Fault Detection and Classification in Electrical Systems: A Machine Learning and Fuzzy Logic-Based System

• Main Authors: Carvalho, Davi, Melo, Jose F. V., Silva, Eraylson G., Wesley, Jose, Passos, Dioeliton, De Mattos Neto, Paulo S. G.
• Format: Conference Proceeding
• Language: English
• Subjects: Cat-Boost; Decision tree; Electrical Systems; Expert Systems; Fault detection; Fuzzy logic; Fuzzy Rules; Knowledge based systems; LightGBM; Machine learning; Machine learning algorithms; Maintenance engineering; Power supplies; XGBoost
• ISSN: 2768-0045
• DOI: 10.1109/WCNPS60622.2023.10345098

In order to ensure a continuous and uninterrupted power supply, the detection and classification of faults in electrical systems play a critical role. This paper introduces an innovative decision support system that combines Machine Learning (ML) and Fuzzy Logic (FL) techniques to detect, classify, and diagnose faults within electrical systems. This comprehensive system consists of two key modules: the Classification Module and the Rules Module. The Classification Module utilizes a Decision Tree algorithm to generate a set of rules that enable the accurate classification of the current status of the electrical system. On the other hand, the Rules Module empowers electrical system operators to manipulate this knowledge base, creating more understandable and interpretable rules by employing FL. The results obtained from experimentation using two datasets demonstrate that the proposed system outperforms most of the evaluated classification models and produces more straightforward and interpretable rules. This advancement holds significant promise for enhancing the reliability and maintenance of electrical systems.