You are on page 1of 12

IEEE SYSTEMS JOURNAL, VOL. 12, NO.

2, JUNE 2018 1329

Intelligent Residential Energy Management System


for Dynamic Demand Response in Smart Buildings
S. L. Arun and M. P. Selvan , Member, IEEE

Abstract—The advancements in renewable energy technologies the consumers’ loads [7], [8]. On the other hand, in the RTP
direct the power sector to focus on power generation from renew- technique, the utilities have a time-dependent pricing scheme to
able energy resources (RER) as an alternative solution for meeting make the system load curve almost flat by enforcing higher price
the future demand. Nowadays residential buildings are becoming
smarter with wide use of smart appliances, integration of infor- during the high-demand period [9], [10]. Due to increasing de-
mation and communication technology, and in-house power gen- mand and depletion of conventional energy resources, the price
eration using RER. In this paper, an intelligent residential energy of energy is increasing continuously. Hence, the utilities are fo-
management system (IREMS) for prosumers of smart residential cusing on renewable energy resources (RER) as an alternative
buildings is proposed, and its benefits are demonstrated through a solution to meet the demand. However, the unpredictability in
case study. The primary objective of IREMS is reduction in elec- the production of power from RER, caused by the intermittent
tricity bills while maintaining the power demand under the maxi-
mum demand limit subjected to the various constraints governing nature of RER, will lead the utility to face additional operational
the operation of household loads and RER. The IREMS achieves challenges. In addition to the large-scale renewable power gen-
the objective by scheduling the schedulable loads during low pric- eration systems, residential consumers are also encouraged by
ing intervals while considering the operational dynamics of non- the government to install small-scale roof-top solar PV and/or
schedulable loads and availability of RER. IREMS also manages small wind turbine based power generation systems to meet out
the battery energy storage in such a way so as to reduce the power their own demand partially or completely. Such small-scale re-
dissipated through the dump load when excess power is available
from RER due to the utility-defined power export limit to grid. newable generation helps in reduction of the electricity bill if the
Further, an optimal resources sizing algorithm is used to choose residential loads are timely operated when the power generation
the size of RER and battery storage for the effective utilization of from renewable sources is more [11], [12]. It is expected that the
available renewable energy. penetration of new generation sources will affect the dynamics
Index Terms—Demand response (DR), demand-side manage- of energy pricing. Further, in recent days, the residential build-
ment (DSM), distributed generation, renewable energy, smart grid. ings are becoming smarter with wide use of smart appliances
and integration of information and communication technology.
I. INTRODUCTION In order to achieve maximum utilization of in-house RER and
EMAND-SIDE management (DSM) is a very attractive maximum benefit of RTP, the operation of smart appliances
D approach in smart grid environment adopted by the util-
ity to minimize the energy consumption at consumer premises
needs to be appropriately time scheduled. Consequently, the re-
quirement of a load scheduler for every residential building is
[1]. DSM scheme includes energy conservation and energy ef- inevitable in the smart grid environment.
ficiency programs, fuel substitution programs and demand re- Hubert and Grijalva [13] proposed a mixed-integer linear pro-
sponse (DR) programs [2]. In the DR programs, consumers are gramming based scheduling algorithm while considering the
encouraged to have direct interaction with the grid by actively objective as reduction in electricity bill for residential electric-
participating in the electricity market. Consumers are expected ity consumers. Chen et al. [14] developed an energy-efficient
to alter the pattern of their energy consumption as a response to scheduling algorithm while considering the uncertainties in the
the variations in electricity price in order to reduce their electric- operating time of household appliances, intermittent renewable
ity bills. It was demonstrated in the literature that a significant generation, and energy storage systems.
amount of price reduction can be obtained by the consumers Adika and Wang [15] presented a prosumer-based DSM. The
through DR programs [3]–[6]. Direct load control and real time proposed scheme concentrates on reduction in electricity bill
pricing (RTP) are two common techniques in DSM proposed and more profit by exporting power to grid in dynamic pricing
by different electricity suppliers. In direct load control, ser- environment. Appliances time of using probability technique
vice providers have direct control on switching ON and OFF of is employed to synthesize the household electricity demand
pattern. Mesari and Krajcar [16] discussed that integration of
renewable energy sources and electric vehicles integrated with
Manuscript received June 28, 2016; revised September 25, 2016 and Novem- proper home DSM can contribute to microgrid stability and
ber 27, 2016; accepted December 27, 2016. Date of publication January 23,
2017; date of current version May 2, 2018. This work was supported by the decrease grid dependence.
Department of Electronics and Information Technology, Government of India. Ratnam et al. [17] proposed a liner programming based
The authors are with the Hybrid Electrical Systems Laboratory, National scheduling algorithm with the objective as maximizing the op-
Institute of Technology, Tiruchirappalli 620015, India (e-mail: slarun2010@
gmail.com; selvanmp@nitt.edu).
erational savings. Wang et al. [18] developed a robust optimiza-
Digital Object Identifier 10.1109/JSYST.2017.2647759 tion approach for household load scheduling considering the

1937-9234 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
1330 IEEE SYSTEMS JOURNAL, VOL. 12, NO. 2, JUNE 2018

