Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.

edu/electricity 1

Consumer Strategies for Controlling Electric

Water Heaters under Dynamic Pricing
Chong Hock K. Goh and Jay Apt

Abstract—Electricity used to heat water represents 9% of

residential demand in the USA and can be 40% in other countries.
Hourly residential use of hot water is often anti-coincident with
the peak generation of electricity, presenting an opportunity for
reducing consumer costs under dynamic pricing during the
afternoon generation peak. We have examined the effects of three
strategies on customer costs under dynamic pricing: timed power
interruption (long used by certain utilities), a price-sensitive
thermostat, and a double period setback timer. Systems which
lower the water temperature set points are as economical as
power interruption systems, and result in higher minimum water
temperature. Our model predicts that a setback thermostat will
keep the tank water warmer than a load interruption timer with
very similar electricity use. The setback thermostat and the more
complex price-sensitive thermostat achieve similar water
temperatures and consumer savings. Fig. 1. Hourly USA residential hot water usage (stepped curve, per [20]) and
representative electric load curve for all sectors (smooth curve, from California
ISO, July 16, 2004).

Index Terms—Economics, Home appliances Power systems, The average household hot water draw at the 4 PM (hour 16)
Power system economics, Power system planning, Power system peak of overall electric demand is one-third of that at the 8-9
operations, Power system reliability.
AM peak of hot water use. Reference [7] presents data from
one U.S. utility on hourly electricity energy use for residential
I. INTRODUCTION
water heaters in both summer and winter. Reference [6] gives

D EMAND for electricity reaches a peak near 4 PM in many

control areas, such as the California ISO, whose load
curve for a July day is shown in Fig. 1. The peak demand often
similar curves for South Africa. Since hot water use is low in
the afternoon, residential hot water use has been identified as a
candidate for electricity peak reduction based on pricing that
can be 25% above the average demand during the day and 70% reflects the high hourly costs of delivering electricity at peak
above the minimum [1]. Reducing the peak load can use times [8]. In addition, since thermal loss causes the
significantly reduce consumer costs when the consumer is thermostat to command the heating element on (a heater cycle)
being charged for the true cost of peak power. once or twice during the low-use period after midnight, minor
Heating of residential water by the 41 million electric water energy savings may be achieved by heater control during this
heaters in the United States out of 107 million total residential period.
water heaters [2] is responsible for 8.7% of residential Detroit Edison began in 1934 using timers to interrupt power
electricity use [3]. The largest reported residential load fraction to residential water heaters for four continuous hours per day,
is 40%, in South Africa [4]. Periods of peak use of water occur selected to coincide with the daily electric system load peak
in the morning and early evening (see Fig. 1, a similar profile [9]. In 1968, the timers were replaced by FM radio control,
measured over a 4-month period by [5], and the 15,000-home with 200,000 heaters switched in ten blocks under control of
data of [6]). the system operator (still constrained by tariff to a maximum of
four hours off time per day). 95.5% of residential water heater
This work was supported in part by the Alfred P. Sloan Foundation and the customers chose to participate, in exchange for a 26% water
Electric Power Research Institute under grants to the Carnegie Mellon
Electricity Industry Center. heater rate reduction. System costs for the radio control system
C. H. K. Goh is with the School of Computer Science, Carnegie Mellon averaged $275 per customer (2005 dollars). After switchover
University, Pittsburgh, PA 15213 USA (e-mail: kelvingoh@cmu.edu). to the radio control system, the utility found that they
J. Apt is with the Department of Engineering and Public Policy and the
Tepper School of Business, Carnegie Mellon University, Pittsburgh, PA 15213
commanded interruptions most frequently to reduce the use of
USA (corresponding author phone: 412-268-3003; fax 412-268-7357; e-mail: oil-fired peak generators or imported power. Utility cost of
apt@cmu.edu). electricity for heating the water after the heaters were re-
Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.edu/electricity 2

