Non-intrusive load disaggregation based on digital signal processing for microcontroller application

This paper presents a low complexity non-intrusive load monitoring (NILM) approach for residential electric powerbased on digital signal processing. The aim is to identify the real operating frames of each household appliance from the aggregated current signal frames. In the methodology, two detection methods are derived. The first method, named direct method, identifies the active frames of each device by identifying the most probable combination between devices in the aggregated signal. The second method, termed indirect method, identifies the active frames of a particular device by means of a projection of the aggregated signal onto a Fourier subspace representing the characteristic footprint of the device. The methodology is tested on 4 datasets collected in Argentina. and high performance metrics are achieved. A pilot test is carried out with an ATSAMD21G18 microcontroller on the Itsy Bitsy M0 Express.