Abstract: The 10 kV distribution line plays an increasingly crucial role in the power grid. However, single-phase ground faults on these lines are often difficult to quickly and accurately locate. To address the issues of slow response and low positioning accuracy associated with traditional methods, this paper presents an in-depth study on AI-based automatic fault location technology for single-phase ground faults in 10 kV distribution lines. An AI-based fault location technique is proposed, which constructs a time-frequency coupling feature measurement model to extract both time-domain abrupt changes and frequency-domain spectral differences of fault signals under various working conditions. The model combines AI regression with a zero-sequence current ratio method, providing a weighted least squares optimization fusion algorithm. Experimental verification through pilot tests demonstrates the high precision and low-latency fault location performance of the proposed method in complex scenarios, offering a technical reference for intelligent operation and maintenance of distribution networks.