Supero: A Sensor System for Unsupervised Residential Power Usage Monitoring Dennis E. Phillips 1 , Rui Tan 2 ; Mohammad-Mahdi Moazzami 1 ; Guoliang Xing 1 ; Jinzhu Chen 1 ; David K. Y. Yau 2,3 1 Michigan State University, USA 2 Advanced Digital Sciences Center, Singapore 3 Purdue University, USA 1 / 23
Outline • Motivation & Approach • Light Sensing • Acoustic Sensing • Implementation & Experiments • Implementation & Experiments 2 / 23
Residential Electricity in U.S. • Residential electricity Industrial Residential 25.5% – Largest sector 36.7% Others Commercial 34.2% • Rising cost • Rising cost Electricity retail sales in – Increase by 75% in 10 years U.S. 2011 [US EIA-861, EIA-923] Appl. Joul % When? • Understanding usage Bed light 5% 7pm-11pm Fridge 8% Every 1h – Real-time power readings Space 30% Jan 1 … heater – Fine-grained usage info …. …. …. 3 / 23
Related Work • Direct sensing – ACme [IPSN’09] Per-appliance inline meter, intrusive [Jiang IPSN’09] 4 / 23
Related Work • Direct sensing – ACme [IPSN’09] Per-appliance inline meter, intrusive [Jiang IPSN’09] • Indirect sensing • Indirect sensing – At-the-flick [UbiComp’07] High-rate ADC, in-situ training 5 / 23
Related Work • Direct sensing – ACme [IPSN’09] Per-appliance inline meter, intrusive [Jiang IPSN’09] • Indirect sensing • Indirect sensing – At-the-flick [UbiComp’07] High-rate ADC, in-situ training – ViridiScope [UbiComp’09] Labor-intensive sensor installation [Kim UbiComp’09] 6 / 23
Objective & Challenge • Fine-grained usage monitoring – Accurate energy disaggregation – Inexpensive and easy-to-install sensors – Training-free, ad hoc system deployment (“place sensor on shelf facing light to be monitored”) sensor on shelf facing light to be monitored”) 7 / 23
Objective & Challenge • Fine-grained usage monitoring – Accurate energy disaggregation – Inexpensive and easy-to-install sensors – Training-free, ad hoc system deployment (“place sensor on shelf facing light to be monitored”) sensor on shelf facing light to be monitored”) • High-degree sensing uncertainty – Noises from environment and human activities – Source appliance identification • A sensor can sense multiple appliances • An appliance can be sensed by multiple sensors 8 / 23
Supero Smart meter Base station Light and acoustic sensors Light + acoustic captures 90% power consumption 9 / 23
Supero Smart meter 100W Base station ‘+1’ Event Correlation Light and acoustic sensors (remove false alarm) Light + acoustic captures Light/acoustic event Power reading 90% power consumption 10 / 23
Supero Smart meter 100W Base station ‘+1’ Event clustering Event Correlation Light and acoustic sensors (remove false alarm) Light + acoustic captures Light/acoustic event Power reading 90% power consumption 11 / 23
Supero Smart meter 100W Base station ‘+1’ Event-Appliance Association Association Event clustering Event Correlation Light and acoustic sensors (remove false alarm) Light + acoustic captures Light/acoustic event Power reading 90% power consumption 12 / 23
Supero Smart meter 100W Base station ‘+1’ Event-Appliance Association Association Event clustering Event Correlation Light and acoustic sensors (remove false alarm) Light + acoustic captures Light/acoustic event Power reading 90% power consumption 13 / 23
Outline • Motivation & Approach • Light Sensing • Acoustic Sensing • Implementation & Experiments • Implementation & Experiments 14 / 23
Event Detection & Correlation light sensor reading exponential diff Time (second) Time (second) • Exponential difference filter – Diff between long-/short-term moving averages 15 / 23
Event Detection & Correlation light sensor reading Light off Report event exponential diff Report event Time (second) Time (second) Light on Light on • Exponential difference filter – Diff between long-/short-term moving averages 16 / 23
Event Detection & Correlation light sensor reading Light off Report event exponential diff Report Human event Time (second) Time (second) Light on Light on movements movements • Exponential difference filter – Diff between long-/short-term moving averages 17 / 23
