A Novel Method for Segment Detection of Cable Insulation Status Based on Improved ESPRIT with Fourth-Order Cumulants

To address the issue of decreased measurement accuracy in cable attenuation coefficient detection caused by Gaussian colored noise interference, this paper proposes an attenuation coefficient estimation method based on fourth-order cumulant improved estimation of signal parameters via rotational invariance technique (FOC-ESPRIT), for reliable insulation condition assessment of cable segments. First, it constructs FOC matrices to substitute the noise-susceptible correlation matrices, leveraging the intrinsic property that Gaussian noise exhibits zero FOC to suppress Gaussian noise interference at the source. Subsequently, equivalent attenuation coefficients are extracted via generalized eigenvalue decomposition of the FOC matrix pairs, and the true attenuation coefficient of each cable segment is derived through systematic decoupling operations. Simulation results under a 10 – 40 dB signal-to-noise ratio (SNR) demonstrate the method’s superior measurement performance, with a maximum estimation error of 10.14%. This significantly outperforms the 15.00% error of the TLS-ESPRIT method, offering notable enhancements in both anti-noise robustness and estimation precision. Experimental validation on multi-segment coaxial cables confirms that all estimation errors are within 6%. Finally, field tests on a 2.2 km/10 kV power cable successfully detection an insulation defect in segments 2 and 3 by accurate measurement of the attenuation coefficient, further verifying the method's technical effectiveness and engineering practicability.