Developing User Centric HEMS Through Automated Appliance Recognition Framework

Abstract

HomeEnergy Management Systems (HEMs) have been proven to help home users manage their power consumption and improve usage habits. With more advanced HEMs incorporating appliance recognition technology to enable tracking of appliances via its unique electrical signature, there still exists the drawback of requiring complex yet time consuming appliance registration stages. To curb this problem, this paper presents the framework required to automate the appliance registration process to create a much more user centric system. By demonstrating the working of the framework using one-class support vector machine with additional principal component analysis feature extraction using 10 household appliances, the classification rate of unregistered appliance into its rightful class was 100% with a recall rate of 67.04% for registered appliances. The results were obtained based on leave-one-out cross validation technique, excluding the results of the training dataset.