uncertainty in power output of household PV system. A game a) operational dynamics of nonschedulable loads;
theoretic consumption scheduling framework based on mixed b) desire and comfort of the user;
integer programming was proposed by Zhu et al. [19]. Also, c) intermittency of the renewable power generation;
Fadlullah et al. [20] proposed a game theoretic energy sched- d) variation in electricity price of utility;
ule method that has an objective to reduce the peak to average e) operational limits proposed by the utility such as
power ratio by optimizing the users’ energy schedules. MDL and PEL;
In the present scenario, the utilities are facing lot of difficulties 3) Cost benefit analysis for showcasing the savings and min-
because of the high penetration of small-scale grid-connected imum cash payback period (CPP) anticipated from the
renewable resources [21], [22]. Hence, the utilities are inter- proposed algorithm.
ested in imposing a time-varying power export limit (PEL) to The remainder of this paper is organized as follows. In Sec-
prosumers. The prosumers can be benefited by injecting power tion II, the architecture of proposed IREMS is described. In
to grid without exceeding PEL. Excess generated power shall Section III, models of different components available in home
be either stored in the battery bank for future use or dissipated are presented. Genetic algorithm (GA) based optimal sizing of
through dump load. Hence, improper sizing of resources during RER and battery bank is deliberated in Section IV. The devel-
installation will lead more power wastage through dump load. oped scheduling algorithm is detailed in Section V. The data
In most of the literature dealing with prosumer based DSM and results of the case study are presented and discussed in Sec-
schemes discussed above, the optimal sizing of RER and en- tion VI. Further, the CPP analysis is also executed and described.
ergy storage devices has not been performed. Further, most of The conclusions are drawn in Section VII.
the earlier works are concentrating only on schedulable loads
and the operating time of schedulable loads for all days was
assumed to be either fixed or based on the probability of use in II. ARCHITECTURE OF THE PROPOSED SYSTEM
the history. The dynamics in operation of nonschedulable loads
The power consumption pattern in residential buildings is
on reduction in electricity bill while scheduling the schedulable
merely based on the availability and comfort requirement of the
loads was not considered.
consumer and is not certainly flat always. Considering the differ-
It is expected that considering the power demanded by the
ent kinds of power consumption pattern of residential loads, they
nonschedulable loads and available in-house renewable power
are categorized into three types, namely, noninterruptible and
generation while scheduling the schedulable loads shall result in
nonschedulable loads (NINSLs), interruptible and nonschedu-
consumption of total power below the Maximum Demand limit
lable loads (INSLs), and schedulable loads (SLs).
(MDL) imposed by utility during most of the time of a day and
The operation of NINSLs is completely dependent on the
higher reduction in electricity bill. In addition, effective man-
interest of the consumers. These loads should respond im-
agement of the battery operation can reduce the power wastage.
mediately when a user initializes them. Essential loads such
Further, updating the changes in operation of schedulable loads
as light and fan; entertainment loads such as television (TV),
in real time shall enhance the consumer’s comfort.
DVD player, speaker, personal computer (PC), printer, and mo-
In this paper, development of an intelligent residential en-
bile/laptop charger; and kitchen loads such as cooking stove,
ergy management system (IREMS) for prosumers of smart
mixer, and toaster are classified under this category. Since the
residential building is attempted. In this paper, it is assumed
operating time of NINSLs is completely dependent on the de-
that the smart residential building is equipped with in-house
sire and comfort of the user, IREMS does not have any control
PV and small wind based power generation systems, battery
on these loads. However, the total power consumption by these
storage, dump loads, and grid-interactive smart meter, in addi-
loads is taken into account for effective scheduling of other
tion to smart appliances enabled with communication feature.
types of loads. Further, the IREMS is envisioned to give an alert
The proposed IREMS attempts to reduce the electricity bill and
message when the total power consumption of NINSLs exceeds
the dependence of consumer on the utility by scheduling the
the consumer’s predefined limit.
schedulable loads while considering the operating dynamics of
The temperature-controlled loads are categorized as INSLs.
nonschedulable loads and available renewable power genera-
Electric water heater, air conditioner, and space heater are a
tion. Further, the IREMS ensures very minimum power wastage
few examples of INSL. The environmental aspects and comfort
through the dump load while satisfying the PEL constraint im-
level of the consumer are the primary reasons for variation in
posed by the utility. The proposed architecture has a facility
the power consumption pattern of INSLs. During operation,
to frequently update the real-time changes in utility parameters
the INSLs maintain the temperature around the user-defined set
such as variations in electricity price, MDL and PEL; changes
point but within the tolerance limit set by the manufacturer.
in operating time of loads opted by the consumer; and intermit-
The INSLs start to consume the rated power if the temperature
tence of renewable power generation.
exceeds this limit, else the INSLs continue their operation in the
The main contributions of this paper are listed as follows:
standby mode.
1) Optimal sizing of RER and battery units to be installed
The loads whose operating time is usually prefixed and flex-
in the residential building to achieve reduction in total
ible are categorized under SLs. Based on the operating nature,
investment cost, electricity bill, and utility dependence;
these loads are further classified into two types. The first type
2) Development of scheduling algorithm for schedulable
is noninterruptible and schedulable loads (NISLs), which need
loads and battery units to minimize the electricity bill
to be operated continuously without any pause. Cloth washer,
and maximize the utilization of renewable energy while
cloth dryer, and grinder are of this type. Whereas the other type,
considering the following:
interruptible and schedulable loads (ISLs) can operate either
ARUN AND SELVAN: IREMS FOR DYNAMIC DR IN SMART BUILDINGS 1331

Smart power converter module stores the battery specifica-


tions such as Ah rating, dc bus voltage and the boundary limits
of state of charge (SOC), charging and discharging currents,
which can be updated by the consumer if needed. During op-
eration this module obtains the SOC of the battery and gives it
to the IREMS processing unit. IREMS directs the PCU through
smart power converter module to control the mode of operation
and charging/discharging current of the battery.
Smart power converter module transfers the information
about the power generation from RER to the IREMS processing
unit. Using this real-time information and the past history
data, IREMS predicts the renewable power generation of
upcoming intervals. Considering the present and predicted
renewable power, IREMS schedules the SLs and battery. When
Fig. 1. Architecture of the proposed IREMS.
the difference between the renewable power generation and
residential demand exceeds the PEL, the IREMS processing
unit instructs the PCU either to dissipate power in the dump
continuously or discontinuously in the user-defined duration.
load or to minimize the extraction of power from RER. The user
Plug-in Hybrid Electrical Vehicle (PHEV), well pump, and dish
interface present in the IREMS reads the maximum limit of
washer are of this type.
power consumption by NINSLs and extended tolerance limits
As discussed in Section I, residential consumers are showing
for INSLs decided by the user. Moreover, the user interface can
interest in the installation of small-scale renewable power gen-
be used to display the warning messages and other information
eration systems to reduce the utility dependence and electricity
such as electricity price, power consumption by different kinds
bill. In this paper, a residential building equipped with solar
of loads, etc.
PV and small wind power generation is considered. The inter-
mittent nature of power generation by RER is managed by the
battery backup. IREMS is proposed to manage the operation of III. MODELING OF HOUSEHOLD COMPONENTS
residential loads and battery in such a way that the power gen- The smart residential building considered for analysis in this
erated from RER is effectively utilized. The battery operation paper consists of different kinds of loads, battery backup, and
(charging, discharging, and floating) is altered by IREMS as per renewable resources. The modeling of renewable power gener-
the grid conditions (electricity price and MDL) and renewable ation systems such as solar PV and wind turbine are done as
power availability. The proposed architecture of IREMS is de- given in [23] and [24], respectively. All the types of residential
picted in Fig. 1. IREMS consists of smart NINSL module, smart loads (NINSL, INSL, and SL) and battery are modeled based on
INSL module, smart SL module, smart power converter module their steady-state performance. The initial switching transients
to have data and control signal transfer between IREMS pro- are not considered.
cessing unit and NINSLs, INSLs, SLs, and power conditioning
unit (PCU) of RER, respectively. Further, it has a smart meter
interface to interact with the grid and a user interface. A. Model of NINSLs
The IREMS receives the timely updates of variation in elec- The operating pattern of the NINSLs are highly related with
tricity price, MDL, and PEL from the utility through the smart the availability and requirement of the consumer. The need for
meter interface. The smart meter has two-way communication switching ON NINSLs in a given day is time dependent. Hence,
with the utility. IREMS considers the total power consumption IREMS aggregates all the NINSLs and considers as a single
of all the NINSLs aggregated to be a single load, whose power load whose power demand is varying continuously.
demand varies continuously. Smart NINSL module aggregates
all NINSLs and deliver an alert message when the NINSLs B. Model of INSLs
demand exceeds the consumer set limit. Smart INSL module
collects INSL parameters such as set point temperature, toler- The power consumption pattern of INSLs is decided by the
ance limit, rated power, standby power, and user status of the variation in operating temperature. INSLs compare the present
load (switch ON/OFF) from all the INSLs. Further, it delivers operating temperature with the reference temperature set by the
the operational instruction (RUN/STANDBY) generated by the consumer. INSLs maintain the STANDBY mode of operation
IREMS processing unit to INSLs. when the difference between these two temperatures is within
Smart SL module obtains the information such as load initial- the tolerance limit. Whenever the difference exceeds the toler-
ization interval (the interval in which the SL is initialized), load ance limit, INSLs toggle to RUN mode. Then INSLs remain
dead interval (interval before which a particular task should be in the RUN mode until the operating temperature reaches the
completed), computation intervals (number of intervals required reference value and this process is repeated.
to complete the task), and pre-emptive status (nature of the task Let A denote the set of INSLs and I denote the set of moni-
whether continuous or discontinuous operation) from all SLs toring intervals of nonschedulable loads (NINSLs and INSLs).
and delivers the control signals to SLs as per the instruction The operating vector representing the ON/OFF status of an INSL
Δ Δ
from IREMS processing unit. a (a ∈ A = [1, 2, ..., A]) in each interval i (i ∈ I = [1, 2, ..., I])
1332 IEEE SYSTEMS JOURNAL, VOL. 12, NO. 2, JUNE 2018

