy market, electricity prices changes generation for each active consumer (R i ) is set at 10 kWh.
in real-time. In the proposed system model, the aggregator The price of electricity can take one of the following values:
can determine the distribution for price of electricity (based (1) 0.1 unit/kWh; (2) 0.5 unit/kWh, and (3)1.0 unit/kWh
on estimated response) and announce mixed strategy to the (ρ = {0.1, 0.5, 1.0}). For this case, the optimal mixed strategy
active consumers. This probability distribution for electricity of the aggregator at the Stackelberg equilibrium is shown in
price can be treated as the preliminary handshake process that Table II. Actions of a consumer in response to the optimal
establishes equilibrium electricity prices. The aggregator will mixed strategy of the aggregator are shown in Fig. 4(a) and
decide and announce the actual price of electricity based on corresponding distribution grid voltage levels are shown in Fig.
the equilibrium probability distribution. The active consumers 4(b). From Fig. 4(a) it can be seen that energy consumptions
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
8
8 35
Action of Consumer (kWh)
P(ρ = 0.1) = 0.2
6 P(ρ = 0.5) = 0.8
25
2
0 20
-2
15
-4
0 10 20 30 40 50 60 70
(a) Consumer Index 10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Coefficient of Rationality
1.05
Fig. 5: Change in active consumer action with coefficient of rationality
Node Voltage (p.u.)
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
9
15 4
10 3
2
5
1
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
0 10 20 30 40 50 60 70
Coefficient of Rationality
(b) (b) Consumer Index
actions of the active consumers are very different for these Fig. 8: Active consumer actions for different coefficient of rationality
scenarios when the active consumers are relatively less rational
(α = 0.5); however, actions of the active consumers are
almost the same for both scenarios as they are highly rational α = 0.7 for all the active consumers. If coefficient of rationality
(α = 0.9) as shown in Fig. 7(b). for all the active consumers is 0.6 or 0.8, actual response of
the active consumers is represented by curve with diamond
The active consumers with different levels of rationality take
and hexagram markers respectively. This shows that incorrect
different actions under the same probability distribution of
estimation of coefficient of rationality can result in actions
electricity price. A scenario is considered when there is an
different from actions expected from the consumers.
error in the aggregator’s estimate of coefficient of rationality
Number of physical voltage violations can be reduced when
(α). The aggregator decides the optimal mixed strategy as-
the active consumers do consider physical voltage while taking
suming αi = 0.7 for all the active consumers. However, actual
actions. To observe change in number of voltage violations
coefficient of rationality for the active consumers is different.
with γ, a scenario is considered where α = 0.8 and β = 0.85 for
In this scenario, values of β = 0.85 and γ = 1 are assumed
all the consumers. Fig. 9 shows change in number of voltage
to be known to the aggregator. Sample space for the price
violation for different values of γ. As value of γ increases
of electricity is {0.1, 0.5, 1.0} and the optimal mixed strategy
number of voltage violations decrease.
of the aggregator is {0.16, 0.18, 0.66} assuming α = 0.7.
Fig. 8 shows response of the active consumers for a given
mixed strategy of the aggregator. Expected response of the VI. Conclusion and Future-work
active consumers by the aggregator is represented by curve This work develops a Prospect Theory based realistic model
with triangle markers, which corresponds to the case where of the active consumers with varied preferences for electricity
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
10
8 [11] C. Chen, J. Wang, and S. Kishore, “A distributed direct load control ap-
proach for large-scale residential demand response,” IEEE Transactions
7 on Power Systems, vol. 29, no. 5, pp. 2219–2228, Sept 2014.
Number of Voltage Violations
6
[12] A. Faruqui and S. George, “Quantifying customer re-
sponse to dynamic pricing,” The Electricity Journal,
5 vol. 18, no. 4, pp. 53–63, 2005. [Online]. Available:
http://EconPapers.repec.org/RePEc:eee:jelect:v:18:y:2005:i:4:p:53-63
4
[13] A. H. Mohsenian-Rad and A. Leon-Garcia, “Optimal residential load
3 control with price prediction in real-time electricity pricing environ-
ments,” IEEE Transactions on Smart Grid, vol. 1, no. 2, pp. 120–133,
2 Sept 2010.
1 [14] E. Vrettos, K. Lai, F. Oldewurtel, and G. Andersson, “Predictive control
of buildings for demand response with dynamic day-ahead and real-
0 time prices,” in 2013 European Control Conference (ECC), July 2013,
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
γ→
pp. 2527–2534.
[15] Z. Wen, D. ONeill, and H. Maei, “Optimal demand response using
Fig. 9: Change in number of voltage violation at equilibrium with γ device-based reinforcement learning,” IEEE Transactions on Smart Grid,
vol. 6, no. 5, pp. 2312–2324, Sept 2015.
