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 are we doing this? The three of us are each involved in academic research on non-intrusive load monitoring (NILM; the aim is to try to estimate the energy consumption of each individual appliance from a whole-house meter reading). Over the years, we have each - independently - had to figure out the various tedious issues involved in NILM. We have each built our own tools for solving these problems. But it seems remarkably inefficient for individual researchers to re-invent these solutions in their own little silos; hence the desire to create an open-source toolkit which solves these problems and allows researchers to focus on the real meat of building better disaggregation algorithms.

How did this project come about? The short answer is that Nipun emailed a few days ago to propose it, and then we naturally wanted to get Oli involved! I was really excited when Nipun suggested the project. It's something that's been in the back of my mind for some time now (and I know I'm not alone in this) but it took Nipun's initiative to kick start the project and move it from a daydream to something that might actually happen!

Some specific, speculative aims of the project are listed on the nilmtk wiki.

Just to emphasise: we are in the very earliest stages! But it feels great to be collaborating which such smart folks on a project that I really believe in. And it's nice to be participating in open science ;)


Good luck! I think your task is close to impossible without intrusive metering but I am not an expert. If you need some raw electricity measurement data to test your theories/software tools, visit There is various amount of 1-3 phase data available which was measured from homes located in Scotland, Iceland and Finland.

Measurement system was using Current Cost (with impulse counter or current transducers) and Trim-Slices to log the data.

Hi Teemu,

It's certainly true that NILM will always provide an estimate. I suppose the game is to make the estimate as good as possible, and to make it as easy as possible for folks to use ;)

Thanks loads for pointing me to the Octes dataset, I hadn't heard of that before! Looks useful!

Thanks, Jack

This is a good effort, Jack, Nipun and Oliver. I will see if/how I can be of some help.


Hi Krishnan, thanks loads for the support! Yes, please do dive in if you fancy it!

Add new comment