enabled were generally two-thirds of the peak costs. resistance water heaters, suggest that we do not currently even
Wisconsin Electric Power Company installed power line enjoy the conservation effects which market forces would
carrier load management of water heaters in Milwaukee and provide.”
surrounding areas beginning in 1979 [10]. By 1985, 100,000 Most previous work on water heater electricity use reduction
heaters were in the program. The load management rate tariff has been undertaken from the point of view of the utility.
permitted heaters to be switched off for up to eight hours per However, consumers have adopted devices such as setback
day. In practice, heaters were de-activated for no more than thermostats for heating and air conditioning as a means of
three consecutive hours or four total hours in one day. Peak reducing energy bills even without tariff incentives or dynamic
load reduction of 2% was achieved, with oil-fired peakers pricing. Here we examine the potential savings to the
being replaced by gas shoulder plants and coal baseload units. residential consumer from both price-sensitive thermostats and
The increase in system load when the heaters were re-enabled setback thermostats for residential electric water heating,
was generally three times the average shed load, but was finding that similar savings can be realized by both methods.
managed by staggering groups of heaters. Although both methods are feasible, implementation of the
Carolina Power & Light Company conducted an experiment setback thermostat method requires fewer infrastructure
interrupting power to 200 residential water heaters as a load changes, and presents an near-term attractive opportunity for
control mechanism in the winters of 1979-80 and 1980-81 [11]. adoption by manufacturers and customers.
Using this field data, they modeled the effects of controlling
200,000 residential heaters, finding that peak system demand II. MODEL
could be reduced by 3%. A significant spike in load was In order to investigate these strategies, we have developed
observed after the heaters were switched back on; optimal and verified a simple model of a residential electric water
control strategies involved staggered control of groups of heater (Fig. 2).
heaters. In this model, the temperature of the mass of water increases
Florida Power & Light uses power line communication in a or decreases evenly without incorporating thermal layers.
load control program for 712,000 customers. FPL currently Commercial electric water heaters have 2 heating elements
pays residential customers monthly incentives of $3.50 for (one near the top and the other at the bottom, with only 1
controlling water heaters and attributes 1 GW of peak demand heating element active at a time); this model heater has only a
reduction to the load control program [12]. single heating element.
If water heater loads are simply cut off at particular periods, Reference [16] describes a more complex simulation,
the utility may experience system stability issues. This was a incorporating two heating elements and heat transfer among six
subject of research in the early 1980s, as reviewed in [5]. Here layers of water within the tank. As we discuss below, our
we compare this technique to two others: lowering the set-point results agree closely with theirs. We also have run our model
of the water heater at specific times (as in a HVAC thermostat) using the U.S. Department of Energy EnergyGuide test
and changing the set-point in response to dynamic prices via a procedure for commercial water heaters [17]. Using that test
price-sensitive thermostat communicating with the load serving protocol, we calculated the total energy used per year, and
entity. compared these results to the EnergyGuide ratings of two
In a recent report, the California Energy Commission commercial 50-gallon water heaters, and two commercial 80-
Demand Response Committee [8] estimated that dynamic gallon water heaters. As discussed in the section below
pricing (real-time pricing or time-of-use pricing are forms of describing model validation, our results showed agreement to
dynamic pricing based on the real-time market prices) could within 0.6 to 4.3 percent for several validation tests.
achieve “short-term peak reduction …between 4.7 and 24
percent of California’s estimated peak load by 2013. The
residential and small commercial customer share of these
estimated peak savings range from roughly 15 to 25 percent
with balance coming from medium to large commercial and
industrial customers. The long-term peak reduction is estimated
to be 3.4 to 15 percent of the projected 2013 peak load.”
A few U.S. utilities have used residential real-time pricing,
notably GPU (beginning in 1997), AEP, and Gulf Power. In
the GPU pilot program, residential summer peak use was 2 kW
in a control group and 1.5 kW in the participating group, with
larger peak shaving during “critical price” events [13], [14].
California’s Statewide Pricing Pilot (SPP) program tested
several forms of real-time pricing in 2003 and 2004. A strong
argument for dynamic pricing is provided in [15]: “consumer Fig. 2. Water heater model
underestimates of hot-water cost, especially for electric
Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.edu/electricity 3

