Exploring The Value of Energy Disaggregation through actionable feedback Nipun Batra , Amarjeet Singh, Kamin Whitehouse 14 May 2016 1
General eco feedback vs Actionable Feedback Eco feedback Misc. 22% Light HVAC 10% 56% Fridge 11%
General eco feedback vs Actionable Feedback Eco feedback Misc. 22% Light HVAC 10% 56% Fridge 11%
General eco feedback vs Actionable Feedback Eco Actionable feedback feedback Fridge consumption over 24 hours Home 1 700 Power (W) Misc. Misc. 525 22% 350 22% 175 0 Light Light HVAC HVAC 10% 10% Home 2 56% 56% 700 Power (W) 525 Fridge Fridge 350 11% 11% 175 0
General eco feedback vs Actionable Feedback Eco Actionable feedback feedback Fridge consumption over 24 hours Home 1 700 Power (W) High power Misc. Misc. 525 state 22% 350 22% 175 0 Light Light HVAC HVAC 10% 10% Home 2 56% 56% 700 Power (W) 525 Fridge Fridge 350 11% 11% 175 0
General eco feedback vs Actionable Feedback Eco Actionable feedback feedback Fridge consumption over 24 hours Home 1 700 Power (W) High power Misc. Misc. 525 state 22% 350 22% 175 0 Light Light HVAC HVAC 10% 10% Home 2 56% 56% 700 Power (W) 525 Fridge Fridge High 350 power 11% 11% state 175 0
General eco feedback vs Actionable Feedback Eco Actionable feedback feedback Fridge consumption over 24 hours Misc. Misc. 22% 22% Your fridge defrosts too much , wasting 30% energy Light Light HVAC HVAC 10% 10% Home 2 56% 56% 700 Power (W) 525 Fridge Fridge 350 11% 11% 175 0
Approach overview- How to give feedback Specific features of trace to infer why energy usage is high 700 Length of duty cycles 525 Power (W) 350 175 0
Approach overview- How to give feedback Specific features of trace to infer why energy usage is high 700 Actual power value 525 Power (W) 350 175 0
Feedback methods on Fridge and HVAC Others 38% HVAC 54% Fridge 8% Both appliances commonly found across homes
Evaluation overview Submetered traces Submeter sensor 700 Power (W) 350 0
Can we give such feedback? Submetered Disaggregated traces traces Smart meter 4000 Household 2000 aggregate 0 Submeter sensor NILM 700 4000 Power (W) 700 350 2000 Power (W) 350 0 0 0
Do disaggregated traces provide features needed for providing feedback? Submetered Disaggregated traces traces Smart meter 4000 Household 2000 aggregate 0 Submeter sensor NILM 700 4000 Power (W) 700 350 2000 Power (W) 350 0 0 0
Fridge is a duty cycle based appliance; compressor turns ON and OFF periodically 500 375 250 125 0
Defrost cycles occurs periodically and consume high amount of power 500 375 250 125 0
Defrost introduces heat increasing ON duration of next cycles 500 375 250 125 0
Fridge usage increases compressor ON durations (and reduce compressor OFF durations) 500 375 250 125 0
Night hours typically have “baseline” usage Baseline duty % = Median 700 duty % in the night 525 350 175 0
Defrost energy Defrost energy = Energy consumed in defrost state + Extra energy consumed in next few compressor cycles 700 525 350 175 0
Defrost energy Defrost energy = Energy consumed in defrost state + Extra energy consumed in next few compressor cycles 700 525 350 175 0
Usage energy Usage energy = Extra energy consumed over baseline 700 525 350 175 0
Experimental setup Wiki Energy data set 1. 97 fridges 2. 58 HVAC
13 out of 95 homes can be given feedback based on usage energy saving upto 23% fridge energy
13 out of 95 homes can be given feedback based on usage energy saving upto 23% fridge energy NILM algorithms show poor accuracy in identifying homes which can be given feedback based on usage energy
17 out of 95 homes can be given feedback on excess defrost saving upto 25% fridge energy
Such feedback can’t be given with disaggregated traces, since these techniques fare poorly on defrost detection.
Benchmark NILM algorithms on our data set give accuracy comparable or better than state-of-the-art Kolter 2012 Parson 2012 Parson 2014 Batra 2014 CO FHMM Hart 0 17.5 35 52.5 70 Error Energy %
“Average” error in energy would be low even if NILM predicted this 500 375 250 125 0
But, we wanted to predict.. 500 375 250 125 0
It’s the details that we care about 500 375 250 125 0
Like fridge, HVAC duty cycles to maintain the set temperature 4000 3000 2000 1000 0
As temperature increases during the day, more energy required to cool the home 4000 3000 2000 1000 0
People typically turn up the temperatures when they leave home 4000 3000 2000 1000 0
EnergyStar.gov recommended HVAC setpoint schedule Sleep Morning Work Evening 85 Recommended Min. Temp (F) 83 81 79 77 2 4 6 8 10 12 14 16 18 20 22 24
Setpoint schedule score Sleep Morning Work Evening 85 Recommended Min. Temp (F) 83 81 79 77 2 4 6 8 10 12 14 16 18 20 22 24
Setpoint schedule score Sleep score = 1 if sleep temp. > 82, (82-temp.)/4 if 78<sleep temp. <82 0 otherwise Sleep 85 Recommended Min. Temp (F) 83 81 79 77 2 4 6 8 10 12 14 16 18 20 22 24
Learning HVAC setpoint Weather Learnt setpoint 85 77 5 1015 20 4000 3000 2000 1000 0 HVAC trace
Giving feedback Don’t need feedback 85 Learnt setpoint 77 85 5 1015 20 77 5 10 15 20 Need feedback 4000 75 3000 69 2000 5 10 15 20 1000 0 Features from HVAC trace
84% accuracy on giving feedback using submetered traces 39
NILM methods give 15-30% worse accuracy for feedback 40
Benchmark NILM algorithms on our data set give accuracy comparable or better than state-of-the-art Batra 2014 CO FHMM Hart 0 7.5 15 22.5 30 Error Energy %
Morning hours which have lesser NILM accuracy are important for HVAC feedback Night Morning 24 minutes of HVAC usage (%) Error in prediction of 18 12 6 0 Hart FHMM CO
Conclusions Appliance level data does enable actionable energy saving feedback
Conclusions Appliance level data does enable actionable energy saving feedback BUT Results show that we need to revisit the metrics by which we measures progress
Recommend
More recommend
