December 18, 2004

1
December 18, 2004
ADRS Load Impact
2
Executive Summary ADRS homes with technology
consume less on-peak energy than comparable
homes on standard rates or the CPP-F the
technology benefit is even stronger on Super Peak
days

• On non-event weekdays from July through
  September, average ADRS homes with technology
  consumed less on-peak energy (between 2 p.m. and
  7 p.m.) than comparable homes on standard
  tiered-rates (A03 subset) or on the SPP CPP-F
  rates (A07 subset)
• ADRS homes with technology used 3.7 kWh less
  on-peak electricity per home (34 lower) than
  comparable homes on standard rates (A03 subset)
• ADRS homes used less on peak than CPP-F homes
  (A07 subset) as well, 1.6 kWh lower on average
  (savings of 18)
• Over the twelve Super Peak days,
  technology-enabled ADRS homes consumed
  considerably less on-peak energy per home than
  their comparable control groups
• ADRS homes consumed 7.4kWh (or 50) less Super
  Peak energy per day than homes on standard rates
  (A03 subset)
• With ADRS technology, participants consumed 2.5
  kWh less super peak electricity per day (26
  savings) than comparable homes in the SPP on
  CPP-F rates (A07 subset)

Note ADRS participants were enrolled on a
first-come, first-served basis results were not
modified to address potential self-selection
bias Homes in the treatment and control groups
are comparable in that they all lie in Climate
Zone 3 and are single-family (detached) units
with central air conditioning further, raw load
data for the A03 and A07 control groups have been
weighted according to the distribution of the
ADRS population with respect to utility and
historical consumption strata
3
Executive Summary Performance of
technology-enabled ADRS homes improved relative
to both control groups from July to September

• ADRS technology enabled homes reduced load by
  50 consistently across the summer Super Peak
  events relative to homes without technology or
  rates (A03 subset)
• Relative to CPP-F homes (A07 subset), ADRS homes
  performance improved throughout the summer. Load
  reduction during the Super Peak hours increased
  from 25 in July and August to 31 in September
• This observed improvement in ADRS performance
  does not seem to be explained by weather
  differences or other variables other than
  occupant behavior
• Technology enabled-ADRS homes reduction of Super
  Peak load decreased over the five-hour Super Peak
  period, but still out-performs the comparable
  subset of A07 homes on the CPP-F rate without
  technology. Performance again improved in
  September, when the load reduction was sustained
  better in the last 1-2 hours of the Super Peak
  events
• Total daily energy consumption of ADRS houses was
  5 lower than that of the comparable subset of
  A03 homes on non-event weekdays and 12 lower on
  Super Peak days. Compared to the comparable
  subset of A07 homes, ADRS homes total daily
  usage was 2 lower on both Super Peak and
  non-event weekdays

4
Executive Summary more granularly, ADRS proved
very useful to pool owners and to
moderate/high-consumption homes less so for
homes with modest consumption

• Where present, pool pumps make a significant
  contribution to reduction of peak load vs. A03
• Relative to a control group of pools (from a
  Nevada Power load management program), ADRS pools
  reduce on-peak / Super Peak consumption by 2.8
  kWh per day
• For the average ADRS home with a pool, this 2.8
  kWh reduction is 48 of the 5.8 kWh total
  reduction on non-event weekdays and 29 of the
  9.5 kWh expected on Super Peak days
• As just 44 of the 175 ADRS have pools,
  reductions from pool loads comprise roughly 20
  of total peak load reduction and 10 of the
  reduction in Super Peak consumption
• Breaking down the population by
  energy-consumption stratum, technology appears to
  be an important driver in reducing Super Peak
  load for high-consumption homes, while the price
  signal appears to be a stronger driver of
  reduction in low-consumption homes
• Household level analysis reveals that the
  majority of ADRS homes (52) actively
  experimented with the technology to control home
  energy use, while an additional 7 made minor
  adjustments. Furthermore, about 10 of the ADRS
  population are Supersavers, reducing load at 2
  p.m. by more than 30 consistently across the
  summer months on a daily basis

Total reduction of on-peak/Super Peak load by
homes with pools is calculated algebraically
rather than by direct measurement
5
Executive Summary ADRS load reductions relative
to both control groups are statistically
significant for the high consumption homes

• As an indication of the statistical quality of
  the results, the coefficient of variation allows
  us to compare relative variation between
  populations
• The coefficient of variation (CV), which allows
  for comparison of the relative variation of
  values between populations, is defined as the
  standard deviation (SD) of a sample divided by
  the samples mean value
• A CV value less than 0.5 implies that the
  statistic is significant within a 95 confidence
  interval
• A CV value greater than 1 implies that the
  statistic is not significant (less than 70
  confidence)

Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
6
High CVs for load-reduction in low-consumption
ADRS homes relative to A07 homes confirms our
hypothesis that those results are not
statistically significant, due to small and
diverse samples

• Variation in total consumption of ADRS and
  control groups is high for both high- and
  (especially) low-consumption homes.
• The CV declines when we look at the difference in
  consumption between ADRS and each of the control
  groups, particularly for Super Peak days this
  suggests that the variations among the ADRS homes
  loads and the control groups are not independent,
  but are correlated ( i.e., relatively high or low
  values tend to occur at similar times in each
  population
• The CV values for load reductions of
  high-consumption ADRS homes relative to both
  control groups are substantially less than 0.5,
  suggesting that results are statistically
  significant at a 95 confidence level
• Results of ADRS load reductions relative to
  control groups for low-consumption homes are
  mixed
• The CV for low-consumption ADRS load reductions
  relative to the A03 control group and between
  low-consumption A03 and A07 homes are 0.6-0.8,
  suggesting that the results are statistically
  significant with 80-90 confidence
• The CV for low-consumption ADRS load reductions
  relative to the low consumption A07 control group
  confirms our hypothesis that these results are
  not statistically significant
• Similarly, the low-consumption A07 control group
  load reductions relative to the low-consumption
  A03 control group on Super Peak days are not
  statistically significant
• The small size of the low-consumption home
  populations seems to limit statistical quality

