Cross-linked polyethylene (XLPE) cables are essential in modern power distribution, yet insulation defects such as localized aging, moisture ingress, and sheath damage threaten grid reliability. Conventional diagnostic methods, including time-domain reflectometry and broadband impedance spectroscopy, lack resolution or require outages, limiting their use in live systems. To overcome these challenges, we developed an advanced online defect localization system, employing frequency-modulated continuous-wave signals with a novel noninvasive capacitive coupling sensor. To mitigate signal attenuation caused by metallic shielding, a specialized hardware interface and a selective shielding removal strategy were implemented. Experimental results demonstrated high spatial accuracy: transmission distances of 97.3 and 101.26 m matched physical cable lengths under core-conductor and optimized capacitive coupling configurations. Capacitive and inductive sensor data exhibited a 4.46% difference in transmission speeds, highlighting distinct coupling characteristics. The system developed in this study successfully localized a 2 m thermal aging defect at 41 m along a 64 m cable, with only 5.10% relative error. Furthermore, notch peak intervals were shown to characterize defect severity, while single-cycle signal mixing mitigated harmonic interference, producing smoother spectral curves. By translating longitudinal delays into frequency differences, the system can be used to develop cable diagnostics from passive monitoring to active, high-resolution live-line detection, laying the foundation for intelligent sensor technologies in smart grid infrastructure.