SEEM Calibration: Phase-2 Adjustments for Failed VBDD Fits RTF Calibration Subcommittee July 31, 2014.

Slides:



Advertisements
Similar presentations
Simple Linear Regression 1. 2 I want to start this section with a story. Imagine we take everyone in the class and line them up from shortest to tallest.
Advertisements

1 Residential Weatherization Calculator Deriving Realistic Savings Estimates for the “Uninsulated” House.
SEEM Calibration: Revisited Revising the regression to use continuous heat loss variable Regional Technical Forum December 17, 2013.
New Construction Calibration Research Results and Request for Decision Regional Technical Forum March 18, 2014.
Topic 12 – Further Topics in ANOVA
Residential Single Family Weatherization and HVAC Measures Progress Reports: 1.Estimating Electric and Supplemental Fuels Savings 2.Estimating Value of.
Statistics: Data Analysis and Presentation Fr Clinic II.
1 Qualitative Independent Variables Sometimes called Dummy Variables.
© 2003 Prentice-Hall, Inc.Chap 14-1 Basic Business Statistics (9 th Edition) Chapter 14 Introduction to Multiple Regression.
Statistics: Data Presentation & Analysis Fr Clinic I.
Ch. 14: The Multiple Regression Model building
Estimating Heating Energy from Utility Bills More on the “Phase 2” Adjustment Ecotope, Inc.
Electric “Grid” Savings and Non-Electric Benefits for Residential HVAC-effected UES Measures Regional Technical Forum March 20, 2013.
DHP Supplemental Fuel Screen Subcommittee Meeting Regional Technical Forum December 10, 2013.
Manufactured Homes Calibration: Existing and New Homes Mohit Singh-Chhabra & Josh Rushton RTF Update March 17, 2015.
Manufactured Homes Calibration: Existing and New Homes Mohit Singh-Chhabra & Josh Rushton RTF Update May 12, 2015.
Research Strategy Review: Heat Pump Water Heaters Thermostatic Restriction Valves Jennifer Anziano RTF R&E Subcommittee July 8, 2015.
1 Quality Control Review of E3 Calculator Inputs Comparison to DEER Database Brian Horii Energy and Environmental Economics, Inc. November 16, 2006.
DHP for Houses with Electric FAF Provisional UES Measure Proposal Adam Hadley, Ben Hannas, Bob Davis Regional Technical Forum January 21, 2015.
Inference for regression - Simple linear regression
Manufactured Homes Calibration: Existing and New Homes Mohit Singh-Chhabra & Josh Rushton RTF MH Calibration Subcommittee February 27, 2015.
Manufactured Homes Calibration: Existing and New Homes Josh Rushton & Mohit Singh-Chhabra RTF Update June 16, 2015.
1 NORTHWEST ENERGY EFFICIENCY ALLIANCE Creating a DHP UES Measure Ecotope, Inc. July 16, 2013.
SEEM 94 Calibration to RBSA Data Progress Report on Phase 2 (Digging a Little Deeper) Subcommittee Regional Technical Forum March 20, 2013.
DHP for Houses with Electric FAF Research Plan: Revisions Adam Hadley, Ben Hannas, Bob Davis, My Ton R&E Subcommittee February 25, 2015.
Sample size vs. Error A tutorial By Bill Thomas, Colby-Sawyer College.
Scientific Irrigation Scheduling provisional analysis and research plan BPA December 2014.
Applying SEEM Updates, Calibration, and Measure Interaction Decisions to: Manufactured Homes Weatherization UES Measures Regional Technical Forum June.
1 NORTHWEST ENERGY EFFICIENCY ALLIANCE Northwest Ductless Heat Pump Pilot Project Impact & Process Evaluation: Billing Analysis Ecotope, Inc. February.
QC Follow-up for Single Family SEEM Calibration, HVAC UES Measures, and Weatherization Measures Adam Hadley Regional Technical Forum May 12, 2015.
SEEM Calibration: Phase-2 Adjustments for Failed VBDD Fits Regional Technical Forum August 12, 2014.
EvergreenEcon.com ESA 2011 Impact Evaluation Draft Report Public Workshop #2 August 7, 2013 Presented By: Steve Grover, President.
DHP Supplemental Fuel Screen Subcommittee Meeting Regional Technical Forum February 12, 2014.
Analysis of Weatherization Measures Options for Savings Methodology and Presentation of Costs from 6 th Plan Regional Technical Forum May 4 th, 2010.
Evaluation Results of the 2004 & 2005 California Statewide ENERGY STAR® New Homes Program Clark Bernier Presented at the October 17, 2007 CALMAC Meeting.