7
Executive Summary An initial comparison with
CRAs results for the SPP of comparable homes
indicate that ADRS savings relative to A07 homes
may be too low

• It appears that the price response of our pilots
  A07 control population (subset of the SPP A07
  population) is greater than the performance
  observed in the 2003 statewide pricing pilot.
• The ADRS study finds a 32 savings for its subset
  group of A07 homes that are single-family with
  central air, relative to comparable A03 control
  sample in the pilot. Charles Rivers Associates
  (CRA),using a different methodology shows in
  their Summer 2003 final report (August 9, 2004),
  Table 5-9 for zone 3, that single family homes
  saved 14.27, and with central AC saved 13.45.
• Whether the A07 population is representative of
  residential customers statewide is still an open
  issue
• The price response performance of the A07
  population continues to be studied in detail in
  the statewide pricing pilot with, a larger sample
  population
• Collaboration with CRA to investigate into the
  nature of these differing results has been
  proposed for 2005.
• For ADRS homes, pretreatment data was not
  adequately available to investigate consumption
  behavior prior to participating in the pilot.

8
Table of Contents

• Executive Summary
• Pilot Background and Overview of Experimental
  Design
• Data Sources
• Analytical Methodology
• Load Impact Results
• Conclusions and Recommendations
• Appendix

9
Project Background

• The Automated Demand Response System (ADRS)
  program is an additional and parallel pilot
  alongside the Statewide Pricing Pilot (SPP)
• ADRS focuses on the further impact of energy
  management technology on residential customers in
  addition to the time-differentiated tariffs
  experienced under the SPPs critical peak pricing
• Rocky Mountain Institute (RMI) was tasked to
  conduct an independent analysis and evaluation of
  the ADRS pilot with respect to additional load
  impact and economic efficacy
• Demand response evaluation of ADRS must answer
  two questions
• What is the range of demand response/load drop
  observed?
• Is the range and average demand drop larger or
  smaller than that observed in the larger
  statewide pilot (SPP), given comparable rates and
  weather conditions?

10
Overview of ADRS Experimental Design

• The ADRS pilot installed full-scale system
  technology, capable of automatically controlling
  the electrical load of multiple appliances in a
  limited number of residential customers across
  the three participating CA IOUs (SCE, PGE,
  SDGE)
• ADRS targeted 175 participants in the 3 major
  California IOU service territories
• 75 SCE
• 75 PGE
• 25 SDGE
• ADRS participants were recruited only from
  climate zone 3 of the four climate zones defined
  by the SPP and were required to have central air
  conditioning
• Participants were placed on the SPPs critical
  peak pricing-fixed (CPP-F) tariff
• Time of use tariff with rates differentiated by
  time of day
• Off peak (weekdays, excluding 2 p.m. 7 p.m.
  all hours weekends and holidays)
• On peak (weekdays 2 p.m. - 7 p.m.)
• Super Peak (select weekdays 2 p.m. 7 p.m.)
• Maximum of 12 super peak days during the summer
  season 15 total annually
• Maximum of three consecutive super peak days

11
Overview of ADRS Technology

• Invensys GoodWatts technology was selected for
  the ADRS pilot
• The ADRS control technology includes
• Two-way communicating interval whole house meter
• Wireless internet gateway and cable modem
• Smart thermostat(s)
• Load control and monitoring device (LCM) to
  manage select loads (e.g., pool pump)
• Web-enabled user interface and data management
  software
• At all times, ADRS displays the current price of
  electricity, both on the thermostat and on the
  Web
• Via the Internet, pilot participants can
• View real time interval demand and trends in
  historical consumption
• Set climate control and pool runtime preferences
• Program desired response to increase in
  electricity price
• Change in thermostat temperature set point
• Reschedule operation of LCM controlled appliance
  (e.g., pool pump)
• Once programmed, technology automatically changes
  operations in response to electricity prices

12
Pilot Design

• Pilot customers were recruited from
  owner-occupied, single-family homes from climate
  zone 3 in geographies served by appropriate cable
  providers and in zip codes identified by the
  participating utilities
• Otherwise, pilot homes were recruited at random
  regardless of historical consumption, although
  homes were screened for eligibility with respect
  to presence of central air conditioning, within
  prescribed zip codes
• Pilot homes were screened for availability of
  other loads (i.e., swimming pool pumps and spas),
  but not disqualified from participation in their
  absence
• Pilot homes were segmented into two strata by
  historical consumption according to the
  methodology established for the SPP
• Modest consumers, those with summer average daily
  usage below 24 kWh, comprised the low stratum
• All other homes, those with Summer ADU above 24
  kWh, fell into the high stratum

13
Recruitment of pilot participants

• Eligible customers were mailed an announcement
  describing the pilot and benefits of
  participation
• Technology and user tools for greater control of
  energy in the home
• Potential to achieve bill savings by managing
  consumption
• Package indicated that customers would be paid
  incentives totaling 100
• 25 for enrollment and completion of the home
  energy survey
• 75 payable at the end of the pilot for
  continuous participation and completion of mid-
  and end-of-pilot customer satisfaction surveys
• Incentive payment parallels structure of offer to
  SPP participants
• Enrollment packages were then mailed to the
  customers the packages included enrollment
  application and informed prospective participants
  of three avenues by which to enroll
• Mail in enrollment application
• Phone call
• ADRS website
• Reminder postcards were sent out noting deadline
  for enrollment
• Third-party call center (Cypress) was contracted
  to handle inbound enrollment calls and for
  outbound calls as needed to fulfill enrollment
  goals by target deadline

