Transformer fault diagnose intelligent system based on DGA methods

Area of Science:

• Electrical Engineering
• Power Systems
• Artificial Intelligence

Background:

• Power transformers are critical infrastructure; malfunctions cause significant economic losses for utilities.
• Transformer failures stem from electrical, thermal, and mechanical stresses on insulation (oil and paper).
• Dissolved Gas Analysis (DGA) is standard for fault detection, but traditional methods (IEC Code, Rogers Ratio, Duval triangle) have limited accuracy.

Purpose of the Study:

• To develop an advanced intelligent system for enhanced power transformer fault diagnosis.
• To improve the diagnostic accuracy of existing Dissolved Gas Analysis (DGA) methods.
• To create a Transformer Fault Diagnosis Intelligent System (TFDIS) by integrating multiple DGA techniques.

Main Methods:

• Developed a Transformer Fault Diagnosis Intelligent System (TFDIS).
• Integrated outputs from four DGA methods: Code Tree 2020, modified IEC, Rogers' Ratio, and Neural Pattern Recognition.
• Compared the diagnostic accuracy of the integrated system against individual methods.

Main Results:

• The developed TFDIS achieved a diagnostic accuracy of 89.12%.
• This accuracy surpasses the highest individual accuracy of 86.01% obtained by Neural Pattern Recognition.
• The integrated approach significantly enhances analytical accuracy in diagnosing transformer faults.

Conclusions:

• The Transformer Fault Diagnosis Intelligent System (TFDIS) offers superior fault diagnostic accuracy compared to individual DGA methods.
• Integrating multiple DGA techniques within an intelligent system framework is effective for reliable power transformer health monitoring.
• The TFDIS represents a significant advancement in ensuring power system reliability and reducing utility losses.