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/ACCESS.2020.2967798, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/[Link]

Dynamic Pricing Mechanism with the

Integration of Renewable Energy Source
in Smart Grid
MUHAMMAD BABAR RASHEED1* , MUHAMMAD AWAIS2 , NADEEM JAVAID3 AND THAMER
ALQUTHAMI4
1*
Department of Electronics and Electrical Systems, The University of Lahore, Lahore, 5400, Pakistan
2
Department of Technology, The University of Lahore, Lahore, 54000, Pakistan
3
Department of Computer Science, COMSATS University Islamabad, 44000, Pakistan
4
Electrical and Computer Engineering Department, King Abdulaziz University, Jeddah 21589, Saudi Arabia
Corresponding author*: Muhammad Babar Rasheed (e-mail: babarmeher@[Link]).

ABSTRACT Day-ahead electricity pricing is an important strategy for electricity providers to improve grid
stability through load scheduling. In this paper, we investigate a general framework for modelling electricity
retail pricing based on load demand and market price information. Without any a priori knowledge, we have
considered a finite time approach with dynamic system inputs. Our objective is to minimize the average
system cost and rebound peaks through energy procurement price, load scheduling and renewable energy
source (RES) integration. Initially, the energy consumption cost is calculated based on market clearing price
and scheduled load. Then, through reformulation and subsequent modification of optimization problem, we
utilize a day-ahead price information to construct individualized price profiles for each user, respectively. To
analyse the applicability of proposed pricing policy, analytical solution is obtained which is further validated
through comparison with solution obtained from genetic algorithm (GA). From results, it is observed that
proposed price policy is non-discriminatory in nature and each user obtained a fair electricity tariff rather
than a day-ahead price, which is based on load demand and consumption variation of other users. We also
show that optimization problem is sequentially solved with bounded performance guarantee and asymptotic
optimality. Finally, simulations are carried in different scenarios; aggregated load and market price, and
aggregated load, individualized load, market price and proposed price. Results reveal that our proposed
mechanism can charge the price to each user with 23.77% decrease or 5.12% increase based on system
requirements.

INDEX TERMS Demand Response, Optimization, Non-discriminatory Prices, Individualized Prices,

Smart Grid, Renewable Energy;

NOMENCLATURE Eunsch energy consumption of unscheduled load

℘ power rating
T total time duration β ON/OFF states of different loads
U total number of users C electricity cost
` set of load ψ real time electricity price
k price factor total
Cuti total cost of utility for selling energy
mrl must run load Spt electricity pricing tariff
dl discrete load ecu1 energy consumption of user 1
cl continuous load f (t) function of time shift of all appliances
Emrl energy consumption of must run load τa,` actual start time of load `
Edl energy consumption of discrete load τs,` start time of load
Ecl energy consumption of continuous load τe,` end time of load
Eus unscheduled energy consumption τsch,` scheduled time of load
Eu total energy consumption of all users

VOLUME 4, 2016 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

τon,` ON time of load in many ways focusing wholesale, distribution, and incen-
τlot,` duty cycle of load tivized sides. In most of DSM techniques, DR programs
Υ peak to average ratio have key importance in order to make them feasible for both
ϕ individualized new price signal utilities and consumers. Wholesale energy market focusses to
γ power consumption limit keep balance between generation and demand, and transmits
ζ fitness of particles energy to end users on variable or flat prices depending on
ηP V energy conversion efficiency (%) of PV sys- the nature of DR program [45]. In DLC techniques, utility
tem has partial access to some loads on consumer sides. During
EP V solar energy generated from PV system high peaks, utility can shut down these loads to avoid possible
AP V area of the solar irradiance (m2 ) blackouts. It can, therefore, improve the grid’s stability, but at
I(r,t) solar irradiance the cost of user comfort. Considering incentivized demand
Ta outdoor temperature side management (DSM) techniques, a major focus is to-
Eg energy obtained form grid wards electricity bill reduction of end users while taking into
consideration grid side [46], [47]. Authors in [48] proposed
I. INTRODUCTION individualised demand aware price policies to calculate sep-
Electricity pricing mechanisms in a day-ahead market charge arate price signals for each user. Initially, a low tariff area de-
a fixed price to residential customers for specific time peri- pending on historical load demand is found, based on which
ods. The objective is to keep a balance between inefficient the load is scheduled. Unlike other pricing schemes, this
pricing system that charge a single price for a long time scheme charges customers depending whether the required
period and complex real time systems that prevails in elec- load falls under low tariff area or not. Consequently, the
tricity market [1]. The goal of dynamic pricing schemes load consumption cost is calculated. In [49], a customised
is to make more efficient utilization of generation capacity retail price has been calculated on the basis of historical load
through load scheduling, optimization and incentivising. This demand data and [50] provides a uniform pricing policy
enables the customers to enhance their consumption level based on dual tariff system. The initial problem was set to
during off and on-peaks hours, without heavily relying on minimize the peak demand, while, the second problem is
costlier generation and other dependences [2]–[7]. set to balance supply-demand. Another way of modelling
In smart grid, various types of demand response (DR) electricity prices is the consideration of time of use (TOU)
programs are available which are specifically designed to pricing tariff in conjunction with thresholding policy, which
manage end user loads and CO2 emissions reduction in re- is reported in [1]. The main objective is the formulation of
sponse to electricity prices [8]–[13]. Generally, there are two problem in such that adaptive prices are obtained which are
types of DR programs being widely used while developing load dependant. In [2], authors used a threshold policy in the
energy management programs; direct load control (DLC) and sense that users can only consume power when price is below
price based [14], [15]. In former, the utility has control to a certain threshold in order to minimize energy consumption
directly turn-off selected load during variation in frequency price. However, none of the work except [1], [2], [48]–[50]
or overload conditions to maintain the power system stability considers the problem of fair pricing distribution, considering
[16]–[19]. Although, these schemes are useful in improving utility and user objectives. Thus, by considering the problem
grid stability, however, this may lead to loss of social welfare of unfair price distribution among all users, we have proposed
and comfort of end users [20]. Thus, in the presence of dy- a novel pricing mechanism by keeping in view the aforemen-
namic load and energy consumption patterns, DLC is rather tioned trade-offs. The proposed work in this paper focusses
considered a passive approach in handling load. In contrast, both the energy retailer and consumer sides. In addition,
the price based schemes [21]–[25] are specifically designed the other constraints of grid and consumer sides are also
for the residential customers to actively participate in DR included, while formulating the cost distribution problem.
programs to reschedule the patterns of loads [26], [27]. In The underlying assumption is that, despite only considering
addition to price based schemes [28]–[34], others offered the cost, comfort, or stability objectives, the affect of low,
attractive incentives and penalties for exceeding pre-specified medium and high energy consumption on electricity bills are
consumption levels. In existing researches, the electricity also studied. It is also worth noting here that the proposed
prices are usually given in advance, hence without optimizing mechanism is equally feasible for both wholesale energy
price structures, the temporal complementarity among end retailers and incentivized consumers. Most importantly, the
users having different load profiles is neglected. Because, utility revenue remains unchanged, which is discussed in
without complementarity, it seems difficult to design a flex- section VIII, while using the proposed individualized pricing
ible pricing tariffs depending on diverse power consumption mechanism. The key contributions of this work are given as:
and market pricing trends. Although, some researchers have
done a preliminary work on pricing mechanism, however, • Load is categorized in such a way to facilitate all types
there required more appropriate model to customize retails of user including those who do not want to schedule
electricity prices in smart grid area. their load in response to market clearing price. In other
In literature [35]- [44], DSM techniques are implemented words, continuous power supply is provided without
2 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

waiting low tariff area which can cause rebound peaks. that overall cost and user discomfort are minimized. It is also
To overcome rebound peaks problem, load of other cat- studied that without coordination between consumer and en-
egories is scheduled in order to balance supply-demand ergy retailer, there are chances of rebound peaks which may
and cost minimization objectives. disturb system stability. In [57], an agent based modelling
• Based on traditional pricing mechanism (section IV-C), and simulation technique is used to measure performance of
a novel pricing mechanism is proposed to charge cus- DR in commercial building. From different perspectives, it is
tomized prices to all types of users without compromis- studied that without price based DR program, commercial
ing the objectives of other users. buildings do not get significant impact. Furthermore, this
• To compare the performance of proposed mechanism, impact differs with different scales of DR participation based
analytical results are obtained and compared with mar- on market condition. In [58], a multi-objective optimiza-
ket pricing technique (Figs. 4,5, table 3). Furthermore, tion of TOU price under multi-model structure is studied
the applicability of proposed pricing model is also as- in order to reduce end user cost and satisfaction. For this
sessed in terms of renewable energy integration. For purpose, load demand data is first pre-processed and then
this purpose, the user-3 is supposed to equipped a pho- divided into different clusters using k-nearest neighbour and
tovoltaic module and then variation on others tariff is adaptive affinity propagation methods. To solve the opti-
analysed. It is found that proposed pricing model is mization problem, non-dominated sorting genetic algorithm
equally feasible in case of RES integration. with probabilistic deviation is used to find optimal solu-
• Based on the mathematical models and respective con- tion to achieve aforementioned objectives. In [59], particle
straints on loads, the optimization problems is formu- swarm optimization algorithm is used to minimize peak-
lated using Knapsack techniques and a heuristic so- valley difference and power loss in regard with TOU pricing.
lution is obtained using GA. It is also observed that To find optimal solution with reduced complexity, multi-
algorithm converged within feasible time to provide objective constrained optimization problem is transformed
global optimum solution. Finally, extensive simulations into single objective unconstrained optimization problem.
are conducted to validate the proposed idea in terms of Results reveal that proposed strategy optimally manages the
cost reduction and fair cost distribution. From results load for different time periods. It can also be seen that without
presented in table 3, it can be concluded that proposed TOU pricing, the load faces more fluctuations. While, the
mechanism charged electricity prices on the basis of load profile seems more stable when employ constrained and
actual consumption levels, rather than only considering unconstrained methods. However, this work is different from
electricity prices and aggregated demand. our proposed work in such a way that real time load patterns
The remaining paper is distributed in the following sec- and price are not considered, which may discourage users to
tions. Section II gives state of the art work, section IV participate in DR programs. In [60], a heuristic DR scheme
discusses system model, where major attentions are given is used to model end user appliances without taking individ-
to highlight the needs and feasibility of proposed method. ualised consumption patterns. In order to develop optimized
Section V presents problem formulation using multiple knap- consumption pattern, a load demand vector is obtained from
sack technique. Section VI then discusses the proposed load energy management controller using stochastic programming
scheduling algorithms. Sections VII and VIII discuss simula- with the objective to PAR minimization.
tion methodology and discussions, respectively. At the end,
conclusion and future work have been presented in section B. DP & CP
IX. The work reported in [51] has considered a DP scheme with
the objective of energy efficiency in SG. With consideration
II. LITERATURE REVIEW of renewable energy buying-back scheme, a DP scheme is
Different optimization techniques [19]–[21], [51]–[56] to reformulated as a convex optimization dual problem, based
optimally control residential load based on market clearing on which a time dependant price is developed. As the scheme
prices are reported in literature. Among these techniques, works in a distributed fashion, so both the utility and end user
[47], [57]–[62] used TOU and real time pricing (RTP), take their benefits. Unlike other schemes, the objective of
[51], [63]–[65] used proposed dynamic pricing (DP) mecha- this work is novel, however, end users having cost reduction
nism, [66] used customized pricing (CP) mechanism, while, as a primary objective are not given priority. In [63], a
other pricing mechanisms are discussed in [67]–[70]. Details Stackelberg game approach is developed to model interaction
of each category are given as follows. between user and electricity producer to find possible trade-
off between a consumer surplus and net-profit. Furthermore,
A. RTP & TOU a renewable energy storage system is also integrated to anal-
In [47], authors have proposed a decentralized approach yse the behaviour, that shows that benefits go to retailer if
to coordinate end users DR. In order to avoid chances of the installed capacity is smaller. On the other hand, the user
rebound peaks, customers’ load profiles have been modified can take benefits of renewable energy if capacity increased
and exchanged with service provider. This process continues from a given threshold level. Authors overcome this problem
until service provider announces the final load profile in such by using Reinforcement learning strategy without any priori
VOLUME 4, 2016 3

