Вы находитесь на странице: 1из 8

This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.


IEEE TRANSACTIONS ON POWER SYSTEMS 1
Development of Physical-Based Demand
Response-Enabled Residential Load Models
Shengnan Shao, Student Member, IEEE, Manisa Pipattanasomporn, Member, IEEE, and
Saifur Rahman, Fellow, IEEE
AbstractIn order to support the growing interest in demand
response (DR) modeling and analysis, there is a need for phys-
ical-based residential load models. The objective of this paper is to
present the development of such load models at the appliance level.
These include conventional controllable loads, i.e., space cooling/
space heating, water heater, clothes dryer and electric vehicle. Val-
idation of the appliance-level load models is carried out by com-
paring the models output with the real electricity consumption
data for the associated appliances. The appliance-level load models
are aggregated to generate load proles for a distribution circuit,
which are validated against the load proles of an actual distri-
bution circuit. The DR-sensitive load models can be used to study
changes in electricity consumption both at the household and the
distribution circuit levels, given a set of customer behaviors and/or
signals from a utility.
Index TermsAppliance-level load model, demand response
(DR) and distribution circuit, electric vehicle, physical-based.
I. INTRODUCTION
T
HERE has been a lot of interest in load modeling in
the past several decades. Scale of load models can vary,
ranging from the appliance level to the power grid level. In
1995, IEEE provided a load model bibliography for power
ow and dynamic performance simulation. This bibliography
covered almost all relevant load modeling work in late 20th
century [1]. The classic load models dened load as a function
of voltage and frequency, which are referred to as static load
models. The dynamic load models were developed for tran-
sient studies. To better guide the load modeling work, authors
in [2] provides the basic denitions and concepts. Obviously,
different simulations need different load model representations.
In order to study demand management, published work
has focused on physical-based load models, especially on
heating, ventilation, and air conditioning (HVAC) loads and
water heating loads [3][7]. Physical-based load models for
residential HVAC were developed in [8], [9] and tested against
utility data. The models captured thermodynamic principles
of building structures and the diversication was created by
random distribution functions when building the distribution
circuit load prole. These load models have been mostly used
for direct load control (DLC) studies [10][12]. Other than
Manuscript received June 06, 2011; revised November 21, 2011 and
February 15, 2012; accepted March 29, 2012. This work was supported in
part by the United States Army Corps of Engineers Humphreys under grant
W912HQ-09-C-0058. Paper no. TPWRS-00536-2011.
The authors are with the Advanced Research Institute, Virginia Tech,
Arlington, VA 22203 USA (e-mail: sshao@vt.edu; mpipatta@vt.edu;
srahman@vt.edu).
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TPWRS.2012.2208232
a physical-based methodology, load models were developed
based on consumers behavior [13]. Load models built from
statistical survey data and historical measurements were pre-
sented in [14][16] with proper random functions designed for
aggregation diversity.
With the development of the smart grid, there is a need for
load models that can facilitate the study of changes in electricity
consumption in response to customer behavior and/or signals
from a utility. Such load models are useful to evaluate the im-
pact of demand response (DR) at a distribution circuit level. For
the load models to represent DR activities, the following char-
acteristics are required, which have not all been all addressed in
the published literature:
Comprehensivethe models should cover all major types
of controllable loads so that DR can be simulated consid-
ering consumer choices instead of simple load curtailment.
Physically basedthe models should be built according to
the physical and operational characteristics of the appli-
ances to reect the real-world situation.
Interactivethe models should allow interfaces for ex-
ternal signals to simulate the DR control actions.
Reasonably aggregatedthe algorithm should provide
reasonable load diversication and aggregation to repre-
sent the distribution circuit load prole.
This paper proposes a set of DR-enabled load models at the
appliance level, including space cooling/heating, water heating,
clothes drying and electric vehicle loads, for a residential dis-
tribution circuit (Comprehensive). The models are designed
based on their physical characteristics (physical-based) and
can be controlled externally (interactive) to reect any desired
DR strategies. At the same time, these models are aggregated
using a stochastic method to create load proles of a distribu-
tion circuit (reasonably aggregated). The resulting load pro-
les are validated against the utility measurement data in both
summer and winter to show representativeness of the proposed
load model. In addition, the developed load models are easy to
use as the models parameters are easily adjustable to t dif-
ferent network sizes present in different locations.
II. LOAD CATEGORIZATION
Hourly load curves of an average household by load type
are available from the RELOAD database [17], which is used
by the Electricity Module of the National Energy Modeling
System (NEMS) [18]. The load data are available for twelve
months (January to December), and three day types (typical
weekday, typical weekend and typical peak day). Based on the
RELOAD database, residential loads are classied by the fol-
lowing load types: space cooling/heating, water heating, cloth
drying, cooking, refrigeration, freezer, lighting and others.
