| Texto completo | |
| Autor(es): |
da Silva, Mauricio Pavani
;
Lopes, Gabriela Nunes
;
Vieira Jr, Jose Carlos de Melo
Número total de Autores: 3
|
| Tipo de documento: | Artigo Científico |
| Fonte: | Electric Power Systems Research; v. 251, p. 7-pg., 2026-02-26. |
| Resumo | |
High Impedance Faults (HIFs) are common events in Electrical Power Distribution Systems (DSs) and are responsible for imposing danger to public safety due to the potential to ignite fires and cause electric shocks. Many recent studies have applied methodologies based on intelligent algorithms (IA) and machine learning (ML) to classify these events. However, these studies did not address feature selection and collinearity analysis to investigate the quality of methods' input metrics, resulting in high-dimensional data inputs that may hinder the method training and generalization, as well as its interpretability. Therefore, the purpose of the present study is to perform a feature selection evaluation for HIFs. Then, the metrics are used as inputs to a newly proposed classification method based on Support Vector Machine (SVM). The main goal is to compose a model for practical implementations, therefore a generalization study was conducted in two stages: training the detection algorithm under normal system conditions and validating it with unseen system conditions and events. The model was validated with a wide variety of tests that indicate an accuracy greater than 97% in the generalization analysis. The dataset for training, testing, and validation was obtained using Alternative Transients Program (ATP) software. This study proposes a robust method for classifying HIF with potential for practical application, in addition to providing a model for developing new studies on HIF classification using intelligent algorithms. (AU) | |
| Processo FAPESP: | 24/16135-7 - Sistema Distribuído para Detecção de Faltas da Alta Impedância em Redes de Distribuição de Energia Elétrica |
| Beneficiário: | Maurício Pavani da Silva |
| Modalidade de apoio: | Bolsas no Brasil - Doutorado |