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

knowledge of both, consumers and retailer sides. In [64], a which are developed based on day-ahead pricing schemes
service provider acts as a broker between utility and customer [66]–[70]. The works discussed in [47], [57], [58] use price
to purchase electricity from utility company and selling back based DR mechanism to reduce cost and rebound peaks.
to end users. The inherent uncertainties due to dynamic load However, the authors do not consider fair distribution of
consumption and aggregator based price signal, a reinforce- electricity price among all users. Similarly, the main focus
ment learning scheme without a priori knowledge is used that of [59] is to minimize power loss minimization. Unlike our
allows each of the service provider to learn its strategy. DP proposed work, the focus of [60] is to schedule appliances on
[64], however, is useful for utilities in stabilizing the electric the basis of aggregated load profiles leading to unfair price
grid. On the other hand, it is difficult to implement due to distribution. The authors in [66] have done a remarkable
lack of knowledge of variation in consumer load demand. work by considering individualized load demand profiles. In
Similarly, consumers also face difficulties in managing their conclusion, the focus of other works is to schedule end user
loads due to price variation. The work [65] uses a TOU load based on market or customized price signal in such a
mechanism to develop a DP based on customers load profiles. way to minimize cost, rebound peaks and system stability.
The load demand data is communicated to system operator. However, to facilitate end users in terms of fair pricing
In a response, a daily price signal comprising low, mid and through individualized load consumption and pricing policies
high pricing tariffs are designed to reduce high peaks. In is still missing and an needs to be considered carefully.
[66], a demand aware price policy has been developed based
on individualized load demand profiles of all users. Unlike III. MOTIVATION
other schemes, this scheme differs in such a way that each As discussed in section II, the DSM techniques are efficient
user receives a separate price signal which is discriminatory in reducing energy consumption in response to dynamic
in nature. It means, the end users are provided the facility to price signals which are often been recognized to reduce high
manage their load demand without heavily relying on market peaks [11], [12]. In these works, end users can change their
price signal only. consumption schedules in response to price signals rather
than static energy optimization techniques [13]. It increases
C. OTHER SCHEMES the stability, flexibility [28], [29] and reliability [30], [31] of
The work presented in [67] is based on a three stage frame- electric grid by managing end user’s loads. DSM is used for
work, where DR aggregator determines the incentives of- its ability to increase the stability of electric grid with reduced
fered to participating customers for joining load management cost [32], [33]. Furthermore, by encouraging end users to
programs. Initially, the aggregator interacts with wholesale participate in DSM programs in response to dynamic price
electricity retailer and end users to model their behaviours signals, the need for more expensive electricity generators to
using satisfying theory. Then based on aspiration level and meet peak demands can be minimized. DSM programs are
disutility, aggregator decides the tariff and incentives. An- also useful in balancing the energy demand and generation
other work presented in [68] used a simplified conduction capacities. Otherwise, these generating facilities would have
heat transfer method to model energy consumption in a been idle during off-peak hours which may lead to reduced
residential unit. In [69], authors used a model predictive revenue. By introducing DSM programs, participating, and
control theory to reduce energy cost through incorporating electricity suppliers all can take benefits. The deregulations
building dynamics. Is is observed that, rather than saving of the electricity pricing markets, and advancements in ICT
energy from building material, up to 15% & 28% energy can [34], [46], have motivated for the development of more
be saved depending on other factors such as outside tempera- optimal DSM programs. As a result, energy retailers, and
ture and efficient load scheduling. A non-cooperative game electricity transmission organizations have been implement-
approach is developed considering customer’s behaviours ing DSM programs to facilitate end users with lower prices,
to decide whether or not it is feasible for a customer to and increase the elasticity from generation facilities to retail
participate in load management programs [70]. Because, in markets [35]. In this way, end users can reduce their electric-
traditional load management programs, it was pre-assumed ity bills alongwith maximum comfort. DSM researchers have
that customers are rational in their decisions and voluntarily addressed the: (i) minimization of energy consumption and
participate in load management programs. In order to reduce user discomfort, (ii) the stabilization of energy generation
electricity cost, a prospect theory (PT) is used to explicitly and electricity prices, using optimization techniques [36],
incorporate the impact of user behaviour on DSM decision. [37], [46]. Moreover, they also considered the integration of
Furthermore, a new algorithm based on fictitious game is RESs, such as solar, electric vehicle, energy storage systems,
used to develop a Nash equilibrium. etc., [38]–[40], centralized and decentralized load manage-
Detailed literature review shows that few researches pro- ment techniques [46], models of consumer behaviours [41],
posed a day-ahead pricing based optimal load scheduling to and the trends of DSM participation [42].
minimize electricity consumption cost [46], [47], [57]–[60], Despite the importance and benefits of DSM programs,
while other works are based on DP and customized pric- utilities and researchers have reported different approaches
ing mechanisms [63]–[65]. Among price based DR mecha- on their objectives. A common approach being adopted by
nism, few authors proposed customized pricing mechanisms, DSM researcher is to maximize the end user comfort in terms
4 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

of electricity cost minimization in response to time varying end users take monetary benefits, encouraging by utilities
price signals. Meanwhile, DSM researchers focussing utility to take part in these programs. As discussed in section III,
side have introduced baseline capacity limits to enhance the DR prices are calculated on the basis of aggregated energy
stability of power grid during peak hours [43]. Furthermore, consumption. In response, the low or medium energy users
they also introduced the peak power plants to compensate can not take monetary benefits according to what have they
the load while charging high prices during particular hours contributed in DR. So, keeping in view the aforementioned
[44]. From the above discussion, we can easily conclude that limitation, the proposed work directly addresses the cost min-
the major objective of DSM and DR techniques is twofold: imization problem to benefits potential users. In this regards,
(i) to minimize the electricity cost of end users alongwith this work has direct applicability to the regions where smart
high comfort level, and (ii) minimize the high peaks along- grid infrastructure is being implemented (i.e., NYISO, PJM,
with electricity cost reduction [71], [72]. Here is a trade- [45], [73]).
off between user comfort and electricity cost reduction. If
the focus is towards electricity cost minimization, it is very IV. SYSTEM MODEL
difficult to achieve high comfort at the same time. However, We consider a smart grid model in which the energy from
some techniques are being proposed which take into account power grid is distributed among several units/homes to fulfil
this trade-off and proposed different solutions. But, no one load demand. Each home is equipped with different type of
considered the DSM by taking into account individualized loads and a smart meter to communicate with the energy
energy consumption patterns regardless of DR signal which management controller (EMC) to collect energy consump-
is based on aggregated energy demand. tion information of all users. In addition, U 3 is equipped
This work builds upon the same concept which is given with a RES to further analyze the impact of RES on the price
in [56], in which the analytical simulations were performed calculation of other users. The rest of the units are assumed
to highlight the impact of individualized price profiles on to get energy from main grid. In day-ahead DR programs,
end user cost. In the proposed work, we first develop a generally the aggregator is responsible for the calculation of
mathematical model for obtaining price profiles of all users. electricity price signals which are transmitted to consumers
Then the well established optimization problem is solved by using AMI [17]. Then utility sets the electricity price on
using non-linear programming (NLP) technique to schedule the basis of total load demand received from smart meter.
connected loads using market clearing price. Here, it is Regarding EMC, it can be considered as a standalone entity
important to understand that the load scheduling is performed or embedded in retailer/aggregator side, where individualized
on the basis of dynamic electricity prices. Furthermore, the electricity prices are calculated on the basis of load demands.
proposed work is not primarily designed to control or de- Then these prices are transmitted to the respective units
sign new electricity prices. Thus, the mathematical model via smart meters. In order to best describe the proposed
presented through Eqs. 1-8, however, does not address the mechanism, we first categorized the loads in each unit. The
individualized energy consumption trends creating discrim- Fig. 1 gives the conceptual diagram of the proposed model.
ination in electricity prices. So, to overcome this limitation,
the mathematical model is further modified to consider both, A. LOAD TYPES
individualized load trends and dynamic prices. Eventually, Based on energy demand and power rating, the loads are
along with fair price distribution among all users, a balanced categorized into the following categories: (i) discrete load
load profile is also obtained which ultimately improves power (dl ), (ii) continuous load (cl ), and (iii) must run load (mrl ),
system stability. In addition, on-site RES is also incorporated (table 1), such that [` ∈ dl , cl , mrl ]. The EMC is responsible
to further visualize the impact of proposed mechanism on end to control the working cycles of dl and cl , respectively.
user cost profile. However, prior to this, loads are categorized While the mrl does not take part in scheduling process, so
on the basis of power rating and user preferences. Finally, the as to increase the user comfort. Because, these loads are
results obtained using traditional and proposed mechanism considered to be turned ON whenever users require. The
are compared to validate the key findings (Table 2). duty cycles of dl can be altered during operation time due to
The optimization and scheduling of residential loads are their flexible nature. In other words, users can bear delay due
of greatest interest in the world due to variations in energy to variations in appliance starting time. Whereas, cl cannot
demand. As the natural energy sources are depleting quickly, be turned OFF during operation time. Because, these loads
and the world is struggling in finding green energy sources to complete their duty cycles once they are turned ON. The load
fulfil future energy demand with reduced carbon emissions. demand of each load is shown in Fig. 2. While the detailed
Furthermore, the additional reliance on green energy and cli- working of these loads are discussed below.
mate control buildings exacerbates reliable delivery of power
with forecasted demands to ensure reliability. Fortunately, the 1) mrl
potential customers use global pricing schemes (i.e., RTP, We assume that mrl does not take part in DR program,
TOU) within a single utility or DSO. So that end user and hence it is not scheduled over the given time. Although, this
utility can take equal benefits with mutual information shar- load has specified scheduling intervals, which is 24 hours, as
ing via AMI. Although DR programs are utility centric where shown in Fig. 2, load 2. So, it is observed that this load needs
VOLUME 4, 2016 5