For the purpose of this study, these load types are classied
into two categories: controllable and critical. Controllable loads
0885-8950/$31.00 2012 IEEE
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE TRANSACTIONS ON POWER SYSTEMS
Fig. 1. Block diagram of the space cooling/space heating load model.
are dened as the loads that can be controlled without noticeable
impacts on consumers life styles. The critical load category
contains loads that are either very important or noticeable when
being controlled.
For the residential load, space cooling/heating, water heating
and clothes drying loads are controllable; and all other loads
are considered critical and cannot be controlled. As a new type
of residential load, EV is not mentioned in RELOAD database.
Considering that the electric vehicle (EV) can usually be con-
trolled without noticeable effect on consumers, this paper puts
it into the Controllable category.
Sections III and IV below present the detailed load models
for each controllable load type (space cooling/heating, water
heating, clothes drying and EV), their validation against the
measurement data from a real house, and the methodology to
perform load aggregation by appliance type. The bottom-up ag-
gregation approach for each load type enables demand response
studies in a large-scale distribution circuit. All load models are
simulated in 1-min intervals. Data from the RELOAD database
are used to construct the critical load prole, which is presented
in Section VII.
III. DR-ENABLED SPACE COOLING/HEATING LOAD MODEL
This section presents the development of a DR-enabled space
cooling/heating load model, together with its validation and ag-
gregation to create the load prole at a distribution circuit level.
A. Model Development for the Space Cooling/Heating Load
Fig. 1 illustrates the block diagram of the DR-enabled space
cooling/space heating load model.
Inputs to the model are the DR control signal , time
series outdoor temperature data , thermostat set point
, allowable temperature deviation or dead band and
time series room temperature data . The model outputs
are the time series electric power consumption in
kilowatts of the space cooling/space heating unit, and the room
temperature at the next time step. The room temperature
output is used as an input to the load model at the next time
step. The model needs additional house parameters, including
house structures, number of people dwelling and the electrical
characteristics of the space cooling/space heating unit.
1) Calculation of the Electric Power Demand : A
central space cooling/space heating unit with a thermostat works
in an on-off mode and will simply run at its rated power when
turned on. In general, the thermostat control is set such that the
room temperature will uctuate around the thermostat set
point within the dead band of . For each time step ,
the demand for electricity of the space cooling/space heating
unit is calculated as
(1)
where is the rated power of the space cooling/space heating
system (kW), and is the status of the space cooling/space
heating unit in time slot , , .
For space cooling, the unit is ON when the room temperature
increases above the set point, plus the threshold. The unit is OFF
when the roomtemperature decreases belowa certain value. The
status of the unit remains the same if the room temperature is
within the acceptable band. This relationship is presented in (2):
(2)
For space heating, the operation is similar to above:
(3)
where is the DR control signal for the space cooling/space
heating unit in time slot , .
The electric power demand also depends on the DR control
signal . This signal is implicitly derived from the
revised thermostat set point preset by a homeowner during a DR
event. This scenario is possible with the use of a home energy
management (HEM) controller. For example, if during a normal
operating condition the homeowner sets the thermostat set point
at , but pre-determine that during a DR event his/her new
set point should be set at , this implies the signal of
. is positive for space cooling and negative for space
heating. Note that with a HEM controller, the value can
be congured to allow demand response implementation that
takes into account the priority of all end-use loads in a house,
and customer comfort levels. This topic is not in the scope of
this paper. More information about a possible DR strategy is
reported in [19].
2) Determination of Room Temperature : For each time
step , the room temperature is calculated as
(4)
where
room temperature in time slot ;
length of time slot (hour);
heat gain rate of the house during time slot ,
positive value results in an increase in room
temperature; and negative value results in a
decrease in room temperature (Btu/h);
cooling/heating capacity, positive for heating and
negative for cooling (Btu/h);
energy needed to change the temperature of the
air in the room by .
3) Calculation of Other Parameters ( and ): For each
time slot , the heat gain rate of the house is calculated as
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
SHAO et al.: DEVELOPMENT OF PHYSICAL-BASED DEMAND RESPONSE-ENABLED RESIDENTIAL LOAD MODELS 3
(5)
where , , and represent the area of the
wall, ceiling, and window of the house, in . , ,
and represent the heat resistance of the wall, ceiling,
and window, in [20].
number of air changes in each time slot (1/h);
house volume, in ;
outdoor temperature in time slot ;
solar heat gain coefcient of windows [21];
solar radiation heat power );
heat gain from people (Btu/h).
