Sequence-to-point learning methods for Non Intrusive Load Disaggregation: A Review

Energy disaggregation, also known as non-intrusive load monitoring (NILM), is a single-channel blind source separation (BSS) problem. It aims to decompose the total energy consumption of a dwelling into individual appliance energy usage. This is achieved by disaggregating the aggregate power signal recorded by the smart meters and into individual appliance readings. Sequence to point methods are a form of Deep neural networks that have shown promise in tackling this problem. These methods use a window of the mains data as input and output a single data point representing the targeted appliance’s energy consumption. This work surveys contemporary sequence-to-point deep learning techniques and outlines our upcoming research in this field.