"The idea is to process the data locally, within the smart meter. So we'll need a chip with enough processing power, but it will have to be cheap, so moving to a smaller sampling frequency with less bit depth would mean less data to crunch".
But 42 Technology has yet a lot of work to do optimizing its algorithms and finding the simplest ways to shift through the data generated by going into the frequency domain.
"We look at a whole bunch of stuff, maybe eighteen or nineteen different parameters, mostly in the frequency domain, but also in the voltage and current domains", explained Carey.
A dedicated team at 42 Technology is recording the signatures of different classes of appliances, building up a library of simplified templates which could be stored on-chip. Winning an Innovate UK Energy Catalyst 4 grant, the company was able to build a number of demonstration units and started collecting real usage data to show to potential development partners.
For now, the company is training its algorithms on discerning between different classes of appliances. "We can't say the difference between a phone charger and a tablet charger, it is still very much class-level, but in the future we could teach our algorithms to identify the individual signatures of different brands, dishwashers have different cycles, but we haven't got there yet" told us Carey.