Impacto do Registro de Consumo de Energia pela Média de Consumo na Detecção de Furto de Energia Utilizando Redes Neurais Artificiais

Abstract

The energy theft generates significant losses for the electricity distribution concessionaires. Research over the past decade has focused on machine learning as a method of detecting energy theft by analyzing consumption logs. The use of convolutional neural networks is one of these methods. Operating procedures can affect the quality of consumption data, impacting the performance of machine learning methods. In this paper is evaluate how the procedure of concessionaries of use the consumption average affect the performance of the energy theft detection method using a convolutional neural network.