Loan Nguyen Southern California Gas Company August 30 2017 Food for Thought Ingredients for a Delicious MeterBased Restaurant Program Agenda Introduction Approach Model Selection Results ID: 646163
Download Presentation The PPT/PDF document "Amy Allen, David Jump, Raphael Vitti, kW..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Amy Allen, David Jump, Raphael Vitti, kW Engineering Loan Nguyen, Southern California Gas Company August 30, 2017
Food for Thought: Ingredients for a Delicious Meter-Based Restaurant Program Slide2
AgendaIntroduction ApproachModel Selection
ResultsConclusionQuestions? Slide3
Meter-Based M&V Concept3Slide4
Motivation
California has ambitious efficiency goals Recent legislation in California has encouraged utilities to pursue programs using “normalized metered energy consumption” in M&V approach
CA gas utility running a pilot program for restaurants
Wanted to determine how well models would meet accuracy criteria, and what level of savings is required Slide5
Overview of Study Analyzed sample of restaurants that participated in SoCal Gas’ energy efficiency programs between 2013 and 2015
Developed models for calculating bldg. natural gas usage as a function of weather conditions, and applied best-fitting model to calculate energy savings using IPMVP Option C Assessed this approach based on level of savings verified, & associated uncertainty Slide6
Key Research Questions What models are most appropriate for M&V analysis? How accurately can we predict energy use and savings?
How much savings for a typical restaurant is required? How do M&V calculated savings compare with “deemed” values? Slide7
Characterization of Sample
By Baseline Gas Usage
Sample of 433 restaurants with complete data available, all located in southern California
Energy efficiency measures grouped in two categories: cooking equipment and water heating Slide8
Development & Selection of Model Weather conditions (degree days or outdoor air temperature) are often used as independent variable
Analyzed four different statistical algorithms (change-point models) to predict monthly natural gas usage over baseline period 2-parameter (linear regression) with HDD and average monthly temperature
3-parameter and 4-parameter with average monthly temperature
Selected model based on accuracy over baseline period, minimizing CV(RMSE), which characterizes the random modeling error Slide9
Distribution of CV(RMSE) Slide10
Calculating Savings & Uncertainty Calculated avoided energy use per IPMVP Option C Used method outlined in ASHRAE Guideline 14 for calculating uncertainty
Set threshold of 50% savings uncertainty at 95% confidence interval, exceeding ASHRAE Guideline 14 recommendation Slide11
Magnitude of Uncertainty at 95% CI vs. Gas Usage for All SitesSlide12
M&V Savings vs. Deemed Savings
M&V savings for vat fryers were lower than deemed values at many sites
Water heating measures were disproportionately represented in the “low savings” subset (not shown)
Further investigation needed: site size and savings compared to deemedSlide13
Results Summary Slide14
Conclusions 2-P model based on heating degree days generally models gas usage well for restaurants in sample
Total energy savings calculated through M&V are significantly less than total deemed savingsWhen baseline model fits well, range of savings <10% can be detected For over 90% of sites in sample, could either determine savings with acceptable uncertainty(41%), or conclude that savings were low (51%)Slide15
Additional Data NeedsAdditional independent variables (meals served, water consumption) could make this analysis more comprehensive
With more meta-data, could draw more conclusions about what type of sites are good candidates for this approach Installation dates for multiple measures would allow consideration of sites with more than one measure installed, and interactive effectsSlide16
Additional Slides for Reference Slide17
Areas for Further Study Slide18
Change-Point Models (2-Parameter) Slide19
Change-Point Models (3-Parameter)Slide20
Baseline Model CV and Gas Usage Slide21
Average CV(RMSE) by Model Type Slide22
Baseline CV(RMSE) for 2-P Model Slide23
Save The Dates
AESP’s National Conference
New Orleans, LA
AESP’s Spring Conference
Atlanta, GA
For more information -
www.aesp.org
February 19-22, 2018
May 21-23, 2018