is defined as Further, the pre-emptive status of the schedulable load l is


represented by ρl and its value is assigned as follows:
Xa = [x1a , x2a , . . . xia , . . . , xIa ] ∀a ∈ A (1)

0, for pre-emptive loads (ISL)


where I is the total
 number
 of nonschedulable load intervals in a ρl = . (7)
day I = 24 · D60N S . Here, DNS is the duration of nonschedu- 1, for non-pre-emptive loads (NISL)
lable load interval in minute set by the consumer. Given the set
point temperature and allowable tolerance limit of an INSL a as D. Model of Battery
Tseta and ΔTtola , respectively, and the actual temperature at
the end of the interval i − 1 as Tacti−1 IREMS decides the mode of operation of battery by con-
a , the desired operating
status of a during the interval i can be described as follows. sidering the available renewable power generation, consumer
If a is a cooling load demand and MDL. IREMS assumes that the battery is an ad-
⎧ ditional SL during charging operation whereas in discharging

⎪ −1, if a is not yet initialized operation, it is considered to be an additional generation. Let


⎨ 0, if Tacti−1 < Tseta K denote the set of battery scheduling intervals. The operating
a
xia = vector representing the mode of operation of battery (charg-

⎪ 1, i−1
if Tacta > Tseta + ΔTtola ing (Bck ), floating (Bfk ), and discharging (Bdk )) in interval k,


⎩ i−1
xa , if Tseta ≤ Tacti−1 a ≤ Tseta + ΔTtola . Δ
(k ∈ K = [1, 2, . . . , K]), is defined as
(2)
If a is a heating load B = [B 1 , B 2 , . . . B k , . . . , B K ] (8)


⎪ −1, if a is not yet initialized k
B = (Bck , Bfk , Bdk ) (9)


⎨ 0, if Tacti−1 > Tseta
a
xia = where K is the total
 number
 of battery scheduling intervals in

⎪ 1, if Tacta < Tseta − ΔTtola
i−1
a day K = 24 · D60B . Here, DB is the duration of battery


⎩ i−1 scheduling interval in minute set by consumer in consultation
xa , if Tseta − ΔTtola ≤ Tacti−1 a ≤ Tseta .
(3) with manufacturer for better utilization of battery. Each element
of the operating vector in interval k is expressed as
C. Model of SLs ⎧

⎨ (1, 0, 0), if charging
Let SL denote the set of SLs. The scheduling vector Sl for k k k k
B = (Bc , Bf , Bd ) = (0, 1, 0), if floating (10)
each SL l is defined as ⎪

(0, 0, 1), if discharging.
Sl = [s1l , s2l , . . . stl , . . . , sM
l ] ∀l ∈ SL (4)
During starting of each battery interval (k), available SOC
where M is the totalnumber of schedulable load intervals in a of the battery is obtained [25]. The battery output power for
day M = 24 · D60S . Here, DS is the duration of the schedu- interval k, (PBk ) is obtained from the charging power (PBC
k
) and
lable load interval in minute set by the consumer. Each element k
discharging power (PBD ) of the battery during the interval and
of scheduling vector, stl describes the status of SL l at the tth is as follows:
schedulable load interval and is given as

PBk = (1 − Bfk ) (Bck · PBC
k
− Bdk · PBD
k
). (11)
0, if load l is OFF
t
sl = ∀ l ∈ SL; t = 1, 2, . . . , M . The renewable power generation is highly intermittent and
1, if load l is ON
(5) site specific. The generalized model of solar PV [23] and wind
The IREMS processing unit should receive information about turbine [24] are considered in this paper. The total renewable
the initialization interval (βl ) and dead time interval (ηl ) opted power (PRj ) is the sum of power generated from solar PV (PSj )
j Δ
by the user for each schedulable load either through the user and wind (PW ), which is expressed in (12). Here, j(j ∈ J =
interface or from the settings available in the load (if it is fully [1, 2, . . . , J]) is the RER interval
 J is total number of RER
and
automated smart load). Scheduling algorithm can schedule any intervals in a day, J = 24 · D R60E R . DRER is the duration of
SL only between these two intervals of that load. Smart loads in updation of power generation from RER in minutes.
recent days are equipped with artificial intelligence to determine
the actual number of intervals required to complete the task (ωl ) PRj = PSj + PW
j
. (12)
during initialization itself based on the initial conditions such
as existing water level in water tank (if the load being a smart Better utilization and higher cost benefit can be obtained by
well pump) and weight of clothes (if the load being a smart the consumer if the power generated from RER is sufficient
cloth washer). The primary condition that should be considered to meet out the expected power demand in all the intervals.
during selection of initialization and dead time intervals by the However, excess power generation can be better managed by
user is given in (6) storing in battery banks for future use. Hence, the renewable
resources and battery should be optimally sized for reducing the
ωl ≤ ηl − βl ∀l ∈ SL. (6) investment cost.
ARUN AND SELVAN: IREMS FOR DYNAMIC DR IN SMART BUILDINGS 1333

IV. OPTIMAL SIZING OF RESOURCES space at the installation site


As discussed in the previous section, the components of RER 0 ≤ NPV ≤ NPVm ax
such as solar PV panels and wind turbines and the batteries
are required to be optimally sized to achieve reduction in total 0 ≤ NW T ≤ NW Tm ax
investment cost, electricity bill, and utility dependence [26]. 0 ≤ NBB ≤ NBBm ax (20)
Hence, the objective of optimal sizing is to minimize the cost
of energy (CoE) and the objective function is formulated as where NPV , NW T , and NBB are the actual number of solar
follows: PV panels, wind turbines, and batteries, respectively, whereas,
NPVm ax , NW Tm ax , and NBBm ax are the consumer defined max-
CC + CR + CO imum number of solar PV panels, wind turbines, and batteries,
min CoE = min (13)
ER respectively.
where CC is the sum of annualized capital cost of all com-
ponents, CR is the sum of annualized replacement cost of all C. Demand Constraint
components, CO is the sum of annualized operation and main- The demand supplied by the RER over a year should be
tenance cost of all components, and ER is the total energy greater than the desired fraction of the total annual energy de-
delivered by RER over a year. The total annualized capital cost mand of the consumer
is obtained from
ER
≥ϑ (21)
CC = (CCS + CCW + CCB ) · Φ(A, T ) (14) EAD

