Rotor Fault Diagnosis of Hydroelectric Generator Units Based on KPCA-GWO-KELM

To achieve precise fault diagnosis of hydroelectric generator units, it proposes a rotor fault diagnosis model integrating Kernel Principal Component Analysis (KPCA), Grey Wolf Optimization (GWO), and Kernel Extreme Learning Machine (KELM). The diagnostic framework proceeds as follows: first, the variational mode decomposition (VMD) method is applied to decompose the rotor vibration signals of hydroelectric generator units, followed by the extraction of time-domain statistical features associated with rotor faults. Subsequently, KPCA is applied to reduce the dimensionality of the extracted time-domain statistical features. Finally, the GWO algorithm is utilized to optimize the Gaussian kernel coefficient and penalty coefficient of KELM. Case study results demonstrate that the KPCA-GWO-KELM model achieves an average diagnostic accuracy of 98.88% with an average computation time of 14.59 seconds, outperforming alternative comparative models in both diagnostic precision and efficiency.