leveraging big data to develop next generation demand side management programs and energy...
TRANSCRIPT
Leveraging Big Data to Develop Next Generation Demand Side Management Programs and Energy Regulations
Daniel Young, Energy SolutionsMike McGaraghan, Energy SolutionsNate Dewart, Energy SolutionsPat Eilert, PG&EDan Hopper, SCE
Moneyball Analogy #1
“Some of the scouts still believed they could tell by the structure of a young man’s face not only his character but his future in pro ball.”-Michael Lewis. 2003. Moneyball.
“All I have is the box scores.” -Bill James, 1977. Baseball Abstract. Sabermetrics statistician.
2
Conclusion: We can do better with more data, and the data we need is out there.
The Need for Data
The development of successful utility programs and energy codes and standards requires a LOT of data:
Base-case product performance
Tech options for higher efficiency/performance
Forecasts of future product performance trends
Incremental cost of improvement3
Example: Where Data Can Help
Image source: DOE study: Incorporating Experience Curves in Appliance Standards Analysis
4
More Stringent
Mor
e Co
st E
ffecti
veHighest NPV?Highest NPV?
Less
Conserva
tive
Case Study: LED Lamps
• Goal: Ensure minimum performance across several operating parameters for LED lamps: • Light color, light quality, efficacy, lifetime,
dimmability, etc.• Opportunity: LEDs and big data • LED technology is rapidly improving, while
costs are rapidly decreasing• Several existing databases to track product
performance• Many existing industry forecasts to calibrate
against• Looking beyond efficiency
5
Moneyball Analogy #2
“A hitter should be measured by his success in that which he is trying to do, and that which he is trying to do is create runs.” -Bill James.1979. Baseball Abstract. Sabermetrics statistician.
Conclusion: Focus on the right metrics and keep the end goal in mind.
6
2012 Analysis
• Approach:• 700 unique price points were manually collected for over 500 unique
lamp models (not new, definitely not big data)• Multi-variable regression model to analyze the dataset (a little new) 7
ENERGY STAR?
CRI
CCT
Power Factor
Wattage
Efficacy
Light Output
Bulb Shape
Dimmability Lifetime
Price Modeling – 2012 Data
8
Note: Results based on online retailer data, which we found to be significantly higher on average than in store prices.
Moneyball Analogy #3
“The power of statistical analysis depends on sample size…a right-handed hitter who has gone two for ten against left-handed pitching, cannot as reliably be predicted to hit .200 against lefties as a hitter who has gone 200 for 1,000.”-Michael Lewis. 2003. Moneyball.
Conclusion: We could use some more data.
9
Next Step: Bringing in Big Data
• Retailer-based web crawler tool:• screen-scraping methods • retailer provided APIs (Application Programming
Interfaces) • Scope of data collection:• Nine online retailers• 3,000 unique price points• 1,000 unique LED lamp models• 50 different manufacturers• Data collected weekly 10
2012 Data vs 2014 Data
11
Watts
Lamp Sh
ape / Typ
e
Lumens
Lumen M
aintenance
Color Tempre
rature
Efficacy
(lpw)
Dimmable (Y
/N)
Energy S
tar Qual. (
Y/N)
Warra
nty CRI
Power Facto
r
Beam Angle
Input Volta
ge R9
Color Consis
tency /
Change
Candlepower (intensit
y)
Product
Weight
Power Typ
e (AC/D
C)0
500
1000
1500
2000
2012 Data - Manually Collected 2014 Data - Web-Crawler
Lamp Property
# Pr
oduc
ts
Note: 2014 data is refreshed every week
Benefits of Big Data
• More data -> improvements to the regression analysis:• Individual models could be created for each lamp type• Additional independent variables analyzed• Comparable or improved explanatory power for each
model• New data is collected each week with minimal effort• Ability to monitor real-time performance and price
changes• Observe trends in performance and price
12
Example Regression Results
Best fit model is based on:• Lumens • Brand• Energy Star Qualified
Metrics not independently impacting price include:• Dimmable• Color Temperature• CRI• Wattage• Beam Angle• Warranty Length• Diameter• Efficacy• Lumen Maintenance
13
Observed Trends
14
Implications on IMC
15
No more inc cost for CRI?
Back to the Future• Key questions for IDSM program development and codes
and standards advocacy/evaluation:
What’s the baseline performance?
How do the best products perform?
How is performance changing over time?
What’s the incremental cost?
16
Summary
Major Benefits
• Major increase in data volume and accuracy
• Better data for more effective programs and codes
• Saves significant time and resources over existing methods
Outstanding Issues
• How to use the data most effectively
• Linking to product performance databases
• Inconsistent retailer info and labeling
• Legality of web-crawling 17
Moneyball Analogy #4
“Statcast, a 3-D tracking system that provides detailed metrics on the locations and movements of the ball, the players, and even the umpires…will proliferate not just through the ranks of all professional sports but to youth sports, affecting everything from how games are taught to the statistical nomenclature of sport”-Billy Beane. July 7, 2014. “A Tech-Driven Revolution.” Wall Street Journal
Conclusion: The opportunities for big data have only just begun.
18