Keeping Down with the Jones

Download Report

Transcript Keeping Down with the Jones

Preliminary Impact Analysis of
SCE’s 04-05 HEES Program
Population Regression using Consumption Trend
of Participants’ Neighborhoods as Covariates
John Peterson, Athens Research
Carol Yin, Yinsight
CALMAC, October 17, 2007
1
SCE’s long standing interest in energy
savings from home audits

RER study (1997) of SCE’s 1993 & 1995
residential audit programs


Ridge study (2002, on CALMAC.org)


Net 1st year savings: In Home (433 kWh), Telephone
(154 kWh), Mail-In (123 kWh)
Net 1st year savings: In Home (440 kWh), Telephone
(185 kWh), Mail-In (123 kWh), Online (123 kWh)
BUT, savings estimates are outdated
2
Benefits of using the ecological
consumption trend

Consumption related variables arguably relatively
homogeneous at block group/dwelling level

Accounts for weather changes and difficulties
linking weather station data to small areas

Accounts for economic differences between areas,
and changes over time within those relatively
homogenous areas

Accounts for demographic features, cultural and
behavioral trends
3
Method

Models estimated at a gross population level: whether or not the
customer has participated in 04-05 HEES


Filtered 04-05 participants on the basis of:



Preliminary analysis does NOT track causal flow of savings from
recommendation adoptions
12 months of pre-audit consumption data, 12 months of post-audit
and post 10 month deadband data (others deadbands were tested).
The ecological consumption trend

Calculated from calendarized residential consumption data from
2002-2007

Averaged by dwelling type (single family/other) within geography
(block group/tract/zip) so that each participant compared with at
least 100 neighbors
Ecological consumption data scaled on participant by
participant basis, so pre-period average consumption of
participant and neighbors matched.
4
Preliminary Results
Based on simple ecological adjustment “meatgrinder”

OVERALL RESULT FOR SCE: < 200 kWh per year. IN HOME AUDIT about 50% more savings than
MAIL-IN.

LONG VERSION ON-LINE: Negligible savings, yet some evidence that install recommendations
lead to savings.
AUDIT - RECOMMENDATION
KWH/DAY
PATTERN
IMPACT
STDERR
KWH/YR
T_VAL
------------------------------------------------------------------------IH_AUDIT, ANY REC
-0.8621 0.0546
-314.7 -15.79
IH_AUDIT,PRACTICE
-0.8621 0.0546
-314.7 -15.79
IH_AUDIT, INSTALL
-0.8693 0.0561
-317.3 -15.50
MI_AUDIT, ANY REC
MI_AUDIT,PRACTICE
MI_AUDIT, INSTALL
-0.5802
-0.5805
-0.8110
0.0203
0.0203
0.0260
-211.8
-211.9
-296.0
-28.58
-28.60
-31.19
OL_AUDIT, ANY REC
OL_AUDIT,PRACTICE
OL_AUDIT, INSTALL
0.1005
0.1010
-0.2850
0.0360
0.0360
0.0554
36.7
36.9
-104.0
2.79
2.81
-5.14
5
Remaining Issues and Recommendations

Ecological trend consumption approach is an effective
way to isolate adjusted gross savings

Applicable to most consumption analyses of
residential and small commercial programs

Still need to track savings that flow causally through
adoptions of HEES recommendations

Need to understand measure-level impacts


But we have extreme collinearity due to…
…Too many measures. Huge numbers of
recommendations were made to every 04-05
participant.
6
More questions?
John Peterson
Athens Research
[email protected]
(626) 798-3147
Carol Yin
Yinsight
[email protected]
(626) 676-2198
7
Ecological Consumption Adjustment Approach: Path Model of Role of
Ecological Trend
PATH MODEL: ANY AUDIT/ANY RECOM, DEADBAND 10,
APPROX 36,000 SERVICE ACCOUNTS
1
3
U_ECOL.i,m
(av g nbhd
kwh)
0.5137.
0.8865
e3
.0.8931
CDD74.i,m
0.0338
5
U.i,m
(sv c acct
kwh
0.1316
-0.5065
2
0.0398
.01821
HDD63.i,m
-.0.01711
.0.0455
4
POST.i,m
(dum m y !)
1.
2.
3.
4.
e5
0.4230
0.9991
e4
Almost no direct impacts of weather upon customer consumption.
Virtually all of the weather impacts “flow through” the ecological consumption term.
The net neighborhood trend term is very strong, 0.8931.
Since weather only accounts for about 21% of neighborhood trend, this means that a great deal of the
neighborhood adjustment involves either non-weather variables or micro-climatic impacts that are not
well captured by weather stations. This unique impact is .8865 * .8931 or 0.792.
8