A Kernel Correlation-Based Enhanced Resolution Fault-Location Method for Lossy Transmission Lines

This article presents a nonlinear fault-location method that leverages a kernel correlation coefficient to map transient fault waveforms into a high-dimensional Hilbert space using a Gaussian radial basis function (RBF) kernel. This approach enhances spatial resolution by nonlinearly amplifying subtle differences between signals that appear similar in the original space. Unlike linear methods, the resolution is not limited by signal wavelength components but is controlled by a tunable kernel scale parameter, enabling a flexible tradeoff between accuracy and robustness, even under strong attenuation. Field validation using real fault data from a 10-kV distribution system shows that the proposed method achieves a location error of only 13 m (0.85%), whereas the error of the conventional natural frequency-based method ranges from 67 to 198 m (6.6% –9.6%). Moreover, a spatial resolution of 60 m based on low frequency components (below 120 kHz) was found, a fivefold improvement over the linear correlation metric. The proposed kernel-based framework provides an effective and scalable solution for high-resolution fault location in distribution networks and long HVDC power lines with complex topologies and strong attenuation characteristics.