14
The analytical approach accounts for changes in
ADRS enrollment

• Installation completed for 175 homes by mid-June
  (76 SCE, 75 PGE, and 24 SDGE)
• With opt outs, total enrollment declined to 164
  active participants by October
• ADRS analysis is executed on a per home basis
• Data from homes that ultimately opted out is
  included in the analysis for the period during
  which they both were subject to the CPP-F rate
  and had use of GoodWatts

Total Program Participation, JuneSeptember, 2004
PGE
72
70
SCE
SDGE
22
15
Design and selection of control group

• The SPP collects interval meter data on many
  customers for purposes of program evaluation.
  Populations selected for the SPP were intended to
  be representative of the statewide residential
  population
• One SPP population, known as A03, is comprised of
  homes that
• Are on standard, tiered rates
• Do not possess ADRS technology, and
• Are unaware of their role as a control group for
  the SPP or ADRS
• A second SPP population, known as A07, is
  comprised of homes that
• Voluntarily enrolled to test the CPP-F
  experimental rate
• Were not provided any additional technology by
  their utility
• The two SPP populations were filtered to only
  single-family homes in climate zone 3 with
  central air, SPP populations both used as control
  groups against the ADRS population
• The subset of single-family, A03 homes with
  central air in climate zone 3 is used to assess
  the total ADRS impact of technology and CPP-F
  rate
• The subset of single-family, A07 homes with
  central air in climate zone 3 is used to assess
  the incremental impact of ADRS technology over
  and above SPP rate impacts

16
Design and selection of control group
Characteristics of ADRS and Control Group
Populations and Distribution of Homes, as of
September 2004
A03
A07
ADRS
Rate
Standard tiered-block pricing
CPP-F
CPP-F
Technology
Not Provided
Not Provided
GoodWatts
Price Response
Monthly billing
Manual shift save
Automated shift Save
Pools Penetration
23.1
23.7
25.6
Participants
PGE
SDGE
SCE
PGE
SDGE
SCE
PGE
SDGE
SCE
Low Stratum
2
3
14
10
1
16
22
15
4
High Stratum
12
3
22
21
5
38
49
7
65
Total
14
6
36
31
6
54
71
22
69
17
Table of Contents

• Executive Summary
• Pilot Background and Overview of Experimental
  Design
• Data Sources
• Analytical Methodology
• Load Impact Results
• Conclusions and Recommendations
• Appendix

18
Sources of Load Data

• Control Groups
• Revenue-grade utility meters measure time of use
  consumption in 15-minute intervals
• Data collected from meters on monthly basis and
  aggregated and distributed to RMI six to eight
  weeks following close of each month
• SCE and SDGE transferred data directly to RMI
• PGE meter data for ADRS homes were posted on a
  secure website for direct download
• ADRS participants
• GoodWatts meters report demand and consumption
  data for all utilities in near real-time
• Although data from Invensys meters proved
  commensurate with utility revenue-grade interval
  meters in pilot testing, utilities chose to rely
  upon utility meters and manual collection of data
  for ADRS participants
• Data was aggregated and reported to RMI with
  control group data
• In an effort to speed availability of data for
  load impact analysis, Invensys data was used for
  SCE service territory for month of September in
  response to administrative issues in scheduling
  an accelerated final read of revenue meter
• GoodWatts load control monitors (LCM) provide
  15-minute interval load data for pool pumps

19
Sources of Additional Data

• Address/zip code information were collected for
  treatment group homes and extracted from SPP
  database for A07 control group homes
• Hourly outdoor temperature data at the zip code
  level were provided by Invensys via weather data
  subscription service
• Data on home characteristics were collected to
  help gain greater insight into impact findings
• Installation survey collected air conditioning
  and pool pump nameplate data
• SPP Customer Characteristics Survey gathered
  information on appliance saturation, house size,
  and demographics for both treatment and control
  group homes

20
Table of Contents

• Executive Summary
• Pilot Background and Overview of Experimental
  Design
• Data Sources
• Analytical Methodology
• Load Impact Results
• Conclusions and Recommendations
• Appendix

21
Methodology for Analysis of Energy Impacts

• Daily 15-minute interval load data used to
  construct average load profiles for homes in the
  pilot and each of the two control groups
• Comparison of load curves is the primary means of
  analysis
• The differences in mean load profiles of the A03
  control group versus ADRS participants reflect
  the overall impact of ADRS enabling technology in
  conjunction with time-varying rate
• The differences in mean load profiles of the A07
  control group versus ADRS participants reflect
  the incremental impact attributed to the ADRS
  enabling technology
• Differences were studied on both Super Peak days
  and non-event weekdays
• Weekends and Holidays are excluded from the
  analysis
• Weekends and holidays are charged only off-peak
  rates within the CPP-F experimental rate
  structure
• Occupancy patterns on these days are distinctly
  different from weekdays there is typically
  higher and more constant occupancy, resulting in
  higher loads relative to weekdays