[16] A. H. Mohsenian-Rad, V. W. S. Wong, J. Jatskevich, R. Schober,
and A. Leon-Garcia, “Autonomous demand-side management based on
consumption and different levels of rationality. The interaction game-theoretic energy consumption scheduling for the future smart
between the active consumers and the aggregator is modeled grid,” IEEE Transactions on Smart Grid, vol. 1, no. 3, pp. 320–331,
Dec 2010.
as a Stackelberg leader-follower game from a techno-economic [17] Z. Chen, L. Wu, and Y. Fu, “Real-time price-based demand response
perspective by including physical grid constraints in terms of management for residential appliances via stochastic optimization and
voltage violation within the analysis. The interaction between robust optimization,” IEEE Transactions on Smart Grid, vol. 3, no. 4,
pp. 1822–1831, Dec 2012.
the active consumers and the aggregator is analyzed using [18] D. Eryilmaz, T. Smith, S. Dhople, E. Wilson, and J. Schmitt, “Demand
simulations on the IEEE 69 bus test system. Impact of con- response for industrial-scale energy users in midwest iso: A dynamic
sumer actions and irrationality on physical grid voltage and the programming approach for curtailing energy use,” in 2014 IEEE PES T
D Conference and Exposition, April 2014, pp. 1–4.
aggregator payoff is quantified using simulation. Developing a [19] D. Arnone, V. Croce, G. Patern, A. Rossi, S. Emma, R. Miceli,
time dependent model of the active consumer behavior along and A. O. D. Tommaso, “Energy management of multi-carrier smart
with correlated load scheduling will be pursued in the future buildings for integrating local renewable energy systems,” in 2016
IEEE International Conference on Renewable Energy Research and
work. Applications (ICRERA), Nov 2016, pp. 845–850.
[20] D. Minoli, K. Sohraby, and B. Occhiogrosso, “Iot considerations,
requirements, and architectures for smart buildings x2014;energy op-
VII. Acknowledgment timization and next-generation building management systems,” IEEE
Internet of Things Journal, vol. 4, no. 1, pp. 269–283, Feb 2017.
The authors would like to thank the NSF for providing [21] Y. Wang, Y. Xu, Z. Dong, and W. Zhang, “Multi-stage planning and
support for this research through award No. CNS-1544705. economic analysis of intelligent energy management system in a smart
building,” in 10th International Conference on Advances in Power
The views expressed in this paper are of the authors. System Control, Operation Management (APSCOM 2015), Nov 2015,
pp. 1–6.
[22] S. L. Arun and M. P. Selvan, “Intelligent residential energy management
References system for dynamic demand response in smart buildings,” IEEE Systems
Journal, vol. PP, no. 99, pp. 1–12, 2017.
[1] “The smart grid: An introduction,” US DOE Publication, 2008. [23] K. Jhala, B. Natarajan, A. Pahwa, and L. Erickson, “Coordinated electric
[2] “Gridwise transactive energy framework version 1.0,” The GridWise vehicle charging solutions using renewable energy sources,” in 2014
Architecture Council, January 2015. IEEE Symposium on Computational Intelligence Applications in Smart
[3] R. B. Jr, J. Romanowiz, and C. Staples, “Energy management and home Grid (CIASG), Dec 2014, pp. 1–6.
automation system,” US Patent 5,544,036, 1996. [24] ——, “Coordinated electric vehicle charging for commercial parking
[4] M. Kuzlu, M. Pipattanasomporn, and S. Rahman, “Hardware demon- lot with renewable energy sources,” Electric Power Components and
stration of a home energy management system for demand response Systems, vol. 45, no. 03, pp. 344–353, 2017. [Online]. Available:
applications,” IEEE Transactions on Smart Grid, vol. 3, no. 4, pp. 1704– http://www.tandfonline.com/doi/abs/10.1080/15325008.2016.1248253
1711, Dec 2012. [25] ——, “Real-time differential pricing scheme for active consumers
[5] M. Pipattanasomporn, M. Kuzlu, and S. Rahman, “An algorithm for with electric vehicles,” Electric Power Components and Systems,
intelligent home energy management and demand response analysis,” vol. 45, no. 14, pp. 1487–1497, 2017. [Online]. Available:
IEEE Transactions on Smart Grid, vol. 3, no. 4, pp. 2166–2173, Dec https://doi.org/10.1080/15325008.2017.1362074
2012. [26] K. Jhala, “Coordinated electric vehicle charging with renewable energy
[6] M. A. A. Pedrasa, T. D. Spooner, and I. F. MacGill, “Coordinated sources,” Master’s Thesis, Department of Electrical and Computer
scheduling of residential distributed energy resources to optimize smart Engineering, Kansas State University, Aug 2015. [Online]. Available:
home energy services,” IEEE Transactions on Smart Grid, vol. 1, no. 2, http://krex.k-state.edu/dspace/handle/2097/19767
pp. 134–143, Sept 2010. [27] I. O. Volkova, M. V. Gubko, and E. A. Salnikova, “Active consumer:
[7] D. Cardwell, “Solar power battle puts hawaii at forefront of worldwide Optimization problems of power consumption and self-generation,”
changes,” New York Times, April 2015. Automation and Remote Control, vol. 75, no. 3, pp. 551–562, 2014.
