With the growing integration of renewable energy sources, the penetration of the distributed generation leads to increasingly dynamic power distribution system topologies. This poses challenges for traditional fault diagnostic methods in accurately classifying and locating faults. This paper develops a deep graph neural network (GNN) for detecting and managing fault events on distribution systems. The 1-D convolutional graph attention networks (1-D GAT) can exploit the spatial-temporal features from both voltages and branch currents to enhance the accuracy of fault diagnostics compared to existing methods. The effectiveness of the proposed method is evaluated on the Potsdam 13-bus and IEEE 123-bus test feeder systems. Results indicate notable improvements in accuracy and other metrics, achieving a 1-2% increase in fault event detection, an average 4% improvement in identifying fault types, and an average 5% improvement in fault location.