Comparison of Peak Demand Shaving Potential of

Demand Response and Distributed Energy Storage in

Residential Buildings
Akintonde Abbas Sumedh Halbe Badrul Chowdhury
Department of Electrical and Computer Department of Electrical and Computer Energy Production & Infrastructure Center
Engineering Engineering University of North Carolina at Charlotte
University of North Carolina at Charlotte University of North Carolina at Charlotte Charlotte, USA
Charlotte, USA Charlotte, USA [email protected]
[email protected] [email protected]

Abstract— Due to a growing power demand and cyclic energy TA Room air temperature [oF]
consumption patterns, thermostatically controlled loads (TCLs) Tcg Battery charging period [hours]
in residential buildings have become major candidates for To Outdoor air temperature [oF]
residential demand response (DR) programs. However, in recent TM Temperature of solid mass [oF]
times, there has been an increased interest in the usage of
tUA Water heater tank thermal coefficient [Btu/°F·h]
distributed energy storage, such as batteries, for grid support
and energy management applications including peak shaving and tSA Water heater tank surface area [ft²]
frequency regulation, which overlaps with typical objectives of UA Building envelope thermal conductance [Btu/ oF]
TCL-centric residential DR programs. In this paper, we present π Electricity price [$/kWh]
a comparison between peak shaving using TCLs and distributed
battery energy storage. A detailed model of 100 residential II. INTRODUCTION
homes, developed using GridLAB-D, is used to simulate demand Electricity has been a major driver of growth and
response strategies for the TCLs as well as battery charging and development in many nations of the world. This explains why
discharging algorithms to achieve peak load reduction while the demand for electricity has increased significantly over the
maintaining predefined customer comfort levels. Different years. For the most part, this increase in demand has translated
customer enrolment scenarios are also considered for more into strategic expansion of electric utility companies. However,
realistic results. An economic analysis of both solutions is also
there are still situations where it is impractical to meet all of
presented. The results obtained show that while TCLs alone may
not be able to realistically eliminate the most expensive peaking
this demand due to limitations in the available capacity of
plants, they can still provide cheaper peak load reductions when electricity generation and transmission facilities and the capital
compared with distributed battery energy storage units, with the intensive nature of expanding these capacities. These
reductions increasing as more customers are enrolled. limitations, among others, were key drivers behind the
development of a load and generation balance methodology
Keywords—Demand response, thermostatically controlled loads, called demand response (DR).
distributed battery energy storage, peak shaving.
In the United States, the residential sector accounts for the
I. NOMENCLATURE highest percentage of total electricity consumption. In 2016,
residential customers consumed about 37.5% of the total
CA Air heat capacity [Btu/ oF] electricity produced in the US [1]. For these residential
Ce Annualized energy capacity cost [$/MWh] customers, about 53% of their electricity consumption is
CM Mass heat capacity [Btu/ oF] attributed to thermostatically controlled loads i.e. HVACs,
CNPC Net present cost [$] water heaters and refrigerators [2]. Although demand response
Cp Annualized power capacity cost [$/MW] programs in the US are directed at all categories of electricity
CRF Capital recovery factor consumers (residential, commercial and industrial), industrial
D Battery usage days loads, which consume the least amount of electricity [1], still
E Battery roundtrip efficiency [%] contribute the highest amount of peak demand savings. In fact,
Hm Mass thermal conductance [Btu/ oF] the peak demand savings from industrial demand response
i Interest rate [%] resources in 2015 was more than twice the peak demand
N Project lifetime [years] savings from residential demand resources participating in
P Annual peak reduction [kW] retail demand response programs [1]. This suggests that it is
QA Air heat flux [Btu/hr] easier to access industrial loads for demand response purposes.
QM Mass heat flux [Btu/hr] However, it can also be inferred that while there are still some
challenges to overcome, the residential sector and by
R Thermal resistance [ft²·°F·h/Btu]
extension, thermostatically controlled loads, still possesses