To change the house temperature by , the energy required
is calculated as
(6)
where is the specic heat capacity of air for a typical room
condition ( or ), and is
the volume of the house .
B. Model Validation for the Space Cooling/Heating Load
To validate this model, the space cooling/space heating model
is run with inputs (outdoor temperature, thermostat set point,
dead band, house structure parameters, and HVAC unit size)
from a real house in Virginia. Using the same inputs, the model
outputs (power consumption and indoor temperature) are com-
pared with the actual measurements. The comparisons in 1-min
intervals are illustrated in Fig. 2 for two 24-h periods: (a) one
in August for cooling demand; and (b) the other in January for
heating demand (kW).
Fig. 2 indicates similarities between the actual power con-
sumption and the model output. The energy difference is 1.4%
for summer and 1.2% for winter. Any discrepancies may result
from the assumptions associated with the house structure pa-
rameters.
C. Aggregation of Space Cooling/Heating Loads
The input parameters for space cooling/space heating model
are divided into three categories: temperatures, building struc-
tures and the space cooling/space heating unit characteristics.
These are parameters needed to be randomized for different
homes in the same distribution circuit.
The temperature category includes outdoor temperatures and
indoor temperature set points. The outdoor temperatures are ac-
quired from the National Climatic Data Center (NCDC) [22],
which should be the same for all houses in the same neighbor-
hood. For the indoor temperature set points, a uniform random
function is used to determine the variation in temperature set
points among different houses in the same distribution circuit.
Fig. 2. Space cooling/space heating model validationComparison of load
prole (kW). (a) Space cooling comparison: The total energy difference = 1.4%,
the total daily energy = 14.5 kWh (model); 14.3 kWh (measurement). (b) Space
heating comparison: The total energy difference = 1.2%, the total daily energy
= 17.2 kWh (model); 17.4 kWh (measurement).
The lower and upper limits for the temperature set points are
determined based on data from ASHRAE [23].
The building structure category includes the oor plan of the
buildings, areas of walls, ceilings and windows as well as the
R-values for each of them. Similar to the indoor temperature, a
uniform random function is used to determine the variation in
R-values and areas of wall, window, ceiling and oor among
different houses in the same distribution circuit. The lower and
upper limits for these values are acquired from a survey of the
service area. Such data for residential houses are obtained from
the American Housing Survey [24].
The space cooling/heating characteristic category includes
the cooling/heating capacities and power consumptions, which
is usually known as the unit sizing. Usually, the sizing is based
on the building oor plans, activities, occupants and environ-
ment. The unit sizing is calculated according to ASHRAE [25].
After getting all three sets of random functions for these input
parameters, the demand aggregation for space cooling/heating is
quantied by running the space cooling/heating model times
with different parameters. is the number of residential houses
in the distribution circuit.
IV. MODELING, VALIDATION, AND AGGREGATION
OF WATER HEATING LOAD
This section presents the development of a DR-enabled water
heating load model, together with its validation and aggregation
to obtain load proles of the water heating load at the distribu-
tion level.
A. Model Development for the Water Heating Load
Fig. 3 shows the block diagram of the developed water heater
model.
For each time step i, the demand for electricity of the water
heating unit is calculated as
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4 IEEE TRANSACTIONS ON POWER SYSTEMS
Fig. 3. Block diagram of the water heater load model.
(7)
where
rated power of the water heater (kW);
efciency factor;
status of the water heater in time slot , ,
;
DR control signal for water heater, ,
.
The water heater status is determined according to
the following rules: when the water temperature in the hot water
tank goes above the set point, it does not operate. When the
water temperature drops below a lower bound, the heating coils
start working again at its rated power until the outlet hot water
temperature reaches the upper bound:
(8)
where
hot water temperature set point ;
lower tolerance ;
mixed water temperature in the tank .
The water temperature in the tank is calculated as
(9)
where
temperature of inlet water ;
room temperature ;
Fig. 4. Water heater model validationload prole comparison. (a) Water
heater power consumption data from TED [26]. (b) Modeled water heater load
prole.
hot water ow rate in time slot (gpm);
surface area of the tank ;
volume of the tank (gallons);
heat resistance of the tank ;
duration of each time slot (minutes).
Note that the unit of the third term in (9) is Btu/lb, which is
equivalent to as 1 Btu is the amount of energy required to
heat 1 lb of water by , i.e., .
The electric power demand also depends on the DR control
signal received from an external source, such as an
in-home controller, or a utility. The DR control signal of 0 will
shut off the unit, and the DR control signal of 1 will turn the unit
on.