22
Methodology for Analysis of Energy Impacts,
continued

• Results are reported in terms of 5-hour averages
  (duration of 2 p.m. 7 p.m. on peak and Super
  Peak periods) and hour-by-hour reductions
• Results are reported state-wide and sample
  average is weighted according to distribution of
  participants by utility for each customer stratum
• Greater granularity is shown as well, with
  results broken out by consumption strata
• Trends in impact across the summer months are
  reported
• Load reduction is also analyzed in the context of
  peak daily temperature for both super peak
  pricing days and typical non-event days to test
  two competing hypotheses
• Controllable load increases with outside
  temperature since air conditioning demand also
  increases
• Controllable load decreases with outside
  temperature as homeowner willingness to
  contribute decreases and rate of overrides
  increases

23
Methodology for Analysis of Energy Impacts,
continued

• A03 and A07 load data were weighted according to
  the distribution of the ADRS population (by
  utility and consumption strata) so as to permit
  direct comparison among populations which vary by
  geography, weather, and baseline consumption
• For each month, A03 and A07 load data were
  recorded per utility and strata (high/low)
• The data were then multiplied by a constant,
  reflecting ADRS population distribution per
  utility and strata (i.e., 32 high-stratum ADRS
  homes in SCE territory)
• Customers that opted out of A07 were included in
  the analysis for the period during which they
  were subject to the CPP-F rate and excluded at
  the point of rate expiration the A07 opt outs
  contributed to less than a one percent increase
  in average A07 monthly load and therefore the
  impact on results was negligible
• A03 and A07 results were not adjusted based on
  pool penetration to match the ADRS population
• Control group pool loads have not been measured
  separately
• Difference in pool ownership between A03 and A07
  homes vs. ADRS homes yielded less than one
  percent decrease in total average control group
  load

24
Table of Contents

• Executive Summary
• Pilot Background and Overview of Experimental
  Design
• Data Sources
• Analytical Methodology
• Load Impact Results
• Conclusions and Recommendations
• Appendix

25
On non-event weekdays, technology enabled ADRS
homes consumed less on-peak energy than homes on
standard tiered rates (A03) or the SPP CPP-F rate
(A07)
Average Non-Event Weekday Load Profile July
through September - All Homes
? On Peak ?
Difference in On-Peak Usage
A03-A07
A03-ADRS
A07-ADRS
0.43 kWh/hr
0.74 kWh/hr
Average
0.31 kWh/hr
2.1 kWh
3.7 kWh
5-hr Total
1.6 kWh
19
34
Reduction
18
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note Non-event weekdays exclude weekends,
holidays and Super Peak event days Summer period
is defined as July through September - June data
was excluded because its results differed
significantly due to unfamiliarity with
technology, lower average temperatures, and lack
of event days.
26
ADRS technology enabled homes further reduced
their load in comparison to standard tiered rate
or SPP CPP-F customers during Super Peak hours
on the 12 event days
Average Event Day Load Profile July through
September - All Homes
? Super Peak ?
Difference in Super Peak Usage
A03-A07
A03-ADRS
A07-ADRS
0.96 kWh/hr
1.47 kWh/hr
Average
0.51 kWh/hr
4.8 kWh
7.4 kWh
5-hr Total
2.5 kWh
32
50
Reduction
26
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note Load data for the A03 and A07 control
groups has been weighted according to the
distribution of the ADRS population with respect
to utility and historical consumption strata.
27
ADRS technology-enabled homes reduced load
consistently across the summer events, though
performance vs. CPP-F homes improved in September
Average Reduction in Super Peak Consumption, All
Homes on Event Days
Consumption (kWh)
Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
28
Reduction of Super Peak load for all ADRS homes
decreased over the five hour peak period, but
continued to out-perform the A07 homes on the
CPP-F rate without technology
Average Percent Reduction in Super Peak
Consumption, All Participants for all Summer
Events
Reduction in Super Peak Consumption
Hour of the Super Peak Period (2 p.m. - 7 p.m.)
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
29
On a month-by-month basis, however, hourly
reduction was more sustained in September,
compared to the A07 homes
Average Reduction in Super Peak Consumption, All
Participants
Reduction from Control Group
Hour of the Super Peak Period (2 p.m. - 7 p.m.)
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
30
As September events days were slightly warmer
than the others, temperature does not explain
Septembers improvement in ADRS performance,
suggesting improvements may be behavioral
Statewide Average Peak Temperature, Non-event
Weekdays
Statewide Average Peak Temperature, Super Peak
Weekdays
Peak Temp (º F)
31
ADRS reduction in Super Peak consumption varied
from 1.1 to 1.9 kWh/hr relative to homes on
standard tiered rates (A03)
Average Reduction In Super Peak Consumption
Relative to Homes on Standard, Tiered Rate, All
ADRS Homes
Reduction in Super Peak Consumption (kWh/hr)
High Temp (o F)
Event Date
Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
32
ADRS technology-enabled homes consistently
reduced Super Peak load by 50 vs. homes on
standard tiered rates (A03)
Average Reduction In Super Peak Consumption
Relative to Homes on Standard, Tiered Rate, All
ADRS Homes
Reduction in Super Peak Consumption
High Temp (o F)
Event Date
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
33
Super Peak consumption relative to homes on the
CPP-F rate without technology (A07) was
consistently lower by 0.5 kWh/hr on average,
improving somewhat in September
Average Reduction In Super Peak Consumption
Relative to Homes on CPP-F Rate, All ADRS Homes
vs. A07
Reduction in Super Peak Consumption (kWh/hr)
High Temp (o F)
Event Date
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
34
On a percentage basis, however, the load
reduction during Super Peak hours of ADRS homes
relative to homes on the CPP-F rate without
technology (A07) varied more, averaging 26 lower
demand
Average Reduction In Super Peak Consumption
Relative to Homes on CPP-F Rate, All ADRS Homes
vs. A07
Reduction in Super Peak Consumption
High Temp (o F)
Event Date
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
35
Load reducing behavior varied across several
consecutive Super Peak events, but was strongest
in September compared to A07
ADRS Homes vs. Standard Tiered Rates (A03)
ADRS Homes vs. Homes on CPP-F Rates without
Technology (A07)
Reduction from Control Group
High Temp (o F)
Event Date
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
36
ADRS homes used technology to lower average daily
energy consumption overall, compared to homes
without technology, both on the CPP-F rate (A07)
and without dynamic rates (A03)
Average Daily Consumption, All Homes on Event
Weekdays
Average Daily Consumption, All Homes on non-Event
Weekdays
Consumption (kWh)
Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
37
It appears that ADRS homes are conserving energy
in addition to shifting some load to off-peak
hours however, most of the conservation effect
occurs during the peak hours
Average Consumption, July-September Non-Event
Weekdays
Average Consumption, July-September Super Peak
Event days
Consumption (kWh)
Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
38
Additional analysis can isolate the portion of
load impact attributable to control of pool pumps

