electrical appliance signature identification by using RF method

[ernest_ckl]

Electrical Appliance Signature Identification by using RF method

Hey, I am doing my FYP project with a topic of Electrical Appliance signature identification by using RF method!


Every electrical appliances emits it's own and unique signal when it consumed energy. I would like to capture and record the signature of electrical appliances This project is more towards signal processing where the recorded signature will be processed and analysed to be used with classifiers to classify the source of the signal which is hope in turn will identify the appliance being used.

The active loop antenna was connected with spectrum analyzer.
The antenna was used to measured the signal emitted from a blender.
so far, I am able to capture the waveform of blender by using spectrum analyzer !

For the signal analysis and classification of the appliance, what is the next step or method that I can used to analyse the waveform from the spectrum analyzer?

For the electrical appliances identification, is it the fuzzy logic or neural network need to be used?

Any expect here can give me some help???? Thanks a lot

 

[mtwieg]

Wait, are you actually sensing the RF generated by appliances, radiated through free space, or the EMI conducted through the AC line?

A while back I saw a presentation by a company who had a system for detecting appliance and device operation by looking at the signals they inject into the AC line. They had some pretty impressive results, especially since it's as simple as plugging a simple spectrum analyzer into an AC outlet. However they are only able to really differentiate between appliances which use switching power supplies (or phase dimming control), and they can only really deduce whether the device is on or off, not how much power it actually consumes.

But I've never heard of doing it with radiated RF. That sounds much more difficult, since the signal will likely be much weaker, and you'll need a very broadband antenna to pick up the interference from a broad range of devices.

 

[ernest_ckl]

ya, my project is sensing the RF generated by appliances, radiated through free space. By using the signature , I can determine the classification of appliance whether the devide is on or off.
so far, I am able to capture the waveform of the blender when it turn on. Do u have any idea of how to process the waveform for making unique feature of the signature? so that, I can make the identification for this devide.

ya, there are other method such as, real time recognition and profiling of appliances through a single electricity sensor(Episensor ZEM-30 Zigbee Energy Monitor)

 

[mtwieg]

ya, my project is sensing the RF generated by appliances, radiated through free space. By using the signature , I can determine the classification of appliance whether the devide is on or off.
so far, I am able to capture the waveform of the blender when it turn on. Do u have any idea of how to process the waveform for making unique feature of the signature? so that, I can make the identification for this devide.

It's hard to give an answer without knowing anything specific about the signal itself. In fact I'm surprised that something like a blender actually gives out detectable RF (at frequencies more than a few KHz, anyways). Unless it's some kind of fancy moden blender with a switching supply and a DC motor or something. Blenders with AC motors should give off practically no EMI in the RF range. Could you show what the signals look like on a signal analyzer?

ya, there are other method such as, real time recognition and profiling of appliances through a single electricity sensor(Episensor ZEM-30 Zigbee Energy Monitor)

The method I was describing is totally different than that. It doesn't measure current through conductors. It simply plugs in anywhere to the household wiring, not in series with any devices, and can sense the operation of any devices on that local distribution network. I'm wondering if it effectively does what you're trying to do, with simpler technology.

 

[ernest_ckl]

Active loop antenna (model 6507)- Range:1Khz-30Mhz
**broken link removed**

Spectrum Analyzer -range: 9Khz-1.5Ghz


Two different blenders was tested and two different type signals was captured on spectrum analyzer. By using this signal, I have no idea of how to analyze it as unique feature of blender.
analyze it by using fourier tranform? or any else?

Blender A




Blender B

 

[ernest_ckl]

Anyone can help?

 

[mtwieg]

The main difference between the two is just the shape of their noise floors, and I don't think there's any way you'll be able to use noise information effectively (at least not with more than one device operating at a time). What you need to see is actual peaks in the plots which are distinguishable from the noise floor. I think I might see a couple in the ones you posted. Try decreasing the bandwidth/increasing the time of the acquisition in order to lower the noise floor. Then you might see something more coherent information. Also you'll want to check multiple devices of the same model to make sure the characteristics are consistent.

In order to recognize the differences between the signals (if there are any), you're likely going to have to pretty much make a spectrum analyzer. Then you'll need some peak detection algorithms to pick out signals above the noise floor, and compare those to profiles of known appliances to figure out what's what.