Appliance Recognition Using a Density-based Clustering Approach with Multiple Granularities

Abstract

Electricity may not be economically stored as other forms of energy such that it would be in short supply during the peak time In view of this difficulty most power suppliers encourage their customers to adopt time-of-use rate plans Consequently it is essential for a user to be able to perceive the real-time information of power consumption With the advancement of Internet of Things technologies smart sockets are becoming a commodity to manage power consumption in a household However current smart sockets merely present the total electricity consumption rather than the individual consumption of household appliances In this work we thus investigate the problem of appliance recognition and implement an unsupervised algorithm on a modular smart socket so as to identify each appliance on the socket Specifically we propose to adopt a density-based clustering approach to perform this appliance recognition task Furthermore appliances with similar load signatures (or power features) can be identified by considering different data granularities in our approach Experimental studies show that our approach is feasible even when there is no prior knowledge of new appliances We also develop a prototype system with graphical user interfaces to present the real-time power consumption of individual appliance With the user interaction our system can learn from the user feedback

Date of Award	2019
Original language	English
Supervisor	Wei-Guang Teng (Supervisor)