Power Factor Signature Analysis

For Device Identification Applications

Dane Kouttron, Rensselaer Polytechnic

To provide detailed usage info to the user via signal disambiguation

Constantly monitors residential power usage for transients and eventsMeters similar characteristics as existing smart meter technology

Per phase voltage Per phase current consumption

Device Identification Overview

Dane Kouttron, Sept. 20082

Example: what does an incandescent lamp look like?

Video clip displays consumption Look at transient characteristicsNotice anything interesting?

Device Identification Overview

Dane Kouttron, Sept. 20083

Basic Process OverviewIncandescent Bulb warm up transient vs CFL bulb warm up transient (current consumption)

Upon ‘event’ occurrence, ‘fingerprints’ can be isolatedFingerprint characteristics are based on P, Q and Harmonic monitoringR-squared statistical analysis used to match ‘fingerprints’ and known files from the ‘encyclopedia’

Device Identification Process Overview

CFL

Incandescent

Event occurrence R Squared probability Device ID

Dane Kouttron, Sept. 20084

Example test applicationUnit constantly searching for ‘events’Displays list of devices in use on circuit (based on known / stored fingerprint characteristics)Able to identify small devices (refrigerator lamp)

Device Identification Overview

Dane Kouttron, Sept. 20085

Analyzing Common OccurrencesHarmonic Response

(CFL turn on transient, greatly reduced speed)Peaks indicate harmonic disturbance and harmonic characteristic of deviceSeparate controller used as harmonic interrupt

Dane Kouttron, Sept. 20086

Device Identification Hardware OverviewAnalog Inputs:

Per phase voltagePer phase currentAmplified Harmonic input

Processing:AVR Xmega series Micro controller

Harmonic InterruptsAVR Mega 168 series Microcontroller

Networking Requirements:Serial interfaced networking platform (pre-existing hardware repurposed for internet communications)

Dane Kouttron, Sept. 20087

Hardware Requirements For PFS Processing- upwards of 70 MIPS (single core)

[equivalent to low end smart-phone processor]

Storage-Onboard of 4mb for cache / swap space

Transmission-Dependant on required resolution / update frequency.

[realtime = ~ 12 mb of data tx / 3 weeks, assuming 100 known devices]

All PFS computations to be completed on client side

Resolution of the data-12 bits minimum sampling resolution

Sampling frequency vs accuracy plot30kHz minimum sampling frequency

Dane Kouttron, Sept. 20088

Device Identification Hardware: AnalogAnalog Inputs:

Current ClampsProvide moderate frequency response of current consumptionRelatively expensive

Voltage InputResistive Voltage divider (no induced phase shift), excellent frequency responseLow cost

Dane Kouttron, Sept. 20089

Device Identification Hardware: DigitalTwo microcontrollers implemented

Hardware (band pass) filtering for harmonics Harmonic interrupt subroutines run in parallel to PFS processing

Shared real time clockSwap space / ‘encyclopedia’ spaceAnalog inputs are digitally adjustable for variability in current clamps used

Dane Kouttron, Sept. 200810

Device Identification Hardware: Communications

Currently using existing commercial WIFI transceivers reprogrammed for communication backboneCan be redesigned to incorporate other communication protocols

Serial connected phone modemBluetooth phone connectionEthernet

Dane Kouttron, Sept. 200811

Device Identification Software:Software Design

Triggering system for event occurrence

Triggering dependant on change in real and reactive powerTriggering also interrupted by harmonic events

(hardware FFT)

ComparisonProbability analysis using R-squared regression

Prototype testing in Labview

Dane Kouttron, Sept. 200812

User FeedbackClient side Feedback

Email updates (tabulated client side) can be sent directly from devicePreset timing interval could trigger sequential updates along with real time data / accessibility

Web based interfaceCell phone updates

Dane Kouttron, Sept. 200813

Questions / Comments

Dane Kouttron, Sept. 200814