An Approach to Adjust ERCOT’s Long-Term Load Forecast for Utility Energy Efficiency Programs Jay Zarnikau Frontier Associates LLC June 10, 2013.
SEEM Input Variables for RBSA Calibration Manufactured Homes RTF Subcommittee on SEEM MH Calibration May 9, 2013.
Path for Multi-Family (MF) Weatherization and NC Measures Christian Douglass 8/18/2015.
Applying SEEM Updates, Calibration, and Measure Interaction Decisions to: Single Family Weatherization and HVAC UES Measures Follow-up from August RTF.
SEEM Calibration: Phase II Single Family Heating Energy Regional Technical Forum September 17, 2013 Presented By: Josh Rushton and Adam Hadley Subcommittee.
SEEM Calibration: Phase II Single Family Heating Energy Regional Technical Forum August 20, 2013 Presented By: Josh Rushton and Adam Hadley Subcommittee.
SEEM Calibration for Manufactured Homes Regional Technical Forum July 15, 2014.
Copyright © 2014, 2011 Pearson Education, Inc. 1 Chapter 19 Linear Patterns.
MGS3100_04.ppt/Sep 29, 2015/Page 1 Georgia State University - Confidential MGS 3100 Business Analysis Regression Sep 29 and 30, 2015.
1 NORTHWEST ENERGY EFFICIENCY ALLIANCE DHP Costs & NEBs Ecotope, Inc. August 20, 2013.
Research Strategy Review: Advanced Power Strips MH HVAC Related Measures Jennifer Anziano RTF R&E Subcommittee August 6, 2015.
2009 Impact Evaluation Concerns ESAP Workshop #1 October 17, 2011.
NEEA DEI Study Data Analysis Plan October 28, 2005 RLW Analytics, Inc. Roger L. Wright, Chairman, and Principal Consultant.
Electric “Grid” Savings and Non-Electric Benefits for Residential HVAC-affected UES Measures Regional Technical Forum March 20, 2013.
Ductless Heat Pumps (DHP) in Single Family Homes with Zonal Electric Heat UES Measure Update Regional Technical Forum March 18, 2014.
1 Proposed Changes to the RTF’s Heat Pump Deemed Savings and Cost- Effectiveness Draft 6th Plan Assumptions April 7, 2009.
Ductless Heat Pumps (DHP) in Single Family Homes with Zonal Electric Heat Proven UES Measure Proposal Regional Technical Forum October 16, 2013.
ANOVA, Regression and Multiple Regression March
Heat Pump Research Project Sponsored by the Heat Pump Working Group April 5, 2005.
Working Towards Better Savings Estimates for HVAC and Weatherization Measures Regional Technical Forum September 16, 2014.
ENERGY STAR and Eco-Rated Homes: Planning Estimates and Research Strategy Regional Technical Forum December 8 th, 2015 Josh Rushton & Mohit Singh-Chhabra.
Multifamily Homes: Calibration Update Regional Technical Forum December 8 th, 2015.
New Construction Calibration Preliminary Research and a Request for Guidance on Issues Raised in December Regional Technical Forum January 22, 2013.
Applying SEEM Updates, Calibration, and Measure Interaction Decisions to: Manufactured Homes Weatherization UES Measures Regional Technical Forum December.
Where did it go? Lost savings found in real-world data SEEM Calibration for the Northwest Power and Conservation Council’s Regional Technical Forum (RTF)
Research Strategy: New-Construction Manufactured Home Measures Research and Evaluation Subcommittee Josh Rushton November 19, 2015.
Idaho and Montana Residential Single Family New Construction Measures Mohit Singh-Chhabra Regional Technical Forum October 20 th, 2015.
RTF New Homes Subcommittee February 11, 2016 Next Step Homes Update.
Extra Slides to Accompany: SEEM 94 Calibration to Single Family RBSA Data Analysis and proposed actions Regional Technical Forum May 21, 2013.
Non-Proven Residential Duct Sealing Measures: Research Strategy Josh Rushton and Adam Hadley Research and Evaluation Subcommittee October 7, 2015.
RTF Management Updates Jennifer Light Regional Technical Forum February 17, 2016.
Ductless Heat Pumps (DHP) in Single Family Homes with Zonal Electric Heat UES Measure Update Regional Technical Forum June 17, 2014.
Research Strategy: Residential Ductless Heat Pumps Research and Evaluation Subcommittee Christian Douglass and Josh Rushton January 8, 2016.
Data Screening. What is it? Data screening is very important to make sure you’ve met all your assumptions, outliers, and error problems. Each type of.
Manufactured Homes: Heat Pump Related Measures Regional Technical Forum Presentation August 18 th, 2015.
Presentation transcript:

SEEM Calibration: Phase-2 Adjustments for Failed VBDD Fits RTF Calibration Subcommittee July 31, 2014

Agenda 1.Problem overview and background 2.Proposal A 3.Proposal B 4.Discussion

1. Problem Overview

Calibration Review Phase I: Estimating total heating energy. – Align SEEM with billing data for homes with strong and clear heating energy signatures and no off-grid fuels. Phase II: Estimating electric heating energy in “typical” program homes. – How is electric heating energy affected by the presence of natural gas and off-grid fuels? – What can we say about electric heating energy in homes with weak or unclear heating energy signatures? 4

Phase II general approach 5 Phase I gives total heating energy estimates for homes with clear VBDD signatures. RTF measure savings needs average electric energy savings for all program homes. Phase II uses regression to find out… – How the presence of non-electric fuels affects electric heating energy, – How heating energy differs in homes with unclear VBDD signatures. Regression focuses on TMY-normalized (VBDD) estimates derived from electric billing data.

Example: SF Phase II regression 6 VariableDefinitionEstimateStd. ErrorP-value Intercept Natural log of UA x HDD Natural log of the area of conditioned floor space Indicator: Has electric forced-air furnace Indicator: Has heat pump Indicator: Non-utility fuel over 40,000 kBtu/yr Indicator: HZ1 and non-utility fuel between 5,000 and 40,000 kBtu/yr Indicator: Gas heating energy over 5,000 kWh/yr Indicator: Failed SEEM Calibration billing analysis filter

Example: MH Phase II regression 7 VariableEstimateStd. Errort valuep value Intercept ln(UA × HDD65) Heat pump Gas Ht. (kWh) 4K to 8K Gas Ht. (kWh) Over 8K Wood (kWh) 6K to 12K Wood (kWh) Over 12K Bad VBDD fit Adjusted R-square: 40%

The issue Regressions show that all else equal, VBDD heating energy estimates tend to be lower in homes with “poor” VBDD fits. But VBDD fits are “poor” for these sites! Should we really conclude that actual heating energy is lower?

What’s at stake? In homes with poor VBDD fits, VBDD estimates are around... o 1 - exp(-0.42) = 34% lower in SF homes; o 1 - exp(-0.33) = 28% lower in MH homes. About 20% of SF homes and 28% of MH homes have poor VBDD fits. Net result is 6% to 7% less heating energy across the full population.

Additional Notes In addition to VBDD-normalized estimates, the RBSA also provides “annualized” energy estimates. Annualized estimates are… Simple sums of 1 year’s billing records (maybe a couple months imputed if necessary) Very reliable (no regression fit, no modelling error) All end uses together – no separate heating energy estimate.

Additional Notes All else equal, homes caught by “poor VBDD” filter have lower annualized energy values. Regression-based quick estimate (accounts for UA*HDD, square footage, etc.) puts them around 2800 kWh lower kWh is about 14% of average total energy; 2800 kWh is about 28% of average heating energy (per VBDD estimates with good fits);

The big question So “poor-VBDD” homes have around 2800 kWh less than their “good-VBDD” peers. How is the missing 2800 kWh distributed across end-uses? Mostly missing heating energy? From all end-uses equally? Something else?

One more bit of data Annualized estimates available for full RBSA but don’t separate out heating energy. RBSA Meter Study has a lot of detail for a small number of sites. – For sites in the meter study, we have pretty good information on actual heating energy. – 40 of these are in the SF Phase II sample. 12 with “poor” VBDD fits 28 with “okay” VBDD fits

2. Proposal A

Summary Proposal A: Take the existing Phase-II regression results at face value. Supporting claim: Homes with poor VBDD fits really do have much lower heating energy (on average) than those with okay fits. It’s mostly heating energy that’s missing from these homes’ annualized energy values. Reasons…

