Enhancing Circuit Breaker Fault Detection With Data-Augmented SVD-XGBoost Methodology

This study introduces a novel approach for fault detection in high-voltage circuit breakers, utilizing a data-enhanced SVD-XGBoost model. It focuses on addressing the limitations of traditional offline diagnostic methods by applying advanced signal processing and machine learning techniques. The paper proposes the use of the Normalized Least Mean Squares (NLMS) technique for data augmentation, coupled with Fast Fourier Transform (FFT) for time-frequency spectrogram conversion. Singular Value Decomposition (SVD) is then employed for dimensionality reduction of the spectrogram data, and the SVD-XGBoost algorithm is used for fault classification. This method aims to improve the reliability, efficiency, and accuracy of fault detection in circuit breakers, contributing significantly to the stability of power systems.