Banca de QUALIFICAÇÃO: FELIPE DUARTE PINTO COELHO

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
DISCENTE : FELIPE DUARTE PINTO COELHO
DATA : 10/07/2026
HORA: 14:00
LOCAL: meet.google.com/qou-wsxy-ogp
TÍTULO:

A Proposal for an Automatic Oscillography Classifier Using Artificial Intelligence Techniques in Systems with High Penetration of Distributed Generation

PALAVRAS-CHAVES:

Electric Power Systems; COMTRADE oscillography; Event classification; Distributed generation; Artificial Intelligence; Feature extraction; Signal processing; Explainable Artificial Intelligence.

PÁGINAS: 72
GRANDE ÁREA: Engenharias
ÁREA: Engenharia Elétrica
SUBÁREA: Sistemas Elétricos de Potência
ESPECIALIDADE: Medição, Controle, Correção e Proteção de Sistemas Elétricos de Potência
RESUMO:

The increasing complexity of Electric Power Systems (EPS), driven by the insertion of new technologies and, especially, by the integration of Distributed Generation (DG), imposes significant challenges on traditional protection and disturbance analysis strategies. In this context, the use of oscillographic records in the COMTRADE format has become an important source of information for the diagnosis of electrical events, enabling the detailed analysis of voltage and current behavior under abnormal operating conditions. Therefore, the development of automated fault classification methodologies, combined with result interpretation, contributes to increasing the reliability of electrical power systems.  This work proposes a methodology for oscillography analysis, encompassing everything from the acquisition and organization of real and simulated data to the application of Artificial Intelligence (AI) techniques for automatic event classification. Initially, data obtained from real numerical relays records, as well as signals generated through engineering software simulations, are considered, enabling the coverage of different operating scenarios, including conditions with and without the presence of distributed generation. Subsequently, the signals undergo preprocessing stages, including standardization, temporal segmentation, and feature extraction, for use in machine learning models. Afterwards, classical machine learning algorithms, such as Decision Trees and Random Forests, are applied with the objective of performing the automatic classification of electrical events. Additionally, the work incorporates an analysis of the impact of distributed generation on electrical signal patterns and classifier performance, exploring different penetration levels and their implications for model decisions. Furthermore, through Explainable Artificial Intelligence (XAI), the study seeks to interpret model decisions, aiming to increase the practical acceptance of the proposed solutions. The obtained results allow not only the evaluation of the methods’ performance, but also provide support for their application in real operating environments.

MEMBROS DA BANCA:
Presidente - 2132806 - AURELIO LUIZ MAGALHAES COELHO
Interno - 2362923 - GIOVANI BERNARDES VITOR
Externo à Instituição - JOÃO RICARDO DA MATA SOARES DE SOUZA - UFMG
Externo ao Programa - 1611588 - RAFAEL FRANCISCO DOS SANTOS - UNIFEIInterno - 1183761 - SANDRO CARVALHO IZIDORO