<scp>VMD</scp> ‐Based Feature Extraction and Enhanced <scp>GWO</scp> ‐ <scp>DBN</scp> for Health Assessment of Automatic Transfer Switching Equipment

This article proposes a method for assessing the health condition of automatic transfer switching equipment (ATSE) during the switching process. The method combines variational mode decomposition (VMD) with deep belief networks (DBN) for non‐invasive monitoring and fault diagnosis. First, the VMD method is introduced to address mode mixing, using sample entropy to determine the decomposition iterations of VMD. Wavelet packet energy entropy is then extracted as the feature for health condition assessment. Subsequently, the Gray Wolf Optimization (GWO) algorithm is enhanced with a nonlinear convergence factor and a dynamic weight strategy to improve performance and avoid local optima. The enhanced GWO is used to optimize the network parameters of the DBN, which then serves as the pattern recognition algorithm for assessing ATSE health. Comparative experimental analysis demonstrates that the proposed method effectively addresses the health condition assessment of ATSE vibration signals, exhibiting high diagnostic accuracy. © 2024 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.