Time-Frequency Feature Extraction and Classification Method of XLPE Cable Insulation Aging Signal

To address the challenges of weak signal extraction and transient disturbance susceptibility in online monitoring of XLPE cable insulation aging, an aging signal identification method based on time-frequency panoramic information was proposed, enabling precise non-stop power-off diagnosis of cable insulation status. First, a nonlinear equivalent model of XLPE cable insulation aging regions was established to reveal fundamental differences in time-frequency distribution between aging signals and transient disturbances (e.g., capacitor switching) from a mechanistic perspective: aging signals were characterized by persistent high-frequency components, while disturbance signals demonstrated exponential decay of high-frequency components. Subsequently, robust masked time-frequency representation decomposition (RMCMD) method was applied to decompose collected signals, and a time-frequency spectral matrix was constructed. For the first time, time-frequency energy entropy (TFEE) was introduced as a quantified feature to characterize the complexity of signal energy distribution. By setting appropriate thresholds, automatic differentiation between aging signals and transient disturbances was achieved. Simulation results demonstrated that the proposed method accurately identified insulation aging signals under varying noise levels, aging severity, and location conditions, with significantly higher recognition accuracy for transient disturbance signals compared to traditional time-frequency analysis methods. The online diagnostic approach based on RMCMD algorithm and TFEE proposed in the study was shown to effectively and robustly identify early-stage insulation aging in XLPE cables, providing a non-stop, easy-to-implement technical pathway for condition-based maintenance.