The temperature change of the water is governed by [18], [19]: temperature of the tank’s water 135°F, thermostat high cutoff
dT temperature 140°F, and low cutoff temperature 130°F.
M ⋅ SHw ⋅ = UA ⋅ (Tamb − T ) + Q (1) We compared our results modeling four commercial heaters
dt
sold by Sears to their published EnergyGuide energy use. All
where
commercial models studied are equipped with two 5500 watt
M = mass of water in the tank (lb)
heating elements, with only one in use at a time; our model
SHw = specific heat of water (BTU/lb/°F)
used a single 5500 watt element. The first heater (Kenmore
T = Temperature of the water in the tank (°F)
model 32756, Sears item 04232756000) has 2.5 inch thick
t = time (hr)
insulation and a 50 gallon capacity. The EnergyGuide test
UA = standby heat loss coefficient • area of the storage tank
energy usage is reported as 4879 kWh/year. Running our
(BTU/°F/hr) (see [17] p. 26014 for definition)
simulation program with the same parameters and procedures
Tamb = Ambient temperature (°F)
used in the Department of Energy test protocol, the energy
Q = rate of heat input to tank from the heater (BTU/hr); zero
usage by our simulated heater is 4912 kWh/year, within 0.7%
when the heater is off.
of the published rating. The only parameter adjusted to achieve
The modeled heater has insulation which affects the UA of
a match was the heat loss factor; all others were set at the
the heater. The UA factor governs the rate at which the water
midrange of the test protocol values as listed above. The best
cools. Commercially available heaters with 2 inch insulation
match (values reported above) was with a heat loss factor 3.6
have an approximate UA factor of 4, and heaters with 3 inch
Btu/°F/hr. Using the same heat loss factor, we simulated a
insulation have an approximate UA factor of 2.5. The UA
similar 80 gallon model with 2.5 inch insulation (Kenmore
values are not directly proportional to the inverse of the
model 32986, Sears item 04232986000); the EnergyGuide
insulation thickness due to end effects and penetrations.
usage is 4721 kWh/year, while our simulated heater’s usage is
As heated water leaves the tank, the tank is refilled with
4750 kWh/year (within 0.6% of the reported rating).
water from the cold water inlet supply. The temperature of the
We also tested the model against two higher efficiency Sears
water after a withdrawal of hot water and addition of inlet
models with 3 inch insulation. We used the same parameters as
water is:
above, except that the heat loss factor was set to 2.5 Btu/°F/hr
Tcurr ⋅ Mcurr + Tinlet ⋅ Minlet
Tnew = (2) to model the thicker insulation. For the 50 gallon Kenmore
Mcurr + Minlet model 32154 (Sears item 04232154000), the reported
where EnergyGuide usage is 4622 kWh/year. Our simulated heater’s
Tnew = temperature of water after inlet water is added to wattage usage was 4423 kWh/year (4.3%). For the similar 80
current water in the tank gallon model with 3 inch insulation (Kenmore model 32184,
Tcurr = current temperature of the water (after water draw) Sears item 04232184000), the reported EnergyGuide usage is
Tinlet = cold water inlet temperature also given as 4622 kWh/year. Our simulation (with the
Minlet = mass of water from cold water inlet parameters unchanged except for the water volume) gave 4708
Mcurr = mass of water left in the tank (after water draw). kWh/year, within 1.9% of the reported rating. These
The rated tank volume for some commercial hot water comparisons are summarized in Table I.
heaters is somewhat larger than the measured volume. We have
used the rated volume in this work. TABLE I
COMPARISON OF SIMULATED AND MEASURED ENERGYGUIDE VALUES
The model was run with a time step of 0.01 hour, for a
Sears Kenmore Sears Simulation Difference
specified length of time (normally 10 days). For the 24-hour
Product Kenmore Result (b) (b-a)/a
EnergyGuide test profile validation runs discussed below, the
EnergyGuid (kWhr/yr)
time step was decreased to 0.001 hour. The ASHRAE water
e Values (a)
use profile [20] (or the EnergyGuide test profile [17]) was (kWhr/yr)
consulted to determine the withdrawal at each time step. When
Model 32756 4879 4912 + 0.7%
the heater was off, the temperature of the tank was calculated
(50 gallon, 2.5”
with equations 1 (with Q = 0) and 2. If the temperature was
insulation)
below the lower thermostat set point, the heater was switched
Model 32986 4721 4750 + 0.6%
on, and the temperature increased per equation 1 until a time
(80 gallon, 2.5”
step when temperature reached the upper set point and the
insulation)
heater was switched off.
Model 32154 4622 4423 - 4.3%
III. MODEL VALIDATION TESTS (50 gallon, 3”
insulation)
We performed two types of validation testing on the model. Model 32184 4622 4708 + 1.9%
We first ran the test profile used in the EnergyGuide tests for (80 gallon, 3”
electric water heaters [17]. The parameters used were: ambient insulation)
temperature 67.5°F, inlet water temperature 60°F, initial
Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.edu/electricity 4

