1. The Self-Programming Thermostat: using occupancy to optimize setback schedules Kamin Whitehouse Joint work with: Gao GeBuildSys University of Virginia Nov 3, 2009

2. Problem Definition Goal in US: Reduce energy usage in existing buildings by 25% by 2020

3. Problem Definition

4. But reducing HVAC energy --> $$ –Insulation, new windows, solar panels, geothermal, HVAC upgrades, etc. –All require $1000’s and take many years for ROI Federal stimulus: $5 billion for weatherization of low-income homes –Small % of target savings We need low-cost energy solutions

5. State of the Art Setback Schedules Widely-accepted Cost-effective But still largely untapped potential! –Why? Occupancy Time Temp Setpoint Setback

6. State of the Art Too much hassle! X X

7. Occupancy State of the Art Occupancy Time Temp Setpoint Setback Start timeEnd time Miss time

8. Self-Programming Thermostat Occupancy Time Temp Setpoint Setback Start timeEnd time Miss time Occupancy Leave DistributionArrive Distribution

9. Self-Programming Thermostat Occupancy Time Temp Setpoint Setback Start timeEnd time Miss time Occupancy Leave DistributionArrive Distribution

10. Self-Programming Thermostat Occupancy Time Temp Setpoint Setback Start timeEnd time Miss time Occupancy Leave DistributionArrive Distribution 1. User specifies miss time 2. Thermostat maximizes setback period wrt miss time

11. Self-Programming Thermostat Occupancy Time Temp Setpoint Setback Start timeEnd time Miss time Occupancy Leave DistributionArrive Distribution 0 Hrs miss time

12. Self-Programming Thermostat Occupancy Time Temp Setpoint Setback Start timeEnd time Miss time Occupancy Leave DistributionArrive Distribution 0.5 Hrs miss time

13. Self-programming Thermostat Occupancy Time Temp Setpoint Setback Start timeEnd time Miss time Occupancy Leave DistributionArrive Distribution 1 Hr miss time

14. Self-Programming Thermostat Miss time knob allows user to navigate the Pareto optimal set of schedules

15. User Interface Three knobs: setpoint, setback, miss time As the user tunes the knobs, the system displays resultant schedule and energy usage Result: explicit energy/comfort trade-off – controllable and predictable – not smart!

16. Sensing Occupancy Time of day from 0 hours (12 AM) Front Door Bathroom Motion Kitchen Motion Bedroom Motion Sense occupancy $50-100 per home No cameras/tags

17. Sensing Occupancy True Positive Rate Event Detection Rate Duration Accuracy UbiComp ‘08

18. Evaluation Two publicly-available data sets – Kasteren – Tulum (not a random sample) – Both ~1 month – Hand-labeled many activities – We used “leave home” and “return home”

19. Evaluation

26. Summary Use sensors to identify occupancy Automatically tune setback schedules Use miss time knob to navigate Pareto set Benefits – Simple interface – More energy savings; same comfort – More comfort, same energy savings – Cheap! $50-$100 per home

27. Other Related Work Reactive Thermostats – Similar to motion-sensor triggered lights Microenvironments – User-controlled local conditioning Facilities management and building operators

28. Future Work More users, deployments and Energy – Spoiler alert! Results still good with 44 users and 8 homes with sensors, w/ heat pump Micro-zoning control Other building types Market penetration: UI & Economics.

29. Questions?

30. Deployment Details for FATS Demonstration Eight homes deployed with wireless X10 sensors for at least 7 days with an X10 receiver to record messages Four diverse single person homes, four diverse multi-person homes