VBDD estimates mostly agree with our “truth set”, except for a few sites. VBDD is actually too high in these sites. Suggests that for bad-VBDD sites, actual heating energy is even lower than what VBDD says. Warning: Note mismatched time periods. Reason 1, VBDD vs “Truth”

Meter study sample probably not representative in all respects we care about. – In the broader RBSA sample, VBDD estimates tend to be lower for poor-VBDD-fit sites than for okay-VBDD-fit sites. – Reverse holds in previous plot. Discrepancy partially explained by confounding variables (UA*HDD, wood heat, etc.); After accounting for these, t-test for difference in means has p = In any event, meter study sample gives no evidence that VBDD is artificially low in poor-VBDD sites; If anything, it suggests actual heating energy is lower still in these sites. Reason 1, continued

Graph uses meter- study data; values are “real”. 9 of the 12 bad-VBDD sites lie below the okay-VBDD trend. Meaning: Relatively small fraction of their total energy is heat- related. Two main exceptions, and 10877; VBDD- and meter- based estimates agree for these. Reason 2, Heating kWh vs Total kWh

Among sites in the meter study, most of those with “poor” VBDD fits use less heating energy, as a fraction of total energy, than those with okay fits. There are two serious exceptions, but VBDD appears to have gotten these sites “right.” (For whatever that’s worth.) As support for Proposal A, Reason 2 is sort of a mixed bag. Reason 2, continued

3. Proposal B

Summary Proposal B: Treat the “poor” VBDD fit sites’ missing 2800 kWh as coming from all end uses equally, so everything is reduced by about 14%. This leads to a net reduction of about 3% to 4% in population-average heating energy (instead of 6% to 7%). Supporting claim: VBDD is totally detached from reality for some of these homes so we shouldn’t trust it at all. We have no clear basis for dividing the missing kWh among end uses, so the conservative move is to assume a uniform allocation. Reasons…

Sometimes a “poor” VBDD fits are total nonsense. VBDD-based estimates in such cases may as well be taken directly from a random number generator. Reason 1, total VBDD failures

Reason 2, heating fractions This is another perspective on heating kWh versus total kWh. We saw that this comparison is a mixed bag as support for Proposal A. This supports Proposal B by highlighting how little we really know.

4. Discussion

Subcommittee notes

Subcommittee recommendation

Supplementary slides

CoefficientDescriptionEstimate*Standard Errorp-val β0β0 Intercept <.001 β1β1 Floor Area (ft 2 ) <.001 β2β2 UA * HDD65 (kWh/yr) <.001 β3β3 Heat Pump (TRUE/FALSE) β4β4 Failed Filter (TRUE/FALSE) < ft 2 higher area  360 higher annualized kWh 1000 kWh higher UA * HDD65  120 kWh higher annualized kWh Presence of Heat Pump  60 kWh lower annualized kWh (not significant) Failed Bill Filter  2800 lower annualized kWh *Estimated w/ RBSA survey weighting Raw kWhAdjusted kWh Passed Filter Failed Filter Difference

Summary of VBDD failures Site IDReason FailedDiagnosisMeter heating kWhVBDD vs Meter 10887Tbal 70F, Low R^2Malfunctioning HP (?)HighAgree 11418Tbal 48FProbably actual TbalNormalAgree 12507Tbal 70FSpaNormalVBDD too high 13912Tbal 48FNo utility heatingLowAgree 14140Tbal 70FShop (?)NormalVBDD much too high 20020Tbal 48FSpaNormalVBDD much too high 20230Tbal 70FPossibly actual TbalNormalAgree 20469Tbal 48FLarge cooling loadNormalAgree 20998Tbal 48FCooling, Spa, & Well PumpNormalAgree 21143Tbal 48FBad ReadLowAgree 23960Low R^2Bad ReadNormalAgree 24203Tbal 48FProbably real TbalNormalAgree 24375Low R^2Vacations & SpaLowAgree 24684Low R^2Possible Bad ReadsNormalVBDD too high

Annotated VBDD vs “Truth” Shop, sump Spa Bad meter read Among “poor” VBDD fit sites in the meter study sample, the sites with large VBDD errors mostly have large additional loads unrelated to space heat.

Annotated Heating kWh vs Total kWh HP Malfunction Possible real T-Bal ≈ 70 Cooling, spa, well pump Cooling Among “poor” VBDD fit sites in the meter study sample… Large heating fractions are associated with a lot of heating equipment usage; Sites with small heating fractions have different causes (little heat or lots of other stuff). No utility heat