To address the difficulties in extracting fault features under high-impedance faults and the low reliability of internal/external fault identification and fault pole selection in flexible DC distribution networks

a fault protection method based on gradient entropy (GE) and reverse dispersion entropy (RDE) criteria is proposed by leveraging the boundary characteristics of current-limiting reactors. First

variational mode decomposition (VMD) combined with kurtosis weighting is employed to extract and reconstruct fault signals

with the impulse factor adopted as the starting criterion. Second

the one-dimensional power signals at each measurement point are transformed into two-dimensional matrices

from which their GE values are calculated. Internal and external faults are then distinguished by comparing the differences in GE values among different points. Furthermore

the RDE values of each component are computed

and a fault pole-selection criterion is established based on the ratio of the RDE values between the positive and negative poles. Finally

a simulation model of a two-terminal flexible DC distribution network is constructed

and fault simulations are performed under different transition resistances and fault locations. The results demonstrate that the proposed method can effectively distinguish fault types even when the transition resistance reaches 1000 Ω

demonstrating its feasibility and anti-interference performance.