Transformer Incipient Fault Prediction Using Combined Artificial Neural Network and Various Particle Swarm...

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Predictive Maintenance

Background:

• Accurate prediction of incipient faults in transformer oil is crucial for effective maintenance and cost reduction.
• Dissolved gas analysis (DGA) is a common method, but existing techniques can be inaccurate due to condition-specific limitations.
• Previous intelligence methods for transformer fault prediction show room for accuracy improvement.

Purpose of the Study:

• To propose a novel approach combining artificial neural networks (ANN) and particle swarm optimization (PSO) for enhanced transformer incipient fault prediction.
• To evaluate the effectiveness of various PSO techniques integrated with ANN.
• To demonstrate the superiority of the proposed methods over existing techniques and prior research.

Main Methods:

• Implementation of a hybrid model integrating ANN with different PSO techniques.
• Validation of the proposed model by comparing its predictions against actual fault diagnoses and existing DGA methods.
• Benchmarking the performance against ANN alone and previously reported fault prediction studies.

Main Results:

• The proposed ANN-Evolutionary PSO method achieved the highest percentage of correct fault type identification.
• The hybrid ANN-PSO approach demonstrated improved accuracy compared to ANN alone and existing diagnosis methods.
• The study confirmed the effectiveness of integrating PSO with ANN for transformer fault prediction.

Conclusions:

• The combination of ANN and PSO, particularly the ANN-Evolutionary PSO variant, offers a highly accurate solution for predicting transformer incipient faults.
• This advanced method enhances the reliability of power transformer maintenance by minimizing errors and costs.
• The findings suggest a significant advancement in intelligent fault diagnosis for power transformers.