On time series representations for multi-label NILM

Given only the main power consumption of a household, a non-intrusive load monitoring (NILM) system identifies which appliances are operating. With the rise of Internet of things, running energy disaggregation models on the edge is more and more essential for privacy concerns and economic reasons. However, current NILM solutions use data-hungry deep learning models that can recognize only one device and are impossible to run on a device with limited resources. This research investigates in-depth multi-label NILM systems and suggests a novel framework which enables a cost-effective solution. It can be deployed on an embedded device, and thus, privacy can be preserved. The proposed system leverages dimensionality reduction using Signal2Vec, is evaluated on two popular public datasets and outperforms another state-of-the-art multi-label NILM system.