A(1 + A)T where EAD is the total annual energy demand of the consumer
Φ(A, T ) = (15) and ϑ is defined as the RER energy factor that varies from 0 to
(1 + A)T − 1
1.
where CCS , CCW , and CCB are the total capital cost of solar
PV, wind turbine, and battery backup, respectively, Φ is the D. Power Export Constraint
capital recovery factor, A is the annual discount rate, and T is
the project life time. The total annualized replacement cost CR In this paper, it is considered that the utility limits the power
is calculated by the following equation: delivered by the prosumers to the grid. Consequently, the power
exported by a prosumer over an interval should be within the
CR = (CRS + CRW + CRB ) · Φ(A, T ) (16) utility defined limit called PEL. Therefore, the power generated
where CRS , CRW , and CRB are the replacement cost of solar PV by RER beyond the demand and export limit needs to be dissi-
panel, wind turbine, and battery backup, respectively. The total pated through dump loads to have power balance in the system.
annualized operation and maintenance cost CO is expressed in However, power dissipation through dump load should be made
the following equation: as minimum as possible for better energy saving. In order to take
into account this issue the power export constraint is defined as
CO = COS + COW + COB (17) follows:
  
where COS , COW , and COB are the total annual cost spend (D R E R )
year ΔPE 60
toward the operation and maintenance of solar PV panels, wind ≤ EDm ax (22)
turbines, and batteries, respectively. The total energy supplied ER
by RER over a year can be expressed as

PE − PLim , if PPLRi m > 1


 ΔPE =
DRER 0, if PPLRi m ≤ 1
ER = (PS + PW ) . (18)
year
60
(23)
The objective function described in (13) is subjected to the
where PLim and PE are utility defined PEL and power generated
following constraints.
by RER beyond the consumer demand for a particular interval,
respectively, and EDm ax is energy dissipation factor chosen by
A. Initial Cost Constraint
consumer and it should be considered as minimum as possible
The sum of initial investment cost of RER and battery storage to ensure effective utilization of RER. The objective function
should be within the maximum allowable initial investment cost (13) subjected to the constraints (19)–(22) is solved using GA
that is decided by the consumer for optimal sizing of RER and battery. The steps involved in this
optimization process are briefed in Algorithm 1.
IC ≤ ICm ax (19)
where IC and ICm ax are the actual and maximum initial invest- V. SCHEDULING OF HOUSEHOLD LOADS
ment cost, respectively.
The main objective of proposed IREMS is to assist the con-
sumers in reducing the electricity bill by properly time schedul-
B. Component Constraint
ing the loads with due considerations to the variations in elec-
The size of each component should not exceed the consumer tricity price, MDL and PEL dictated by the utility as well as the
defined maximum limit, which is decided based on the available power availability from the renewable and battery resources. The
1334 IEEE SYSTEMS JOURNAL, VOL. 12, NO. 2, JUNE 2018

Algorithm 1: Optimal sizing of RER and battery using GA.


Step 01: Start the project.
Step 02: Collect the annual wind speed, solar irradiation,
ambient temperature, and load demand data
averaged over a desired interval.
Step 03: Consider random population as initial
population.
Step 04: Start iterative process from the first interval.
Step 05: Compute the available solar PV and wind power
and compare with demand in the present
interval.
Step 06: If the RER generation is more than the demand
then charge the battery as per the SOC and
charging current limits.
Step 07: If the RER generation is less than the demand
then discharge the battery as per the SOC and
discharging current limits.
Step 08: Calculate amount of RER power that is not
utilized and to be dissipated through the dump
load.
Step 09: Repeat steps 5–8 in all the intervals of a year.
Step 10: Repeat steps 4–9 for all population.
Step 11: Formulate the population of the next generation
from the present generation using GA operators
such as crossover and mutation.
Step 12: Repeat steps 4–11 till the generation count
reaches predefined maximum value.
Step 13: Export the optimal size of RER and battery
satisfying all the annual constraints with
minimum CoE.
Fig. 2. Flow chart depicting the functioning of IREMS.

full time horizon of IREMS (24 h) is divided into pricing interval


duration (DP ) decided by the utility, schedulable load interval the IREMS provides the optimal values of decision variables:
duration (DS ) and nonschedulable load interval duration (DNS ) xia , stl , B t , PBC
t t
, and PBD , for all the scheduling intervals so
decided by the consumer, battery scheduling duration (DB ) sug- as to reduce the electricity bill of the consumer. The basic steps
gested by the manufacturer, and RER interval duration (DRER ) involved in the functioning of the proposed IREMS is depicted
decided by the consumer as per the required accuracy. The effec- in Fig. 2 as a flowchart.
tive operation of IREMS is possible only when the duration of
intervals are maintained as DRER ≤ DNS ≤ DB ≤ DS ≤ DP . A. Scheduling of INSLs
However, these durations need not be an integer multiples. Dur- Since the functioning of NINSLs depends on consumers com-
ing the pricing interval electricity price remains constant. Using fort, the IREMS does not control these loads. However, IREMS
the price information of the present interval and historical data, provides additional control on INSLs based on the thermal dy-
the price of electricity in future intervals can be predicted. The namics of the building and present status of INSLs. In the pro-
power demand of nonschedulable loads is considered to be con- posed IREMS, the tolerance limit of INSLs is extended by a
stant during the interval i. The IREMS reserves the expected consumer defined value. If the temperature varies in between
power demand of nonschedulable loads for all the scheduling the tolerance limit and extended tolerance limit (ΔTela ), then
intervals. For this reservation, the power demand of NINSLs the IREMS makes a decision to either turn ON the INSL or put
during the last nonschedulable interval (i − 1) is assumed to be in STANDBY mode according to MDL. When the number of
same for the next immediate scheduling interval t and the user- working INSLs increases, IREMS controls them by consider-
defined value is considered for future scheduling intervals of a ing the priority of each load. The priority of each INSL will be
day in order to schedule SLs without exceeding the MDL. Also, defined as follows: If a is the cooling load
IREMS predicts the renewable power generation for all schedul- ⎧

ing intervals from the past history of data. During the scheduling ⎪
⎨ if INSL a is not yet initialized
interval t, the IREMS computes the expected demand of INSLs 0,
Pricia = else if Tacti−1 < Tseta
using thermal dynamics of the building. The battery parameters ⎪

a

such as mode of operation and amount of power exchange dur- 1 − ΔPrica , otherwise
i

ing interval t are optimally decided by the IREMS. Succinctly, (24)


ARUN AND SELVAN: IREMS FOR DYNAMIC DR IN SMART BUILDINGS 1335

Tseta + ΔTtola + ΔTela − Tactia tth scheduling interval if t is not present in interval [βl , ηl ]. This
ΔPricia = . (25) constraint can be expressed as follows:
ΔTtola + ΔTela
If a is the heating load stl = 0, t < βl ∀l ∈ SL

stl = 0, t > ηl ∀l ∈ SL. (31)