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

Unit-1 Unit-3

User

User
User

User
User

User
n
ecU1 = å ecu
u =1

Retailer

cu
n
åe
1
u=
=
User

User
User

User
User

User

User

User
3
U
ec
ecT = ecU 3 + ecU 4
Unit-2 Utility Retailer AC Unit-4
User

User
User

User
User

User
ec U
4
=
n
u=
å
ec u
Retailer

1
n
ecU 2 = å ecu
User

User
User

User
User

User

User

User
u =1

A B
Power Line LAN

FIGURE 1: Conceptual diagram for energy cost calculation: (A) traditional scheme, (B) proposed scheme.

TABLE 1: Appliance energy consumption and duty cycle

requirements. 3 4
Load 2
E (KWh)

E (KWh)
2 Load 1
3
Duty cycle Energy consumption
Load Type Loads 1
(Hour) (kW)
0 2
Load 1 20 2.5 0 5 10 15 20 25 0 5 10 15 20 25
mrl
Load 2 24 3 Time (hours) Time (hours)

Load 3 5 2 3
dl 3
Load 3 Load 4
Load 4 7 2.5
E (KWh)

E (KWh)
2
2
Load 5 8 3.5
cl 1 1
Load 6 8 3
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (hours) Time (hours)
4
Load 5 3 Load 6
continuous supply of power in order to fulfil the required task
E (KWh)

E (KWh)

2
2
during given time interval. The energy demand and power 1

rating of mrl can be denoted by Emrl and ℘mr , respectively. 0 0

0 5 10 15 20 25 0 5 10 15 20 25
The total energy consumption of mrl is calculated by using Time (hours) Time (hours)

the following equation:

FIGURE 2: Energy demand profile over 24 h period.
XX 
Emrl (t) = ℘mrl (t) × βmrl (t) , (1)
mrl t∈T
consumption cost, however, user comfort in terms of schedul-
where, βmrl denotes the ON/OFF state of mrl in the given ing delay has to be compromised. The power consumption of
time slot t. dl is represented by Edl . Let, ℘dl denotes power rating of dl ,
 then total power demand is calculated as:
1 If load is ON
βmrl (t) = (2)
0 If load is OF F. XX 
Edl (t) = ℘dl (t) × βdl (t) , (3)
dl t∈T
2) dl
We assume that power demand of dl can be scheduled from where, βdl is the state of dl in a particular time slot t and
given time slots to any other time slots such as to reduce is given as follows:
overall cost. In other words, this load has deferrable nature
and its normal working can be shifted or changed regardless

1 If load is ON
to user specified time intervals. This ensures minimum power βdl (t) = (4)
0 If load is OF F
6 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

3) cl Energy Demand Taken

from Smart Meters Aggregated
We assume that, unlike dl , the power demand of cl can Energy Demand
(KW/h) Control Unit
be scheduled, however, the normal working can not be in-
terrupted once scheduled. For example, if a load has 5kW RTP Taken from Utility
via Smart Meters
Step-I Electricity Cost

power demand, then scheduling algorithm fulfils the required Electricity Prices
($/MWh)

demand by rescheduling the load in different time slots, based Individualised

Step-II
on minimum priced intervals (Fig. 2, loads 3&4). However, Energy Demand Taken
Prices

from Smart Meters Individualised

the customers have to bear more scheduling delay. Eventu- Energy Demand
(KW/h)
Energy Flow Cost Information
ally, this load would contribute in achieving better optimized RTP

results, as compared to other types of load. Because, their

normal working can not be disturbed once they turned ON, FIGURE 3: A conceptual diagram of proposed system model
i.e., they must complete their scheduled duty cycles. The total for the calculation of individualized price profiles.
energy consumption and power rating of cl are denoted by
Ecl and ℘cl , respectively. The total energy consumption of cl
For example, if electricity pricing signal is totally based on
is calculated by using following equation:
aggregate energy consumption, then it might not be feasible
XX  for all types of consumers. Because, low, medium and high
Ecl (t) = ℘cl (t) × βcl (t) , (5) energy consumers are getting the same prices, irrespective to
cl t∈T their individualized energy consumption patterns. Moreover,
where, βcl denotes the state of cl and can be written as utilities are getting the bill in response to what they are
follows: selling. But, the low and medium energy consumers affects
 due to this strategy. This is the underlying problem in DR
1 If load is ON programs focussing user benefits which is highlighted and
βcl (t) = (6)
0 If load is OF F. solved in this research work. In the following sections, the
traditional method used to calculate energy cost is given.
The total energy consumption Eu (t) of all users is equal to Then based on this, the individualized prices in response to
the sum of aggregated energy consumption of individualized energy demand of associated users are calculated (Fig. 3).
users which is given as:
C. TRADITIONAL METHOD
X X 
In recent DR programs, used in wholesale electricity market,
Eu (t) = Emrl ,u (t) + Edl ,u (t) + Ecl ,u (t) ,
u∈U t∈T
day-ahead electricity prices (hourly prices) are calculated on
(7) the basis of aggregated energy consumption of any specific
region. Whereas the electricity unit price for next hour de-
B. ELECTRICITY PRICE MODEL pends on the current energy consumption level, which can be
This section provides brief information about the day-ahead calculated using the following expression:
electricity market. As it is understood that electricity price
signal is dynamic in nature, due to variations in end user X X 
total
energy demand. Consequently, it is very difficult rather im- Cu (t) = Eu (t) × ψ(t) ≡ Cuti (8)
u∈U t∈T
practical for regulatory authorities to fix the electricity prices
for certain time period. So, it can be expected that electricity Eq. (8) shows that the energy consumption cost of [u ∈ U ]
price signal varies as demand varies. However, in day-ahead must be equal to utility revenue calculated on the basis of
market, the price can be fixed for certain interval of time if the load consumed. In case of TOU or day-ahead RTP, the ψ(t)
load remains within acceptable limits. But, irrespective of all is assumed to be fixed for the time duration t. This expression
the information, there is a probability that loads can be varied Eq. 8 is generic and used for electricity cost calculation.
which can eventually disturb the stability of power grid. To However, the drawback associated with this method is that
avoid all these problems and risks, the flexible DR programs the energy prices set by utilities apply to any specific re-
are being widely used by encouraging the consumers to take gion are not based on individualized energy consumption
part in these programme. Users can take part in DR programs patterns. But these are rather on the basis of aggregate energy
by modifying their energy consumption schedules in regards consumption of that particular region [6]- [63]. Due to this
to the day-ahead price signal. This mechanism also allows the strategy, the low, medium and high energy consumers are
end users to increase or decrease their energy consumption charged equally which penalizes the low and medium energy
in different time slots. In this way, users can get mone- users. To overcome this problem, there must be a mechanism
tary benefits and utilities avoid the risk of grid instability which charged electricity prices to the associated consumers
during high peak hours. This approach, however, is quite on the basis of their consumption level rather than on aggre-
efficient and being widely used now a days. But there are gate basis. To further highlight the problem and its effect on
certain problems associated with these types of techniques. electricity bill, the section VIII provides the details.
VOLUME 4, 2016 7

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

D. PROPOSED METHOD
5
The details given in section IV-C reveal the possible draw- P=2.5 kW
P=3.0 kW
backs and limitations regarding electricity cost saving per- 4.5 P=3.5 kW
P=2.5 with RES
spective. It is also discussed that aggregated price policies 4
P=3.5 with RES
P=3.0 with RES

are feasible for market retailers, reducing user benefits. To

3.5
overcome such types of problems, we have proposed a novel
mechanism to calculate electricity sub-prices in accordance 3

to: (i) the individualized energy consumption patterns, and

k
2.5
(ii) aggregated electricity price signal. Because, it is very
difficult for utility to design/calculate a separate electricity 2

price signal for each user. Consequently, it can further in- 1.5

crease the communication overhead based on which the DR

1
programs work. In order to calculate individualized prices
for potential users, we develop a mathematical expression 0.5

01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
24:00
which calculates these prices for each user/unit according to
Time (hours)
their consumption levels. For this purpose, we need; a day-
ahead electricity price signal obtained from retailer/utility, FIGURE 4: The dynamic variation in the values of k against
and individualized energy consumption of U obtained from different load and RES integrated load profiles, Eq. (9).
smart meters. Then based on these parameters, the price
factor k for each user/unit is calculated by using the following
300 140
equation:
120
P P 250
t∈T u∈U (Eu (t)×ψ(t)
ku (t) = P P 2 . (9)
t∈T u∈U (Eu (t))
100
200

