A novel swarm budorcas taxicolor optimization-based multi-support vector method for transformer fault diagnosis

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Transformer faults pose significant risks to power system reliability.
• Accurate fault detection is crucial for preventing cascading failures and ensuring grid stability.
• Existing transformer fault diagnosis methods often struggle with low recognition accuracy.

Purpose of the Study:

• To develop a novel and highly accurate method for transformer fault detection.
• To enhance the classification performance of transformer fault diagnosis models.
• To provide a robust technical solution for improving the operational reliability of power systems.

Main Methods:

• A Multi-Support Vector (MSV) model was developed for multi-classification of transformer faults using dissolved gas analysis data.
• A Swarm Budorcas Taxicolor Optimization (SBTO) algorithm was employed to optimize MSV model parameters during training.
• The proposed SBTO-MSV method integrates optimization and classification for effective transformer fault diagnosis.

Main Results:

• The SBTO-MSV method achieved a top average accuracy of 98.1% on the IEC TC 10 dataset.
• Experimental results show significant outperformance compared to traditional and state-of-the-art machine learning algorithms.
• Validation on additional datasets confirmed the model's excellent classification performance and generalization capabilities.

Conclusions:

• The SBTO-MSV method offers superior performance for transformer fault diagnosis.
• The SBTO algorithm demonstrates excellent parameter searching capabilities for machine learning models.
• This advancement provides strong technical support for enhancing transformer fault detection and ensuring power system reliability.