An intelligent system based on optimized ANFIS and association rules for power transformer fault diagnosis

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Materials Science

Background:

• Power transformer failures can lead to significant economic losses and grid instability.
• Accurate fault diagnosis is crucial for timely maintenance and preventing catastrophic failures.
• Dissolved Gas Analysis Method (DGAM) provides essential data for transformer health assessment.

Purpose of the Study:

• To develop an intelligent method for accurate diagnosis and classification of power transformer faults.
• To enhance fault detection and recognition precision using optimized machine learning algorithms.
• To improve the robustness and efficiency of power transformer fault diagnosis systems.

Main Methods:

• Utilized 14 attributes from Dissolved Gas Analysis Method (DGAM) as input for Adaptive Neuro-Fuzzy Inference System (ANFIS).
• Applied Association Rules Learning Technique (ARLT) for selecting the most informative DGAM attributes.
• Employed Black Widow Optimization Algorithm (BWOA) for optimal training of the ANFIS model.

Main Results:

• The proposed method demonstrated high accuracy in fault detection and classification.
• The system exhibited robust performance and short run times when evaluated on industrial datasets.
• Attribute selection and optimized ANFIS training significantly improved diagnostic precision.

Conclusions:

• The developed intelligent diagnosis system effectively identifies and classifies power transformer faults.
• Optimizing ANFIS with BWOA and selecting informative DGAM attributes are key contributions.
• This research offers a reliable and efficient solution for power transformer fault diagnosis.