In Eq. (8), Cu (t) denotes the total energy consumption 80

cost of n users in response to day-ahead electricity price 150
k

k
signal ψ. In contrast, Eq. (9) shows that k depends on the 60

energy consumption of each unit. In other words, if any 100

unit consumes more energy, then the values of k will be 40

higher and vice versa. The value of k in accordance with

50
the load profiles of individualized users is shown in Fig. 20

4. While the behaviour of k with respect to electricity cost

0 0
and energy consumption is shown in Fig. 5. The figure also 5 10 15 20 5 10 15 20
E (KW) C ($)
shows the variation in k when RES is utilized. The variations
in value of k shows the individualized energy consumption FIGURE 5: The behaviour of k as a function of E and C.
trends and its possible impact on electricity price signal
(Fig. 11). Because the proposed mechanism has the objective
to distribute the electricity prices in accordance with per
slot energy consumption of the users, while preserving the
k1t1
 t1
ϕtu11 Eut21 Eut31 EuT1
    
Eu1 ···
retailers objective.  ϕtu22   k2t2  Eut2
  t2 Eut22 Eut32 ··· EuT2 
Eut23 Eut33 EuT3 

ϕtu33
  

=
  k3t3  Eu1
× 3 ···
1) Non-discriminatory Prices
 
 ..   ..   .. .. .. .. .. 
Intuitively, in order to have consumers agreeing on paying
 .   .   . . . . . 
electricity bills on the basis of individualized prices, it is ϕTU T
kU EUT EUt1 EUt2 EUt3 ···
(10)
mandatory that these prices must be non-discriminatory.
where, ϕ1 (t), ϕ2 (t), ϕU (t) are respective prices for buying
These are also important for retailers and/or DSO, as discrim-
electricity form retailers. The total electricity cost Cu of U
inatory prices may reduce the potential users accepting these
users is calculated as follows:
prices. In the proposed work, the individualized prices cal-
culated on the basis of load profiles are non-discriminatory. XX 
Where, the individualized prices depend on the respective CU (t) = Eu (t) × ϕu (t) . (11)
load profiles as discussed in section IV-D. In contrast, the t∈T u∈U

traditional method (section IV-C) uses market based prices The steps involved to calculate individualized DR signals
irrespective of individualized load profiles. We define a pric- for user U are given as:
ing tariff Spt = (ecu1 , ecu2 , ecu3 , ψu , k), the individualized
prices ϕ(t) on the basis of k, for each user:
8 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

1) a set of u users u ∈ U directly connected to retailers as primary objectives [4], [5], [46]. Although, cost reduc-
via AMI. In traditional DR program, u (i.e., unit-1 and tion objective satisfies the end-users which can bear extra
unit-2 in Fig. 1) are directly connected to retailers. delay due to load scheduling process. In contrast, comfort
While in the proposed method, u (i.e., unit-3 and unit- maximization may reduce electricity cost and vice versa.
4) are connected to the AC entity which is responsible According to the literature, user comfort can be defined as
for the calculation of DR signals for individual users. follows [46], [79]:
Then the aggregated cost on the basis of net load is sent XX
to utility vis retailer. Here, the accumulator AC (Fig. f (t) = (τn (t)), (12)
1, i.e., centralized EMC) is considered as a separate t∈T u∈U
entity. However, it can also be embedded in retailer where, f (t) is function of time shift of all scheduled loads
unit. under user preferences and lifestyle constraints. In Eq. (12),
2) a time slot t ∈ T , typically with a horizon of 24 equal the variable τn is equal to |sn − αn |, defines the time shift
length time slots. of appliance n. In other words, the discomfort level is equal
3) the desired energy demand profile (kW) of u obtained to the total number of hours shifted from specified time slot.
from AC. So, the objective function is to reduce this time in order to
4) the per-unit electricity tariff ψt (i.e., TOU, RTP) com- provide maximum comfort alongwith minimum cost. Here,
ing from electricity retailers. one important point is that user comfort maximization and
5) then on the basis of ψt and load profiles of u (these electricity cost reduction are contradictory objectives and
profiles may be computed on the basis of historical data difficult to achieve [46].
of [u ∈ U ] in T ), AC calculates the DR signals in
regards to individualized energy consumption profiles.
F. RENEWABLE ENERGY INTEGRATION
6) then consumers reschedule their loads on the basis of
To further analyze the performance of proposed mechanism,
updated DR signals. Please note that maximum energy
we assume that U 3 is equipped with photovoltaic RES.
consumption limit is also implemented in optimization
The proposed mechanism utilizes the renewable energy as a
algorithm to reduce the chances of high peaks on grid
primary source to reduce the electricity prices (i.e., individu-
side. Otherwise, the grid stability could be affected.
alized prices). The remaining units use the utility energy as a
7) in each time slot [t ∈ T ], AC updates the price signal
primary source and receive the prices on the basis of energy
for each u on the basis of individualized load profiles.
consumption. The solar PV energy used in the proposed work
is obtained from the model demonstrated in [74], [75], Eq.
13. The output power is shown in Fig. 6:
2) Discussion about k
As discussed in section IV-D, the proposed scheduling mech-
anism for residential DR depends on factor k (Eq. 9). Where, E P V = η P V × AP V × Ir (1 − 0.005(T a − 25)) . (13)
the k further depends on two factors; Eu and ψ. We provide
numerical results (Fig. 4 to show the dependence of Eu and
ψ on k, which is a key factor for the calculation of individ- 7

ualized prices. The left side of Fig. 5 shows the relationship RES (User 3)

between energy consumption and k, while the Fig. 4 shows 6

the variation in k using: (i) energy demand pattern without

Renewable Energy Generation (KW)

RES, and (ii) energy demand pattern with RES. The consid- 5

ered load and respective categories have been given in table

4
1. Fig. 4 shows that when energy consumption increases, the
value of k decreases. Because, the energy consumption factor 3
Eu2 is in denumerator in Eq. (9). In contrast, the right side of
Fig. 5 shows the relationship between C and k. It is clear 2

from the figure that cost increases as we increase the value of

k, which depends on E. In conclusion, it is obvious from Fig. 1

5 that individualized prices depend on k which is the base of

proposed work. 0
01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
24:00

Time (hours)

E. USER COMFORT
FIGURE 6: Renewable energy profile over the period of 24 h
From detailed literature review [21]- [38], it can be concluded used in U-3.
that most of the energy management schemes have cost
reduction as a major objective. Whereas, other schemes have Keeping in view the aforementioned points, the proposed
considered both the cost reduction and comfort maximization work not only considers the delay based comfort, but also
VOLUME 4, 2016 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

provides end user comfort with extra cost savings to low and
XX 
middle class users. For this purpose, sections IV-A and IV-B min Eu (t) × ψu (t) (14)
discus different types of loads alongwith preferences and t∈T u∈U
energy price model used in the proposed work, respectively.
s.t:
Then, section VIII discusses the real time benefits in terms of
cost reduction and comfort maximization.
τs,` ≤ τa,` ≤ τe,` , (14a)
V. PROBLEM FORMULATION τsch,` = τon,` , ∀ mrl , (14b)
In this work, multiple knapsack (MK) problem formulation
τsch,` = τon,` ≤ τe,` − τlot,` , ∀ del , (14c)
technique has been used to formulate load scheduling prob-
lem [80]. In general, knapsack is a formulation technique τsch,` = τon,` ≤ τsch,` ≤ τon,` + τlot,` , ∀ cel , (14d)
in which the “knapsack” is filled with multiple “objects” β(t) = τlot . (14e)
considering its “values” to take maximum benefits. The MK
technique has been used to formulate the proposed load Where, Eq. (14) is cost minimization objective function,
scheduling problem using individualized pricing. However, Eq. (14a) denotes upper and lower limits on load start and
prior to the formulation, following assumption are consid- end times, Eqs. (14b, 14c, 14d) give scheduling horizons for
ered: mrl , del and cel , respectively. Eq. (14e) demonstrates that
total working hours of any load, that must be equal to its duty
1) Multiple time slots t are considered as a “MK”.
cycles.
2) Number of loads have been considered as “objects”.
3) Energy consumed by each user is taken as “weight”.
B. CENTRALIZED ENERGY MANAGEMENT PROBLEM
4) Energy consumption cost of each user is considered as
FORMULATION
“value”.
5) Energy consumption limit in each time slot t is consid- In this mechanism, all users share their energy consumption
ered as “capacity”. profiles in neighbourhood area network. In first step, the
electricity cost in t is calculated using electricity prices ψ
By taking into consideration the aforementioned assump- obtained directly form utility. In second step, the energy
tions, we formulate the objective function with the aim at consumption of each user and their respective cost have been
reducing electricity cost in association with different con- calculated. In third step, the individualized electricity prices
straints. On the other hand, for reliable and un-interruptible on the basis of energy consumption data in previous hours
energy supply, the power grid must not overburdened due to are calculated as shown in Fig. 11. It is worth noting here
heavy load. To achieve this objective, a limit on total energy that for the calculation of individualized prices ϕ, electricity
consumption in each time slot t and RES are also considered. price signal obtained form utility and energy demand patterns
So the total power consumption of each user does not exceed are used as input parameters. In response, ϕ for each unit
the maximum power capacity in each t. This mechanism u are calculated. In this way, electricity prices are fairly
is being widely used in different DSM programs [4], [5] distributed among all units while preserving utility revenue.
and has provided efficient results. Although, it increases the Furthermore, the total energy consumption limit is imposed
stability of power grid, however, it can disturb end user in optimization problem to lower high peaks on grid side.
comfort to some extent. But, the major focus of this work The centralized problem is now formulated as multiobjective
is towards facilitating end users with significant amount of optimization as written below:
bill reductions.
The objective function is divided into two parts:
XX 
1) in the first part, the load scheduling problem is formu- min [Eu (t) × ϕu (t) − E P V (t)] + f (t)
lated and cost reduction objective is achieved by using t∈T u∈U
day-ahead pricing signal. (15)
2) in second part, the load scheduling problem is formu- s.t:
lated and cost reduction objective is achieved using
individualized prices. β(t) = τlot,` , (15a)
X X 
(Eu (t) × β(t)) ≤ γu (t) , ∀, [` ∈ mr, ce, de]
A. DISTRIBUTED ENERGY MANAGEMENT PROBLEM
u∈U t∈T
FORMULATION (15b)
In distributed energy management mechanism, the load has XX
total


been controlled on the basis of RTP signal obtained from Cu (t) − Uuti (t) = 0. (15c)
u∈U t∈T
utility as shown in Fig. 7. In first phase, energy consumption XX
Eu3 (t) = Eg (t) + E P V (t) ∀[` ∈ mr, ce, de].


limit is not imposed, because the EMC works in a distributed
way and is independent to the other units. The optimization u∈U t∈T
problem is then formulated as follows: (15d)

10 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

28 140
RTP Signal Hour-1
Hour-2
26 130
Hour-3

24 120

22 110
Electricity price ($)

20 100

C ($)
18 90

16 80

14 70
Feasible Points

12 60

10 50

8 40
01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
24:00
100 200 300 400 500 600 700 800 900 1000
No of Iterations
Time (hours)

FIGURE 7: RTP signal used in the proposed work which is FIGURE 8: Behaviour of GA w.r.t., cost and total number of
obtained from day-ahead electricity market [45]. iterations in achieving global optimal solution.

