Inteligência artificial para avaliação da qualidade da energia do sistema elétrico de um prédio da Universidade do Estado do Amazonas

Abstract

This research presents a study on the application of a system based on artificial intelligence to evaluate the quality of electricity at the Escola Normal Superior. An artificial intelligence model consisting of two scripts was developed. The first script uses a fuzzy controller that employs rules in the if-then format, based on the Distribution Rules and Procedures (PRODIST) of the National Electric Energy Agency (ANEEL). The result of the first script is an assessment of the power quality as good or bad. This information is then inserted into a dataframe to provide the target for the next script. The second script is based on machine learning models. The available samples were divided into a portion of 70% for training and 30% for testing, in order to select the best classifier for future online use. The machine learning models that showed the best precision and accuracy rates were DecisionTree and RandomForest, achieving metrics of over 99%. The study was validated with system data collected by an electrical magnitude analyzer installed at the ENS. The data obtained indicates poor energy quality in the ENS building, but has enabled the creation of an artificial intelligence model that not only classifies the energy quality, but can also contribute to improving it in the future.