B. Model Validation for the Water Heating Load
To validate the developed water heater model, the following
data from a real house were used as inputs to the water heater
model: inlet water temperature, ambient temperature, water
heater temperature set point, tank sizes and rated power of the
water heating unit. The water heater load prole was measured
in 1-min intervals for a 24-h period using the Energy Detective
Device (TED), and compared with the model output. Fig. 4
illustrates this comparison.
The comparison is to show the power consumption character-
istics of a water heater, which are similar to the model output.
This implies that the developed model can be reasonably used
for simulations and analysis.
C. Aggregation of Water Heating Loads
The input parameters for water heater model are divided into
three categories: temperature prole, water heater characteris-
tics and hot water usage.
The temperature prole includes tank ambient temperatures,
inlet water temperatures and hot water temperature set points.
Tank ambient temperature is assumed to be the same as the room
temperature, which can be acquired from the space cooling/
space heating model. Inlet water temperature is assumed to be
the same as the ground temperature, which can be acquired from
Soil Climate Analysis Network (SCAN) [27]. For the hot water
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
SHAO et al.: DEVELOPMENT OF PHYSICAL-BASED DEMAND RESPONSE-ENABLED RESIDENTIAL LOAD MODELS 5
Fig. 5. Clothes dryer model validationload prole comparison.
temperature set points, a uniform random function is used to de-
termine the variation in temperature set points among different
houses in the same distribution circuit.
The lower and upper limits for the hot water temperature set
points are determined based on data from [28], which speci-
es typical residential hot water temperature set points between
and .
The water heater characteristics include the R-values, tank
sizes and rated power. Similar to the hot water temperature, a
uniform random function is used to determine the variation in
R-values, water heater tank sizes and rated power among dif-
ferent houses in the same distribution circuit. The typical ranges
for these values are acquired from [28] and [29].
For hot water usage, the hourly fraction data fromCalifornias
Hourly Water Heating Calculations [30] are taken as a reference.
At the same time, hot water usage is categorized into different
types in percentage of the daily household hot water usage [31].
Therefore for each type of hot water usage, the water consump-
tion duration in a minute is the hot water demand in gallon di-
vided by the ow rate in gallon per minute (gp). Monte Carlo
method is used to decide when the hot water is consumed based
on the hot water hourly usage fraction shown in Fig. 5 for dif-
ferent residential houses.
V. MODELING, VALIDATION, AND AGGREGATION
OF CLOTHES DRYER LOAD
This section presents the development of a DR-enabled
clothes dryer load model, together with its validation and
aggregation at the distribution level.
A. Model Development for the Clothes Drying Loads
The power consumption of a typical clothes dryer is from the
motor and the heating coils. The power demand of the motor
part is usually in the range of several hundred watts, but that of
the heating coils can be several kilowatts.
For each time slot , the demand for electricity of the clothes
drying unit is calculated as
(10)
where
rated power of clothes-dryer heating coil (kW);
drying level ;
total number of drying levels;
power consumption of the dryers motor (kW);
status of the clothes-dryers heating coils in time
slot , , ;
DR control signal for clothes dryer in time slot
, , .
The electric power demand also depends on the DR control
signal received from an external source, such as an
in-home controller, or a utility. For the clothes dryer load, when
a DRcontrol signal is received, only the heating coil will be con-
trolled (ON/OFF) but the motor part will not be controlled. This
implies that the clothes dryer will be spinning during the con-
trol period, thus consuming only a fraction of the overall load
(several kW).
B. Model Validation for the Clothes Drying Loads
For validation, the rated power of a real clothes dryer and its
usage prole were used as inputs to the clothes dryer model.
The clothes dryer load prole was measured in 1-min intervals
during its operation using TED, and compared with the model
output. Fig. 5 illustrates this comparison.
The close match between the actual and modeled power con-
sumption characteristics of a clothes dryer implies the useful-
ness of the clothes dryer model for simulation and analysis.
C. Aggregation of Clothes Drying Loads
The clothes dryer load prole is developed based on a prob-
ability distribution function that is similar to the dryer power
consumption prole obtained from the measurement of a home
in Florida [32]. The dryer running time generally starts from
late morning to midnight, with the mean in the evening. After
nding out the dryers running time, the demand aggregation of
clothes dryers is acquired by running the developed model to
obtain power consumption of all dryers in the distribution cir-
cuit.
VI. MODELING AND AGGREGATION OF EV LOAD
This section presents the development of a DR-enabled EV
charging load model, together with its aggregation.
