Appliance recognition system for ILM using AGILASx — Dataset of common appliances in the Philippines

Abstract

This study presents the development of a system which can automatically recognize home appliances based on a dataset of electric consumption profiles. The authors report the creation of AGILASx, a dataset of 50 common home appliances and devices in the Philippines. The dataset is populated with 100 appliance signatures in .XML format acquired using plug-based sensors. Each appliance signature consists of the following electric characteristics: real power (W), apparent power (VA), reactive power (var), RMS current (A), RMS voltage (V) and Power Factor (PF). A machine learning approach was utilized for the recognition experiment following a set of test protocols — intersession and unseen instances. The baseline recognition algorithm used was the k-Nearest Neighbor (k-NN) for both test protocols and accuracy levels were collected over three different acquisition frequencies. Using results of the confusion matrices, best results were observed at acquisition frequency of 10−1 Hz for intersession (99%) and unseen instance (99%) test protocols. Lastly, to integrate the dataset and the recognition algorithm, a web application was developed adapting a Web-of-Things architecture to present a smart of recognized appliances and their corresponding consumption.