DISCLOSURE : This project was something I worked on, a year ago in a different electric energy hackathon but I wanted to see whether there is a business appetite for such a solution in the water utilities industry since I learnt that smart meters are being introduced at an hourly level.
My electric meter (can be used with water or gas too) gives an overall monthly or daily consumption. However, this is not actionable information.
How it works
Using a concept called energy disaggregation, one can take an aggregated signal and decompose it to the energy consumption per appliance automatically without any instrumentation. Its a prediction of when the appliance is active, so there is an error margin. The input is the detailed energy consumption on a per hour basis and energy signatures of various major appliances. With the information of which appliances are used, we can pattern match which appliances were active at what time and derive the energy consumption at a appliance level.
Once you have the appliance level energy use, we open up a whole new data layer that can then be used to power a whole new set of actionable visualizations and recommendations such as whether to use tiered vs peak pricing or other more energy efficient appliances.
Challenges I ran into
Confidence of the prediction. How do we get the data around the residence about what appliances they have.
Accomplishments that I'm proud of
Creating a demo that reflects the proof of concept and vision.