DC Microgrid Arc Fault Diagnosis Based on Feature Frequency Extraction and Kernel Extreme Learning Machine

To address the challenge of series arc fault detection in DC microgrid systems, this research proposes an arc fault diagnosis approach based on feature frequency extraction and kernel extreme learning machine (KELM). The arc fault characteristic frequency is chosen by means of the signal's variational mode decomposition (VMD) and Hilbert marginal spectrum (HMS) analysis. The singular value, sample entropy, and energy entropy of the modal component that corresponds to the eigenfrequency are computed in order to generate the eigenvector. Subsequently, the KELM is utilized for arc fault diagnosis. Experimental results demonstrate that the proposed method effectively extracts fault characteristics, achieving a recognition accuracy of <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">98.26%</sup> in arc fault diagnosis.