• Other than penetration rates, there is little
  information available on the contribution of pool
  pump loads to the A03 and A07 aggregate load
  profiles
• Energy consumption by ADRS pools can, however, be
  compared against consumption of pools from a
  demand response program being conducted by Nevada
  Power
• Since there is no financial incentive for pool
  owners to shift load away from peak in Nevada
  Powers ACLM program, operation of pools in Las
  Vegas is presumed to provide an appropriate load
  shape for comparison purposes
• Reported load is average of all pools and
  reflects load diversity in scheduling
• The aggregate load profile from Nevada is scaled
  down to reflect the smaller operational load of
  pools participating in ADRS (from 1.8 kW in
  Nevada to 1.6 kW among ADRS participants)

39
ADRS homes reduce pool load during peak hours on
all weekdays, regardless of whether or not a
Super Peak event is called
Average Pool Pump Load (July-September, 2004)
? On Peak ?
Nevada Power (n 78)
Average Pool Load (kWh/hr)
ADRS (n 44)
Time of Day
Source Invensys GoodWatts Reports Server
40
Where present, shifting of pool loads to off peak
is a significant contributor to reduction of
on-peak consumption

• The average Nevada pool consumes 2.8 kWh between
  2 p.m. 7 p.m. (on a scale adjusted basis)
• By scheduling pools to operate outside of the 2
  p.m. to 7 p.m. period, ADRS homes effectively
  reduce on-peak or Super Peak consumption by 2.8
  kWh each day
• 2.8 kWh is roughly 48 of the 5.8 kWh total
  on-peak reduction for a house with a pool
• With the further reduction of other loads on
  Super Peak days, 2.8 kWh constitutes 29 of the
  homes 9.5 kWh total super peak reduction
• Since only one out of approximately each four
  ADRS homes has a pool, pools in aggregate
  comprise about 20 of peak load reduction and 10
  of Super Peak load reduction

Average Reduction of On-Peak / Super Peak Load
Non-Event Weekday
Super Peak Day
ADRS Segment
Pool
Other
Total
Pool
Other
Total
No Pool (131)
--
3.0 kWh
3.0 kWh
--
6.7 kWh
6.7 kWh
With Pool (44)
2.8 kWh
3.0 kWh
5.8 kWh
2.8 kWh
6.7 kWh
9.5 kWh
Weighted Avg. (175)
0.7 kWh
3.0 kWh
3.7 kWh
0.7 kWh
6.7 kWh
7.4 kWh
0.7 kWh / 3.7 kWh 20
0.7 kWh / 7.4 kWh 10
Reduction of other loads calculated
algebraically from total average load reduction
and average pool load reduction rather than
direct measurement
41
Stratified results suggest that technology is a
significant driver of behavior among moderate
high consumption homes for lower-consumption
homes, price signals appear to be the primary
driver

• High-consumption homes use the ADRS technology to
  further reduce load during Super Peak hours
• On non-event weekdays, average load is reduced by
  2.5 kWh vs. CPP-F rate homes without technology,
  compared to 1.6 kWh for the overall population
• On Super Peak event days, average load is reduced
  by 3.8 kWh vs. CPP-F rate homes without
  technology, compared to 2.5 kWh for the overall
  sample
• Compared to low-consumption ADRS homes,
  low-consumption CPP-F rate homes without
  technology (A07) have consistently lower loads at
  all hours of the day on both Super Peak and
  Non-Event Weekdays
• Low-consumption homes appear to more sensitive to
  price signalsthe CPP-F rate structure alone
  changes their load profile significantly and
  technology appears to add little incremental
  benefit for this (albeit small) population sample
• This suggests that high-consumption homes should
  be targeted for the ADRS technologylow-consumptio
  n homes may be less likely to be cost effective

