To improve the stability and response speed of ground fault location devices in 10 kV distribution networks, it is necessary to investigate more effective fault location methods and design corresponding hardware devices. First, the transient zero-sequence current characteristics of single-phase and two-phase ground faults are analyzed for different neutral point grounding modes (i.e., ungrounded and grounded through an arc suppression coil). Subsequently, the superior performance of fuzzy entropy in identifying changes in zero-sequence current is verified through theoretical derivation and Simulink simulations, demonstrating its strong noise immunity due to a continuous fuzzy membership function, low parameter sensitivity, and strong capability in analyzing nonlinear dynamic characteristics. Then, based on the calculated entropy ratio between adjacent detection points, combined with the principle of fuzzy entropy fault section location and an internal non-fault threshold (0.75~1.2), efficient and stable identification of the fault section is achieved. Finally, a device centered on the STM32F103C8T6 development board is developed, and two operating conditions (ungrounded neutral point and neutral point grounded through an arc suppression coil) are set up on a 10 kV distribution network simulation platform. Experimental results show that the device can accurately identify single-phase and two-phase ground faults in the first section, the last section, and between detection points. The device exhibits fast operation, high stability, strong noise immunity, and high location accuracy with minimal influence from communication transmission errors.