978-1-7281-0137-8/19/$31.00 ©2019 IEEE

Authorized licensed use limited to: University College London. Downloaded on July 05,2020 at 10:13:22 UTC from IEEE Xplore. Restrictions apply.
harnessable demand response potentials for cost effective The house model in GridLAB-D provides the possibility of
management and operation of the US power grid. specifying a myriad of parameters to capture the thermal
dynamics of a residential building and the desired
On another note, battery energy storage technologies have characteristics of each appliance within the building. However,
received increased attention over the past several years due to the water heaters and HVAC units are the main focus of this
decreasing costs and different potential value streams. work as the demand response program being investigated is
Different researchers have studied various grid level value directed at these two appliances. Other appliances were
streams of battery energy storage using both singular and modeled as ZIP loads with schedules. A more detailed
cascaded approaches. Ref. [3] presents an analysis of battery description of the model is provided below.
energy storage in a load leveling application and [4] discusses
the results of a study aimed at determining the extent to which A. HVAC System
batteries can be used to reduce renewable resource curtailment. Although many parameters can be specified for a detailed
Also, [5] presents an approach for peak load shaving using representation of an HVAC system in GridLAB-D, the major
battery storage and [6] considers data center-owned battery parameters influencing the thermal mass and heat flow path are
storage systems for combined peak shaving and frequency the floor area, thermal integrity level and the coefficient of
regulation applications. performance (COP) of the HVAC system. The thermal
While most of the aforementioned works focus on large- integrity level captures the thermal resistance and the air
scale battery storage systems, others postulate that the bulk of exchange rates which are functions of the thermal resistance of
future battery storage capacity will be from smaller distributed the building [12]. The COP is a measure of the efficiency of
battery storage systems owned by individual customers [7]. the HVAC system. The heating and cooling setpoints were also
This suggests that electric utilities might consider investing in specified. To achieve load diversity in the model, each of these
distributed battery storage thereby reducing their investment parameters were randomized based on data obtained from
costs via economies of scale and reaping other benefits surveys conducted by the authors and information from real
associated with the proximity of these batteries to the end estate websites about typical characteristics of homes in
users. However, before these investments can be made, the Charlotte, NC [13]. A summary of the random distributions
potential of existing demand resources to provide these same used is given in Table I.
value streams should be considered. Hence, this work presents B. Water Heater
a comparison between using residential thermostatically
controlled loads (TCLs) and distributed battery storage units The water heaters were sized using a family size of four
for the peak shaving application. adults. The specifications of the water heater including heating
element capacity, tank volume, tank height and tank diameter
After investigating the generation mix of a local utility were obtained from the AO Smith residential water heater
company, an observation can be made that approximately 60 to product catalog [14]. The thermal coefficient, tUA, which is a
65% of the generation resources support the base load [8], measure of the heat retention capacity of the water heater, was
while 35 to 40% support the peak load, out of which the top
calculated using (1). The thermal resistance, R, was specified
20%, typically, natural gas combustion turbine units are the
as 12 ft²·°F·h/Btu which is the minimum storage water heater
most expensive [9]. Hence, 20% peak reduction is used as a
reference point for the DR programs being considered. thermal resistance value recommended by DOE [15].

III. RESIDENTIAL LOAD MODEL  tUA = tSA / R 

To compare peak load reductions using TCLs and
distributed battery energy storage-based DR programs, a The hot water consumption profile used in the model is
detailed residential load model was developed for 100 homes based on the ASHRAE standard 90.2 which provides a
using GridLAB-D. The residential house model in GridLAB-D reliable typical hot water consumption profile [16]. The
is based on the simplified equivalent thermal parameters (ETP) profile is shown in Fig. 2. For each home, the hourly hot water
model which represents the thermal dynamics of a building consumption was then calculated by multiplying the ASHRAE
using a series-parallel combination of passive elements profile values with normally distributed random numbers.
analogous to an electric circuit as shown in Fig. 1 [10]. This
simplified model is suitable for modelling and simulating TABLE I. HVAC SYSTEM PARAMETERS
residential buildings and small commercial buildings [11].
Parameter Range

Floor area 1000 to 1500 ft2

Heating setpoint 68 to 70oF

Cooling setpoint 71 to 73oF

Thermal integrity level NORMAL

Fig. 1. ETP model of a typical home in GridLAB-D

Authors gratefully acknowledge funding support received from the

Energy Production & Infrastructure Center at the University of North Carolina
at Charlotte.

