An IoT deep learning-based home appliances management and classification system
Zahra Solatidehkordi, Jayroop Ramesh, A. R. Al-Ali, Ahmed Osman, Mostafa F. Shaaban
Abstract
The rise in household energy consumption globally has increased the necessity for effective electricity consumption management and load monitoring. Smart meters can facilitate fine-grained analysis by providing consumption insights even at the level of individual appliances, for detecting deterioration of appliances, anomalous behavior, and demand response. In this work, we propose a smart home appliance classification that utilizes the deep learning architecture of Long Short-Term Memory (LSTM) trained on the latest version of the Plug-Load Appliance Identification Database (PLAID). The model achieves competitive precision, recall and F1-scores across 16 different home appliances manufactured by 330 vendors. The model is then deployed on a Raspberry Pi micro-controller and interfaced with smart meters in a home to generate almost real-time classification of appliances and transmit this to a cloud database. The results and insights are made accessible to the end user or utility provider through a mobile application connected to the same database.