Where, Eq. (15) is cost minimization objective function, [64]. Because, to find global optimum solution in the search
Eq. (15a) shows that operating hours of any load must satisfy space when many local optimum solutions are present is a
its duty cycle requirements, Eq. (15b) denotes maximum classical problem. In this regard, the GA has the ability to find
energy consumption limit considered to minimize high peak global optimum with high accuracy through parallelizable
during overload conditions. Eq. (14c) provides the cost bal- search. It is evident from the Fig. 8 that proposed algorithm
ance expression (i.e., utility revenue must be equal to total is designed in such a way that it always converged within
power dispatch). The Eq. (15d) shows that (U 3) utilizes RES certain bounds and limits. For example, in each time slot, the
as a primary energy source along with grid energy. While the algorithm searches global optimum solution from the entire
rest of the users fulfil their energy demand from electric grid search space, where other local optimum solutions exist. In
only. Initially, the energy consumption limit can be calculated order to avoid GA being trapped in a local optimum solution,
using the expression: we have expanded the search space (i.e., 400 population
X  size). It is then ensured that the optimal solution is always
1
γ (t) = max (β(t) × Eus (t)) . (16) feasible. Because, if we consider minz∈S f (z) as a standard
t∈T minimization problem, where, S ⊂ IR denotes the feasible
For u, the above equation would become: set. Any z ∈ S is considered to be feasible point and
X conversely, any z ∈ IR\S := {z ∈ IR : z ∈ / S} is infeasible.
γu (t) = γ 1 (t) − β(t) × Eu (t). (17) In another sense, a solution minz1 ∈S f (z) is known as Pareto
t∈T optimal solution if and only if there does not exist another
Where, Eq. 17 gives power consumption capacity limit for solution in the same search space which dominates it. In our
u, which depends on Eq. 16. case, all solutions are Pareto optimal (Fig. 8).

VI. PROPOSED ALGORITHMS A. DISTRIBUTED ALGORITHM FOR EMC

The DR strategy is applied to both distributed and centralized In a distributed manner, we assume a user u ∈ U has a variety
schemes. For each unit, there is a distributed EMC to control of loads with different parameters. The EMC communicates
load in response to the traditional RTP. On the other hand with different types of loads, and take energy consumption,
the AC (Fig. 1) calculates ψ on the basis of aggregate duty cycle and priorities of interruption as input parameters
energy consumption of each user u ∈ U , which results in the algorithm. After that, the EMC performs scheduling for
final energy consumption pattern of the user. Each user is respective user in order to reduce electricity bill. In EMC the
equipped with several types of loads with different control traditional price signal is used, in which each user has equal
parameters. To deal with all these parameters, i.e., duty cycle, price independent to their energy consumption.
energy consumption, priorities of interruption according to The working steps of proposed distributed EMC algorithm
load type, is a challenging task. The evolutionary algorithm are given as follows (algorithm 1):
such as genetic algorithm (GA) [37], [76]–[78] has the ability
to solve load scheduling problem in association with all B. CENTRALIZED ALGORITHM FOR C-EMC
constraints. GA has good computational ability and high con- In a centralized manner, the EMC manages the load on the
vergence rate as compared to other mathematical approaches basis of energy consumption patterns and assigns different
VOLUME 4, 2016 11

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

Algorithm 1 Distributed EMC Algorithm the individual energy consumption cost may differ in this
1: Required Unscheduled pattern, population size, max. case, and is depends on their energy consumption. High
generations, N, ψ. energy consumers always pay their bills according to the
2: Initialize random population which represents the pat- load consumed and vice versa. The factor by which the price
terns of appliances. among the users’ is distributed is calculated by using Eq. (9).
3: for t=1:24 do This mechanism assures grid stability in terms of reduced
4: EMC checks for available RES source of user 3. peaks by restricting consumers to limit their consumption.
5: if g(t) ≥ esu3 (t) then The working steps of proposed centralized EMC algorithm
6: Turn ON the appliance i of user 3. are given as follows (algorithm 2):
7: else
8: for i=1:popsize do Algorithm 2 Centralized EMC Algorithms
9: Evaluate fitness function eq. (14) 1: Required Scheduled pattern from distributed EMC, pop-
10: ζ=f itness ulation size, max. generations, N, ψ.
11: if (ζ(i) < ζ(i − 1)) then 2: Initialize random population which represents the pat-
12: ζ(i) = ζ(i) terns of appliances and t = 1.
13: if cei (t − 1) == 1 then 3: Calculates k on the basis of distributed EMC energy
14: if (τi ≤ τlot ) then consumption data using Eq. (9).
15: c(t) = 1 4: Evaluate ϕ for each user u ∈ U using Eq. (10).
16: end if 5: Communicates this new ϕ to its respective user u ∈ U .
17: end if 6: Evaluate fitness function using Eq. (15).
18: if dei (t) == 1 then 7: From step 6, select optimal pattern which satisfies Eq.
19: if (τi ≤ τlot ) then (14a-14e) and (15b-15d).
20: de(t) = 1 8: Turn ON the loads according to optimal pattern and
21: end if calculates τ and Ω for the next evaluation.
22: end if 9: Generates new population to calculate fitness until con-
23: else vergence criteria is met.
24: ζ(i) = ζ(i − 1) 10: t= t+1, go to step 3, till t=24.
25: end if
26: end for
27: scheduled-load(1,:)=popnew(1,i) VII. SIMULATION METHODOLOGY
28: if scheduled − loadi == 1 then The optimization models discussed in the section IV-D are
29: τlot = τi − 1 tested and evaluated on a hypothetical test case where differ-
30: end if ent homes/units are considered. It is assumed that each home
31: has different types of loads having different duty cycles, and
32: end if power ratings as given in (table 1). To further check the
33: Generate new population. Select crossover pair a, b validity of the proposed mechanism, U 3 is equipped with on-
34: if Pc > rand then site RES which can be utilized during peak hours to reduce
35: crossover(a, b) electrify cost. Each home is equipped with various types of
36: end if loads having different power ratings and variable duty cycles
37: if Pm > rand then (table 1). The optimization program runs for a complete day
38: mutate(a, b) where [t ∈ T ] time slots are further divided into sub-time
39: end if slots having equal length of one hour. The simulations are
40: popnew(popsize,N) performed for two different cases: (i) load scheduling using
41: end for market based clearing prices (i.e., TOU, RTP), and (ii) using
individualized prices as discussed in section IV-D.
The former one uses traditional RTP signal taken from [45]
and is shown in Fig. 7. Then on the basis of RTP signal,
prices for each user u ∈ U . The EMC communicates with algorithms solve load management problem [6], [7], [46],
utility and local EMC to achieve cost minimization and PAR [63], [64], [81] to minimize electricity cost and high peaks on
reduction goals. For this purpose, each user shares its sched- grid side. Although the traditional energy management mech-
uled energy consumption pattern provided by the distributed anisms work on the basis of price based DR programs and are
EMC to AC. Then on the basis of these patterns, AC assigns efficient for cost, comfort and grid stability. But, the discrimi-
different prices to each user. Here, it is assumed that the price natory price policies do not ensure the consumers’ savings. In
is distributed among all users in such a way that the aggregate contrast, the proposed work uses non-discriminatory prices
energy consumption cost of all users is same as in the case which are calculated as described in section IV-D. These
of distributed DSM strategy discussed earlier. In contrast, cases clearly depict the impact of individualized prices on
12 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

electricity cost of end users. Then the final optimization

300
program is solved by using GA providing the optimal results C us

given in table 2. As the load optimization problem is well C s (U1)

C s (U2)
250
behaved. So the existing optimization engine is sufficient for C s (U3)
C s with RES (U1)
the numerical solution. The MATLAB (Matrix Laboratory) C s with RES (U2)

language is used to solve the problem with the objective 200 C s with RES (U3)

of cost minimization. The code was executed on Macbook

C ($)
Laptop, 1.7GHz Intel Core i7 processor. 150

100

18
Eus
ec u1 50
16
ec u2
ec u3
14 ec u1 with RES
ec u2 with RES 0

01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
24:00
ec u3 with RES
12
Time (hours)

10
E (KWh)

FIGURE 10: Cost incurred by different users over the period

8 of 24 h using OP.
6

4 obvious that scheduled cost is always less than unscheduled

2
cost alongwith utility objective i.e., reduced PAR.
Figs. 11 and 12 elucidate another case where energy prices
0
are calculated on the basis of proposed mechanism discussed
01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
24:00