[8] D. Kahneman and A. Tversky, “Prospect theory: An analysis of [Online]. Available: http://dx.doi.org/10.1134/S0005117914030114
decision under risk,” Econometrica, vol. 47, no. 2, pp. 263–291, 1979. [28] S. Ghosh, X. A. Sun, and X. Zhang, “Consumer profiling for demand
[Online]. Available: http://www.jstor.org/stable/1914185 response programs in smart grids,” in IEEE PES Innovative Smart Grid
[9] S. Salinas, M. Li, and P. Li, “Multi-objective optimal energy consump- Technologies, May 2012, pp. 1–6.
tion scheduling in smart grids,” IEEE Transactions on Smart Grid, vol. 4, [29] N. Zhang, Y. Yan, S. Xu, and W. Su, “Game-theory-based electricity
no. 1, pp. 341–348, March 2013. market clearing mechanisms for an open and transactive distribution
[10] S. Althaher, P. Mancarella, and J. Mutale, “Automated demand response grid,” in 2015 IEEE Power Energy Society General Meeting, July 2015,
from home energy management system under dynamic pricing and pp. 1–5.
power and comfort constraints,” IEEE Transactions on Smart Grid, [30] N. Zhang, Y. Yan, and W. Su, “A game-theoretic
vol. 6, no. 4, pp. 1874–1883, July 2015. economic operation of residential distribution system with
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2018.2810819, IEEE
Transactions on Smart Grid
11
Appendix A
Concavity Analysis
Balasubramaniam Natarajan (SM’08) received
Lemma 1. Optimal action of an active consumer a ∗i , given by the B.E. degree (Hons.) in electrical and electronics
equation (6), is convex decreasing function of P. engineering from Birla Institute of Technology and
Science, Pilani, India, and the Ph.D. degree in elec-
Proof. Numerator of equation (6) is linear and decreasing trical engineering from Colorado State University,
Fort Collins, CO, USA, in 1997 and 2002, respec-
function of P̂. Denominator of (6) is y − y2 − 4xz, where tively. Since Fall 2002, he has been a Faculty Mem-
x, y, and z are given by equation (7), (8),and (9). Here y is ber of the Department of Electrical and Computer
linear and decreasing function of P̂; and y2 − 4xz is convex Engineering, Kansas State University, Manhattan,
KS, USA, where he is currently a Clair N. Palmer
and decreasing function of P̂, which is proved in lemma 2. As and Sara M. Palmer Endowed Professor and the
reciprocal of a positive convex function is convex, it can be Director of the Wireless Communication and Information Processing Research
shown that √−12 is convex and increasing function of P̂ Group. His research interests include statistical signal processing, stochastic
y− y −4xz modeling, optimization, and control theories. He has worked on applications
and product of such two functions is concave. spanning wireless communications, sensor networks, biomedical, and cyber
physical systems. He has published nearly 180 refereed journal and conference
articles along with a book titled Multicarrier Technologies for Wireless
Lemma 2. Action of an active consumer that maximizes its Communications.
perceived payoff given by equation (6).
y2 − 4xz, (20)
Anil Pahwa (F’03) received the B.E. (honors)
where, x, y, and z is given by equation (7), (8), and (9) is degree in electrical engineering from BITS-Pilani,
convex and increasing function of P. India, in 1975, the M.S. degree in electrical engineer-
ing from University of Maine, Orono, in 1979, and
Proof. Let b = y2 − 4xz. Therefore, the Ph.D. degree in electrical engineering from Texas
√ A&M University, College Station, in 1983. Since
b = 2γi S(i, i)(v(i) + S(i, i)â i − 1 − S(i, i)(Ri + 1)) 2 + P̂2 1983, he has been with Kansas State University,
Manhattan, where presently he is University Distin-
− 4P̂γi S(i, i)(v(i) + S(i, i)â i − 1 − S(i, i)(Ri + 1)) + 8γi S(i, i)2 guished Professor and holds the Logan-Fetterhoof
0.5 Chair in the Electrical and Computer Engineering
[βi − P̂(Ri + 1) + 2γi S(i, i)(v(i) + S(i, i)â i − 1)(Ri + 1)] . Department. The National Academy selected him as
(21) a Jefferson Science Fellow in 2014 to serve as a Senior Science Advisor
in the U.S. State Department for one year. He worked in the East Asian and
Taking first and second order derivative: Pacific Affairs Bureau on international policies to facilitate higher deployment
∂ √ 1 −1 ∂b of renewable energy. His research interests include distribution automation,
b= b2 (22) distribution planning, renewable energy integration into power systems, and
∂ P̂ 2 ∂ P̂ intelligent computational methods for distribution system applications.
1949-3053 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.