Partial discharge pattern recognition based on adaptive multi-resolution analysis of higher-order Singular Spectrum Entropy

the method Singular Spectrum Entropy Analysis (SSEA) studies the complexity and irregular extent of Partial Discharge (PD) signals but cannot fully reflect the intrinsic nonlinear characteristic. In this paper, we apply higher-order statistics to substituting for the co-variance matrix and introduce the Ensemble Empirical Mode Decomposition theory (EEMD) to realize multi-scale as well. The signal energies is also used to weight the entropy values. This method is called Multi-resolution Higher Order Singular Spectrum Entropy Analysis (M-HSSEA) of PD signal herein. Three typical insulation defect models of composite insulator are designed and detected by UHF. The method presented not only enhances the robustness of the phase space reconstruction parameters and improves the noise immunity, but it also fully reflects the nonlinear characteristics of the PD signal.