Fig. 3 shows the temperature of the water in the tank during our simulations) as in Fig. 6. This is similar to double setback
a simulation of the 24-hour EnergyGuide test for the 50-gallon thermostats routinely used for residential heating and air
model with 2.5 inch insulation. The test protocol calls for six conditioning control.
water draws at hourly intervals at the start of the test period, (4) The limits are controlled in response to the price of
after which the heater is allowed to maintain temperature electricity. The controller is given the range of daily price
within its temperature control band without water draws. variation, and adjusts the lower limit so that it reaches its
minimum value when the price is at a maximum. We have
modeled a dynamic price sensitive thermostat by varying the
target water temperature (Ttarget) with the price of electricity
(P), and fixing the deadband (difference between the high and
low setpoints), as shown in Fig. 7:

P − Pmin
Tt arg et = Tmax − ( Pmax − Pmin ) × (Tmax − Tmin ) (3)

where Tmax is the desired water temperature at low electricity

price (for example, 120 °F), Tmin is the minimum temperature
the consumer will accept (100 °F in our work here), and Pmax
and Pmin are the maximum and minimum electricity prices
during the day.
Fig. 3. Simulated 24 hour EnergyGuide test. The solid line is the modeled We used a summer Massachusetts wholesale price schedule,
temperature of the water in the heater tank, which decreases during the hourly scaled to give the 2002 average retail price for the 48
water draws mandated in the first six hours of the test. Water temperature also
decreases due to thermal leakage. The heater is switched on when the lower set
contiguous states of 8.41 ¢/kWh. The maximum scaled price is
point (130 °F) is reached, and off at the upper set point (140). Simulation test 12.39 ¢/kWh at 4-6 PM, while the minimum is 6.34 ¢/kWh at
conditions: Kenmore model 32756, 50 gallon capacity, 2.5" insulation, ambient 4-5 AM. The model calculates the cost of electricity at each
temperature 67.5 °F, inlet water temperature 60 °F, 5500 watt heater, initial time increment, noting whether the heater is on (as controlled
tank water temperature 135 °F, heater cut-in temperature 130 °F, heater cut-out
by the thermostat) and the price at that time (selected either as
temperature 140 °F, heat loss factor (UA) 3.6 BTU/ °F/ hr.
a flat rate or as a dynamic price).
Second, we used our model to calculate thermal efficiency as For consistency, we adopted a 5500 watt heater element, a ±
defined in [16], using their operations schedule A and constant 2.5 degree temperature control band, and the EnergyGuide test
volume hot water draws. Simulating their 82 gallon “Tank A” protocol parameters for ambient temperature (67.5 °F) and inlet
and using parameters given in their Table 4, we calculate a water temperature (60 °F). We note that inlet water
thermal efficiency of 0.801, in agreement to within 2.6% of temperature has a strong influence on the energy used, and that
their value of 0.822. opportunities for savings may exist in regions where the cold
Comparison of our model’s results to the EnergyGuide water pipes are not buried to a depth which minimizes seasonal
ratings and to the model of [16] show agreement to within the change. As an example, increasing the inlet temperature from
experimental variability noted in [21]. 45 to 55 °F decreases energy use by 12% in a simulation using
the residential hot water use profile of Fig. 1, the 50 gallon, 2.5
IV. MODELING FOUR STRATEGIES inch insulation model, and a 120 °F set point.
A conventional water heater thermostat has a lower (cut-in)
and upper (cut-out) limit, providing a control band around the
desired temperature. The model permits the selection of the
upper and lower limits, and allows investigation of the results
of three mechanisms by which those limits are controlled:
(1) No changes to the limits. The model heater maintains the
temperature of the water as in Fig. 4, with the frequency of
heater cycles affected by the ASHRAE schedule water draw.
(2) Power to the water heater is turned off at specific times.
Interruption control of this type has been used in most of the
load control programs discussed above. Fig. 5 shows the
results for the model heater when power is interrupted from six
hours, beginning at 1 PM.
(3) The limits are controlled by a two-period timer with Fig. 4. Conventional thermostat. water temperature, with high and low
selectable periods and selectable change to the lower limit (the thermostat limits shown on the left scale. hourly water use in gallons shown on
upper limit is a fixed offset above the lower, by 5 degrees in the right scale. High use causes frequent heater cycles.
Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.edu/electricity 5

