Indian Dataset for Ambient Water and Energy

In the summer of 2013, we decided to instrument a home in New Delhi India with an aim to characterize the unique aspects of energy monitoring an consumption in India. In total we collected about 73 days of data.

Electricity was measured at all 3 levels:

  1. Electricity meter: Using Schneider Electric EM6400 sensor
  2. Circuit panel
  3. Appliance level

Power outages are still common in India! Yes, even in Delhi. During the deployment, we observed power outage lasting up to 9 hours a day.

Water supply is not available 24X7. Infact, usually only available for a few hours a day. To overcome this, almost everyone has an overhead storage tank where water is stored. We measured the volume of water flow from the supply and usage from the overhead tank.

Voltage can be observed to fluctuate a lot. While the rated voltage is 230 V, we observed voltages ranging from 180-260V.

What about the network? We observed that internet is not the most reliable. Packet drop is very common. We quantified that and found upt 25% packet drop a day.

How did we resolve this packet drop? We introduce a notion of local storage in our arhitecture which also compensates for other failures.

Github repo

Take a look at our Github repository which contains all the issues we faced and how we resolved them.

What about DB schemas?

Right here! Making it easy for you to replicate such deployments.

Starter IPython notebooks

Why take trouble and code from scratch? Start with our IPython notebooks which illusrate common functionalities


Wouldn't it be useful if the rated power of all appliances was known? Or which appliance is mapped to which circuit? Check out our metadata to know more.

Installation instructions

How to download the data set and quickly set up the environment

Why so many sensors?

Take a look here! You will realize the additional value by summing up sensor data.



    CSV, nilmtk HDF5


    Compressed SQL (15 MB) , Compressed PIR CSV (3 MB) , Compressed Light, Temp CSV (14 MB)


    Compressed SQL (90 MB) , Compressed CSV (75 MB)

    System information

    Compressed SQL (18 MB) , Compressed CSV (18 MB)


Paper & Slides

Author version

Slides presented at Buildsys 2013