⎨ if INSL a is not yet initialized
0,
Prihia = else if Tacti−1 > Tseta 2) Computational Interval Constraint: The SLs should be


a
scheduled only for the predetermined number of computational
1 − ΔPrihia , otherwise
intervals. The total number of intervals that SL l be in ON sta-
(26)
tus computed at tth interval should be equal to the remaining
Tactia − (Tseta − ΔTtola − ΔTela ) number of intervals required to complete the task (λtl ) by load l
ΔPrihia = . (27) from tth interval
ΔTtola + ΔTela
The IREMS will immediately start the INSLs that are having ηl
sgl = λgl ∀l ∈ SL. (32)
priority greater than or equal to one. To avoid excess payment,
g =t
the IREMS will consider the MDL and generation from RER
for operating the remaining INSLs in the order of high priority 3) Pre-emptive Constraint: The NISLs are non-pre-
to low. When the demand exceeds the MDL, the operation of emptable. These loads should operate continuously till the com-
low priority INSLs will be postponed to upcoming intervals. pletion of the task. When the NISL l is started, the ωl number of
computation intervals should be reserved continuously within
B. Scheduling of SLs [βl , ηl ]. However, the ISLs are pre-emptable and the ωl number
of computation intervals for any ISL l can be reserved either
When the utility is proposing RTP, a consumer may prefer to continuously or in a discrete manner within [βl , ηl ]
shift the SLs from peak intervals (having high-energy price and
υ −1
 
β l +ω l +θ −1
less MDL) to nonpeak or mid peak intervals (having moderate
energy price and enough MDL) in order to reduce the price of sμl ρl = ρl (33)
energy consumption and to avoid excess payment. Hence, the θ =0 μ=β l +θ

objective of IREMS is not only effective utilization of RER but υ = ηl − βl − ωl + 2. (34)


also maintaining the power demand below MDL and to reduce
the electricity bill as well as decrease the utility dependence. 4) Power Demand Constraint: Power demanded by the con-
Hence, the IREMS should schedule the SLs during the inter- sumer during every interval should be within the MDL (Pmt ax )
vals at which power generation from RER is more and utility set by the utility in order to avoid excess payment. The IREMS
electricity price is less. The mathematical formulation of this should schedule the SLs while considering power demanded by
constrained optimization problem is described below. NINSLs and INSLs. Further, The IREMS considers the battery
Let us define a dynamic set H, whose number of elements as additional SL during charging mode, whereas in the discharg-
vary dynamically with respect to the present operating interval ing operation, it is considered to be an additional generation.
(t) and the number of initialized SLs. Hence, the demand constraint can be written as
⎧ t

t
PNINSLs + PINSLs t
+ PSLs + PBt − PRt ≤ Pmt ax
H = [t, t + 1, . . . , tm d ] (28) ⎪
⎪  


⎨ R t1+ 1 PSLs
t+1
+ PBt+1 − PRt+1 ≤ Pmt+1
where tm d is maximum of the dead time intervals of initialized ax
(35)
SLs. Now, the objective function can be written as given in (29) ⎪

..
  ⎪
⎪ .
 ⎪
⎩ 1  tm d 
min EDG · C
h h
∀h ∈ H (29) Rt m d
PSLs + PBt m d − PRt m d ≤ Pmt maxd
h where Rt is the reservation factor for nonschedulable loads

DS (NINSLs and INSLs) at tth interval. The expected power de-
h
EDG h
= (PNINSLs h
+ PINSLs h
+ PSLs + PBh − PRh ) · mand of NINSLs for the present schedulable load interval (t)
60
(30) is considered to be that of last nonschedulable load interval
(i − 1). The expected power demand of INSLs for the present
h
where h is the element of H and EDG is the net energy utilized schedulable load interval is computed by IREMS using thermal
h h
during the interval h. PNINSLs and PINSLs are the expected dynamics of the building. Either nonconsideration of demand of
power demand by NINSLs and INSLs during interval h, re- nonschedulable loads or considering same power demand in all
h
spectively. PSLs is reserved power for SLs during interval h by future schedulable load intervals will lead to ineffective schedul-
IREMS. PB and PRh are the net battery power and expected RER
h
ing of SLs. Hence, the reservation factor is used to reserve a part
power during interval h, respectively. The electricity price (C h ) of MDL for meeting out the varying nonschedulable demand in
during the interval h is obtained from the utility. The objective the upcoming intervals. This factor varies between 0 and 1; and
function defined in (29) is optimized subject to the following it is either based on the history of power consumption data of a
constraints: day or the consumer requirement and desire.
1) Load Idle Constraint: The element of the scheduling vec- The mode of operation and the power exchange of battery
tor (stl ) representing the status of the SL l should be zero in the is to be maintained same for that schedulable load interval if
1336 IEEE SYSTEMS JOURNAL, VOL. 12, NO. 2, JUNE 2018

TABLE I TABLE II
NONINTERRUPTIBLE AND NONSCHEDULABLE LOADS INTERRUPTIBLE AND NONSCHEDULABLE LOADS

S. No. Load Power rating (kW) S. No. Load Power rating (kW)

1 Fan 0.10 1 Air conditioner-1 1.5


2 Fluorescent lamp 0.06 2 Water heater 2.0
3 Compact fluorescent lamp 0.04 3 Refrigerator 0.5
4 TV 0.25 4 Air conditioner-2 1.0
5 Mobile/laptop charging 0.05

scheduling is represented in the following equation:


the duration of schedulable load interval and battery scheduling ⎧
k −1
interval are equal. When the battery scheduling interval is less ⎪
⎪ PE > PEk −1


max
than the schedulable load interval, IREMS may reschedule the ⎪

(1, 0, 0), if k

⎪ SOC < SOC m ax
battery operation with due consideration to the dynamics of ⎪



nonschedulable loads and the power generation from RER as ⎪


⎪ PEk −1 > PEk −1
indicated in (42). ⎪
⎪ max


5) Battery Operation Constraint: The mode of operation of ⎪
⎨ SOC k
≥ SOC m ax
battery is always unique, which is expressed as (Bck , Bfk , Bdk ) = (0, 1, 0), if or (42)

Bct + Bft + Bdt = 1. ⎪


⎪ k −1
> Pmk −1
(36) ⎪

Ptot ax



⎪ SOC ≤ SOCm in
k
6) Battery Boundary Constraint: The battery parameters ⎪



such as SOC, charging power, and discharging power should ⎪


⎪ k −1
Ptot > Pmk −1
remain between the minimum and maximum limits defined by ⎪
⎪ ax

⎩ (0, 0, 1), if
the manufacturer. These constraints are formulated as follows: k
SOC > SOC m in