Authorized licensed use limited to: University College London. Downloaded on July 05,2020 at 10:13:22 UTC from IEEE Xplore. Restrictions apply.
 Fig. 3. Aggregated HVAC consumption and outdoor temperature profile
Fig. 2. Typical 24-hour residential hot water consumption profile
The water heaters are curtailed for a maximum of 3 hours
C. Other Loads to further reduce the peak demand and limit the rebound effect
The other residential loads were modeled using the basic occurring as a result of the recovery of the HVAC units after
ZIP load model with schedules based on residential the peaking period. The 3-hour duration falls within the 2 to 4-
consumption data provided by the United States Office of hour curtailment period typically used in water heater DR
Energy Efficiency and Renewable Energy (EERE) which is programs e.g. the Wiregrass Electric Cooperative, which is also
available on the OpenEI website. Again, parameter a southeastern electric utility, curtails water heaters for 2 to 4
randomization was used to achieve load diversity in the hours in its H2O+ program [19].
aggregated model. V. DEMAND RESPONSE STRATEGY FOR DBES
The residential load model was designed to reflect typical
conditions in Charlotte, NC. Hence, weather files for Charlotte For peak shaving with distributed battery energy storage
obtained from the NREL database were used and where (DBES), lithium-ion batteries are considered as this battery
possible, model parameters were selected to reflect typical technology is fast becoming the dominant storage technology
for residential applications. Also, the specifications of the Tesla
conditions in Charlotte [17].
Powerwall 2 [20] are used, as this is a commercially available
To verify the performance of the model, a typical hot
battery designed for distributed storage applications. It is
summer day (27th July) was simulated, and the aggregated assumed that the batteries are completely owned by the electric
load profile of the HVAC units was compared with the actual utility and will be used for peak shaving purposes. The
temperature profile for the same day. This test is deemed batteries are simply charged at night with cheaper electricity,
sufficient as the other components of the model are based on and then discharged during the peak period occurring later in
data from veracious sources. As depicted in Fig. 3, the results the day to achieve a 20% peak reduction. The base load profile
obtained show that as the temperature increased, the HVAC (section VI) is used to determine the discharge duration
units consumed more energy to maintain the required indoor required to achieve the 20% peak reduction goal.
temperature in each of the homes. Also, as the outdoor
temperature decreased, the energy consumption of the HVAC VI. SIMULATION RESULTS AND DISCUSSIONS
units decreased as expected. A. Base Line Consumption
IV. DEMAND RESPONSE STRATEGY FOR TCLS To compare the peak shaving effects of DR and energy
storage, a typical peak summer day was simulated without
As mentioned earlier, the DR program being considered TCL or distributed battery storage demand response. The
focuses on HVAC and water heating units as these appliances resulting load profile, shown in Fig. 4, is taken as the base
make up the majority of residential electricity consumption. It profile. It should be noted that the authors are aware that Duke
is assumed that typical heating and cooling setpoints are known Energy Carolinas, which is the electric utility serving the
for each customer participating in the DR program. During Charlotte area, has recently filed an Integrated Resource Plan
peak events (with a maximum of 4 hours), the participating which designates its system as winter peaking [21]. However,
customer’s typical cooling setpoint is offset by 6oF which the data used in the model are from periods with summer
complies with ASHRAE standard 55-2010, which specifies peaks. Hence, the summer peak assumption is still valid in this
that indoor temperature changes in residential buildings should context. From the base profile, the peak period is assumed to
not exceed 6oF within a 4-hour time window [18]. This be between 5 pm and 9 pm, with the absolute peak occurring at
guarantees that the customer’s indoor temperature does not go 6 pm.
beyond a predefined comfort limit.

