Energy disaggregation is software that reads the electricity use of a whole home from its smart meter and works out which appliances are running. It is also called non-intrusive load monitoring. A utility can then show each customer where the power goes: how much for the air conditioner, how much for the water heater, how much for everything else.
A smart meter measures only the total power of the whole home. It does not know about individual appliances. But every appliance leaves a pattern in that total. When an air conditioner starts, the total jumps by a certain amount and then settles. A refrigerator switches on and off in a regular cycle. A water heater draws a large, steady load for a short time. The software learns these patterns from the readings, which are time series data. No sensor is fitted on any appliance.
The readings must be frequent enough. If the meter reports only once in 15 or 30 minutes, small appliances blur together. A reading every minute or faster shows the jumps clearly. So the method works best where meters report often.
These services run mainly for electricity utilities in the United States and Europe, where smart meters were installed earlier. So the example here is American, in dollars. Say a family in Houston, Texas, receives a monthly bill of $160. The utility's app shows that $70 came from air conditioning, $25 from the water heater, $20 from the refrigerator and $45 from everything else. The family sees that running the heater for two hours less each day would save money. The utility benefits too, because customers who understand their bills call the complaint desk less often.
Utilities use the same information for themselves. Knowing which homes have large flexible loads, such as a heater or an electric car charger, helps them plan a virtual power plant. The data is handed to apps through a utility data API, with the customer's permission.
India's smart-meter roll-out creates the same opportunity for Indian discoms, the companies that distribute electricity in each state. The results are estimates, not measurements. Two appliances of similar size can be confused, so the app should say that the split is a good guess.