NIALM for Smart Buildings LoReMA Load Recognition Monitoring and - - PowerPoint PPT Presentation

NIALM for Smart Buildings LoReMA Load Recognition Monitoring and Acting

1st International Workshop on Non-Intrusive Load Monitoring Carnegie Mellon University, Pittsburgh, PA May 7th, 2012 Stephan Tomek Head of Energy Efficiency Applications iHomeLab Research Center Lucerne University, Switzerland

7th May 2012 2

• iHomeLab
• LoReMA

– Data acquisition – Data analysis – Data transmission

• Learnings / Outlook

Agenda

iHomeLab, Switzerland

7th May 2012 3 iHomeLab, Switzerland

Intelligent living begins here!

7th May 2012 4 iHomeLab, Switzerland

Research Areas

Building Intelligence „The Building as a System“ Ambient Awareness & Intelligence, Sensornetworks, Contentmanagement Human Building Interaction Energy Efficiency Ambient Assisted Living

7th May 2012 5 iHomeLab, Switzerland

18 Engineers

• Software and firmware development
• Embedded system and PCB design
• Simulation and modeling
• Fast prototyping and testing

Domain knowledge

• Low power wireless sensor networks
• Networking and routing mechanisms
• Sensor fusion and context processing
• Building intelligence and machine learning
• Building automation

The iHomeLab Research Team

7th May 2012 6

Requirements

• Identify all appliances in households and offices using NIALM

for load-disaggregation and increase of energy efficiency

• Decentralized wireless measurement nodes
• Actuating of appliances from the node
• Low power design
• Retrofit and zero installation

Goals

• Proof of the technical concept
• Increase transparency of energy usage
• Development of energy efficiency strategy
• Increase energy efficiency

LoReMA: Research project

iHomeLab, Switzerland

7th May 2012 7

LoReMA: System overview

iHomeLab, Switzerland

7th May 2012 8

• Standard components for the measurement unit
• Active power / Reactive power / I RMS and Individual waveform

samples on chip

• Sampling rate of 3.2 kHz, 64 measurement at 50 Hz
• Data Set: active power, reactive power, RMS value of the

current and a FFT calculation of: I1, I3, I5, I7, I9, I11

Measurement concept

Spannungsmessung (Netzspannung) Strommessung (Shunt Widerstand) ADC Energiemesschip DSP (Energiemessung) MCU (Vorverarbeitung, FFT) Energiemess- werte (P, Q, I1-I11) Analoge Eingangssignale

iHomeLab, Switzerland

7th May 2012 9

Measurement range

iHomeLab, Switzerland

7th May 2012 10

• 1. Load change triggers (P, Q)
• 2. Wait for steady state
• 3. Waveform sampling
• 4. FFT analysis
• 5. Data set transfer to the central unit for post processing

Runtime behavior

iHomeLab, Switzerland

7th May 2012 11 iHomeLab, Switzerland

Data set comparison with best matching algorithm

10 20 30 40 50 60 70 80 90 100

P P, Q P, Q, Irms P, Q, Harmonische P, Q, Harmonische normiert P, Q, Samples

42 95 96 96 95 84

32 3 3 1 1 8 1 2 2 2 7 1 1 25 1 1 1

Lasterkennung richtig Lasterkennung falsch Höchste Wahrscheinlichkeit richtig Höchste Wahrscheinlichkeit falsch Last unbekannt

7th May 2012 12 iHomeLab, Switzerland

Best matching algorithm with switched power supplies

5 10 15 20 25 30 35 40

P P, Q P, Q, Irms P, Q, Harmonische P, Q, Harmonische normiert P, Q, Samples

6 31 34 31 36 33

22 13 14 19 14 17 1 2 2 1 24 8 3 1

Lasterkennung richtig Lasterkennung falsch Höchste Wahrscheinlichkeit richtig Höchste Wahrscheinlichkeit falsch Last unbekannt

7th May 2012 13 iHomeLab, Switzerland

Robustness of algorithms when injecting noise according EN50160

5 10 15 20 25 30

P P, Q P, Q, Irms P, Q, Harmonische P, Q, Harmonische normiert P, Q, Samples

13 29 25 15 13 25

7 4 4 7 4 1 5 4 6 3 2 8 8 14 2 3 2

Lasterkennung richtig Lasterkennung falsch Höchste Wahrscheinlichkeit richtig Höchste Wahrscheinlichkeit falsch Last unbekannt

7th May 2012 14

ZigBee Profile: Smart Energy Profile (SEP) Network topology: Mesh-Network Smart Energy Devices:

• Metering Device
• Load-Control Device
• Energy Service Portal (ESP)

Coordinator (ESP) as Cluster Head

• Network configuration
• Channel selection
• Assignment of the node ID’s (short address)

ZigBee Network

iHomeLab, Switzerland

7th May 2012 15

Miniaturization

• 3. LoReMA Steering Committee Meeting

4.5*5*3 [cm]

7th May 2012 16 iHomeLab, Switzerland

End-user information, Load-Disaggregation

7th May 2012 17 iHomeLab, Switzerland

End-user information, Energy efficiency tips

7th May 2012 18 iHomeLab, Switzerland

End-user information,

• n the SmartPhone

7th May 2012 19 iHomeLab, Switzerland

• The various NIALM algorithms have their strengths

and weaknesses

• Choosing the correct algorithm depends on the

application

• NIALM combined with ambient information can lead

to 100 % identification of the appliances

• There is no standard data set definition for energy

fingerprints available

• Norming bodies didn’t picked up NIALM until now

Lessons learned

7th May 2012 20 iHomeLab, Switzerland

• We don’t have a technology issue with NIALM
• The issues are missing standards and norms which

will make it hard for NIALM products to reach the public

• NIALM offers new business and can become a gold

mine for system developers, the danger is proprietary standards

• The IEA is interested to explore NIALM in more

details in order to understand the necessary next steps required

Conclusion / Outlook

7th May 2012 21 iHomeLab, Switzerland

NIALM can bring transparency to the end-users and increase energy efficiency with out big investment.

7th May 2012 22 EuZDC – Munich 22/16

Thank you for your attention

Contact: Stephan Tomek info@iHomeLab.ch www.iHomeLab.ch