Irish researchers have a way to tell which appliance is burning your electricity without putting a sensor on a single one of them.
The system splits a household's half-hourly smart meter readings into nine appliance categories - things like water heating, cooking, and lighting - using event detection for big power spikes (think kettles or immersion heaters) combined with estimates drawn from a short questionnaire about which appliances a household actually owns. No new hardware, no submetering, just existing meter data plus a few survey questions. The team tested it against two existing disaggregation systems across four datasets: a calibration home checked against a commercial comparator, the UK-DALE and REFIT public benchmarks with real per-appliance submetering, and a smart meter dataset covering more than 4,800 household-years from 2,968 Irish consumers. The questionnaire-guided method came out with the lowest overall decomposition error on every dataset and beat the rivals across 54 paired months, a result the researchers report as statistically significant.
Ireland's smart meter rollout already covers over 80% of households, mostly to support time-of-use billing, but 30-minute readings are too coarse to see individual appliances. A method that infers appliance-level use from data utilities already collect, plus a one-time questionnaire, is a cheaper path to personalized energy advice than installing submeters or smart plugs in every home.
It is a less flashy pitch than the smart-plug startups selling appliance-level insight one gadget at a time, but it works with hardware already bolted to the wall.