A. Model Development For the Electric Vehicles (EV)
To model EV charging proles, three parameters are essen-
tial: the rated charging power, the plug-in time and the battery
state-of-charge (SOC). Equation(11) shows the calculation of
the EV charging prole:
(11)
where
EV charge power in time slot (kW);
EV rated power (kW);
EV connectivity status in time slot , 0 if EV is
not physically connected to the outlet and 1 if EV
is connected;
uncontrolled EV charging status in time slot ,
which depends on the battery SOC as shown in
(12): 0 if EV is not being charged and 1 if EV
is being charged;
DR control signal for EV in time slot , ,
.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6 IEEE TRANSACTIONS ON POWER SYSTEMS
TABLE I
EVS IN THE U.S. MARKET [33][37]
(12)
The battery SOC at time slot is a function of the SOC at the
previous time slot, the energy used for driving and the battery
rated capacity, which is determined by
(13)
(14)
where
battery SOC when EV is plugged in;
length of time slot (minute);
energy used for driving (kWh);
battery rated capacity (kWh).
The EVpower demand also depends on the DR control signal
received from an external source, such as an in-home
controller, or a utility. The DR control signal of 0 will stop the
charging of EV while the DR control signal of 1 will allow the
EV to start charging. The developed EV model will extend the
EV charging complete time if the EV charging is interrupted by
the DR signal. That is, the EV will continue to be charged until
it is fully charged.
B. Aggregation of EV Load
To determine the EV eet charge prole in a distribution cir-
cuit, three parameters (vehicle rated charging power, plug-in
time and the battery SOC) have to be reasonably diversied.
For EV rated charging power, Table I shows the basic battery
charge data of three popular EVs in the US market. The charge
power presented in Table I is used to determine the charge power
of EV eet with a reasonable mix of different EV makes and
models.
The EV plug-in time is derived from the National House-
hold Travel Survey [38]. The EV plug-in time is modeled using
a normal distribution with a mean and a variance derived ac-
cording to the data presented in [39].
EV driving patterns are used to determine the EV energy
storage status, i.e., the battery SOC. The daily driving distance
for each EV in a distribution circuit is determined based on the
data presented in [38] using Monte Carlo simulation. The bat-
tery SOC of each EV when plugged in is determined by (13).
Fig. 6. EV eet charging prole.
To illustrate the aggregation of EV load at a distribution cir-
cuit level, Fig. 6 shows an example charging prole of a 100-EV
eet with the mix of 70% Chevy Volt, 20% Nissan LEAF and
10% Tesla Roadster. The EV eet mix (70/20/10) is chosen
as a representative mix to show a sample aggregation of dif-
ferent EVtypes. As EVmarket share changes over time and new
EV models enter the market, EV eet mix can change, but the
methodology will stay the same. The recommended charging
power rates in Table I are chosen for the simulation. The EVs
are assumed to come back home and plugged in at different time
according to a normal distribution with mean at 18:00 and 1-h
variance.
Note that since the daily driving distance of each EV is de-
termined by Monte Carlo simulation, it is unlikely that all EVs
will come home with empty battery. This is the reason for the
narrow EV loading time shown in Fig. 6.
VII. EXAMPLE LOAD PROFILES FOR A RESIDENTIAL
DISTRIBUTION CIRCUIT AND VALIDATION
To demonstrate the aggregation of the developed load model
at the distribution circuit level, a distribution circuit in the Vir-
ginia Tech Electric Service (VTES) service area is taken as a
case study. This circuit has 34 laterals with 117 transformers
serving 780 residential/small commercial customers.
A. Controllable Loads
The aggregated load prole of controllable loads in a dis-
tribution circuit is created using the load models presented in
Sections III and V. The parameters in Table II are used for load
diversication. The percentage of the households owning each
electric appliance, known as ownership rate, is taken into ac-
count according to the method presented in [40]. The owner-
ship rates can be found from the survey data in [24]. EV is not
included in the analysis due to the lack of EV penetration and
measurement data.
B. Critical Loads
The aggregated load prole of critical loads is created based
on the historical data of the industry-accepted database, i.e.,
RELOAD database. According to the categories in Section II,
ve load types cooking, lighting, refrig-
erator, freezer, and othersare considered critical loads.
Assuming in the distribution circuit, there are residential
houses. The critical load prole in a house in each time slot
(in 1-min interval) is derived based on (15). Note that the own-
ership rate has already been taken into account in the RELOAD
database:
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
SHAO et al.: DEVELOPMENT OF PHYSICAL-BASED DEMAND RESPONSE-ENABLED RESIDENTIAL LOAD MODELS 7
TABLE II
PARAMETERS AND FUNCTIONS FOR LOAD MODEL DIVERSIFICATION
(15)
where
number of residential houses;
power consumption of critical load in time
slot ;
hourly fraction of annual residential load
for load type in the hour of time slot ;
average annual household electricity
consumption (kWh) for load type [44].
