Revo

Abstract

The prediction of energy consumption is important for the management of the systems, contributing to the stable and reliable operation of the matrix. The use of neural networks to model electrical systems is highlighted. The aim of this systematic review is to evaluate 15 articles that used neural networks to forecast electricity consumption, and, through the analysis of the input variables, network configuration and performance, identify whether this is an appropriate method for predicting electricity. It was observed that the selected articles are current and have been published in journals of good quality. The most popular input variables are related to consumption and population growth. In network configuration, networks with only one hidden layer are typically used. The number of neurons varies, but the highest value found was 10 neurons. For performance evaluation, in addition to various types of errors, statistical tools can be used, or even comparison with other models. It was concluded that the use of neural networks was adequate in 87% of the articles analyzed in this study.