Abstract: To enhance the robustness and accuracy of fault diagnosis in distribution networks, this paper proposes an intelligent diagnosis method that integrates voltage phase-angle features with artificial neural networks (ANN). The voltage phase angle is selected as the key feature for fault identification. An IEEE 33-node system and a 58-node distribution network model in an actual area are established. Different types of fault simulations are carried out, and the node phase angle data under various fault types, including single-phase-to-ground, two-phase-to-ground, and three-phase symmetrical short-circuit faults, are obtained. An ANN-based classification model optimized by an adaptive weighted loss function is constructed to improve diagnostic performance under imbalanced fault samples. The test results show that the proposed method exhibits advantages in fast response, strong generalization capability, and robust noise immunity. It can serve as a valuable reference for fault diagnosis in complex distribution networks with a high penetration of distributed energy resources and demonstrates promising engineering application potential.