Deokwoo Jung Estimating Building Consumption Breakdowns using ON/OFF State Sensing and Incremental Sub-Meter Deployment Deokwoo Jung and Andreas Savvides.

Deokwoo Jung

Estimating Building Consumption Breakdowns using ON/OFF State Sensing

and Incremental Sub-Meter DeploymentDeokwoo Jung and Andreas Savvides

Embedded Networks & Applications Lab (ENALAB)Yale University

http://enalab.eng.yale.edu

Nov 4, 2010

Deokwoo JungNov 4, 2010

Sensing Loads on Electricity Network

Living Room KitchenBed Room

Breaker Box

Electricity NetworkElectric Meter

Electrical Outlet

How to Estimate Electrical Loads of Appliances ?

Deokwoo Jung

Electricity Energy Monitoring Systems

• Direct Monitoring : Expensive and brute-force method

• Watts up? .Net – $ 230 – Internet enabled – Power switching

• Watts up? – $100-$130 – Data Logging

• Kill-A-Watt EZ– $45 – Data display only

• Indirect Monitoring – Total Load Disaggregation + Load Signature Detection

NALM (Hart.et.al): Nonintrusive Appliance Load Monitoring

A ElectriSense (Sidhant et.all) :Single-Point Sensing Using EMI for Electrical Event Detection and Classification

in the Home

Nov 4, 2010

Deokwoo Jung

Load Disaggregation Data Flow

Event Detection

Load Disaggregation

Partial Load Information

High frequency electromagnetic

interference

Edge detection

Heat

Vibration

Light intensity

Voltage and current

waveforms

at Electrical outlets or

Power entry point

How do we compute the load disaggregation?

ON/OFF state e.g. Total Power consumption

Nov 4, 2010


The Diverse Nature of Loads

Resistive vs. Inductive -> Short-term propertyStationary vs. Non-stationary -> Long-term property

Inductive Resistive

Non-Stationary

Stationary

Short-term property

Lo

ng

-ter

m

pro

per

ty

Refrigerator

Bulb

heater

Washing Machine

TV Air Conditioner

Dehumidifier

Electric

Kettle

Water Pump

Hard to measure power consumption

Hard to estimate energy breakdown

Laptop

DVD Player


Our Approach: Energy Breakdown per Unit Time

Actual Power Consumption Profile

Actual Average Power ConsumptionEstimated Average

Power Consumption

Estimation Error

Example appliance: LCD TVEstimation Period k-1 Estimation Period k Estimation Period

k+1

Instead of instantaneous measurements, use average consumption over a time window


Problem Setup

Goal: Estimate the average power consumption for a time window

Select an appropriate time window to get the best estimate of energy consumption

1010101011

010.0001111

1011100110Time

Appliance

Consumption fluctuation properties

Three Tier Tree Network


Prototype System Implementation

TED 5000 Monitor

BehaviorScope Portal

Active RFIDDry Contact Sensor

One Energy Meter and ON/OFF Sensors

Consumption measurements

Appliance ON/OFF Information


Main Idea ON/OFF sequence of appliances occurs between the worst

(Perfectly Synch) and the best case (Perfectly Desynch)

appliance A

appliance B

Worst Case Best CaseObserved Binary Data

Approach – Variant of Weighted Linear Regression– Accounting for Diversity

• Design Optimal Weight Matrix, W – Metric Driven Data Selection

• Regression data set is adaptively chosen according to active power consumption property, stationary vs. non-stationary

• Using Prediction Metric for Estimation Error


Problem Formulation

Sample Index

On/Off state of TV

On/Off state of Microwave

On/Off state of Lamp

The Average of Power meter measurement

(Watt)

# of samples observed

1 0 0 1 59.3 3

2 0 1 1 369.3 3

3 1 0 0 120 1

4 1 0 1 160 1

5 1 1 1 469 1

Objective Function:

YWXPWX)XPP

(minargˆ0

Solve Opt. Problem:

YX

nnXPXPXPMin 2211-Power TotalW


Designing Weights and Selecting Appropriate Time Window

Samples of #

No Weight Unit Sum Matrix Estimated Variance Sum Matrix

Exact Variance Sum Matrix

W

Appliances All

StateOn

Samples of #

Appliances All

Power Active of VarianceStateOn

Samples of #

• Optimal Choice of Weight Matrix, W

• Account for (Non-) Stationary Property– Stationary Load : larger window of measurements is better– Non-Stationary Load: small window of measurements is better– Automatically select to use either of the entire estimation periods

(Cumulative Data) or only the current period (Current Data)


Evaluation - Case StudyA small electricity Network with single power meter

• Collecting data from 12 appliances in one-bedroom Apt from Thu-Sat• A large variation of energy load

– the heater accounts for more than 60% of the total energy consumption– the laptop consumed the least, less than 1% of the total load.

0

0.5

1

1.5

2

2.5

6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6

Time, (hour)

Energ

y C

onsu

mption,

(kW

h)

Dehumedifier Ceramic Heater Laptop LCD TV(32")

Top Freezer Refrigerator Halogen Desk Lamp Microwave Oven Fluorescent Desk Lamp

Desktop+Monitor Standard Bulb Table lamp Compact Refrigerator Two- Way Floor Lamp

Thursday Friday Saturday

The hourly energy consumption ground truth in one-bedroom apartment from an experiment from Thursday to Saturday

0

5

10

15

20

25

Thursday Friday Saturday

Ene

rgy

Cons

umptio

n, (

kWh)

Dehumedifier Ceramic Heater LaptopLCD TV(32") Top Freezer Refrigerator Halogen Desk LampMicrowave Oven Fluorescent Desk Lamp Desktop+MonitorStandard Bulb Table lamp Compact Refrigerator Two- Way Floor Lamp

Daily energy consumption ground truth in one-bedroom apartment from an experiment from Thursday to Saturday

440 460 4800

5

10Dehumedifier

500 1000 15000

20

40Ceramic Heater

20 30 400

5

10Laptop

60 80 1000

10

20LCD TV(32")

100 120 1400

10

20Top Freezer Refrigerator

38 40 420

10

20Halogen Desk Lamp

0 500 10000

10

20Microwave Oven

32 33 340

10

20Fluorescent Desk Lamp

100 120 1400

10

20Desktop+Monitor

30 35 400

20

40Standard Bulb Table lamp

65 70 750

20

40Compact Refrigerator

100 200 3000

10

20Two-Way Floor Lamp

Histogram of power consumption of appliances during their On state

0 50 100 150 200 250 300 3500

1

2

3

4

5x 10

4

Decimal representation of composite binary states, xk

Th

e nu

mbe

r of

sam

ple

s, n k

Compact Refrigerator only

All appiances are off

Top Freezer Refigerator + Ceramic Heater

Compact Refigerator + Ceramic Heater

The number of meter samples observed given composite binary states


Evaluation - Case Study :A small electricity Network with single power meter

• Estimated hourly energy consumption profile of each appliance– Average 10% of relative error


Performance over Estimation Periods

• With different weight matrix

0 0.5 1 1.5 2 2.5 30

20

40

60

80

100

Estimation Periods, Test

( hour)

Ave

rag

e A

ctiv

e P

ow

er R

ela

tve

Err

or, (

%)

Opt. Data.Sel + Est.Var.Sum.Wgt (Algorithm Performace)Opt. Data.Sel + Unit.Sum.WgtOpt. Data.Sel + No.Wgt.Oracle. Data.Sel + Exact.Var.Sum.Wgt (Lower Bound)

Lower bound

Algorithm performance

No Weight

Unit Sum Weight


Performance over Estimation Periods

• With different data selection schemes

0 0.5 1 1.5 2 2.5 30

50

100

150

200

Estimation Periods, Test

( hour)

Ave

rag

e A

ctiv

e P

ow

er R

ela

tve

Err

or, (

%)

Est.Var.Sum.Wgt + Opt. Data.Sel (Algorithm Performace)Est.Var.Sum.Wgt + Cur. Data.SelEst.Var.Sum.Wgt + Cma. Data.SelExact.Var.Sum.Wgt + Oracle. Data.Sel (Lower Bound)

