Approach for home appliance recognition using vector projection length and Stockwell transform

Recognition of residential electrical appliances has been a highlighted area of research in recent years due to increasing energy consumption in residential sectors around the world. The use of instant current from 17 commonly presented electrical appliances in homes is proposed. The Stockwell transform, a time-frequency-domain transform, is applied to current collected data to be used as a load signature. Then, a novel method for identification of electrical appliances was employed to analyse the data. This method is an image pattern recognition technique, called vector projection length (VPC). In this case, the idea is to recognise patterns in appliances transformed into a 2D image. An extensive analysis is presented, analysing the trade-off between the training set database, the number of current cycles used and the variable k from the VPC projection equation, searching for optimal parameters for this application. Results have shown successful identification employing VPC, close to 90% of devices being identified.