V. RESULTS
Consumers may adopt strategies for reducing their electric
water heating cost which are rational under existing flat rate
tariffs. We quantified the effects of three of these: adjustment
of the conventional water heater thermostat, purchase of well-
insulated heaters or after-market insulation, and use of a
thermostat which lowers the set point. In areas which employ
dynamic pricing, consumers may wish to use setback timed
thermostats or closed-loop price-sensitive thermostats (which
change the set point in response to price signals received from
the load-serving entity). Most earlier work has considered
timed power interruption to the heater. This work examines set
point adjustment to a minimum level set by the consumer.
We have run simulations with both 120 and 140 °F set
points. Significant scalding risks exist at tap discharge
Fig. 5. Power is interrupted to the water heater from 1-7 PM. Same water use
profile as in Fig. 4.
temperatures greater than 120 °F [22], [23]. A few states
require set points in the 120-125 degree range. For example,
§704.06 of the Wisconsin State Code requires a residential
landlord to set water heater temperature no higher than 125 °F.
Our model runs show that for all four water heaters described
previously, setting the thermostat to 120 °F uses 75% of the
energy required to maintain a 140 °F set point. In high-price
states such as New York, the yearly savings would be $315,
greater than the purchase price of a 50 gallon electric water
heater. Most water heater thermostats are not calibrated, so
outlet temperature must be measured with a device such as a
household meat thermometer to adjust the set point. This
strategy is effective and can be important system-wide. The
lower temperature range does not greatly increase the risk of
bacterial growth; the OSHA Technical Manual notes that the
optimum growth range for Legionella pneumophila bacteria is
Fig. 6. Two-period setback thermostat. Same water use profile as in Fig. 4. 95-115 °F and that stagnant water with amoebae and protozoa
is generally required to promote growth [24].
Our simulation shows, as expected, that the cost of running
the water heater increases linearly with values of UA. The
simulation predicts that the difference in yearly operating cost
between an 50 gallon model with UA of 3.6 and one with UA
of 2.5 at the 8.41 ¢/kWh average residential price at a set point
of 140 °F is $21.90. The retail capital price difference between
the two models is $30. Many models are warranted for 12
years, so the extra insulation is a rational purchase with a 2
year pay back period.
As discussed previously, some utilities have interrupted
power to water heaters, and homeowners can purchase timers
to interrupt power to the heater. Multi period setback
thermostats are common in household heating and air
conditioning control systems; they change set points rather
than interrupting power. A similar control might be built into
Fig. 7. Dynamic price-sensitive thermostat. Hourly price shown on right scale.
Setpoints adjusted per equation (3). Same water use profile as in Fig. 4.
water heaters. Table II shows that the yearly savings of the two
strategies is predicted to be very similar, but the power
interruption device allows the water temperature to fall to
levels which consumers may find less desirable than a
minimum temperature that they select with a setback
thermostat.
The reduction of electricity use during peak periods by the
Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.edu/electricity 6

