Open repository for appliance power signatures: computer science group project proposal

Here's another draft proposal for a computer science group project that I'm thinking of submitting...


Introduction

As energy costs increase, there is increasing pressure to use energy as efficiently as possible. The first step towards reducing energy consumption is often to measure your existing consumption. There is good evidence that people are best able to manage their energy consumption if they are given an itemised energy bill describing appliance-by-appliance energy consumption information (e.g. "your fridge cost you £50 this month and your TV cost you £20").

Automatically estimating an itemised energy bill by "disaggregating" a whole-house smart meter signal is an active area of research. One approach to this problem requires that the disaggregation system be trained on existing appliance "signatures" (recordings of the power consumption of an individual appliance). A web service which allows for appliance signatures to be programmatically retrieved and submitted would be very useful both to enable automatic disaggregation systems and also to encourage research into disaggregation.

The Challenge

The aim of this project would be to build a web application to allow appliance signatures to be submitted, categorised, analysed and retrieved.

Some features you could play with:

  • allow users to submit new signatures using a web interface
  • allow users to search for and display signatures using a simple web interface. It would be especially useful to be able to compare multiple instances of the same type of appliance (e.g. to see common features in washing machine signatures, sorted by criteria such as country of origin).
  • An API would be essential to allow disaggregation systems to programmatically retrieve signatures from the DB and to insert new ones.
  • Accept a range of sample rates and data dimensions (e.g. collect not just active power but also reactive power for each appliance and whether the reactive power is capacitive or inductive)
  • It would be useful to collect some metadata for each appliance signature, e.g.:

    • Appliance make, model, age
    • Appliance category ("TV", "fridge", "desktop PC", "washing machine" etc). This could be auto-completed if a previous appliance with the same make & model was given a category).
    • Mode (e.g. "40 degree spin wash", "eco-mode" etc)
    • rough geo location (so external temperatures can be obtained from the metoffice; and so country can be determined: UK washing machines behave quite differently to US washing machines)
    • timestamp (so signatures can be aligned with aggregate data, where available); and so time-of-day dependencies can be analysed.
  • This appliance DB could also be used to store anonymised whole-home aggregate datasets along with continuous recordings of individual appliances.

  • Could also calculate and present (graphically and programmatically) usage patterns for each category, e.g.:

    • average daily pattern of usage
    • seasonal variation
    • correlation with weather variables

Further reading

  • TraceBase is the nearest that we currently have to an online database of appliance signatures. It's a great start. But it doesn't allow users to submit new signatures or allow disaggregation systems to programmatically download signatures.

Comments

Interesting. I did something like this on my home. Hoping to release it soon.
BTW, http://fredjiang.com/papers/MobiSys2013Auditeur.pdf accepted at MobiSys 2013 and its strategies might be useful.

Thanks loads for the info, Nipun. Please do let me know when you release your code ;) It's always nice to build on existing work rather than re-inventing the wheel.

Sorry for the confusion. I meant i manually annotated 100 odd electrical events in isolation in my home. The Auditeur paper looks pretty aligned. IBM Research India had conducted a crowdsourcing challeng to detect potholes etc. Maybe that could also be helpful.

Ah, cool, thanks for the clarification! Data is at least as exciting as code, though ;)

There's been some work in Soton about labelling appliance loads in an aggregate feed which might be relevant to this project: http://eprints.soton.ac.uk/338804/1/fe.pdf

Also, don't forget about the wealth of appliance signatures on Cosm!

That's a really interesting paper, thank you! And, of course, you're right about Cosm ;)

Add new comment