A lightweight efficient model for household edge-side non-intrusive load monitoring

Abstract

Non-Intrusive Load Monitoring (NILM) gets the total electricity consumption data of a household through meters and separates the individual loads by learning the patterns of their energy consumption. Due to considerations of cost-effectiveness in household scenarios and the development of smart meters, NILM models should be deployed to edge devices with limited computational capability. Therefore, the size, complexity, and accuracy of these models are the key points of this paper’s focus. In this paper, we propose a lightweight model that can be deployed on edge devices. In this model, a feature extraction module is proposed to fully mine the load features for accurate disaggregation. Experiments on a public dataset show that the accuracy of the proposed model reaches the level of currently available complex models, and the size of the model is only 0.18 MB.