Fenella Carpena
My name is Fenella, I'm a PhD student in Economics at the University of California - Berkeley, focusing on issues in economic development. It's my first time posting to this group and I'm really excited to join! .
I am not a computer science/engineer by training, but I recently became interested in disaggregation because of a project we are working on in India. We are interested in understanding how poor households consume electricity, and we were able to get data from the household's meters. The meter data has real power, and we have readings every 1 minute. We also have survey data on the household's appliance holdings, from a one-time survey that we did back in July. Since households in our study are quite poor, they only have 3 types of appliances: light bulbs, a fan, cellphone charger (a few have a TV as well).
My goal is to be able to disaggregate these meter data, to examine the electricity use patterns of poor households. Since I'm new to the disaggregation field, I was wondering if anyone might kind enough to share suggestions to the following questions:
(1) Are there existing/off-the-shelf NILM algorithms that I might be able to use/adapt for our setting? I see from the literature that Hart (1985) is the seminal work on this field, and I see that there is an implementation of this in the NILM toolkit. However, I am not sure if it is a good fit for us because of appliances such as chargers. There seems to be a lot of NILM papers in the literature and I also have read the recent review by Ziefman and Roth, but it has been difficult for me to be able to compare these different algorithms to see which one might be the most useful for my setting.
(2) How to remove noise from the power readings? I checked the existing NILM papers and it seems that most use median filtering, but I worry that I'll loose too much information by doing this, since households in our sample have low-powered appliances.
Thank you very much in advance for any information! Have a good weekend!
Fenella
Ahmed Sagarwala
There are some systems you could post the data to, for example Plotwatt and Bidgely. I'm sure other people on the group will have good recommendations for you. I think Plotwatt would work best since it allows for multiple homes within a single account.
Research shows getting disaggregation on low power items (chargers) would require a higher resolution. Have a look at Armel et al, 2012 (Is disaggregation the holy grail of energy efficiency?). Therefore, you'll only get access to large appliances. If I remember correctly, there is some information on removing noise in the paper.
Regarding algorithms, I haven't found anything you can just use 'off the shelf'. There are some algorithms available in R, Matlab, and Torch. There's a learning curve to all of these tools and I've had a lot of trouble getting things to work as I would expect.
Best of luck!
Jack Kelly (aka Daniel)
Hi Fenella,
Another welcome to the group :)
Phone chargers
As Ahmed says, all disaggregation algorithms are unlikely to be able to identify a phone charger in 1 minute data. Phone chargers draw too little power to be identifiable within the noise. Although, if your meters are accurate enough and if you can accurately disaggregate the fan and lights and if you can be sure that homes only have a fan, lights and a phone charger then you could label all non zero power demand which isn't attributable to the other appliances to the phone charger. But that's a lot of 'if's!
NILM folks often don't worry too much about disaggregating phone chargers for at least two reasons:
1) Their total energy consumption is, at least for western homes, a tiny fraction of the total household consumption.
2) Their total consumption can be estimated in other ways. e.g. assume that each phone is charged once per day.
If you really need to accurately identify when phones are charged then I suspect that NILM (on 1-minute data) won't give you a satisfactory answer, so you'll need an alternative approach. Could you install an individual power meter on each phone charger? Or install an app on each phone which records when the phone is charged (assuming the phones in your study have that capacity)? Or maybe the phone already logs when it is charged?
Fans
Is there any chance you could also record reactive power for ? That should make it a lot easier to spot the fan.
NILMTK
You are correct that, right now, NILMTK is a research tool rather than an off-the-shelf disaggregation system. For example, NILMTK currently doesn't do a great job of allowing you to train and test on different houses (that is possible but it's a little more fiddly than it needs to be). I'm optimistic that one day NILMTK might mature into something you can use off-the-shelf. But probably not for a while.
One approach available to you might be to try to collaborate with a NILM researcher :)
Regarding "noise": I'm personally very cautious about removing 'noise'.
Good luck,
Thanks,
Jack
--
You received this message because you are subscribed to the Google Groups "Energy Disaggregation" group.
To unsubscribe from this group and stop receiving emails from it, send an email to energy-disaggreg...@googlegroups.com.
To post to this group, send email to energy-dis...@googlegroups.com.
Visit this group at http://groups.google.com/group/energy-disaggregation.
To view this discussion on the web, visit https://groups.google.com/d/msgid/energy-disaggregation/af315668-f81a-4113-9f8f-05362ffad5e6%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.