Event Detection & Correlation light sensor reading Light off Report event exponential diff Report Human event Time (second) Time (second) Light on Light on movements movements • Exponential difference filter – Diff between long-/short-term moving averages • Event correlation – Simultaneous events have same source – False alarm if no power reading change 18 / 23
Light Event Clustering Light 3 Light 1 Sensor 2 Sensor 1 Sensor 1 Light 2 Floor plan Clustering on intensity changes • Feature: change of light intensity 19 / 23
Light Event Clustering Cluster A Light 3 Cluster B Light 1 Sensor 2 Sensor 1 Sensor 1 Cluster C Cluster C Light 2 Floor plan Clustering on intensity changes • Feature: change of light intensity 20 / 23
Light Event Clustering Cluster A Light 3 Cluster B Light 1 Sensor 2 Sensor 1 Sensor 1 Cluster C Cluster C Light 2 Floor plan Clustering on intensity changes • Feature: change of light intensity • {Cluster A, B, C} ↔ {Light 1, 2, 3}? 21 / 23
Power Law Decay of Light log (measurement) (intensity) lo log (distance from light source) measurement = β � power � d - α 22 / 23
Power Law Decay of Light log (measurement) (intensity) lo log (distance from light source) measurement = β � power � d - α distance from light source 23 / 23
Power Law Decay of Light log (measurement) (intensity) α = 3.5 lo Path loss exponent log (distance from light source) α ∈ [2, 5] measurement = β � power � d - α distance from light source 24 / 23
Power Law Decay of Light log (measurement) (intensity) α = 3.5 lo Path loss exponent log (distance from light source) α ∈ [2, 5] measurement = β � power � d - α distance from Scaling factor light source 25 / 23
Cluster-Light Association • Error of associating cluster m and light j − − ∑ = β ⋅ ⋅ α µ e P d , , , m j m i j i m ∈ i R m Discrepancy between model prediction and observation for sensor i 26 / 23
Cluster-Light Association • Error of associating cluster m and light j − − ∑ = β ⋅ ⋅ α µ e P d observed intensity , , , m j m i j i m change of sensor i ∈ i R in cluster m m Discrepancy between model prediction and observation for sensor i 27 / 23
Cluster-Light Association • Error of associating cluster m and light j − − ∑ = β ⋅ ⋅ α µ e P d observed intensity , , , m j m i j i m change of sensor i ∈ i R in cluster m m model-predicted intensity change of sensor i Discrepancy between model prediction and observation for sensor i 28 / 23
Cluster-Light Association • Error of associating cluster m and light j − − ∑ = β ⋅ ⋅ α µ e P d observed intensity , , , m j m i j i m change of sensor i ∈ i R in cluster m m sensors that model-predicted can detect event intensity change of in cluster m sensor i Discrepancy between model prediction and observation for sensor i 29 / 23
Cluster-Light Association • Error of associating cluster m and light j − − ∑ = β ⋅ ⋅ α µ e P d observed intensity , , , m j m i j i m change of sensor i ∈ i R in cluster m m sensors that model-predicted can detect event intensity change of in cluster m sensor i Discrepancy between model prediction and observation for sensor i Light 1 (50W) Sensor 1 30 / 23 Light 2 (50W)
Cluster-Light Association • Error of associating cluster m and light j − − ∑ = β ⋅ ⋅ α µ e P d observed intensity , , , m j m i j i m change of sensor i ∈ i R in cluster m m sensors that model-predicted can detect event intensity change of in cluster m sensor i Discrepancy between model prediction and observation for sensor i Light 1 (50W) Cluster 1 Cluster 2 Intensity change 0 P 1 = 49W P 2 = 52W of sensor 1 Sensor 1 31 / 23 Light 2 (50W)
Cluster-Light Association • Error of associating cluster m and light j − − ∑ = β ⋅ ⋅ α µ e P d observed intensity , , , m j m i j i m change of sensor i ∈ i R in cluster m m sensors that model-predicted can detect event intensity change of in cluster m sensor i Discrepancy between model prediction and observation for sensor i Light 1 (50W) Cluster 1 Cluster 2 Intensity change 0 P 1 = 49W P 2 = 52W of sensor 1 e m,j Light 1 Light 2 Sensor 1 Cluster 1 3000 100 Cluster 2 110 4000 32 / 23 Light 2 (50W)
Recommend
More recommend