Authorized licensed use limited to: University College London. Downloaded on July 05,2020 at 10:13:22 UTC from IEEE Xplore. Restrictions apply.
B. Demand Response using TCLs It is clear that within the peak period (5 pm to 9 pm), the
Three customer enrolment levels were considered: 25%, demand was significantly reduced, with the 75% scenario
50% and 75%. For each of these enrolment scenarios, the water recording the highest reductions as expected. However, after
heaters and HVAC units were totally curtailed to get a sense of the peak period, the rebound effect created significant peaks at
the maximum peak load reductions achievable. The results 11 pm for 50% and 75% enrolment scenarios, with the 75%
obtained are shown in Fig. 5 and Table II. peak increasing beyond the base profile peak.
The temperature profile for a particular home under the
total curtailment scenario is shown in Fig. 6. From this figure,
it can be seen that the temperature change is beyond the 6oF
specified in the ASHRAE standard 55-2010.
Since it is often unacceptable to totally curtail HVAC units
of residential customers for the entire peak period, the demand
response strategy discussed earlier was then applied. Again,
the three customer enrolment scenarios were simulated to
observe how much of the 20% peak load capacity could be
reduced during each of these enrolment scenarios. The load
profiles obtained are shown in Fig 7. The percentage
reductions in peak demand and energy consumption are also
presented in Table III.

From Fig. 7, it may be observed that for 25% customer

enrolment scenario, the system peak still occurred at 6 pm.
However, for the 50% enrolment scenario the peak shifted to 7
Fig. 4. Base load profile of a typical home on summer day. pm. For 75% enrolment, the system peak occurred at 5 pm
which is the same time at which the demand response event
started. This implies that during the peak period, the system
peak demand was significantly reduced. Hence, the 75%
scenario is selected for comparison with the distributed battery
energy storage scenario.

To examine the impact of the DR program on the

customers, the simulated indoor air temperature profile of one
of the homes was recorded, and is shown in Fig 8. For the
selected home, the typical cooling setpoint is 73oF. This
implies that during the peak period, the indoor air temperature
of the home should not exceed 79oF. Fig 8 shows that this
condition is satisfied.

Fig. 5. Total curtailment of HVACs and water heaters

TABLE II. PEAK DEMAND AND ENERGY REDUCTIONS FOR DIFFERENT

COMPLETE CURTAILMENT SCENARIOS
Peak Demand
Peak Demand
Reduction
Scenario Reduction
(Peak Period
(Including Rebound)
Only)
25% Enrolment 14.6% 14.6%

50% Enrolment 22% 10.3%

75% Enrolment 27% -2%

Fig. 6. Temperature profile under total curtialment

Authorized licensed use limited to: University College London. Downloaded on July 05,2020 at 10:13:22 UTC from IEEE Xplore. Restrictions apply.
 demand from 1 am to 5 am, as shown in Fig. 9, represents the
charging duration of the batteries. After the charging period,
the demand returned to the same level as the base profile and
remained equal to the base profile till 4 pm. From 4 pm to 9
pm, the demand level was significantly reduced due to the
discharging of the batteries. Table IV shows that a peak
demand reduction of 19.7% was achieved.
Also, since the batteries are providing additional energy
during the peak period, the comfort levels of the customers are
not affected.

Fig. 7. Load shapes with 6of setpoint offset

TABLE III. PEAK DEMAND AND ENERGY REDUCTIONS FOR DIFFERENT

DR SCENARIOS

Peak Demand Peak Demand

Scenario
Reduction (kW) Reduction (%)
25% Enrolment 35.05 kW 8.9%

50% Enrolment 53.53 kW 13.5%

75% Enrolment 64.82 kW 16.4%

Fig. 9. DR using distributed battery energy storage

TABLE IV. PEAK DEMAND REDUCTION WITH BATTERY ENERGY

STORAGE
Max Demand
Max Demand
During Peak Peak Period
During Peak
Scenario Period after Demand
period in Base
Deploying Reduction
Case (KW)
Storage (KW)

