Enhancing Transformer Protection - A Machine Learning Framework for Early Fault Detection

Aug 4, 2026

Upload by Noah

Reliable power transformer operation is vital for grid stability, yet many existing fault-detection methods produce inaccuracies and high false alarms. This study proposes a machine learning framework that uses voltage signals for early fault detection. Fault scenarios covering single line-to-ground, line-to-line, turn-to-ground, and turn-to-turn faults are simulated on a three-phase laboratory transformer. Decision trees deliver 99.90% accuracy in 5-fold cross-validation and 95% on 400 unseen samples, with a 0.47% false alarm rate on a separate healthy dataset, supporting cost-effective, robust, and scalable grid reliability.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

Introduction

Motivation and limitations of traditional methods

Machine learning opportunities and challenges

Proposed framework

Data and voltage-signal approach

Model comparison and evaluation

Article Enhancing Transformer Protection: A Machine Learning Framework for Early Fault Detection Mohammed Alenezi 1,*, Fatih Anayi 1, Michael Packianather 2 and Mokhtar Shouran 3,4 Citation: Alenezi, M.; Anayi, F.; Packianather, M.; Shouran, M. Enhancing Transformer Protection: A Machine Learning Framework for Early Fault Detection. Sustainability 2024, 16, 10759. [https://doi.org/](https://doi.org/)[ ](https://doi.org/)[10.3390/su162310759](10.3390/su162310759) Academic Editors: Najib El Ouanjli and Said Mahfoud Received: 3 November 2024 Revised: 26 November 2024 Accepted: 6 December 2024 Published: 8 December 2024 Copyright: © 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . 1 Wolfson Centre for Magnetics, School of Engineering, Cardiff University, Cardiff CF24 3AA, UK 2 High-Value Manufacturing Group, School of Engineering, Cardiff University, Cardiff CF24 3AA, UK 3 The Libyan Center for Engineering Research and Information Technology, Bani Walid 00218, Libya 4 Department of Control Engineering, College of Electronics Technology, Bani Walid 00218, Libya * Correspondence: [alenezim1@cardiff.ac.uk](alenezim1@cardiff.ac.uk) Abstract: The reliable operation of power transformers is essential for grid stability, yet existing fault detection methods often suffer from inaccuracies and high false alarm rates. This study introduces a machine learning framework leveraging voltage signals for early fault detection. Simulating diverse fault conditions—including single line-to-ground, line-to-line, turn-to-ground, and turn-to-turn faults—on a laboratory-scale three-phase transformer, we evaluated decision trees, support vector machines, and logistic regression models on a dataset of 6000 samples. Decision trees emerged asthe most effective, achieving 99.90% accuracy during 5-fold cross-validation and 95% accuracy on a separate test set of 400 unseen samples. Notably, the framework achieved a low false alarm rate of 0.47% on a separate 6000-sample healthy condition dataset. These results highlight the proposed method’s potential to provide a cost-effective, robust, and scalable solution for enhancing transformer fault detection and advancing grid reliability. This demonstrates the efficacy of voltage-based machine learning for transformer diagnostics, offering a practical and resource-efficient alternative to traditional methods. Keywords: power transformer; fault detection; machine learning; decision trees; voltage analysis; classification algorithms 1. Introduction Transformer failures in electrical power systems can lead to costly outages, significant equipment damage, and compromised grid stability. Timely and accurate fault detection is essential to prevent these failures. Traditional methods like dissolved gas analysis (DGA) and frequency response analysis (FRA) have been widely used but face limitations in real-world applications [1–4] . These techniques can be time-consuming, require specialized equipment, and lack precision in detecting early-stage faults. Moreover, they rely on expert interpretation and predefined thresholds, increasing the risk of misdiagnosis or delayed responses. This underscores the need for more efficient and reliable fault detection methods. Machine learning (ML) offers a promising, data-driven alternative for transformer fault diagnosis [5] . However, current ML approaches are often limited by their reliance on manually selected features and simplistic models, which struggle to capture the complex behavior of transformer faults. Furthermore, these models are less effective when dealing with imbalanced datasets—a common issue in fault detection where normal operating conditions

FAQs

What problem does the framework address in transformer protection?

It targets inaccurate early fault detection and high false alarm rates that reduce reliability in existing methods.

Which fault types are considered for early detection?

The study simulates single line-to-ground, line-to-line, turn-to-ground, and turn-to-turn faults.

Why do decision trees perform best in the reported results?

Decision trees achieve the highest accuracy in 5-fold cross-validation (99.90%) and strong performance on a separate test set (95%), while maintaining a low false alarm rate (0.47%).

Aug 4, 2026

Upload by Noah