Residential Non-Intrusive Load Monitoring Overview
Presented by Matt Smith, SDG&E, Emerging Technologies
E Source Forum 2014
September 29–October 2, 2014
How much energy does your refrigerator use?
Why does it matter?
Utilities • Refine customer rebate
and incentive programs • Qualify new products for
residential rebates • Improve relationships
with customers • Understand customer
behavior to improve capacity planning
• Identify/verify appliances that could participate in Demand Response
Customers • Understand your bill • Plan your monthly budget • Identify/repair/replace
energy “hogs” • Take advantage of rate
structures (TOU/Tiers) • Be able to make a
financial decision for when to use an appliance
How much energy can we really save? ACEEE analyzed results from 36 different studies between 1995-2010 and found energy savings varied based on the data given to customers
How do we get that data to residential customers?
Circuit/Plug Monitoring Smart Appliances
Home Area Network / Smart Meter
Electricity Bill Green Button Data
Can we use smart meter data more intelligently?
wattseeker.com
What companies are pursuing NILM?
Belkin Enetics
Smappee Detectent
Bidgley Onzo
Blue Line Innovations PlotWatt
Energy Aware Navetas
Intel EEme
? LoadIQ
Verdigris ?
What did SDG&E do?
• Outfitted 10 Homes • Monitored all of the circuits in the home to get true
energy-usage values • Collected 10-second meter data via an approved HAN
device
• Solicited Disaggregation Vendors • Provided data to the vendors at different frequencies
(10s, 1 minute, 15 minute, 1hr) • Assessed accuracy of their disaggregation
What did we see in our study?
• Disaggregation accuracy improves with increased input frequency (HAN data better than GB)
• Home survey data is useful but not required • Zigbee & Internet signals are critical • Large and regular loads (pool pumps and electric
vehicles) are the easiest to disaggregate
EV Disaggregation
What are the challenges we saw?
• Vendors don’t all detect the same end uses Vendor 1 Vendor 2 Vendor 3 Vendor 4
Fridge X X X Microwave X X EV X X X Pool Pumps X X “Pool” Oven X Laundry “Washer/Dryer” “Cooking/ Laundry” “Tumble Dryer” “Dryer” Dishwasher X Base Load “Background Loads” “Always On” ”Other” ”Base” HVAC X “Heating/Cooling” “HVAC/Water Heating” “AC” WH “Water Heating” “Water Heater” Other “Solar (experimental)” ”Lifestyle”, “Heat”
• Vendors choose to report energy usage on different time scales Vendor 1 Vendor 2 Vendor 3 Vendor 4
Minute X Hour X X Day X X X Month X X
Additional challenges
• Whole-building measurement frequency needs to be high to disaggregate the smaller appliances
• Higher-frequency data is more expensive • In addition to being more expensive, high-
frequency measurements are also more sensitive to connectivity issues
• We saw 92% connectivity with a HAN gateway connected to the meter reading the meter every 10 seconds. (i.e. 8 out of every 100 readings was missing)
What’s next for NILM?
• Large-scale field testing: 100s or 1,000s of homes • Include AC run-times from smart thermostat data • Test with sub-one-second active power, apparent
power, and voltage loggers • Evaluate actual energy savings after installing or
repairing equipment • Test disaggregation of gas appliances
Thank you! Contact: Matt Smith, SDG&E Emerging Technologies
Accuracy results table