Core Concepts
Deep learning and edge computing are utilized to solve the non-intrusive load monitoring (NILM) problem and appliance identification in smart homes.
Abstract
The article proposes a novel approach using deep learning and edge computing to address NILM and appliance identification challenges. It introduces a seq2-[3]-point CNN model for home-NILM, site-NILM, and appliance identification. The study focuses on energy disaggregation, edge computing, smart homes, appliance identification, and transfer learning. Results show high accuracy levels for home-NILM (up to 94.6%), site-NILM (81%), and appliance identification (88.9%). Various datasets like REDD and REFIT are used for training the models.
Stats
Maximum accuracy of 94.6% achieved for home-NILM.
Site-NILM accuracy reached 81%.
Appliance identification accuracy of 88.9% with ResNet-based model.