C. Distribution Circuit Load Proles and Validation
The aggregation of the controllable loads, together with the
critical loads, of a distribution circuit with 780 customers is pre-
sented in Fig. 7 as solid bold lines for both (a) summer and (b)
winter seasons. The actual hourly load proles of the same cir-
cuit for several days are also presented for comparison purposes.
As shown in Fig. 7, the aggregation of the proposed load models
reects the real load prole of a residential distribution circuit
reasonably well.
VIII. CONCLUSIONS
This paper presents a methodology to develop the bottom-up
DR-enabled load models of space cooling/heating, water
heating, clothes drying and electric vehicle loads. The models
are developed at the appliance level taking into account the
physical and operational characteristics of different load types.
A methodology is also discussed showing how to aggregate
each load type to obtain the distribution circuit load prole
using a stochastic method. Each load model is validated against
Fig. 7. Validation of the distribution circuit load proles. (a) Summer load pro-
le comparison. (b) Winter load prole comparison.
the real electricity consumption data of the associated load
type. The comparison of the aggregated load prole at the
distribution circuit level also shows similarity with the actual
distribution circuit load proles. This similarity proves that the
proposed load modeling methodology can be used to create
realistic distribution circuit load prole for DR studies.
With the DR-enabled load models developed and presented
in this paper, researchers in academia and/or electric utilities
can use the developed models to perform detailed DR studies.
For example, studies of the impact of different demand response
strategies and different customer behaviors on distribution cir-
cuit load shape changes can be conducted; and observations can
be made at both the appliance level and the distribution circuit
level.
ACKNOWLEDGMENT
The authors would like to thank UVA for providing the data
for HVAC model validation.
REFERENCES
[1] IEEE Task Force on Load Representation for Dynamic Performance,
Bibliography on load models for power ow and dynamic perfor-
mance simulation, IEEE Trans. Power Syst., vol. 10, no. 1, pp.
523538, Feb. 1995.
[2] IEEE Task Force on Load Representation for Dynamic Performance,
Load representation for dynamic performance analysis, IEEE Trans.
Power Syst., vol. 8, no. 2, pp. 47282, May 1993.
[3] S. Ihara and F. C. Schweppe, Physically based modeling of cold
load pickup, IEEE Trans. Power App. Syst., vol. PAS-100, no. 9, pp.
41424150, Sep. 1981.
[4] E. Agneholm and J. Daalder, Cold load pick-up of residential load,
Proc. Inst. Elect. Eng., Gen., Transm., Distrib., vol. 147, no. 1, pp.
4450, Jan. 2000.
[5] L. Paull, D. MacKay, H. Li, and L. Chang, A water heater model for
increased power systemefciency, in Proc. Canadian Conf. Electrical
and Computer Engineering, May 36, 2009.
[6] A. Gomes, C. H. Antunes, and A. G. Martins, Physically-based load
demand models for assessing electric load control actions, in Proc.
IEEE Bucharest PowerTech, Jul. 2009.
[7] R. E. Mortensen and K. P. Haggerty, A stochastic computer model for
heating and cooling loads, IEEE Trans. Power Syst., vol. 3, no. 3, pp.
12131219, Aug. 1988.
[8] M. Chan, E. N. Marsh, J. Y. Yoon, and G. B. Ackerman, Simula-
tion-based load synthesis methodology for evaluating load-manage-
ment programs, IEEE Trans. Power App. Syst., vol. PAS-100, no. 4,
pp. 17711778, Apr. 1981.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8 IEEE TRANSACTIONS ON POWER SYSTEMS
[9] R. E. Mortensen and K. P. Haggerty, Dynamics of heating and cooling
loads: Models, simulation, and actual utility data, IEEE Trans. Power
Syst., vol. 5, no. 1, pp. 243249, Feb. 1990.
[10] K. Y. Huang and Y. C. Huang, Integrating direct load control with
interruptible load management to provide instantaneous reserves
for ancillary services, IEEE Trans. Power Syst., vol. 19, no. 3, pp.
16261634, Aug. 2004.
[11] D. C. Wei and N. Chen, Air conditioner direct load control by multi-
pass dynamic programming, IEEE Trans. Power Syst., vol. 10, no. 1,
pp. 307313, Feb. 1995.