Time (hours) in section IV-D. For validation purpose, the energy consump-
FIGURE 9: Energy consumption profile of all users over 24 tion patterns given in (Fig. 9) remain same. However, the
h period. cost incurred by different users is different. This is because
the individualized electricity prices are calculated on the
basis of energy consumed by each user. t is also discussed
in section IV-D that the individualized prices of all users
VIII. RESULTS AND DISCUSSIONS depend on factor k. While the other energy management
Fig. 9 shows the energy consumption profile of all users over mechanisms use market based pricing signal irrespective to
the period of 24 h. The optimization program successfully the energy consumption patters. So, prior to the calculation
solves the load scheduling problem in response to respective of electricity in case of traditional and proposed mechanisms,
constraints fulfilling the energy demand. In case of U 1&U 2, first we compute DR signals for all users on the basis of load
the energy source provided by retailers acts as “first choice” consumed in [t−1] h. Fig. 11 shows the individualized prices
coming into operation to fulfil the energy demand. However for all users over the period of 24 h. As a first observation,
in case of U 3, the RES acts as first choice to fulfil energy the electricity prices seem different/fluctuating, which show
demand. If the energy demand during particular hours ex- their dependency on individualized demand. In other words,
ceeds, the main energy source is then utilized. In unscheduled these prices are calculated on the basis of individualized
case, optimization algorithm schedules the loads without demand patterns in associated units. As the energy consump-
considering associated constraints. In response, the load is tion patterns of U 3 seems lower (Fig. 9). Consequently, the
even turned ON during peak hours (i.e., 10:00-12:00h, Fig. individualized price signal is comparatively low.
9). It is worth noting here that the energy consumption of U 3 Fig. 12 further demonstrates the cost of residential units
is comparatively less due to RES integration. In scheduled in terms of individualized energy consumption patterns
case, optimization program turns ON the load in various time over the period of 24 h. In early morning 01:00-6:00h, it
slots and avoids the peak hours. The total cost of U 1&U 2 is observed that the average cost using proposed method
seems lower during 1:00-06:00h (Fig. 10). The average cost (Cs (U1 ), Cs (U2 ), Cs (U3 )) with RES is comparatively lower.
during 07:00-15:00h is comparatively high then the cost This is due to the fact that individualized prices depend
during 1:00-06:00h. From 16:00-24:00h, again the average on the energy consumption pattern of each home, rather
cost of U 1&U 3 is almost same. While the electricity cost than aggregated energy demand. While during 07:00-14:00h,
of U 3 during all hours is less. The reason behind reduced electricity cost of all users except U 3 is much higher due
cost is the integration of on-site RES. These typical results to high individualized electricity prices (Fig. 11). It can also
originate form the fact that optimization algorithm follows be seen that electricity cost of U 3 is comparatively less
the maximum energy consumption limit Eq. (14b) leading due to incorporation of RES. In traditional case (Fig. 10),
to significant reductions in hourly cost. In doing so, it is U 3 incurs more cost during 7:00-10:00h. While, the cost in
VOLUME 4, 2016 13

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

TABLE 2: A comparison of electricity cost using OP and NP.

Without RES With RES
Users Total load (kW) Cost-OP ($) Cost-NP ($) % age Total load (kW) Cost-OP ($) Cost-NP ($) % age
U1 201.50 2337.08 2306.08 -1.34 201.50 2363.40 2491.10 +5.12
U2 201.50 2415.08 2470.04 +2.22 201.50 2437.60 2569.20 +5.12
U3 201.50 2337.04 2313.08 -1.03 139.03 1349.90 1090.60 -23.77
Total 604.50 7089.02 7089.02 – 542.03 6150.90 6150.90 –
OP=Old Price, NP=New Price.

Another important aspect of the proposed scheme is that the

35
ψ
prices are calculated on hourly basis. In other words, if any
φ (U1)

30
φ (U2) user is consuming more energy e1 in h1 hour, that user will
φ (U3)
φ with RES (U1)
φ with RES (U2)
be charged the price according to the proposed mechanism.
25
φ with RES (U3)
The role of OP and N P schemes is further demonstrated in
table 2 which gives the comparison of energy consumption
20 and electricity cost. In case when RES is not utilized, the
Price ($)

electricity cost of U 1&U 3 using N P is reduced. While the

15
electricity cost of U 2 is 2.22% increased. This is due to high
value of k. While the rest of the users U 1&U 3 consumed
10
1.34% and 1.03% less energy and consequently, they receive
5
less bill. In contrast when RES is incorporated in U 3, then
the energy demand is primarily fulfilled by using RES. If
0
the energy demand exceeds from RES capacity, the grid
01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
24:00

energy is then utilized as a secondary choice. It is worth

Time (hours)
noting here that the value of k (Eq. 9) depends on energy
FIGURE 11: Individualized prices for all users over the consumption factor which ultimately changes the electricity
period of 24 h (Eq. 9). cost. So the variation in k gives 5.12% more electricity
cost in case of U 1&U 2. Furthermore, the total amount of
energy obtained form grid is also reduced by 11.52% when
300 RES is used contributing towards grid stability. Hence it is
C s (U1)
C s (U2)
demonstrated that the individualized prices depend on the
250
C s (U3)
C s with RES (U1)
energy consumption patterns of all users/homes.
C s with RES (U2)
C s with RES (U3)
Finally, the table 3 provides a comparison of proposed
200 work with counterpart techniques regarding cost and peak
load minimization. Work reported in [48]–[50] used dif-
ferent pricing policies for peak load minimization, through
C ($)

150

optimized load scheduling. Similarly, the work presented in

100
[1], [63] significantly minimized end user cost when load
is scheduled in accordance with other than market clearing
50
prices such as TOU and CP. However, the proposed work
focusses on devising price profiles for all users such that
each user must be provided a separate price signal based
0
on demand consumption and RTP obtained directly from
01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
24:00

Time (hours) electricity market. Unlike other works, the proposed work
FIGURE 12: Cost incurred by different users over the period focusses to provide a mathematical model to construct load
of 24 h using NP. dependant cost profiles rather than minimize the end user
cost through optimized load patterns. However, in order to
provide the justification of obtained achievement, we used a
proposed scheme is comparatively lower. Therefore, the cost GA to obtained optimal load patterns, so as to compare the
reductions can be attained if the value of the k is lower, which cost against unscheduled case as shown in Table 2.
is demand dependant. Finally, table 2 shows the net revenue
obtained by utility in centralized and distributed algorithms. IX. CONCLUSION
It is evident from the table that using proposed scheme, the In this paper, we have proposed a novel pricing mechanism
utility objective remains same. However, the individualized for demand management using individualized price policies
users may take benefits if they are consuming less energy. in conjunction with energy demand, electricity price and RES
14 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

TABLE 3: A comparison of proposed results with some relevant works.

Reference Pricing Optimization Objective Limitation Comparison
Technique Technique
[48] Individualised DAPP algorithm Rebound peaks and average Price is calculated based on his- (2.08&5.32%),
price policy cost minimization torical demand, while the demand peak demand
is dynamic. For RTP, which is not minimization
known in advance, it seems infea-
sible
[49] TOU and Cus- MINLP Customized electricity retail Proposed price is based on histor- 4% peak load
tomized prices prices based on end users’ ical load demand, while the ac- minimization
load features tual demand is dynamic. So, there
could be variation in price levels
[50] Uniform pricing Greedy and Peak demand is minimized Energy loss problem, as compared (1.3–9%)
approach Sliding-Window & power demand matches to TOU pricing mechanism. It peak demand
the supply means, τlot problem is there minimization,
depending on
W
[1] TOU Monte Carlo and Minimize the mean electric- Focus is to determine TOU pric- (7.23-7.73%)
numerical search ity price ing based contract capacities. Cus- reduction in
tomer centric approach end users’ cost
[2] Market price sig- Dynamic Prediction of the aggregate Supply-demand mismatch as some –
nal aggregate response of a large number loads are deferred to minimize cost
model based on of individual loads
threshold policy
[63] DAHP Stackelberg Reduces user anxiety of TOU and CP performed worse due 10.7%>TOU
game theory price uncertainty to flexibility problem and
10.17%>CP
Proposed RTP based pro- Binary GA Minimized cost, less CMS between utility and con- See results in
posed price pol- chances of rebound peaks, sumer is required Fig. 1. Extra cost Table 2
icy supply-demand balance and computational burden
DAHP=Day ahead hourly price, CP=Critical price, CMS=Community management system.

to incentivise low energy users. The major contribution of REFERENCES

this work is to facilitate low energy consumers by providing [1] Y. Hung and G. Michailidis, “Modeling and Optimization of Time-of-Use
incentives in the form for reduced electricity bill. In addition, Electricity Pricing Systems," in IEEE Transactions on Smart Grid, vol. 10,
no. 4, pp. 4116-4127, July 2019. doi: 10.1109/TSG.2018.2850326
electricity cost is reduced in such a way that high peaks [2] M. I. Ohannessian, M. Roozbehani, D. Materassi and M. A. Dahleh,
are also reduced while maintaining the utility revenue, even “Dynamic estimation of the price-response of deadline-constrained elec-
though the individualized prices are non-discriminatory. For tric loads under threshold policies," 2014 American Control Conference,
Portland, OR, 2014, pp. 2798-2803. doi: 10.1109/ACC.2014.6859473
true analysis, the results of the proposed scheme are com- [3] C. O. Adika and L. Wang, “Demand-side bidding strategy for residential
pared to the traditional scheme where price based DR signal energy management in a smart grid environment,” IEEE Trans. Smart Grid,
is used by the consumers operated under same DSO. The vol. 5, no. 4, pp. 1724–1733, Jul. 2014.
[4] M. B. Rasheed, N. Javaid, A. Ahmad, Z. A. Khan, U. Qasim, and
analysis conducted in this paper illustrates the advantages N. Alrajeh, “An efficient power scheduling scheme for residential load
of the individualized prices thorough extensive simulations management in smart homes,” Applied Sciences, no. 4 (2015): 1134-1163.
considering the data given in table 1. It was also demon- [5] M. B. Rasheed, N. Javaid, A. Ahmad, M. Jamil, Z. A. Khan, U. Qasim,
and N. Alrajeh, “Energy optimization in smart homes using customer
strated in section VIII that the proposed scheme distributes preference and dynamic pricing,” Energies, 9(8), p.593.
electricity prices among all users/units on the basis of load [6] X. Chen, T. Wei, and S. Hu, “Uncertainty-aware household appliance
consumed while keeping utility objectives high. The further scheduling considering dynamic electricity pricing in smart home,” IEEE
Trans. Smart Grid, vol. 4, no. 2, pp. 932âĂŞ941, Jun. 2013.
cost reductions can be obtained if the coordination among [7] P. Harsha, M. Sharma, R. Natarajan, and S. Ghosh, “A framework for the
all the users are incorporated. This objective can easily be analysis of probabilistic demand response schemes,” IEEE Trans. Smart
achieved using multiagent technology. So, in future, we will Grid, vol. 4, no. 4, pp. 2274–2284, Dec. 2013.
[8] Outlook, BP Energy. “Acceptable at: [Link] bp. com/energyoutlook
use multiagent technology for real time coordination and 2035.” January 2014: 53.
control to further reduce the energy consumption costs of [9] [Link] Last accessed, 05
individualized users. September, 2019
[10] K. Ng and G. Sheble, “Direct load control - A profit based load manage-
ment using linear programming,” IEEE Trans. Power Syst., vol. 13, no. 2,
pp. 688–694, May 1998.
[11] K. Dietrich, J. M. Latorre, L. Olmos and A. Ramos, “Demand Response
ACKNOWLEDGEMENT in an Isolated System With High Wind Integration,” in IEEE Transactions
on Power Systems, vol. 27, no. 1, pp. 20-29, Feb. 2012. doi: 10.1109/TP-
This Project was funded by the Deanship of Scientific Re- WRS.2011.2159252.
search (DSR), King Abdulaziz University, Jeddah, under [12] M. N. D. Macedo, J. Galo, L. Almeida, and A. Lima, “Opportunities
and Challenges of DSM in Smart Grid Environment,” in The Third
grant No. (DF-307-135-1441). The authors, therefore, grate- international conference on smart grids, green communications and IT
fully acknowledge DSR technical and financial support. energy-aware technologies; 2013. p. 156–60.