Consistently lower loads suggests the
potential for a systemic bias between treatment
and control group homes in the low consumption
stratum that is not accounted for in the pilot
otherwise we would expect low stratum A07 homes
to have higher demand in the off-peak period to
catch up for their reduction during peak hours
42
High-consumption, technology-enabled ADRS homes
reduce load further than the overall population
Average Non-Event Weekday Load Profile -
High-Consumption Homes
? On Peak ?
Difference in On-Peak Usage
A03-A07
A03-ADRS
A07-ADRS
0.37 kWh/hr
0.87 kWh/hr
Average
0.50 kWh/hr
1.8 kWh
4.3 kWh
5-hr Total
2.5 kWh
15
35
Reduction
24
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
43
Super Peak Event days also see greater load
reductions for high-consumption ADRS homes vs.
the overall population
Average Super Peak Event Day Load Profile
High-Consumption Homes
? Super Peak ?
Difference in Super Peak Usage
A03-A07
A03-ADRS
A07-ADRS
0.94 kWh/hr
1.70 kWh/hr
Average
0.77 kWh/hr
4.7 kWh
8.5 kWh
5-hr Total
3.8 kWh
28
51
Reduction
32
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
44
Low-consumption, CPP-F rate (A07) homes have
lower load than technology-enabled ADRS homes on
non-event weekdays, suggesting price signals
drive their load reduction
Average Non-Event Weekday Load Profile -
Low-Consumption Homes
? On Peak ?
Difference in On-Peak Usage
A03-A07
A03-ADRS
A07-ADRS
0.61 kWh/hr
0.38 kWh/hr
Average
-0.2 kWh/hr
3.1 kWh
1.9 kWh
5-hr Total
-1.2 kWh
44
28
Reduction
-30
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
45
Price signal is again suggested as the stronger
driver of load reduction in low-consumption homes
on Super Peak daysCPP-F rate (A07) homes
demonstrate lower loads than ADRS homes
Average Super Peak Event Day Load Profile - Low
Consumption Homes
? Super Peak ?
Difference in Super Peak Usage
A03-A07
A03-ADRS
A07-ADRS
1.05 kWh/hr
0.81 kWh/hr
Average
-0.2 kWh/hr
5.2 kWh
4.0 kWh
5-hr Total
-1.2 kWh
55
43
Reduction
-27
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
46
Household level analysis reveals that the
majority of ADRS homes (52) actively
experimented with the technology to control home
energy use, with an additional 7 made minor
adjustments

• For each month (June-September), the
  instantaneous load drop at 2 p.m. for each ADRS
  home was calculated, for Super Peak and non-Super
  Peak weekdays
• This instantaneous reduction was categorized into
  High (gt30 drop), Medium (15-30 drop), and
  Low (lt 15 drop) categories
• Trends in load reductions (high, medium, low)
  were observed across the months for both Super
  Peak and non-Super Peak weekdays
• The majority of homes (52) varied their 2 p.m.
  load reductions widely across the summer months
  and between Super Peak and non-Super Peak
  weekdays
• An additional 7 of ADRS homes made some minor
  adjustments with the technology
• Approximately 41 of homes did not change their 2
  p.m. load reductions significantly from month to
  month or between Super Peak and non-Super Peak
  weekdays.

47
Furthermore, about 10 of the ADRS population are
Supersavers, reducing load at 2 p.m. by more
than 30 consistently across the summer months on
a daily basis

• The Supersaver ADRS homes contributed 20 of
  Super Peak reduction and 24 non-Super Peak
  reduction across the summer months, in terms of
  instantaneous load shed at 2 p.m.
• The Supersavers were not the only ones reducing
  significant load at 2 p.m., however.
  Approximately 50 of the population saved more
  than 30 of their 2 p.m. load on Super Peak
  weekdays, while 25 of the ADRS population saved
  more than 30 of their 2 p.m. load on non-Super
  Peak weekdays
• The range of load reduction at 2 p.m. for high
  performance homes ranged from 30 to almost 100
  on both Super Peak and non-Super Peak days
• Ten percent of homes improved their performance
  across the summer months, gradually increasing
  their 2 p.m. load shed July-September on all
  weekdays
• Seven percent of the population showed declining
  performance.
• Three percent even increased their consumption
  during peak hours and on Super Peak days relative
  to off-peak hours and non-Super Peak days
• Additional research is needed to determine
  whether these homeowners are consciously
  increasing their consumption during peak hours or
  whether, out of confusion, they are using the
  technology incorrectly

48
As an indication of the statistical quality of
the results, the coefficient of variation allows
us to compare relative variation between
populations

• The coefficient of variation (CV), which allows
  for comparison of the relative variation of
  values between populations, is defined as the
  standard deviation (SD) of a sample divided by
  the samples mean value
• For ADRS, the mean and standard deviation was
  calculated for each time interval over the 5-hour
  peak period. A coefficient of variation was
  calculated for each consumption stratum according
  to Event and Non-Event days.
• The mean, SD, and CV were calculated for total
  consumption of each group of homes ADRS and
  control populations (A03 and A07)
• The mean, SD, and CV were calculated for the load
  reduction between the control group (A03 or A07)
  and ADRS homes
• A CV value greater than 1 implies that the
  statistic is not significant (less than 70
  confidence)
• A CV value less than 0.5 implies that the
  statistic is significant within a 95 confidence
  interval

49
Variation in total consumption is high across all
groups of homes for both high- and (especially)
low-consumption homes ADRS load reductions
relative to both control groups are statistically
significant for the high consumption homes
Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
50
High CVs for load-reduction in low-consumption
ADRS homes relative to A07 homes confirms our
hypothesis that those results are not
statistically significant, due to small and
diverse samples

• Variation in total consumption of ADRS and
  control groups is high for both high- and
  (especially) low-consumption homes.
• The CV declines when we look at the difference in
  consumption between ADRS and each of the control
  groups, particularly for Super Peak days this
  suggests that the variations among the ADRS homes
  loads and the control groups are not independent,
  but are correlated ( i.e., relatively high or low
  values tend to occur at similar times in each
  population
• The CV values for load reductions of
  high-consumption ADRS homes relative to both
  control groups are substantially less than 0.5,
  suggesting that results are statistically
  significant at a 95 confidence level
• Results of ADRS load reductions relative to
  control groups for low-consumption homes are
  mixed
• The CV for low-consumption ADRS load reductions
  relative to the A03 control group and between
  low-consumption A03 and A07 homes are 0.6-0.8,
  suggesting that the results are statistically
  significant with 80-90 confidence
• The CV for low-consumption ADRS load reductions
  relative to the low consumption A07 control group
  confirms our hypothesis that these results are
  not statistically significant
• Similarly, the low-consumption A07 control group
  load reductions relative to the low-consumption
  A03 control group on Super Peak days are not
  statistically significant
• The small size of the low-consumption home
  populations seems to limit statistical quality