[12] M. W. Gustafson, J. S. Baylor, and G. Epstein, Direct water heater
load control-estimating program effectiveness using an engineering
model, IEEE Trans. Power Syst., vol. 8, no. 1, pp. 137143, Feb.
1993.
[13] C. F. Walker and J. L. Pokoski, Residential load shape modelling
based on customer behavior, IEEE Trans. Power App. Syst., vol. PAS-
104, no. 7, pp. 17031711, Jul. 1985.
[14] I. C. Schick, P. B. Usoro, M. F. Ruane, and J. A. Hausman, Residential
end-use load shape estimation from whole-house metered data, IEEE
Trans. Power Syst., vol. 3, no. 3, pp. 986991, Aug. 1988.
[15] A. H. Noureddine, A. T. Alouani, and A. Chandrasekaran, On the
maximum likelihood duty cycle of an appliance and its validation,
IEEE Trans. Power Syst., vol. 7, no. 1, pp. 228235, Feb. 1992.
[16] A. Capasso, W. Grattieri, R. Lamedica, and A. Prudenzi, A bottom-up
approach to residential load modeling, IEEE Trans. Power Syst., vol.
9, no. 2, pp. 957964, May 1994.
[17] RELOAD Database Documentation and Evaluation and Use in NEMS,
Electric Power Research Institute, Jul. 9, 2001. [Online]. Available:
http://www.onlocationinc.com/LoadShapesReload2001.pdf.
[18] The National Energy Modeling System: An Overview, Energy Infor-
mation Administration, 2003. [Online]. Available: http://www.eia.doe.
gov/oiaf/aeo/overview/.
[19] S. Shao, M. Pipattanasomporn, and S. Rahman, Demand response as
a load shaping tool in an intelligent grid with electric vehicles, IEEE
Trans. Smart Grid, to be published.
[20] Insulation Fact Sheet, Oakridge National Laboratory. [Online]. Avail-
able: http://www.ornl.gov/sci/roofs+walls/insulation/ins_05.html.
[21] Solar Heat Gain Coefcient (SHGC) FAQs, Department of En-
ergyBuilding Energy Codes Program. [Online]. Available:
http://www.energycodes.gov/support/shgc_faq.stm.
[22] National Climatic Data Center. [Online]. Available: ftp://ftp.ncdc.
noaa.gov/pub/data/asos-onemin/.
[23] 2008 ASHRAE Handbook: Heating, Ventilating, and Air-Conditioning
Systems and Equipment. Atlanta, GA, American Society of Heating,
Refrigerating and Air-Conditioning Engineers, Inc. (ASHRAE), 2008.
[24] American Housing Survey National Tables, 2009. [Online]. Available:
http://www.census.gov/hhes/www/housing/ahs/ahs09/ahs09.html.
[25] 2009 ASHRAE Handbook: Fundamentals. Atlanta, GA, American
Society of Heating, Refrigerating and Air-Conditioning Engineers
(ASHRAE). Knovel (Firm), 2009.
[26] TED 5000 Live Demo, The Energy Detective Website, May 2011.
[Online]. Available: http://www.theenergydetective.com/ted-5000/
live-demo.
[27] Soil Climate Analysis Network (SCAN). [Online]. Available: http://
www.wcc.nrcs.usda.gov/scan/.
[28] Department of EnergyEnergy Savers Tips. [Online]. Available:
http://www1.eere.energy.gov/consumer/tips/water_heating.html.
[29] Conventional Storage Water Heaters, Department of EnergyEnergy
Efciency and Renewable Energy. [Online]. Available: http://www.en-
ergysavers.gov/your_home/water_heating/index.cfm/mytopic=12980.
[30] Domestic Hot Water System Modeling for the Design of Energy Ef-
cient Systems, NAHB Research Center Inc., 2002.
[31] Hourly Water Heating Calculations, PG&E 2005 Title 24 Building
Energy Efciency Standards, May 15, 2002. [Online]. Available:
http://www.energy.ca.gov/title24/2005standards/archive/documents/
2002-05-30_workshop/2002-05-17_WTR_HEAT_CALCS.PDF.
[32] D. S. Parker, J. R. Sherwin, J. K. Sonne, S. F. Barkaszi, D. B. Floyd, and
C. R. Withers, Measured Energy Savings of a Comprehensive Retrot
in an Existing Florida Residence, FSEC-CR-978-97, Florida Solar En-
ergy Center (FSEC), Dec. 1997, [Online]. Available: http://www.fsec.
ucf.edu/en/publications/html/FSEC-CR-978-97/,
[33] My Chevy Volt, Mar. 2011. [Online]. Available: http://www.my-
chevroletvolt.com/.