Lower bound

Algorithm performance

Current Data Selection

Cumulative Data Selection


Performance by Data Selection, Weight Matrix, and Estimation Period

The maximum, minimum, and average value of relative error of active power consumption for all estimation periods with various combination of weighted matrix and data selection schemes

0

50

100

150

200

250

Ave

rag

e A

ctiv

e P

ow

er R

ela

tve

Err

or, (

%)

Est.Var.Wgt + Opt.Data.Sel (Algorithm Performace)

No.Wgt +Opt.Data.Sel

Unit.Sum.Wgt+ Opt.Data.Sel

Est.Var.Wgt + Cma.Data.Sel

Est.Var.Wgt + Oracle.Data.Sel

Exact.Var.Wgt +Oracle.Data.Sel(Lower Bound)

Exact.Var.Wgt+ Opt.Data.Sel

Est.Var.Wgt+ Cur.Data.Sel


Increasing Accuracy on Larger Networks with Additional Meters

• How many power meters we need and where should place them?– Tree Decomposition Problem

• Depending on sensor duty cycles

– Combinatorial Optimization Problem• Use Stochastic Search Algorithm :

Simulated Annealing

• Cost function of Simulated Annealing– Evaluated against the initial solution,

• Z0=(1,1…,1) : Placing meters on all available electrical outlets.

Node Efficiency

0

0 )1()|(

)|()(

Z

Z

ZP

ZPZ

MSE

MSEc

Estimation QualityWeight Coefficient: # of meters vs performance

10 yy 1y

1x 2x 3x 4x 5x

2y20 yy

1x 2x 5x 3x 4x

Topology 1

Topology 2

Unsynchronized

00y

1y

1x 3x 4x 5x2x

2y

Synchronized

10 yy 1y

1x 2x 3x 4x 5x

10 yy 1y

1x 2x 3x 4x 5x

2y20 yy

1x 2x 5x 3x 4x

2y20 yy

1x 2x 5x 3x 4x

Topology 1

Topology 2

Unsynchronized

00y

1y

1x 3x 4x 5x2x

2y

Synchronized

?


Evaluation - Case Study 2:A large scale electricity network with meter deployment

• Performance evaluation by increasing the number of Apt units from 1 to 12

• With a single power meter for a large electricity network

• Meter Deployment by Algorithm • Compared by random deployment– For λ= 0.5, x10 in performance – Or reduce x 2~3 in # of meters– λ = 0 Single power meter – λ = 1 Full deployment

0 2 4 6 8 10 12 1410

3

104

105

106

The Number of Electricity Meters

RS

S(P

)

Random DeploymentB-SEND Algorihtm

=0.5

=0

=1

Max

Mean

Min

=0.9

=0.3

=0.7

0 20 40 60 80 100 1200

50

100

150

The number of appliances

Rela

tive E

rror,

(%

)

Deokwoo Jung

Conclusions and Future Work

Developed an energy breakdown estimation algorithm for a single power meter and the knowledge of ON/OFF states• 10% of relative error for 12 home appliances and a single power

meter

Developed an algorithm for optimally placing additional power meters to improve estimation accuracy in large networks• Deployment algorithm can reduce 3-4 times of the number of

power meter for the simulation of 12 households

Future work:- Experimental deployment on a Yale building in January 2011 - Handle incomplete binary state sensing

- Leverage history information and user inputs

Nov 4, 2010

Deokwoo Jung

Discussion & Comparison with Related Work

• The question on high frequency systems makes some sense. Assuming that you can detect signatures, if the frequency of measurement is high enough you may have enough information to computer itemized consumption.

• The key argument to make is that this approach could work today with existing low-frequency meters. The central meter in a home only has to same using 1Hz. Also, in the home, we may be able to do this without any additional hardware by just completing forms on a GUI.

• While we work out details for a journal version it is important to identify and propose the next problem to solve on load disaggregation

Nov 4, 2010

Deokwoo Jung Estimating Building Consumption Breakdowns using ON/OFF State Sensing and Incremental Sub-Meter Deployment Deokwoo Jung and Andreas Savvides.

Documents