51
Table of Contents

• Executive Summary
• Pilot Background and Overview of Experimental
  Design
• Data Sources
• Analytical Methodology
• Load Impact Results
• Conclusions and Recommendations
• Appendix

52
Conclusions ADRS homes with technology consume
less on-peak energy than comparable homes on
standard rates or the CPP-F the technology
benefit is even stronger on Super Peak days

• On non-event weekdays from July through
  September, average ADRS homes with technology
  consumed less on-peak energy (between 2 p.m. and
  7 p.m.) than comparable homes on standard
  tiered-rates (A03) or the SPP CPP-F (A07)
• ADRS homes with technology used 3.7 kWh less
  on-peak electricity per home (34 lower) than
  comparable homes on standard rates (A03)
• ADRS homes used less on peak than CPP-F homes
  (A07) as well, 1.6 kWh lower on average (savings
  of 18)
• Over the twelve Super Peak days,
  technology-enabled ADRS homes consumed
  considerably less on-peak energy per home than
  their comparable control groups
• ADRS homes consumed 7.4kWh (or 50) less Super
  Peak energy per day than homes on standard rates
  (A03)
• With ADRS technology, participants consumed 2.5
  kWh less super peak electricity per day (26
  savings) than comparable homes in the SPP on
  CPP-F (A07)

Note ADRS participants were enrolled on a
first-come, first-served basis results were not
modified to address potential self-selection
bias Homes in the treatment and control groups
are comparable in that they all lie in Climate
Zone 3 and have central air conditioning
further, raw load data for the A03 and A07
control groups have been weighted according to
the distribution of the ADRS population with
respect to utility and historical consumption
strata
53
Conclusions Performance of ADRS homes with
technology improved relative to both control
groups from July to September

• ADRS technology enabled homes reduced load by
  50 consistently across the summer Super Peak
  events relative to homes without technology or
  rates (A03)
• Relative to CPP-F homes (A07), ADRS homes
  performance improved throughout the summer. Load
  reduction during the Super Peak hours increased
  from 25 in July and August to 31 in September
• This observed improvement in ADRS performance
  does not seem to be explained by weather
  differences or other variables other than
  occupant behavior
• Technology enabled ADRS homes reduction of Super
  Peak load decreased over the five-hour Super Peak
  period, but they still out-performed A07 homes on
  the CPP-F rate without technology. Performance
  again improved in September, when the load
  reduction was sustained better in the last 1-2
  hours of the Super Peak events
• Total daily energy consumption of ADRS houses was
  5 lower than A03 homes on non-event weekdays and
  12 lower on Super Peak days. Compared to A07
  homes, ADRS homes total daily usage was 2 lower
  on Super Peak and non-event weekdays

54
Conclusions ADRS proved particularly useful to
pool owners and to moderate/high-consumption
homes less so for homes with modest consumption

• Where present, pool pumps make a significant
  contribution to reduction of peak load vs. A03
• Relative to a control group of pools (from a
  Nevada Power load management program), ADRS pools
  reduce on-peak / Super Peak consumption by 2.8
  kWh per day
• For the average ADRS home with a pool, this 2.8
  kWh reduction is 48 of the 5.8 kWh total
  reduction on non-event weekdays and 29 of the
  9.5 kWh expected on Super Peak days
• As just 44 of the 175 ADRS have pools,
  reductions from pool loads comprise roughly 20
  of total peak load reduction and 10 of the
  reduction in Super Peak consumption
• Breaking down the population by
  energy-consumption stratum, technology appears to
  be an important driver in reducing Super Peak
  load for high-consumption homes, while the price
  signal appears to be a stronger driver of
  reduction in low-consumption homes
• Household level analysis reveals that the
  majority of ADRS homes (52) actively
  experimented with the technology to control home
  energy use, while an additional 7 made minor
  adjustments. Furthermore, about 10 of the ADRS
  population are Supersavers, reducing load at 2
  p.m. by more than 30 consistently across the
  summer months on a daily basis

Total reduction of on-peak/Super Peak load by
homes with pools is calculated algebraically
rather than by direct measurement
55
Executive Summary ADRS load reductions relative
to both control groups are statistically
significant for the high consumption homes

• As an indication of the statistical quality of
  the results, the coefficient of variation allows
  us to compare relative variation between
  populations
• The coefficient of variation (CV), which allows
  for comparison of the relative variation of
  values between populations, is defined as the
  standard deviation (SD) of a sample divided by
  the samples mean value
• A CV value greater than 1 implies that the
  statistic is not significant (less than 70
  confidence)
• A CV value less than 0.5 implies that the
  statistic is significant within a 95 confidence
  interval

Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
56
High CVs for load-reduction in low-consumption
ADRS homes relative to A07 homes confirms our
hypothesis that those results are not
statistically significant, due to small and
diverse samples