VOLUME 4, 2016 15

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

[13] H-P. Chao, “Demand response in wholesale electricity markets: the choice Power and Comfort Constraints,” in IEEE Transactions on Smart Grid,
of customer baseline,” J Regul Econ 2011;39(1):68–88. vol. 6, no. 4, pp. 1874-1883, July 2015. doi: 10.1109/TSG.2014.2388357.
[14] N. Ruiz, I. Cobelo, and J. Oyarzabal, “A Direct Load Control Model [34] F. Ye, Y. Qian and R. Q. Hu, “A Real-Time Information Based Demand-
for Virtual Power Plant Management,” IEEE Transactions on Power Sys- Side Management System in Smart Grid,” in IEEE Transactions on Parallel
tems,24(2), 2009, pp.959-966. and Distributed Systems, vol. 27, no. 2, pp. 329-339, Feb. 1 2016. doi:
[15] B. Jiang, A. M. Farid, and K. Youcef-Toumi, “Demand side management 10.1109/TPDS.2015.2403833.
in a day-ahead wholesale market: A comparison of industrial & social [35] PJM State & Member Training Dept., PJM Demand Side Response; 2011.
welfare approaches. Applied Energy, 156, pp.642-654, 2015. [36] M.B. Rasheed, N. Javaid, M. Awais, Z.A. Khan, U. Qasim, N. Alrajeh, Z.
[16] K. Al-jabery, Z. Xu, W. Yu, D. C. Wunsch, J. Xiong and Y. Shi, “Demand- Iqbal, and Q. Javaid, “Real Time Information Based Energy Management
Side Management of Domestic Electric Water Heaters Using Approximate Using Customer Preferences and Dynamic Pricing in Smart Homes,”
Dynamic Programming,” in IEEE Transactions on Computer-Aided De- Energies, 9(7), p.542, 2016.
sign of Integrated Circuits and Systems, vol. 36, no. 5, pp. 775-788, May [37] Y. H. Lin and M. S. Tsai, “An Advanced Home Energy Management
2017. doi: 10.1109/TCAD.2016.2598563 System Facilitated by Nonintrusive Load Monitoring With Automated
[17] L. Park, Y. Jang, S. Cho; J. Kim, “Residential Demand Response Multiobjective Power Scheduling,” in IEEE Transactions on Smart Grid,
for Renewable Energy Resources in Smart Grid Systems,” in IEEE vol. 6, no. 4, pp. 1839-1851, July 2015. doi: 10.1109/TSG.2015.2388492.
Transactions on Industrial Informatics , [Link], no.99, pp.1-1. doi: [38] M. B. Rasheed, N. Javaid, A. Ahmad, M. Awais, Z. A. Khan, U. Qasim,
10.1109/TII.2017.2704282 and N. Alrajeh, “Priority and delay constrained demand side management
[18] O. Erdinc, A. Tascikaraoglu, N. G. Paterakis, Y. Eren and J. P. S. Catalao, in real-time price environment with renewable energy source,” Int. J.
“End-User Comfort Oriented Day-Ahead Planning for Responsive Resi- Energy Res., 40: 2002–2021, 2016. doi: 10.1002/er.3588.
dential HVAC Demand Aggregation Considering Weather Forecasts,” in [39] M. Mazidi, A. Zakariazadeh, S. Jadid, P. Siano, “Integrated scheduling
IEEE Transactions on Smart Grid, vol. 8, no. 1, pp. 362-372, Jan. 2017. of renewable generation and demand response programs in a microgrid,”
doi: 10.1109/TSG.2016.2556619. Energy Convers Manage 2014;86:1118–27.
[19] [Link] Last ac- [40] N. G. Paterakis, O. Erdinc, I. N. Pappi, A. G. Bakirtzis and J. P. S. Catalao,
cessed, 05 June, 2019. “Coordinated Operation of a Neighborhood of Smart Households Com-
[20] S. J. Darby, “Load management at home: Advantages and drawbacks of prising Electric Vehicles, Energy Storage and Distributed Generation," in
some active demand side options,” J. Power Energy, vol. 227, no. 1, pp. IEEE Transactions on Smart Grid, vol. 7, no. 6, pp. 2736-2747, Nov. 2016,
9-17, 2012. doi: 10.1109/TSG.2015.2512501.
[21] A. Gholian, H. Mohsenian-Rad and Y. Hua, “Optimal Industrial Load [41] N. Adilov, R. E. Schuler, W. D. Schulze, and D. E. Toomey, “The effect
Control in Smart Grid,” in IEEE Transactions on Smart Grid, vol. 7, no. 5, of customer participation in electricity markets: an experimental analysis
pp. 2305-2316, Sept. 2016. doi: 10.1109/TSG.2015.2468577. of alternative market structures,” In System Sciences, Proceedings of the
[22] N. G. Paterakis, O. Erdinc, A. G. Bakirtzis and J. P. S. Catalao, “Op- 37th Annual Hawaii IEEE International Conference on (pp. 10-pp).
timal Household Appliances Scheduling Under Day-Ahead Pricing and
[42] The Cadmus Group, Inc./ Energy Services, “Assessment of Long-Term,
Load-Shaping Demand Response Strategies," in IEEE Transactions on
System Wide Potential for Demand-Side and Other Supplemental Re-
Industrial Informatics, vol. 11, no. 6, pp. 1509-1519, Dec. 2015. doi:
sources,” 2013-2032 Volume I (March 2013).
10.1109/TII.2015.2438534.
[43] J. Ma, H. H. Chen, L. Song and Y. Li, “Residential Load Schedul-
[23] O. Erdinc, “Economic impacts of small-scale own generating and storage
ing in Smart Grid: A Cost Efficiency Perspective,” in IEEE Transac-
units, and electric vehicles under different demand response strategies for
tions on Smart Grid, vol. 7, no. 2, pp. 771-784, March 2016. doi:
smart households”. Applied Energy, 126, pp.142-150, 2014.
10.1109/TSG.2015.2419818.
[24] S. L. Arun; M. P. Selvan, “Intelligent Residential Energy Management
[44] S. Althaher, P. Mancarella and J. Mutale, “Automated Demand Response
System for Dynamic Demand Response in Smart Buildings,” in IEEE Sys-
From Home Energy Management System Under Dynamic Pricing and
tems Journal, [Link], no.99, pp.1-12, doi: 10.1109/JSYST.2017.2647759.
Power and Comfort Constraints,” in IEEE Transactions on Smart Grid,
[25] O. Erdinc, N. G. Paterakis, T. D. P. Mendes, A. G. Bakirtzis and J. P.
vol. 6, no. 4, pp. 1874-1883, July 2015. doi: 10.1109/TSG.2014.2388357.
S. Catalao, “Smart Household Operation Considering Bi-Directional EV
and ESS Utilization by Real-Time Pricing-Based DR,” in IEEE Trans- [45] [Link]
actions on Smart Grid, vol. 6, no. 3, pp. 1281-1291, May 2015. doi: _\\data/[Link]. Last accessed, 05 August, 2019.
10.1109/TSG.2014.2352650. [46] N. Yaagoubi and H. T. Mouftah, “User-Aware Game Theoretic Approach
[26] ENERGY STAR Qualified Products. Accessed on Apr. 11, 2016. [Online]. for Demand Management,” in IEEE Transactions on Smart Grid, vol. 6,
Available: [Link] Last accessed on 08 October, 2019. no. 2, pp. 716-725, March 2015. doi: 10.1109/TSG.2014.2363098.
[27] J. H. Yoon, R. Baldick, and A. Novoselac, “Dynamic demand response [47] A. Safdarian, M. Fotuhi-Firuzabad and M. Lehtonen, “Optimal Residential
controller based on real-time retail price for residential buildings,” IEEE Load Management in Smart Grids: A Decentralized Framework,” in IEEE
Trans. Smart Grid, vol. 5, no. 1, pp. 121-129, Jan. 2014. Transactions on Smart Grid, vol. 7, no. 4, pp. 1836-1845, July 2016. doi:
[28] S. Karnouskos, D. Ilic and P. G. D. Silva, “Using flexible energy infras- 10.1109/TSG.2015.2459753.
tructures for demand response in a Smart Grid city,” 2012 3rd IEEE PES [48] B. Hayes, I. Melatti, T. Mancini, M. Prodanovic and E. Tronci, “Resi-
Innovative Smart Grid Technologies Europe (ISGT Europe), Berlin, 2012, dential Demand Management Using Individualized Demand Aware Price
pp. 1-7. doi: 10.1109/ISGTEurope.2012.6465859. Policies," in IEEE Transactions on Smart Grid, vol. 8, no. 3, pp. 1284-
[29] A. Rosso, J. Ma, D. S. Kirschen and L. F. Ochoa, “Assessing the 1294, May 2017. doi: 10.1109/TSG.2016.2596790
contribution of demand side management to power system flexibil- [49] J. Yang, J. Zhao, F. Wen and Z. Dong, “A Model of Customizing Elec-
ity,” 2011 50th IEEE Conference on Decision and Control and Eu- tricity Retail Prices Based on Load Profile Clustering Analysis," in IEEE
ropean Control Conference, Orlando, FL, 2011, pp. 4361-4365. doi: Transactions on Smart Grid, vol. 10, no. 3, pp. 3374-3386, May 2019. doi:
10.1109/CDC.2011.6161236. 10.1109/TSG.2018.2825335
[30] S. Vandael, B. Claessens, M. Hommelberg, T. Holvoet and G. Deconinck, [50] Z. Almahmoud, J. Crandall, K. Elbassioni, T. T. Nguyen and M. Roozbe-
“A Scalable Three-Step Approach for Demand Side Management of Plug- hani, "Dynamic Pricing in Smart Grids Under Thresholding Policies," in
in Hybrid Vehicles,” in IEEE Transactions on Smart Grid, vol. 4, no. 2, pp. IEEE Transactions on Smart Grid, vol. 10, no. 3, pp. 3415-3429, May
720-728, June 2013. doi: 10.1109/TSG.2012.2213847. 2019. doi: 10.1109/TSG.2018.2825997
[31] L. A. C. Lopes, and M. Dalal-Bachi, “Economic Dispatch and Demand [51] T. C. Chiu, Y. Y. Shih, A. C. Pang and C. W. Pai, “Optimized Day-Ahead
Side Management via Frequency Control in PV-Diesel Hybrid Mini- Pricing With Renewable Energy Demand-Side Management for Smart
Grids,” In 6th European Conference on PV-Hybrid and Mini-Grids, Grids,” in IEEE Internet of Things Journal, vol. 4, no. 2, pp. 374-383,
Chambery, pp. 266-273. 2012. April 2017. doi: 10.1109/JIOT.2016.2556006.
[32] F. Ye, Y. Qian and R. Q. Hu, “A Real-Time Information Based Demand- [52] C. Wang, Y. Zhou, J. Wang, J. and P. Peng, “A novel traversal-and-pruning
Side Management System in Smart Grid,” in IEEE Transactions on Parallel algorithm for household load scheduling”. Applied Energy. 102, pp.1430-
and Distributed Systems, vol. 27, no. 2, pp. 329-339, Feb. 1 2016. doi: 1438, 2013.
10.1109/TPDS.2015.2403833. [53] D. Setlhaolo, X. Xia, and J. Zhang, “Optimal scheduling of household
[33] S. Althaher, P. Mancarella and J. Mutale, “Automated Demand Response appliances for demand response”. Electric Power Systems Research. 116,
From Home Energy Management System Under Dynamic Pricing and pp.24-28, 2014.

