Detecção de novidades em aparelhos eletrônicos através do monitoramento do consumo de energia

Abstract

Electricity in Brazil is mostly generated by hydroelectric plants that depend on the volume of water in their reservoirs. Due to the fact that rainfall is decreasing, other methods with higher costs to generate energy are required. These costs are passed to users, increasing the energy bill. Futhermore, the wasting of energy and overconsumption also contribute to increase the energy bill. At the same time, the wasting of energy are not noticed by the user. To avoid such wasting, an alert could be sent as soon as an anomalous event is detected. In this way, we propose a system that sends an alert of any novelty detection in energy consumption through the analysis of the methods Sliding Window, Exponentially Weighted Moving Averages, Clustering, Average per Circle, Average per Stage, Gauss Distribution and Self-Organizing Novelty Detection. Results demonstrate that the methods evaluated are efficient in real time detection of novelties, presenting 90% of accuracy and 10% of recall, besides a low delay to send the alert.