A High-Resolution Localization Method for Power Cable Defects Based on Pisarenko Harmonic Decomposition

Due to the non-periodicity of the reflection coefficient spectrum (RCS) and the presence of measurement noise, spurious peaks exist in the frequency domain reflectometry (FDR) location spectrum of cable defects, which can easily lead to serious misjudgments in defect localization. This paper proposes a new parameter estimation method with higher resolution for cable defect localization based on Pisarenko harmonic decomposition (PHD). First, we use higher-order statistics to estimate the third-order cumulant of RCS and extract the diagonal slice of the third-order cumulant. Then, a linear difference model is used to fit the probability density distribution of the diagonal slice and construct an autoregressive moving average (ARMA) model to replace the fast Fourier transform (FFT) in traditional FDR. Afterwards, all frequencies in the diagonal slice are estimated based on PHD, achieving further high-resolution defect localization. This method effectively overcomes the impact of attenuation, multi-reflection, and dispersion effects in cable defect localization. Finally, the feasibility of proposed method is verified by simulation and the measurement of a 58 km submarine cable on site. The result indicates that the PHD method can not only weaken the influence of spurious peaks on defect localization, but also has strong noise resistance capability.