Non-intrusive load monitoring (NILM) is a technique that determines the load composition of a household, i.e. the activity of the individual appliances which constitute the load, through a single point of measurement, e.g. the main electrical service entry point of the home. This technique has the great advantage to not require installing meters on each individual appliance.
The proposed approach aims at reducing electrical usage of households by providing to the end-user a detailed description of its electrical consumption, allowing him to understand and modify its consumption habits accordingly. The method can also be used by the energy supplier in order to predict the local electrical consumption to better plan energy requirements and provision.
A model for five different types of appliance was implemented:
- the fridge
- the heat pump
- the base component
- the washing machine and
- the dishwasher
And a graphical interface to display the estimated energy of each appliance, e.g. with a pie chart, was also developed (see figures).
The algorithm was evaluated on a database of aggregated load signals, obtained from a low-frequency (1 Hz) acquisition system able to measure the three phases of a standard household, as well as on a database of synthetic signals obtained by an appliance simulator.
Experiments on real and synthetic mixtures showed very promising results, with an energy estimation error less than 1% for the main contributors of the household.