In flexible DC distribution networks with low-current grounding mode, single-pole-to-ground faults exhibit weak fault characteristics, making fault identification challenging. To enhance fault detection capability, this paper proposes a single-pole fault line selection scheme for flexible DC distribution networks based on K-means++ multi-criterion fusion. First, modulus decomposition of electrical quantities during line faults is carried out, and the characteristics of zero-mode current, zero-mode voltage and zero-mode power for both normal and faulty lines are obtained. Then, according to the differences in the Spearman correlation of zero-mode current, the cosine of the angle between zero-mode current and voltage, and the absolute value of the zero-mode power integral, three fault line selection criteria are constructed. Finally, the K-means++ algorithm is used to analyze the results of the three criteria for each line, enabling accurate identification of the fault line. Simulations in PSCAD show that the proposed method has strong resistance to transition resistance, noise interference, and communication delay, significantly improving the accuracy and reliability of single-pole-to-ground fault line selection.