SOCm in ≤ SOC ≤ SOCm ax


t
(37) where PEk −1 is the power generated by renewable resources
PBCmin ≤ PBC
t
≤ PBCmax (38) beyond the consumer own demand during the battery scheduling
k −1
interval (k − 1) and Ptot is the total demand, which includes
PBDmin ≤ t
PBD ≤ PBDmax (39) the power demand by the NINSLs, INSLs, and SLs, the power
where SOCt is the available battery SOC at starting of interval exchange from battery, and the renewable power generation
t; SOCm in and SOCm ax are the minimum and maximum SOC during the interval (k − 1).
limits, respectively; PB C min and PB C max are the minimum and
maximum power limits for charging and PBDmin and PBDmax VI. CASE STUDY
are the minimum and maximum power limits for discharging, This section validates the proposed IREMS through a case
respectively. study involving various residential loads, renewable resources,
7) Power Export Constraint: The amount of power export and battery storage. Possibility in reduction of electricity bill
to grid is limited by the utility to avoid stability issues. This to be paid to the utility by the residential consumer is also as-
adds a new constraint stated as power export to grid (PEt ) at any certained by considering the real-time domestic load data, solar
interval t should be less than or equal to the maximum value irradiation, and wind velocity data collected from the residen-
permitted by the utility (PEt max ). tial locality of National Institute of Technology Tiruchirappalli,
India.
PEt ≤ PEt max (40)
PEt = PRt − t
(PNINSLs + t
PINSLs + t
PSLs + PBt ). (41) A. Case Study Environment
The residential loads considered for this study are categorized
C. Scheduling of Battery
into NINSLs, INSLs, and SLs, which are tabulated in Tables I–
The scheduling algorithm schedules the mode of operation of III, respectively, along with their power ratings. The residential
battery and amount of power exchange from battery in a schedu- building is equipped with lead acid battery bank. This type of
lable load interval for minimizing the objective function defined battery is commonly used in household energy storage appli-
in (29) subjected to the constraints (31)–(40). IREMS controls cations because of its relatively low price, low investment cost,
the battery by delivering the scheduling information to PCU. high availability, reasonable performance, and life cycle char-
Further, the operation of battery is fixed over a schedulable load acteristics [27]. The battery data are given in Table IV.
interval if the duration of schedulable load interval and battery The RER considered in this paper consist of solar PV pan-
scheduling interval are same. Else, the IREMS reschedules the els and small wind turbines, which can be installed in the roof
battery operation considering the mode of operation in the last top. Each solar PV panel is of 0.1 kW capacity with a derating
interval, total demand, power dissipated in the dump load, and factor of 0.8. The rated power of each wind turbine is 1 kW
the SOC limitation. The mathematical formulation of battery and rated wind speed is 10 m/s. The energy price variations are
ARUN AND SELVAN: IREMS FOR DYNAMIC DR IN SMART BUILDINGS 1337

TABLE III TABLE VI


SCHEDULABLE LOADS OPTIMAL SIZE OF RER AND BATTERY

S. No. Load Power rating (kW) Pre-emptive status S. No. Resources Count Aggregated rating

1 Cloth washer 0.8 1 1 Solar PV panel 24 2.4 kW


2 Cloth dryer 2.7 1 2 Wind turbine 2 2 kW
3 Dish washer 2.1 0 3 Battery 10 in series 20 Ah
4 Well pump 1.5 0
5 PHEV charging 2.3 0
6 Grinder 1.5 1

TABLE IV
BATTERY SPECIFICATIONS

S. No. Parameter Rating

1 Capacity 20 Ah
2 Voltage 12 V
3 Charging efficiency 85%
4 Discharging efficiency 95%
5 SOC limit 30–90%
6 Charging current limit 5–20% of rated capacity
7 Discharging current limit 0–20% of rated capacity

Fig. 3. Time horizon subdivisions.


TABLE V
COMPONENT RATING AND COST
The optimal sizing problem is solved using GA with population
size of 150 and maximum generation count of 100. The sizes of
S. No. Components Rating Life I* R* OM*
time cost cost cost ($/ RER and battery obtained at the end of the optimization process
(year) ($) ($) year) are tabulated in Table VI.
1 Solar PV panel 0.1 kW 20 100 100 8
2 Wind turbine 1 kW 20 1000 800 200 C. Cost Benefit by IREMS
3 Battery 12 V, 20 Ah 5 35 30 1
In this study, the proposed IREMS considers the nonschedu-
I*-Initial, R*-Replacement, and OM*-Operation and maintenance. lable loads in 1 min basis, i.e., duration of nonschedulable load
interval (DNS ) is 1 min. Further, it assumes that the power con-
sumption by any nonschedulable load over a particular interval
collected from [28]. The renewable resources data needed for the
i remains constant. However, the SLs are considered in 15 min
estimation of power generation from RER are measured at Na-
basis, i.e., duration of schedulable load interval (DS ) is 15 min.
tional Institute of Technology Tiruchirapplli, India (10.7611 N,
It is also assumed that the operation of SL during an interval t
78.8139 E) using the weather monitoring station installed in the
is uninterrupted if it is scheduled and started to run during that
Hybrid Electrical Systems Laboratory. The proposed IREMS
interval. The battery scheduling interval is considered as 5 min,
architecture is implemented in MATLAB and tested using a
i.e., DB is 5 min. It is also assumed that the mode of operation
computer with the following configuration: Intel core i5, clock-
of battery and power exchange are fixed for a particular battery
ing at 3.47 GHz and 4 GB RAM. Real-time energy prices can be
interval k. The RER dynamics (solar irradiation, temperature,
obtained from the utility through smart meter and future prices
and wind speed) are considered in one minute basis, i.e., du-
can be predicted by IREMS using any one of the price predic-
ration of RER (DRER ) is 1 min. Further, the solar irradiation,
tion algorithms [29], [30]. However, the case study performed
temperature, and wind velocity over a RER interval j are as-
in this paper presumes that the real time and predicted energy
sumed to be constant. The utility price variation is considered
prices are provided by the utility through smart meter.
in an hourly basis, i.e., the pricing interval (DP ) is 60 min. This
time horizon subdivision over a day is indicated in Fig. 3.
B. Optimal Sizing of RER and Battery
The electricity price variation over a day is shown in Fig. 4.
The RER and battery are optimally sized in such way that the The MDL is considered to be 5 kW and assumed to be con-
objective function defined in (13) is minimized while satisfying stant over a particular day. The excess payment for exceeding
the constraints given in (19)–(22). In this optimal sizing process, the MDL is taken as 2.5 times of the electricity price assigned
the duration of all the intervals (DP , DS , DNS , DB , DRER ) are by the utility. The PEL that is assigned by the utility is con-
taken as 60 min. The rating of individual unit of RER and sidered as 0.5 kW and the power export cost is considered
battery along with their life time and various costs are tabulated to be same as the base price. Based on the nature of prob-
in Table V. lem, special features of the solution algorithm, and the users
The maximum allowable initial investment cost and RER comfort, the scheduling algorithm proposed along with the
energy factor are considered as $10 000 and 0.3, respectively. IREMS discussed in Section V has been realized using GA with
1338 IEEE SYSTEMS JOURNAL, VOL. 12, NO. 2, JUNE 2018

Fig. 4. Energy price variation. Fig. 6. Comparison of maximum demand.

Fig. 7. Comparison of renewable power utilization. (a) Without IREMS. (b)


With IREMS.

