Abstract: The frequency domain reflection (FDR) method, known for excellent capabilities in detecting and localizing minor impedance changes in faults, is widely applied in cable fault diagnosis. However, recent field experience has shown that this method is still inadequate in determining the types and severity of faults. To address these issues, an optimized algorithm based on the frequency-domain to space-domain integral transform is proposed in this paper to enhance the cable fault localization capability and to assess the fault type and severity. Initially, spectral information is modulated into the cable's frequency information using the soliton's good self-sustaining properties, as demonstrated in optical fiber communication. A decay operator is then defined to perform attenuation compensation on the spectral information and the integral kernel function, and a Chebyshev window function is employed to diminish the low-frequency components of the spectrum. This method not only maintains high detection sensitivity and precision but also enables discrimination of impedance mismatches and severity assessment based on reflection amplitude. Research has been conducted with simulations on a 10 kV cable simulation line involving inductive faults, capacitive and joints at different mismatch levels; real measurements have been carried out on coaxial cables, a 10 kV experimental cable line, and a 110 kV operational cable line. The results show that a localization precision error of less than 2% can be achieved, accurately determining fault types and assessing their severity, thereby verifying the effectiveness and feasibility of the improved algorithm.