Recurrent neural networks RNNs and decision tree DT machine learning-based approaches for transmission system faults diagnosis

Tools

Abuanwar, Sayed and Saeed, Mohammed and Mosalem, Hanan and Khater, Hatem (2023) Recurrent neural networks RNNs and decision tree DT machine learning-based approaches for transmission system faults diagnosis. Journal of Engineering Research, 7 (5). 67–76. 10. ISSN 2735-4873 (https://doi.org/10.21608/erjeng.2023.233286.1224)

[thumbnail of Abuanwar-etal-Recurrent-Neural-Networks-RNNs-and-Decision-Tree]
Preview
 Text. Filename: Abuanwar-etal-Recurrent-Neural-Networks-RNNs-and-Decision-Tree.pdf
Final Published Version
Download (1MB)| Preview

Abstract

Accurate and prompt detection of system faults are crucial to maintain sufficient protection of system equipment, avoid false tripping, and cascaded failures. This paper presents a comprehensive study on the effectiveness of machine learning techniques for electrical fault detection and classification. Specifically, a comparative analysis is conducted between two prominent algorithms: Recurrent Neural Networks (RNNs) and Decision Tree (DT). The study employs a dataset comprising realworld electrical fault scenarios to evaluate the performance of RNNs and DT in identifying and categorizing faults. While DT algorithm showed slightly better accuracy in some cases, the RNN exhibited better generalization capabilities and a lower risk of overfitting. The analysis involves various performance metrics such as accuracy, precision, recall, and confusion matrices to comprehensively assess the algorithms' capabilities. The findings for analyzing the operational behavior of such systems. Various studies exist on waveforms characterization with the central premise of obtaining pronounced markers for accurate fault identification [7-14]. An enhanced Fast Fourier Transform-based method is devoted for extracting distinguished voltage features for identifying voltage dips is introduced in [7-9]. provide valuable insights into the strengths and limitations of each approach in the context of electrical fault management. This paper contributes to the selection of suitable techniques based on specific application requirements, advancing the field of predictive maintenance and fault mitigation in electrical systems.

ORCID iDs

Abuanwar, Sayed ORCID logoORCID: https://orcid.org/0000-0002-3396-4020, Saeed, Mohammed, Mosalem, Hanan and Khater, Hatem;
CORE (COnnecting REpositories)