16 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]
 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/ACCESS.2020.2967798, IEEE Access

[54] L. Martirano et al., “Demand Side Management in Microgrids for Load [74] Z. M. Salameh, B. S. Borowy and A. R. A. Amin, “Photovoltaic module-
Control in Nearly Zero Energy Buildings,” in IEEE Transactions on site matching based on the capacity factors,” IEEE Trans. Energy Convers.,
Industry Applications, vol. 53, no. 3, pp. 1769-1779, May-June 2017, doi: vol. 10, no. 2, pp. 326-332, 1995.
10.1109/TIA.2017.2672918. [75] B. S. Borowy and Z. M. Salameh, “Optimum photovoltaic array size for a
[55] M. Yu, S. H. Hong, Y. Ding and X. Ye, “An Incentive-Based Demand hybrid wind/PV system,” IEEE Trans. Energy Convers., vol. 9, no. 3, pp.
Response (DR) Model Considering Composited DR Resources," in IEEE 482âĂŞ488, 1994.
Transactions on Industrial Electronics, vol. 66, no. 2, pp. 1488-1498, Feb. [76] Z. Zhao, W. C. Lee, Y. Shin, and K. Song, “An optimal power scheduling
2019. doi: 10.1109/TIE.2018.2826454 method for demand response in home energy management system”, IEEE
[56] M. B. Rasheed, N. Javaid, M. Awais, M. Akbar and Z. A. Khan, “A Novel Trans. Smart Grid, vol. 4, no. 3, pp. 1391-1400, 2013.
Pricing Mechanism for Demand Side Load Management in Smart Grid,” [77] T. Logenthiran, D. Srinivasan, T. Z. Shun, “Demand Side Management in
31st International Conference on Advanced Information Networking and Smart Grid using Heuristic Optimization”, IEEE Transactions on Smart
Applications Workshops (WAINA), Taipei, Taiwan, 2017, pp. 283-290. Grid, vol. 3, no. 3, pp. 1244-1252, 2012.
doi: 10.1109/WAINA.2017.119 [78] N. Kinhekar N. P. Padhya H. O. Gupta, “Multiobjective demand side
[57] Z. Zhou, F. Zhao, and J. Wang, “Agent-based electricity market simulation management solutions for utilities with peak demand decit,” Int J Electr
with demand response from commercial buildings,” IEEE Trans. Smart Power Energy Syst 2014;55(1):612-9.
Grid, vol. 2, no. 4, pp. 580–588, Dec. 2011. [79] M. C. Bozchalui, S. A. Hashmi, H. Hassen, C. A. Canizares and K.
[58] Y. Hu, Y. Li and L. Chen, “Multi-Objective Optimization of Time-of- Bhattacharya, “Optimal Operation of Residential Energy Hubs in Smart
Use Price for Tertiary Industry Based on Generalized Seasonal Multi- Grids,” in IEEE Transactions on Smart Grid, vol. 3, no. 4, pp. 1755-1766,
Model Structure," in IEEE Access, vol. 7, pp. 89234-89244, 2019. doi: Dec. 2012, doi: 10.1109/TSG.2012.2212032.
10.1109/ACCESS.2019.2926594 [80] P. C. Chu, and J. E. Beasley, “A genetic algorithm for the multidimensional
[59] H. Yang, L. Wang and Y. Ma, “Optimal Time of Use Electricity knapsack problem”. Journal of heuristics, 4(1), pp.63-86, 1988.
Pricing Model and Its Application to Electrical Distribution System," [81] K. McKenna and A. Keane, “Residential load modeling of price-based
in IEEE Access, vol. 7, pp. 123558-123568, 2019. doi: 10.1109/AC- demand response for network impact studies,” 2016 IEEE Power and
CESS.2019.2938415 Energy Society General Meeting (PESGM), Boston, MA, 2016, pp. 1-1.
[60] E. S. Parizy, H. R. Bahrami and S. Choi, “A Low Complexity and doi: 10.1109/PESGM.2016.7741224.
Secure Demand Response Technique for Peak Load Reduction,” in IEEE
Transactions on Smart Grid, doi: 10.1109/TSG.2018.2822729.
[61] A.-H. Mohsenian-Rad and A. Leon-Garcia, “Optimal residential load
control with price prediction in real-time electricity pricing environments,"
IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 120-133, Sep. 2010.
[62] J. S. Vardakas, N. Zorba, and C. V. Verikoukis, “A survey on demand
response programs in smart grids: Pricing methods and optimization
algorithms," IEEE Commun. Surveys Tut., vol. 17, no. 1, pp. 152âĂŞ178,
First Quarter 2015.
[63] L. Jia and L. Tong, “Dynamic Pricing and Distributed Energy Management
for Demand Response,” in IEEE Transactions on Smart Grid, vol. 7, no. 2,
pp. 1128-1136, March 2016. doi: 10.1109/TSG.2016.2515641.
[64] B. G. Kim, Y. Zhang, M. van der Schaar and J. W. Lee, “Dynamic Pricing
and Energy Consumption Scheduling With Reinforcement Learning,” in
IEEE Transactions on Smart Grid, vol. 7, no. 5, pp. 2187-2198, Sept. 2016.
doi: 10.1109/TSG.2015.2495145.
[65] S. Al-Rubaye, A. Al-Dulaimi, S. Mumtaz and J. Rodriguez, “Dynamic
Pricing Mechanism in Smart Grid Communications is Shaping Up,” in
IEEE Communications Letters. doi: 10.1109/LCOMM.2018.2822798.
[66] B. Hayes, I. Melatti, T. Mancini, M. Prodanovic and E. Tronci, “Resi-
dential Demand Management Using Individualized Demand Aware Price
Policies,” in IEEE Transactions on Smart Grid, vol. 8, no. 3, pp. 1284-
1294, May 2017, doi: 10.1109/TSG.2016.2596790.
[67] F. Salah, R. Henriquez, G. Wenzel, D. Olivares, M. Negrete-Pincetic
and C. Weinhardt, “Portfolio Design of a Demand Response Aggregator
with Satisficing Consumers,” in IEEE Transactions on Smart Grid. doi:
10.1109/TSG.2018.2799822
[68] J. W. Black, “Integrating demand into the U.S. electric power system:
Technical, economic, and regulatory frameworks for responsive load,”
Ph.D. dissertation, Eng. Syst. Div., Mass. Inst. Technol., Cambridge, MA,
USA, Jun. 2005.
[69] J. Siroky, F. Oldewurtel, J. Cigler, and S. Privara, “Experimental analysis
of model predictive control for an energy efficient building heating sys-
tem,” Appl. Energy, vol. 88, no. 9, pp. 3079–3087, Sep. 2011.
[70] Y. Wang, W. Saad, N. B. Mandayam and H. V. Poor, “Load Shifting in the
Smart Grid: To Participate or Not?,” in IEEE Transactions on Smart Grid,
vol. 7, no. 6, pp. 2604-2614, Nov. 2016. doi: 10.1109/TSG.2015.2483522.
[71] N. G. Paterakis, A. TascÄśkaraoglu, O. Erdinc, A. G. Bakirtzis and J.
P. S. Catalao, “Assessment of Demand-Response-Driven Load Pattern
Elasticity Using a Combined Approach for Smart Households,” in IEEE
Transactions on Industrial Informatics, vol. 12, no. 4, pp. 1529-1539, Aug.
2016. doi: 10.1109/TII.2016.2585122.
[72] K. McKenna and A. Keane, “Residential load modeling of price-based
demand response for network impact studies,” 2016 IEEE Power and
Energy Society General Meeting (PESGM), Boston, MA, 2016, pp. 1-1.
doi: 10.1109/PESGM.2016.7741224.
[73] [Link] Last accessed, 05 July,
2019.

VOLUME 4, 2016 17

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see [Link]