Improved Power Transformer Incipient Fault Detection Using a New Gas Ratio and Decision Tree Algorithm based on Dissolved Gas Analysis

Shafiei Asl, Milad; Haghjoo, Farhad

doi:10.48308/ijrtei.2024.236235.1051

Improved Power Transformer Incipient Fault Detection Using a New Gas Ratio and Decision Tree Algorithm based on Dissolved Gas Analysis

Document Type : Research Article

Authors

Milad Shafiei asl

Farhad Haghjoo

Faculty of Electrical Engineering, Shahid Abbaspour School of Eng., Shahid Beheshti University, Tehran, Iran

10.48308/ijrtei.2024.236235.1051

Abstract

Power transformers are critical and costly components of power systems, necessitating effective fault detection methods to ensure their reliability and longevity. This paper introduces a new gas ratio (NGR) method utilizing a decision tree (DT) algorithm to detect incipient faults in transformers through dissolved gas analysis (DGA). By incorporating ten gas ratios, the proposed method enhances fault detection accuracy while maintaining simplicity and ease of use. The DT was developed and evaluated using a dataset from the Egyptian Electricity Holding Company, demonstrating superior performance compared to traditional methods. The results indicate that the NGR method significantly improves fault detection accuracy, especially in identifying partial discharge (PD) and high-temperature overheating (T3) faults. This approach offers a promising solution for practical transformer maintenance and fault diagnosis. This research highlights the potential of integrating advanced machine learning techniques with traditional DGA methods to improve fault detection accuracy, reduce maintenance costs, and enhance the reliability of power transformers.

Keywords

Power transformers

Fault detection

Dissolved gas analysis

Gas ratios

Machine learning

Decision tree

Subjects