25% homes with

395.82 317.98 19.7%
Storage

D. Economic Analysis
In this section, we determine the Levelized Cost Of
Electricity (LCOE) for the two cases: TCL-based DR, and
DBES-based load leveling.
Fig. 8. Temperature profile under DR program
a) LCOE for TCL-based DR: The LCOE for the TCL-
C. Demand Response using DBES based DR program over a 10-year period was calculated using
From the base load profile, it is evident that the demand is (2). The interest rate, i, was taken as 5% and P was taken as
greater than 80% of the peak for a period of 6 hours, from 4 pm 64.82 kW (from Table III). It is assumed that smart
to 10 pm. Hence, the batteries were discharged during this thermostats and load control switches for the HVACs and
period to achieve the 20% peak reduction. From the Powerwall water heaters respectively are freely given to participating
datasheet, the rated depth of discharge is 100%, which provides customers. The customer compensations were based on Duke
a justification for using an initial state of charge of 0% for the Energy’s Power Manager program rates for North Carolina,
batteries in the simulation. which includes an annual bill credit of $32 spread over 4
months (July to October) [22].
Assuming that 25% of the homes were fitted with the
batteries, we determined the peak load reduction. The results
 LCOE = CRF (i, N) * CNPC / P 
obtained are shown in Fig. 9 and Table IV. The increase in