• Variation in total consumption of ADRS and
  control groups is high for both high- and
  (especially) low-consumption homes.
• The CV declines when we look at the difference in
  consumption between ADRS and each of the control
  groups, particularly for Super Peak days this
  suggests that the variations among the ADRS homes
  loads and the control groups are not independent,
  but are correlated ( i.e., relatively high or low
  values tend to occur at similar times in each
  population
• The CV values for load reductions of
  high-consumption ADRS homes relative to both
  control groups are substantially less than 0.5,
  suggesting that results are statistically
  significant at a 95 confidence level
• Results of ADRS load reductions relative to
  control groups for low-consumption homes are
  mixed
• The CV for low-consumption ADRS load reductions
  relative to the A03 control group and between
  low-consumption A03 and A07 homes are 0.6-0.8,
  suggesting that the results are statistically
  significant with 80-90 confidence
• The CV for low-consumption ADRS load reductions
  relative to the low consumption A07 control group
  confirms our hypothesis that these results are
  not statistically significant
• Similarly, the low-consumption A07 control group
  load reductions relative to the low-consumption
  A03 control group on Super Peak days are not
  statistically significant
• The small size of the low-consumption home
  populations seems to limit statistical quality

57
Recommendations for possible future extension of
the pilot

• Due to the lack of depth in data populations,
  particularly for the low stratum, additional
  recruiting should be performed to increase
  confidence of results
• Standard tiered rates (A03) low consumption PGE
  and SDGE homes and high consumption SDGE homes
• CPP-F homes without technology (A07) low
  consumption and high consumption SDGE homes
• ADRS homes with technology low consumption SCE
  homes and high consumption SDGE homes
• Continue to provide information and educational
  materials to the ADRS participants, in order to
  provide a test of whether performance can improve
  in subsequent summers
• Because ADRS home load reduction decreases
  relative to A07 homes in the later hours of the
  Super Peak period, a shorter duration event or
  later start may improve the consistency of load
  reductions and the cost effectiveness of the
  program

58
Table of Contents

• Executive Summary
• Pilot Background and Overview of Experimental
  Design
• Data Sources
• Analysis Methodology
• Load Impact Results
• Conclusions and Recommendations
• Appendix

59
Appendix

• Meter Data Comparison
• Low-Strata Issues
• Load Curves
• Methodology Coefficient

60
Invensys meter data serves as a good proxy for
utility meter data from ADRS homes, as
demonstrated by the July 22nd Event Day below
July 22nd Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
61
Weighting the raw data from A07 by the
distribution of the ADRS sample runs the risk of
skewing the results by placing undue emphasis on
the behavior of a single contributor
ADRS Population by Utility - Low Consumption
A07 Population by Utility - Low Consumption
of ADRS Population
of Homes
Utility
Source Utility Data, RMI analysis
Note ADRS and A07 strata classification based on
historical ADU.
62
However, exclusion of potentially skewed SDGE
data does not change the result among modest
energy users the remaining technology-enabled
homes still do not outperform A07
Average July Event Day - All Low Consumption Homes
With a single SDGE home comprising 32 of the
weighted load, the sample is sensitive to a
potential outlier (e.g., the demand spike at 815
a.m.)
? Super Peak ?
Low Stratum Electric Load per Home (kWh/hr)
Average July Event Day - SDGE Homes Omitted
Yet, even excluding the SDGE home, homes without
GoodWatts (A07) consume less than homes with it
(ADRS) at nearly every hour of the day
Time of Day
Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
63
July 14th Event Day
July 14th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
64
July 22nd Event Day
July 22nd Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
65
July 26th Event Day
July 26th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
66
July 27th Event Day
July 27th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
67
August 9th Event Day
Aug 9th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
68
August 10th Event Day
Aug 10th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
69
August 11th Event Day
Aug 11th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
70
August 27th Event Day
Aug 27th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
71
August 31st Event Day
Aug 31st Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
72
September 8th Event Day
Sept 8th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
73
September 9th Event Day
Sept 9th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
74
September 10th Event Day
Sept 10th Event Day Load Profile - All Homes
? Super Peak ?
Electric Load per Home (kWh/hr)
Time of Day
Source Utility Data, Invensys GoodWatts Reports
Server, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
75
MethodologyAs an indication of the statistical
quality of the results, the coefficient of
variation allows us to compare relative variation
between populations

• The coefficient of variation (CV), which allows
  for comparison of the relative variation of
  values between populations, is defined as the
  standard deviation (SD) of a sample divided by
  the samples mean value
• For ADRS, the mean and standard deviation was
  calculated for each time interval over the 5 hour
  peak period. A coefficient of variation was
  calculated for each consumption stratum according
  to Event and Non-Event days.
• The mean, SD, and CV were calculated for total
  consumption of each group of homes ADRS and
  control populations (A03 and A07)
• The mean, SD, and CV were calculated for the load
  reduction between the control group (A03 or A07)
  and ADRS homes
• A CV value greater than 1 implies that the
  statistic is not significant (less than 70
  confidence).
• A CV value less than 0.5 implies that the
  statistic is significant within a 95 confidence
  interval

76
Variation in total consumption is high across all
groups of homes for both high- and (especially)
low-consumption homes ADRS load reductions
relative to both control groups are statistically
significant for the high consumption homes
Source Utility Data, RMI analysis
Note A07 and A03 data scaled to match the ADRS
customers population by utility and strata
population.
77
High CVs for load-reduction in low-consumption
ADRS homes relative to A07 homes confirms our
hypothesis that those results are not
statistically significant, due to small and
diverse samples

• Variation in total consumption of ADRS and
  control groups is high for both high- and
  (especially) low-consumption homes.
• The CV declines when we look at the difference in
  consumption between ADRS and each of the control
  groups, particularly for Super Peak days. This
  suggests that the variations among the ADRS homes
  loads and the control groups are not independent,
  but are correlated, i.e., relatively high (or
  low) values tend to occur at similar times in
  each population.
• The CV values for load reductions of
  high-consumption ADRS homes relative to both
  control groups are substantially less