timer or setback thermostat can be important for electric system heating setback thermostat ($30), the payback period under a
operators. System margins of electric capacity above demand fixed tariff of 8.41 ¢/kWh would be one year.
are generally no more than 20%, and can be much less on hot Under a fixed-price tariff, load interruption timers or setback
summer afternoons or in specific areas. The estimate for the thermostats reduce average electricity used by the heater by 5-
2004 July capacity margin was 6.9% in the New York 8%, saving $30 annually using the U.S. average residential
Independent System Operator (ISO) region and 8% for the price. Our model predicts that a setback thermostat will keep
eastern Wisconsin and Illinois region covered by the Mid- the tank water 9 to 16 °F warmer than a load interruption timer
Atlantic Interconnected Network [25]. The achieved peak load with very similar electricity use. Both can have significant
reduction of 2% in the Wisconsin water heater program is a benefits for system operators, since the load reduction during
significant fraction of total reserve margin, and reductions of peak use time can be a significant fraction of system capacity
this order can be crucial when equipment fails. margin.
Dynamic pricing has been proposed as a means to reduce The introduction of two-way communication with the
peak load by having consumers pay the lower cost of running electric utility to convey dynamic prices to thermostats requires
baseload units at off-peak periods and the higher cost of significant infrastructure changes. Although interval meters are
peaking units during the few hundred hours per year these units being introduced, price communication to the appliance
run (e.g., [8]). We have used our model to investigate the cost appears to be some ways off. However, most of the benefits to
and performance of strategies under the time-of-day price both the consumer and load-serving entity would be realized by
curve of Fig. 7. Table III gives the results for a 50 gallon water the adoption of setback thermostats, with use schedules which
heater equipped with a conventional thermostat, a power can be altered from a factory default by the consumer to fit
interruption timer, a setback thermostat, and an adaptive price- their needs.
sensitive thermostat responding to prices communicated from
the load-serving entity. All strategies under a dynamic price
tariff require interval meters which record the use of power at ACKNOWLEDGMENT
small time steps. A price-sensitive thermostat requires The authors thank Lester Lave, Scott Matthews, Hadi
communication of prices to the consumer and thermostat in Dowlatabadi, Mike Griffin and Rahul Tongia for helpful
real time, or seasonally-averaged prices which are loaded into discussions.
the controller periodically.
In order to quantify the cost and performance under an
hourly price tariff, we have modeled prices in New York City.
The New York ISO publishes day-ahead prices for each hour
in 15 geographic markets [26]. Since the day-ahead markets
represent roughly 90% of the power traded, their prices are
representative of wholesale prices. We compiled the hourly
data for New York City for 2003. We added a fixed 8 ¢/kWh
to these wholesale costs for distribution, billing, and tax, to get
an estimate of hourly residential prices under a dynamic
pricing tariff (the average yearly wholesale cost was 6.37
¢/kWh, so the average residential cost for the year for New
York City customers under our assumptions would be 14.37
¢/kWh, a good match to bills obtained from customers in New
York City). We then used our model to calculate the cost and
performance of strategies under such a tariff with these New
York City prices. Table IV gives the results for a 50 gallon
water heater equipped as for Table III.
One component of California’s Statewide Pricing Pilot
(SPP) program (Critical Peak Pricing – Fixed) allowed rates to
rise to 78 ¢/kWh during peak times on up to 15 days of the
year. While complexities of the tariff make this program
difficult to model, savings at least as large as those in Table III
are likely.

VI. CONCLUSION
Use of a setback thermostat reduces annual electricity use by
320 kWh (at either set point). If we estimate that the sales price
of such a device will be roughly the same as that for a low-end
Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.edu/electricity 7

TABLE II
COST AND PERFORMANCE UNDER A FIXED ELECTRIC PRICE TARIFF

120 ±2.5 °F set point 140 ±2.5 °F set point

Yearly Cost @ Lowest Tank Yearly Cost @ Lowest Tank

Strategy 8.41 ¢/kWh Temperature 8.41 ¢/kWh Temperature
Conventional Thermostat $554 117.4 °F $740 137.4 °F
Interruption Timer Schedule A $522 94.3 °F $698 107.3 °F
Interruption Timer Schedule B $510 91.4 °F $683 103.2 °F
Setback Thermostat Schedule A $528 99.9 °F $712 119.9 °F
Setback Thermostat Schedule B $520 99.9 °F $705 119.9 °F
50 gallon water heater with UA=3.6 (2.5 inch insulation). Fixed electric price tariff of 8.41 ¢/kWh. Both the interruption timer and the setback
thermostat have been programmed for lower electricity use, in Schedule A from 1-4 AM and 1-6 PM. Schedule B extends the second period to 7
PM, and might be used by consumers without large early evening hot water consumption. The interruption timer removes power from the heater
during those times. The setback thermostat lowers the set point by 20 °F during those periods.