Fig. 5. Battery SOC variation. In order to further validate the effectiveness of proposed
IREMS, the simulation study is extended for one-year period
maximum generation and population size of 150 and 100, re- considering the dynamics of consumer behavior, RER, and the
spectively. The results of this case study are presented and com- utility. In order to present the results with better clarity, duration
pared with the daily and annual electricity bill obtained without of all the intervals (DP , DS , DNS , DB , DRER ) are consid-
the proposed IREMS. It is found that the total electricity bill of ered to be 1 h. The proposed scheme reduces yearly electricity
a particular day considered in this case study is reduced from bill from $1550 to $1159, which accounts to be a reduction
146 to 114 cents by employing the IREMS. This indicates a of 25.23% in annual electricity bill. The total annual energy
reduction of 21.9% in daily electricity bill. dissipated through dump load is reduced from 496.5 kWh to
The variation in SOC of the battery for that particular day 191.5 kWh while employing IREMS, which confirms 61.43%
is observed and illustrated in Fig. 5. For better comparison the improved utilization of RER by the proposed IREMS. The de-
SOC of the battery at the starting of the day is maintained at mand variation for a particular week in the months of January
its minimum limit while performing this case study without and and May without and with IREMS are presented in Fig. 8(a) and
with IREMS. While using IREMS, it can be observed that at (b), respectively. The energy cost per day for all components in
the end of the day the SOC of the battery reaches its maximum the building considered for this case study during the months of
limit due to proper scheduling of SLs and effective utilization January and May without and with IREMS are given in Fig. 9(a)
of RER. Further, it is understood from Fig. 6 that by employing and (b), respectively.
IREMS the maximum demand of consumer is maintained below
the MDL in almost all the schedulable load intervals during a
day. D. Cash Payback Period
The power generation from RER and its utilization are pre- Net present value (NPV) and CPP are the commonly used
sented in Fig. 7. The part of energy obtained from RER, which is tools to assess decision of any investment [31]. The same tools
unused but dissipated through dump load due to PEL constraint are used in this section to determine the CPP of the investment
of the utility, is reduced from 4.16 to 2.18 kWh per day while on RER, battery, and the smart components of IREMS. CPP
employing the proposed IREMS. This ensures the effective uti- can be calculated by evaluating the NPV of proposed IREMS
lization of the RER and battery. with RER and battery for every year of the project life time as
ARUN AND SELVAN: IREMS FOR DYNAMIC DR IN SMART BUILDINGS 1339

TABLE VII
COST OF IREMS COMPONENTS

S. No Components Life Q* I* R* OM*


time (year) cost ($) cost ($) cost ($/year)

1 IREMS main unit 20 1 100 80 10


2 Smart NINSL module 20 1 30 20 5
3 Smart INSL module 20 1 50 45 5
4 Smart SL module 20 5 20 20 2
5 Smart PC module 20 1 30 30 5

Q*-Quantity, I*-Initial, R*-Replacement, and OM*-Operation and Maintenance.

be 20 years and 8%, respectively. The initial cost, replacement


cost and operation and maintenance cost of RER and battery
given in Table V along with the cost of PCU has been used in
the CPP analysis. In this paper, a 5-kVA PCU having a life time
of 20 years is considered. The initial cost, replacement cost,
and operation and maintenance cost of PCU are $500 , $500,
and $50/year, respectively. Also, the cost of different IREMS
components are tabulated in Table VII.
Net cash flow for the year i (NCFi ) can be expressed as
NCFi = ECSi − OMi (44)
Fig. 8. Demand variation over a week in month of January and May.
(a) Without IREMS (Consumer behavior). (b) With IREMS. where ECSi is the profit gained during year i by proposed
IREMS through reduction in electricity bill without any RER,
battery and IREMS. The annual electricity bill without RER,
battery and IREMS is obtained as $2337, whereas, the annual
electricity bill during first year of the project with the proposed
system is found to be $1159 in Section VI-C. This shows a sav-
ing of (ECSi ) $1178. The inflation rate of the electricity price is
considered as 8% per year for determination of CPP. OMi is the
expenditure during the year i toward the operation and mainte-
nance of smart IREMS components, RER, and battery. A 3%
inflation per year on operation and maintenance cost given in
Tables V and VII are considered for CPP analysis. NPV is calcu-
lated for every year of the project life time (n = [1, 2, . . . , 20])
using (43). The CPP is the year n, when the NPV crosses zero.
For the proposed system, the CPP is found to be 11 years, which
is 50% of the project life time.
From the case study, it is confirmed that the proposed scheme
is practicable with any number of household appliances and
profitable. Generally, the reduction in electricity bill and ef-
fective utilization of RER will be high as the number of SLs
increases.

VII. CONCLUSION
Fig. 9. Components of energy cost per day in month of January and May.
(a) Without IREMS. (b) With IREMS. In the smart grid environment, DSM and DR techniques de-
liver more benefits particularly in distribution systems. In this
explained below. NPV can be expressed as paper, an IREMS for smart buildings with smart appliances,
integration of information and communication technology and
n
NCFi in-house power generation from RER is presented. The pro-
NPV = −(CCIREM S + CCRER ) + (43) posed IREMS schedules the schedulable loads by considering
i=1
(1 + r)i
the operating dynamics of nonschedulable loads and intermittent
where CCIREM S and CCRER are the capital cost for the IREMS behavior of renewable resources. The resources are optimally
components and RER, respectively. These include the sum of sized using GA. Case study results demonstrate that the pro-
initial cost and replacement cost. n is the total number years in posed IREMS is able to handle a large number of schedulable
project life time and r is discount rate, which are considered to and nonschedulable loads and yields considerable savings by
1340 IEEE SYSTEMS JOURNAL, VOL. 12, NO. 2, JUNE 2018

