Naive Bayes applied impacts harmonic analysis in industrial electrical systems

Abstract

The research of this paper was supported by a technique that made use of a data-mining tool and probability theory to group the data and find the relationship in each set. The Naive Bayes technique sort the data by representing them with directed acyclic graphs in which node represent random variables and the arcs represent the direct probabilistic dependencies between each other. The sorting data method is very important when it is necessary to identify the influence, occurrence and importance of a data among a set of data (database). This data mining technique applied at this paper was used to determine the impact of the harmonic distortion of the current due the nonlinear loads present during the production process of a manufacturing facility located at Manaus industrial park. The process of data collecting was conducted by a measurement campaign during a week where the voltage and current of each production process and the sub process that compose the main like the temperature control process were collected and stored for more detailed analysis. The parameters used to analyze the energy quality were based on the module 8 from the Energy Distribution Proceedings (PRODIST), which rules the energy quality at the distribution network. The results of this paper can be used to conduct the countermeasures necessary to fix the current harmonic distortion influence at the voltage and consequently the active power consume reduction of the overall company.