[34] Voltech 240 V Charge Station, SPX Corporation. [Online]. Available:
https://www.homecharging.spx.com/volt/pdf/GM10-463.pdf.
[35] Nissan LEAF Features Specications. [Online]. Available: http://
www.nissanusa.com/ev/media/pdf/specs/FeaturesAndSpecs.pdf.
[36] Nissan LEAF Ofcial Website Q&A, Mar. 2011. [Online]. Available:
http://www.nissanusa.com/leaf-electric-car/faq/list/charging#/leaf-
electric-car/faq/list/charging.
[37] Tesla Roadster Ofcial Website, Mar. 2011. [Online]. Available: http://
www.teslamotors.com/goelectric/charging.
[38] National Household Travel Survey, Oakridge National Laboratory,
2001. [Online]. Available: http://nhts.ornl.gov/index.shtml.
[39] J. Taylor, A. Maitra, M. Alexander, D. Brooks, and M. Duvall, Eval-
uation of the impact of plug-in electric vehicle loading on distribution
system operations, in Proc. IEEE PES General Meeting, Jul. 2630,
2009.
[40] V. Hamidi, F. Li, and F. Robinson, Demand response in the UKs do-
mestic sector, Elect. Power Syst. Res., vol. 79, no. 12, pp. 17221726,
Dec. 2009.
[41] Basic Health and Safety Requirements. All about Homes Residential
Inspection Service, North American Insulation Manufacturers Associ-
ation, May 2011. [Online]. Available: http://www.allabouthomes.com/
Residential%20Construction%20Brochure%20515.pdf.
[42] Residential Home InsulationHow Much Insulation Should Be
Installed, North American Insulation Manufacturers Association,
May 2011. [Online]. Available: http://www.naima.org/insula-
tion-knowledge-base/residential-home-insulation/how-much-insu-
lation-should-be-installed.html.
[43] The Facts about Windows & Heat Loss, National Fenestration Rating
Council, Jan. 2005. [Online]. Available: http://www.nfrc.org/docu-
ments/U-Factor.pdf.
[44] EIA-826 Database Monthly Electric Utility Sales and Revenue Data,
Energy Information Administration. [Online]. Available: http://www.
eia.doe.gov/cneaf/electricity/page/eia826.html.
Shengnan Shao (S08) received the B.S. and M.S. degrees in electrical engi-
neering from Tsinghua University (THU), Beijing, China, in 2005 and 2007,
respectively. She is pursuing the Ph.D. degree in the Department of Electrical
and Computer Engineering at Virginia Tech, Arlington.
She is now a research assistant at the Advanced Research Institute of Vir-
ginia Tech. She is a member of the team working on Intelligent Distributed
Autonomous Power Systems (IDAPS) project at the Virginia TechAdvanced
Research Institute. Her elds of interest include smart grid, demand response,
electric vehicle, power distribution, and renewable energy systems.
Manisa Pipattanasomporn (S01M06) received the B.S. degree from the
Electrical Engineering Department, Chulalongkorn University, Thailand, in
1999, the M.S. degree in energy economics and planning from the Asian
Institute of Technology (AIT), Thailand, in 2001, and the Ph.D. degree in
electrical engineering from Virginia Tech, Arlington, in 2004.
She joined Virginia Techs Department of Electrical and Computer Engi-
neering as an Assistant Professor in 2006. She is working on multiple research
grants from the U.S. National Science Foundation, the U.S. Department of De-
fense, and the U.S. Department of Energy, on research topics related to smart
grid, microgrid, energy efciency, load control, renewable energy, and electric
vehicles. Her research interests include renewable energy systems, energy ef-
ciency, distributed energy resources, and the smart grid.
Saifur Rahman (S75M78SM83F98) is the director of the Advanced Re-
search Institute at Virginia Tech, Arlington, where he is the Joseph Loring Pro-
fessor of electrical and computer engineering. He also directs the Center for
Energy and the Global Environment at the university. He has published over
300 papers on conventional and renewable energy systems, load forecasting,
uncertainty evaluation, and infrastructure planning.
Professor Rahman is currently the chair of the U.S. National Science
Foundation Advisory Committee for International Science and Engineering. In
2011, he is serving as the vice president for New Initiatives and Outreach of
the IEEE Power & Energy Society and a member of its Governing Board. He is
a member-at-large of the IEEE-USA Energy Policy Committee. He served as
a program director in engineering at NSF between 1996 and 1999. In 2006 he
served as the vice president of the IEEE Publications Board, and a member of
the IEEE Board of Governors. He is a distinguished lecturer of the IEEE PES.