MELS Project Wrap Up
Rich Brown, David Culler, Stephen Dawson-Haggerty, Steven Lanzisera, Jay Taneja
Computer Science Division, University of California, BerkeleyEnvironmental Energy Technologies Division, Lawrence Berkeley National Lab
MELS: Miscellaneous Electric Loads
• Large, rigorous study of miscellaneous electric loads (mostly plugs)– Roughly 1/3 of building energy consumption– Difficult to study due to large number of small
consumers• Test methods to accurately describe
energy use of plug-in devices from individual device to whole building
September 13, 2011 LBNL EETD Seminar Series 2
DOE MELS => Appliance Energy
8/2/2011 CPS-PI-11 3
• 450 plug-load meters• 7 edge routers• 650m data points
Study Phases
Hardware (2009)
• Calibration• Safety
testing
Testing and Setup
• Software validation
Installation
• Device inventory• Stratified sampling
Operation (2011)
• Maintenance and debugging
September 13, 2011 LBNL EETD Seminar Series 4
System Architecture
September 13, 2011 LBNL EETD Seminar Series 5
• IPv6 tunnel from building network to data closet• Data in raw UDP• Configurable metering application
Multipoint Calibration
September 13, 2011 LBNL EETD Seminar Series 6
• Automated 20-point calibration on every meter• Generate a 3-part piecewise calibration
• 90th percentile error is <2 watts across devices
HYDRO Principles
• Maintain multiple next-hop options• Trickelize density-sensitve state propagation• Horizontally scalable with multiple LBRs
September 13, 2011 LBNL EETD Seminar Series 8
Emergent network dynamics
September 13, 2011 LBNL EETD Seminar Series 9
• Verify dynamics results on large scale– Link and device churn are prevalent
• Mean network degree is at least 16, diameter is about 4.5 hops
Exploration is ongoing
September 13, 2011 LBNL EETD Seminar Series 10
• Path length and router degree show clear diurnal and weekly variation
• Exploration of new potential candidate links is a continuous process
• With only “stable” links, diameter increases by factor of 2
ENERGY ANALYSIS
September 13, 2011 LBNL EETD Seminar Series 11
Can we identify when metered devices change over time?
12
From data, found change from older 20” LCD to new 24” LCD
Increase screen area 44%; reduce energy 33%.
13
How Common is Computer-Display Power Management?
40 Hour Work Week
PM w/ breaks
Rarely power down monitor
• 83% of monitors use power management
• 15% use it with breaks for days at a time
• 2% do not use it
14
What is the Distribution of LCD Computer Display Energy Use?
N=118
15
How Common is Desktop Computer Power Management?
40 Hour Work Week
39% rarely powered down
44% managed
How Much of Whole Building is Plugs?
16
All Building Electricity
40% of Building Electricity
3 month weekday average:March, April, May
Note: no cooling during these months
Projected based on full inventory and sample weights
All Plugs
17
What Makes Up Bldg 90 Plugs Energy?
Computers50% of energy
Displays10% of energy
Task Lighting 7% Networking 6%Other 7%
Imaging10% of energy
Misc. HVAC10% of energy
Timer controlled plug strips?75 MWh/year30% of non-computer plug total 6% of building total
Computer power management?150 MWh/year60% of computer total 12% of building total
18
Building 90 Device Inventory & Energy
19
Findings and Next Steps
• Bldg 90 network demonstrated large-scale, end-to-end WSN and collected a lot of useful data– IT equipment should be focus of office energy management
programs– Using data for LBNL-wide plug-load management
• Inventory and meter installation are labor-intensive– Exploring using public (homeowner & building occupant)
participation for data collection– Integrate metering & communications into products
• Robust sensor network needs more engineering– ? Evaluate commercial products now available
• Electricity only part of buildings energy problem– Developing low-cost WSN for gas and water metering
QUESTIONS
September 13, 2011 LBNL EETD Seminar Series 20