Authorized licensed use limited to: University College London. Downloaded on July 05,2020 at 10:13:22 UTC from IEEE Xplore. Restrictions apply.
 The calculated LCOE is equal to $85/kW-year. If the DR Systems," 2013 North American Power Symposium (NAPS),
Manhattan, KS, 2013, pp. 1-5.
resources are used for 300 hours annually, LCOE becomes
[6] Y. Shi, B. Xu, D. Wang and B. Zhang, "Using Battery Storage for Peak
$0.28$/kWh. If they are used for 4 hours daily from July to Shaving and Frequency Regulation: Joint Optimization for Superlinear
October annually (equivalent to 492 hours), then the LCOE Gains," in IEEE Transactions on Power Systems, vol. 33, no. 3, pp.
becomes $0.17/kWh. This implies that the DR cost reduces as 2882-2894, May 2018.
the total demand response hours per year increases. [7] K. Worthmann, C. M. Kellett, P. Braun, L. Grüne, S. R. Weller,
"Distributed and decentralized control of residential energy systems
incorporating battery storage", IEEE Trans. Smart Grid, vol. 6, no. 4, pp.
b) LCOE for Distributed Storage: The LCOE for energy 1914-1923, Jul. 2015.
storage was calculated using the methodology presented in [8] Energy Information Administration, “Levelized cost and levelized
[23]. Equations (3) and (4) represent the formulae used for the avoided costs of new generation resources in the annual energy outlook
2018,” March 2018. EIA. [Online]. Available:
LCOE calculation. https://www.eia.gov/outlooks/aeo/pdf/electricity_generation.pdf.
[9] Duke Energy Corporation, “2016 Duke Energy Carolinas Integrated
 LCOE = [(Cp + Ce*Tcg) + (π*Tcg*D)] / [E*Tcg*D]  Resource Plan,” Sept. 2016. [Online]. Available:
http://www.energy.sc.gov/files/view/DEC%20IRP%202016%20Correct
ed%2010-2016%20Clean%20Copy.pdf
 Annualized Costs, Ce = CRF (i, N) * CNPC  
[10] W. Zhang, J. Lian, C. Chang and K. Kalsi, "Aggregated modeling and
control of air conditioning loads for demand response," 2014 IEEE PES
Cp and Ce were based on the price of Tesla Powerwall with General Meeting | Conference & Exposition, National Harbor, MD,
operation and maintenance costs set at 2% of the initial capital 2014, pp. 1-1.
costs. E was taken as 90%, and π was set at $0.087/KWh - [11] N. Lu, "An Evaluation of the HVAC Load Potential for Providing Load
Duke Energy Carolinas’ billing rate for residential customers Balancing Service," in IEEE Transactions on Smart Grid, vol. 3, no. 3,
pp. 1263-1270, Sept. 2012.
[24]. Tcg and D were taken as 4 hours and 365 days
[12] P. Gotseff and B. Lundstrom, "Data-Driven Residential Load Modeling
respectively. The calculated LCOE for DBES is equal to and Validation in GridLAB-D," 2017 Ninth Annual IEEE Green
$0.389/KWh. Technologies Conference (GreenTech), Denver, CO, 2017, pp. 20-25.
[13] Rent Café, “Charlote, NC rental market trends.” [Online]. Available:
https://www.rentcafe.com/average-rent-market-trends/us/nc/charlotte/.
VII. CONCLUSIONS AND FUTURE WORK [Accessed: Jan. 07, 2019].
TCLs and distributed battery energy storage have both been [14] A. O. Smith Corporation, “Residential product catalog.” [Online].
Available: https://www.hotwater.com/lit/catalogs/aosrb80000.pdf
used for peak shaving. While HVACs and water heaters alone
[15] U. S. Department of Energy, “Storage water heaters.” [Online].
may not be able to realistically shave off 20% of peak demand Available: https://www.energy.gov/energysaver/water-heating/storage-
and consequently replace the most expensive peaking plants, water-heaters. [Accessed: Jan. 07, 2019].
they can still provide cheaper peak shaving when compared [16] A. Gholizadeh and V. Aravinthan, "Benefit assessment of water-heater
with distributed battery energy storage. Also, if the DR management on residential demand response: An event driven
resources are used more often, the costs become lower and the approach," 2016 North American Power Symposium (NAPS), Denver,
electric utility can benefit more from its DR investments while CO, 2016, pp. 1-6.
the customers can further reduce their energy costs. However, [17] NREL, “National solar radiaition database.” [Online]. Available:
https://rredc.nrel.gov/solar/old_data/nsrdb/1991-2005/tmy3/.
the value streams provided by battery energy storage are often
[18] Thermal Environmental Conditions for Human Occupancy. ASHRAE,
cascaded to justify investments. Hence, the insights derived pp. 55-2010, 2010.
from this work will serve as a foundation for combining
[19] Wiregrass Electric Cooperative, “H2O+ program.” [Online]. Available:
different value streams of battery energy storage and assessing https://www.rentcafe.com/average-rent-market-trends/us/nc/charlotte/
the potentials of TCLs to produce these same combined value [Accessed: Jan. 07, 2019].
streams. [20] Tesla Powerwall, [online]. Available: https://www.tesla.com/powerwall,
[Accessed: Jan. 09, 2019].
REFERENCES [21] J. Wilson, “Seasonal electric demand in southeastern United States,”
[1] Energy Information Administration, “Electric power annual 2017,” March 2017. [Online]. Available: https://www.cleanenergy.org/wp-
October 2018, EIA. [Online]. Available: content/uploads/Seasonal-Electric-Demand-in-SE-SACE-March-
https://www.eia.gov/electricity/annual/pdf/epa.pdf 2017.pdf
[2] Energy Information Administration, “2015 Residential energy [22] Duke Energy Corporation, “Power manager.” [Online]. Available:
consumption survey,” May 2018, EIA. [Online]. Available: https://www.duke-
https://www.eia.gov/consumption/residential energy.com/home/products/power%20manager?jur=NC01. [Accessed:
[3] R. J. Kerestes, G. F. Reed and A. R. Sparacino, "Economic analysis of Jan. 14, 2019].
grid level energy storage for the application of load leveling," 2012 [23] Lotfi, Hossein & Majzoobi, Alireza & Khodaei, Amin & Bahramirad, S
IEEE Power and Energy Society General Meeting, San Diego, CA, & A Paaso, E, “Levelized Cost of Energy Calculation for Energy
2012, pp. 1-9. Storage Systems”, Grid of the Future Symposium (CIGRE),
[4] C. Root, H. Presume, D. Proudfoot, L. Willis and R. Masiello, "Using Philadelphia, PA. 2016.
battery energy storage to reduce renewable resource curtailment," 2017 [24] Duke Energy Carolinas Bill Rates, [Online]. Available:
IEEE Power & Energy Society Innovative Smart Grid Technologies https://www.duke-energy.com/home/billing/rates, [Accessed: Jan 10,
Conference (ISGT), Washington, DC, 2017, pp. 1-5. 2019].
[5] A. Rahimi, M. Zarghami, M. Vaziri and S. Vadhva, "A simple and
effective approach for peak load shaving using Battery Storage

Authorized licensed use limited to: University College London. Downloaded on July 05,2020 at 10:13:22 UTC from IEEE Xplore. Restrictions apply.