Neural Network Assisted Walsh Matrix-Based Fault Detection Technique for Power Transformer

A new fault detection technique for power transformers using Walsh matrix analysis and a feedforward neural network is presented. Initially, Walsh coefficients are computed from the Walsh matrix using three-phase current signals from both sides of the transformer. Features are extracted from these coefficients and processed by a feedforward neural network, effectively distinguishing internal fault from non- or external-fault conditions. Test cases are generated through simulation and a laboratory setup containing two different types of customized three-phase transformers for training, testing, and validation. Hyperparameter tuning is conducted to select the best-fit model. The authenticity of the proposed method is also validated using field-recorded data. Finally, the performance of the proposed method is compared with other existing techniques in terms of operating characteristics, computational complexity, and ease of implementation. Results demonstrate the superiority of the proposed approach in detecting internal faults in power transformers compared to other methods.