Early faults in cables are characterized by short-duration discharges and weak fault signatures, making them difficult to trigger conventional overcurrent protection and complicating fault detection. If not detected and addressed in time, such faults can gradually deteriorate into permanent failures. To address the challenge of sensitively and reliably detecting early-stage cable faults that are prone to occurring in medium-voltage distribution networks, this paper proposes a cable early fault detection method based on the sheath grounding wire current and the segmented Teager-Kaiser energy operator (TKEO). First, optimal variational mode decomposition (OVMD) is applied to decompose the sheath grounding wire current signal. The optimal decomposition number is determined using the central frequency method, yielding a set of intrinsic mode function (IMF). Then, the pulse factor and cosine similarity of each mode are calculated, and the mode component with the most pronounced fault characteristics is selected. Next, the energy variation characteristics of the signal are analyzed using the derivative of the segmented TKEO. Finally, the ratio of the maximum derivative value of each TKEO segment to that of the previous segment is calculated, and a suitable threshold is set for fault detection. Simulation results demonstrate that this method can accurately detect cable conditions, enabling early fault detection and ensuring both sensitivity and reliability in detection.