TABLE III
COST AND PERFORMANCE UNDER A DYNAMIC ELECTRIC PRICE TARIFF

120 ±2.5 °F set point 140 ±2.5 °F set point

Yearly Cost Lowest Tank Yearly Cost Lowest Tank

Strategy Temperature Temperature
Conventional Thermostat $554 117.4 °F $740 137.4 °F
Interruption Timer Schedule A $516 94.3 °F $693 107.3 °F
Interruption Timer Schedule B $468 91.4 °F $628 103.2 °F
Setback Thermostat Schedule A $529 99.9 °F $715 119.9 °F
Setback Thermostat Schedule B $497 99.9 °F $687 119.9 °F
Dynamic Thermostat $498 100.6 °F $683 120.6 °F
Dynamic electric price tariff which averages 8.41 ¢/kWh (Fig. 7 and note 2 above). Both the interruption timer and the setback thermostat have
been programmed for lower electricity use, in Schedule A from 1-4 AM and 1-6 PM. Schedule B extends the second period to 7 PM. The
interruption timer removes power from the heater during those times. The setback thermostat lowers the set point by 20 °F during those periods.
The dynamic thermostat lowers the set point by 20 °F during the daily price maximum, and by an amount proportional to the difference between
the daily maximum and minimum price at other times. Same water heater parameters as Table II.

TABLE IV
COST AND PERFORMANCE UNDER A DYNAMIC TARRIF USING HOURLY PRICES IN NEW YORK CITY DURING 2003

120 ±2.5 °F set point 140 ±2.5 °F set point

Yearly Cost Lowest Tank Yearly Cost Lowest Tank

Strategy Temperature Temperature
Conventional Thermostat $988 117.4 °F $1320 137.3 °F
Interruption Timer Schedule A $932 94.3 °F $1257 106.5 °F
Interruption Timer Schedule B $918 91.4 °F $1226 102.5 °F
Setback Thermostat Schedule A $939 100.0 °F $1274 119.9 °F
Setback Thermostat Schedule B $924 100.0 °F $1258 119.9 °F
Dynamic Thermostat $925 104.9 °F $1257 121.8 °F
Dynamic electric price tariff adds fixed costs of 8 ¢/kWh to the actual wholesale hourly cost of power in New York City during the 8760 hours of
2003 (Fig. 8). Same water heater parameters as Table II. All timers and thermostats operate as described for Table III.
Carnegie Mellon Electricity Industry Center Working Paper CEIC-04-02 www.cmu.edu/electricity 8

