Introducing NILMTK: an open source toolkit for non-intrusive load monitoring

Today, Nipun Batra, Jack Kelly and Oliver Parson are really pleased to announce the release of NILMTK: an open source toolkit for non-intrusive load monitoring. The toolkit will allow researchers to easily develop algorithms which disaggregate a household’s total electricity consumption into individual appliances.

Specifically, the toolkit includes:

Non-Intrusive Load Monitoring ToolKit (nilmtk)

Nipun, Oli and I have just started work on an open source toolkit for non-intrusive load monitoring called nilmtk. We're pretty excited about it! It's only in the very, very earliest stages (the code repository currently has precisely zero lines of code in it!) although we've started to flesh out the design on the project's wiki.

Why don't academics discuss research before starting the work?!

Warning: this is a hopelessly idealistic proposal...

As far as I can tell, scientists of all types tend to be secretive about their work prior to publication. I guess the fear is that, without secrecy, another lab might steal the idea and publish it first. But this secrecy comes with several significant costs, one of which is the huge length of time between first having an idea and then getting feedback from peers. First you have your bright idea, then you do the necessary research (which might take years), then you write up a paper, then you submit the paper to a journal or conference and a while later you get your first bit of peer feedback.

This feels extraordinarily inefficient. If an idea is bad then it needs to fail fast (before resources have been expended). And if an idea needs improving then it's far better to improve the proposal before doing all the work rather than trying to retrofit a fix after the bulk of the work has been done.

How about this as an alternative:

Wiki and online community for electricity disaggregation researchers

If you're a researcher working on electricity disaggregation (aka NALM / NILM / NIALM) then would you find a wiki and/or online community useful?

The wonderful MSc students I worked with last year and their award-winning projects ;)

This spring and summer I had the opportunity to work with some exceptional MSc students at Imperial and I just wanted to write a summary of their achievements.

Visualisation of machine learning algorithms: computer science group project proposal

Here's another computer science group project that I have submitted for consideration by students this coming year. As always, comments are very welcome!

Algorithms used in machine learning can often feel quite complex when you first come across them. As you gain experience with the algorithm you begin to be able to visualise each step and then realise that the algorithm is actually quite intuitive. Wouldn't it have been far easier to learn the algorithm if you had seen a good visualisation of the algorithm to begin with!

For example, here's a visualisation of selection sort (taken from WikiPedia):

The aim of this project is to produce interactive, animated visualisations of a set of machine learning algorithms. (Don't worry if you don't know any machine learning algorithms yet; this project would be a good opportunity to gain intimate knowledge of a few algorithms).

Below are some suggestions to get your ideas flowing.

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...


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.

UK energy infrastructure simulation game: computer science group project proposal

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

Could we power the entire of the UK electricity grid on renewables alone? What about a mixture of nuclear and renewables? How much would this scenario cost to build? Or what about going to the other extreme and switching our power generation to 100% coal? What implications would that have?

Discussions about energy are becoming quite common in the media. Questions like those listed above are frequently asked but it's hard to find good answers. One of the big problems in communicating energy issues to the wider public is that people have little sense of the scales involved.

The aim of this project will be to build a kind of "SimCity" game to allow members of the public to explore different energy scenarios for the UK (or for the whole world if you're feeling very ambitious!). Users would be allowed to input any scenario and the game would simulate the consequences. Points are lost for triggering power blackouts or excessive environmental damage.

Dimmers and sparse-sampling home energy monitors don't mix

I have been recording the power consumption of my whole home using two systems: a Current Cost CT clamp and my home-brew "sound card power meter". The latter samples my home's voltage and current waveforms at 44.1kHz and calculates the apparent and active power once a second.

Most of the time, the signal from the Current Cost CT clamp and my SCPM agree remarkably well.

One situation where the two systems disagree wildly is when our kitchen ceiling lights are on. These lights consist of 10 x 10W Philips dimmable LEDs driven by a cheap TRIAC dimmer. Here's a graph showing a time period where the lights start off, then turn on, and then turn off again. The Current Cost reports a rapidly varying power consumption. The SCPM, on the other hand, reports a very steady power consumption. I have also plotted the mains voltage to demonstrate that the mains voltage wasn't doing anything odd during this period:

What's going on? Why does the Current Cost think the kitchen light's power consumption wobbles up and down wildly?


Subscribe to RSS - PhD