maintaining the total household demand below the MDL. Fur- [21] J. M. Carrasco et al., “Power-electronic systems for the grid integration of
ther, the IREMS ensures the effective utilization of RER by renewable energy sources: A survey,” IEEE Trans. Ind. Electron., vol. 53,
no. 4, pp. 1002–1016, Jun. 2006.
optimally controlling the battery operation and properly time [22] M. Liserre, R. Teodorescu, and F. Blaabjerg, “Stability of photovoltaic and
scheduling the schedulable loads. The cost analysis demon- wind turbine grid-connected inverters for a large set of grid impedance val-
strates that the proposed system provides significant savings ues,” IEEE Trans. Power Electron., vol. 21, no. 1, pp. 263–272, Jan. 2006.
[23] Y. A. Katsigiannis, P. S. Georgilakis, and E. S. Karapidakis, “Hybrid
with minimum CPP. simulated annealing-tabu search method for optimal sizing of autonomous
power systems with renewables,” IEEE Trans. Sustain. Energy, vol. 3,
REFERENCES no. 3, pp. 330–338, Jul. 2012.
[24] O. A. Lara, N. Jenkins, J. Ekanayake, P. Cartwright, and M. Hughes, Wind
[1] B. P. Esther and K. S. Kumar, “A survey on residential demand side Energy Generation: Modeling and Control. New York, NY, USA: Wiley,
management architecture, approaches, optimization models and methods,” 2009.
Renewable Sustain. Energy Rev., vol. 59, pp. 342–351, Jun. 2016. [25] R. Belfkira, L. Zhang, and G. Barakat, “Optimal sizing study of hy-
[2] G. Strbac, “Demand side management: Benefits and challenges,” Energy brid wind/PV/diesel power generation unit,” Solar Energy, vol. 85, no. 1,
Policy, vol. 36, no. 12, pp. 4419–4426, Dec. 2008. pp. 100–110, Jan. 2011.
[3] M. Chaabene, M. B. Ammar, and A. Elhajjaji, “Fuzzy approach for opti- [26] S. Kumaravel and S. Ashok, “An optimal stand-alone biomass/solar-
mal energy-management of a domestic photovoltaic panel,” Appl. Energy, PV/pico-hydel hybrid energy system for remote rural area electrification
vol. 84, no. 10, pp. 992–1001, Oct. 2007. of isolated village in western-ghats region of india,” Int. J. Green Energy,
[4] P. Palensky and D. Dietrich, “Demand side management: Demand re- vol. 9, no. 5, pp. 398–408, Jul. 2012.
sponse, intelligent energy systems, and smart loads,” IEEE Trans. Ind. [27] D. Setlhaolo and X. Xia, “Optimal scheduling of household appliances
Informat., vol. 7, no. 3, pp. 381–388, Aug. 2011. with a battery storage system and coordination,” Energy Buildings, vol. 94,
[5] M. A. Zehir and M. Bagriyanik, “Demand side management by controlling pp. 61–70, May 2015.
refrigerators and its effects on consumers,” Energy Convers. Manage., [28] Electricity, “Price,” 2016. [Online]. Available: https://www.comed.com/
vol. 64, pp. 238–244, Dec. 2012. Pages/default.aspx. Accessed on: Apr. 28, 2016.
[6] V. Pradhan, V. S. K. M. Balijepalli, and S. A. Khaparde, “An effective [29] D. Huang, H. Zareipour, W. D. Rosehart, and N. Amjady, “Data mining
model for demand response management systems of residential electricity for electricity price classification and the application to demand-side man-
consumers,” IEEE Syst. J., vol. 10, no. 2, pp. 434–445, Jun. 2016. agement,” IEEE Trans. Smart Grid, vol. 3, no. 2, pp. 808–817, Jun. 2012.
[7] N. Ruiz, I. Cobelo, and J. Oyarzabal, “A direct load control model for [30] S. Fan, C. Mao, and L. Chen, “Next-day electricity-price forecasting using
virtual power plant management,” IEEE Trans. Power Syst., vol. 24, no. 2, a hybrid network,” IET Generation, Transmiss. Distrib., vol. 1, no. 1,
pp. 959–966, May 2009. pp. 176–182, Jan. 2007.
[8] D. D. Weers and M. A. Shamsedin, “Testing a new direct load control [31] M. M. R. Singaravel and S. A. Daniel, “Sizing of hybrid PMSG-PV system
power line communication system,” IEEE Trans. Power Del., vol. PD-2, for battery charging of electric vehicles,” Frontiers Energy, vol. 9, no. 1,
no. 3, pp. 657–660, Jul. 1987. pp. 68–74, Mar. 2015.
[9] K. Herter, “Residential implementation of critical-peak pricing of elec-
tricity,” Energy Policy, vol. 35, no. 4, pp. 2121–2130, Apr. 2007.
[10] P. Centolella, “The integration of price responsive demand into regional
transmission organization (RTO) wholesale power markets and system
operations,” Energy, vol. 35, no. 4, pp. 1568–1574, Apr. 2010. S. L. Arun received the B.E. degree in electrical and
[11] Y. Wu, V. K. N. Lau, D. H. K. Tsang, L. P. Qian, and L. Meng, “Optimal electronics engineering from the Institute of Road and
energy scheduling for residential smart grid with centralized renewable Transport Technology, Erode, India, in 2010, and the
energy source,” IEEE Syst. J., vol. 8, no. 2, pp. 562–576, Jun. 2014. M.Tech. degree in power systems from the National
[12] A. Y. Saber and G. K. Venayagamoorthy, “Resource scheduling under Institute of Technology Calicut, Kozhikode, India,
uncertainty in a smart grid with renewables and plug-in vehicles,” IEEE in 2013. He is currently working toward the Ph.D.
Syst. J., vol. 6, no. 1, pp. 103–109, Mar. 2012. degree in computer applications in power systems at
[13] T. Hubert and S. Grijalva, “Modeling for residential electricity optimiza- the National Institute of Technology Tiruchirappalli,
tion in dynamic pricing environments,” IEEE Trans. Smart Grid, vol. 3, Tiruchirappalli, India.
no. 4, pp. 2224–2231, Dec. 2012. His research interests include smart grid, demand-
[14] X. Chen, T. Wei, and S. Hu, “Uncertainty-aware household appliance side management, demand response, and distributed
scheduling considering dynamic electricity pricing in smart home,” IEEE generation.
Trans. Smart Grid, vol. 4, no. 2, pp. 932–941, Jun. 2013.
[15] C. O. Adika and L. Wang, “Autonomous appliance scheduling for house-
hold energy management,” IEEE Trans. Smart Grid, vol. 5, no. 2, M. P. Selvan (M’13) received the B.E. degree in
pp. 673–682, Mar. 2014. electrical and electronics engineering from Manon-
[16] P. Mesari and S. Krajcar, “Home demand side management integrated maniam Sundaranar University, Tirunelveli, India,
with electric vehicles and renewable energy sources,” Energy Buildings, in 1999, the M.E. degree in power systems from
vol. 108, pp. 1–9, Dec. 2015. the National Institute of Technology, Tiruchirap-
[17] E. L. Ratnam, S. R. Weller, and C. M. Kellett, “Scheduling residential palli, India, in 2000, and the Ph.D. degree in com-
battery storage with solar PV: Assessing the benefits of net metering,” puter applications in power systems from Indian
Appl. Energy, vol. 155, pp. 881–891, Oct. 2015. Institute of Technology Madras, Chennai, India, in
[18] C. Wang, Y. Zhou, B. Jiao, Y. Wang, W. Liu, and D. Wang, “Robust opti- 2006.
mization for load scheduling of a smart home with photovoltaic system,” He is currently an Assistant Professor with the
Energy Convers. Manage., vol. 102, pp. 247–257, Sep. 2015. Electrical and Electronics Engineering Department,
[19] Z. Zhu, S. Lambotharan, W. H. Chin, and Z. Fan, “A game theoretic op- National Institute of Technology Tiruchirappalli, Tiruchirappalli, India, where
timization framework for home demand management incorporating local he is also part of the Hybrid Electrical Systems Laboratory. He has 16 years of
energy resources,” IEEE Trans. Ind. Informat., vol. 11, no. 2, pp. 353–362, teaching and research experience in the field of power systems. He has published
Apr. 2015. more than 95 technical research papers in various national, international confer-
[20] Z. M. Fadlullah, D. M. Quan, N. Kato, and I. Stojmenovic, “GTES: An ences and journals. His research interests include distribution system analysis,
optimized game-theoretic demand-side management scheme for smart distributed generators, microgrid, smart-grid, demand-side management, cus-
grid,” IEEE Syst. J., vol. 8, no. 2, pp. 588–597, Jun. 2014. tom power devices, and power quality.

You might also like