Transformer windings defects identification using frequency response analysis and advanced data visualization...

Area of Science:

• Electrical Engineering
• Power Systems Analysis
• Diagnostic Techniques

Background:

• Transformers are vital in power networks, facing stresses that can cause subtle winding faults.
• Early detection of these faults is critical to prevent major failures and ensure grid stability.
• Frequency Response Analysis (FRA) is a common diagnostic tool, but its complex results hinder interpretation.

Purpose of the Study:

• To develop advanced data visualization techniques for interpreting Frequency Response Analysis (FRA) results.
• To enhance the early prediction and diagnosis of transformer winding faults.
• To simplify the interpretation of FRA data for engineers and operators.

Main Methods:

• Employed Factor Analysis to identify complex, hidden fault factors.
• Utilized Fuzzy Clustering Analysis to detect combined fault conditions with uncertainty.
• Applied Principal Component Analysis for data dimensionality reduction and enhanced interpretability.
• Developed a two-stage identification model for healthy/faulty distinction and subsequent fault classification.

Main Results:

• The proposed techniques effectively extracted frequency response features for accurate fault identification.
• Experimental results demonstrated high accuracy in diagnosing transformer winding faults.
• The methods successfully distinguished between healthy and faulty transformer conditions.
• Fault classification under faulty conditions was achieved with significant accuracy.

Conclusions:

• Advanced data visualization techniques simplify FRA interpretation for transformer fault diagnosis.
• The developed methods reduce the need for specialized expertise in fault diagnosis and classification.
• This approach offers a more efficient and accessible way to diagnose transformer winding faults.