[22] Feldman, K.W., R.T. Schaller, J.A. Feldman, and M. McMillon (1978).
“Tap water scald burns in children.” Pediatrics 62(1): 1-7.
REFERENCES [23] Moritz A.R. and F.C. Henriques (1947). “Studies of Thermal Injury 2.
[1] PJM (http://www.pjm.com/markets/jsp/loadhryr.jsp), NYISO (http:// The relative importance of time and surface temperature in the causation
www.nyiso.com/oasis/load_scuc.html#monthly), and CAISO (http:// of cutaneous burns.” American Journal of Pathology 23(5): 695-720.
www.caiso.com/SystemStatus.html). [24] Occupational Safety and Health Administration (OSHA) Technical
[2] Energy Information Administration, U.S. Department of Energy, 2001 Manual, Section III, Chapter 7, “Legionnaires' Disease”, Paragraph
Residential Energy Consumption Survey: Housing Characteristics III.A.2. Available: http://www.osha-slc.gov/dts/osta/otm/otm_iii/
Tables. Available: ftp://ftp.eia.doe.gov/pub/consumption/residential/ otm_iii_7.html
2001hc_tables/appl_household2001.pdf [25] North American Electric Reliability Council (NERC) “2004 summer
[3] Energy Information Administration, U.S. Department of Energy, assessment: reliability of the bulk electric supply in North America.”
Residential Energy Consumption Surveys, 2001 Consumption and Available:
Expenditures Tables - Water-Heating Consumption Tables, Table CE4- ftp://www.nerc.com/pub/sys/all_updl/docs/pubs/summer2004.pdf.
9c, Available: ftp://ftp.eia.doe.gov/pub/consumption/residential/ [26] NYISO (2004), “Day ahead market LBMP – zonal”, New York
2001ce_tables/waterheat_consump2001.pdf, and U.S. Electric Power Independent System Operator. Available: http://mis.nyiso.com/public/P-
Industry Summary Statistics, Table 2. http://www.eia.doe.gov/cneaf/ 2Alist.htm
electricity/epm/epmt02p2.html
[4] Lemmer, E.F. and G.J. Delport (1999). “The influence of a variable
volume water heater on the domestic load profile.” IEEE Transactions
on Energy Conversion 14(4): 1558-1563.
[5] Reed, J.H., J.C. Thompson, R.P. Broadwater, and A. Chandrasekaran
(1989). “Analysis of water heater data from Athens load control
experiment.” IEEE Transactions on Power Delivery 4(2): 1232-1238.
[6] Lane, I.E. and N. Beute (1996). “A model of the domestic hot water
load.” IEEE Transactions on Power Systems 11(4): 1850-1855.
[7] Gustafson, M.W., J.S. Baylor, and G. Epstein (1993). “Direct water
heater load control – estimating program effectiveness using an
engineering model.” IEEE Transactions on Power Systems 8(1): 137-
143.
[8] California Energy Commission Demand Response Committee (2003).
“Feasibility of implementing dynamic pricing in california.” Report
Number 400-03-020F. Available: http://www.energy.ca.gov/reports/
2003-10-31_400-03-020F.PDF.
[9] Hastings, B.F. (1980). “Ten years of operating experience with a remote
controlled water heater load management system at Detroit Edison.”
IEEE Transactions on Power Apparatus and Systems PAS-99(4): 1437-
1441.
[10] Bischke, R.F. and R.A. Sella (1985). “Design and controlled use of water
heater load management.” IEEE Transactions on Power Apparatus and
Systems 104(6): 1290-1293.
[11] Lee, S.H. and C.L. Wilkins (1983). “A practical approach to appliance
load control analysis: a water heater case study.” IEEE Transactions on
Power Apparatus and Systems 102(4): 1007-1013.
[12] Andreolas, M. “Mega load management system pays dividends”,
Transmission & Distribution World, February 1, 2004. Available:
http://tdworld.com/mag/power_mega_load_management/.
[13] Braithwait, S. D. (2000). “Residential TOU price response in the
presence of interactive communications equipment,” in Pricing in
Competitive Electricity Markets, edited by A. Faruqui and K. Eakin,
Kluwer Academic Publishers, Boston, MA.
[14] Braithwait, S. D. and K. Eakin, (2002). “The role of demand response in
electric power market design.” Edison Electric Institute. Available:
http://www.eei.org/industry_issues/retail_services_and_delivery/wise_en
ergy_use/demand_response/demandresponserole.pdf.
[15] Kempton, W. (1988). “Residential hot water: a behaviorally-driven
system.” Energy 13(1): 107-114.
[16] Fanney, A. H. and B. P. Dougherty (1996). “The thermal performance of
residential electric water heaters subjected to various off-peak
schedules.” ASME Journal of Solar Energy Engineering 118(2): 73-80.
[17] U.S. Department of Energy (1998), Office of Energy Efficiency and
Renewable Energy, 10 CFR Part 430, Energy Conservation Program for
Consumer Products: Test Procedure for Water Heaters; Final Rule
(1998). Federal Register 63(90): 25995-26016 (May 11, 1998).
[18] Kern, D.Q. (1990). Process Heat Transfer, McGraw-Hill.
[19] Chopey, N.P. (1994), Handbook of Chemical Engineering Calculations,
2nd edition, McGraw-Hill.
[20] American Society of Heating, Refrigerating and Air-Conditioning
Engineers, Inc. (ASHRAE) Handbook – Heating, Ventilating, and Air-
Conditioning (HVAC) Applications (1995). Typical Residential Family’s
Hourly Hot Water Use. Fig. 11. page 45.10.
[21] Healy, W.M., J.D. Lutz, and A.B. Lekov (2003). “Variability in Energy
Factor Test Results for Residential Electric Water Heaters.” HVAC&R
Research 9(4): 435-449.