Abstract: The rapid location and self-healing of distribution network faults are the core propositions to ensure reliable power supply. The large-scale integration of distributed power sources has made the traditional centralized fault handling mode increasingly inadequate in terms of latency and perception accuracy. The article introduces edge intelligence into the fault handling system and deploys edge nodes with local inference capabilities on the distribution side. Then, an improved graph convolutional neural network is used to construct a fault localization model, combined with a distributed self-healing control strategy, aiming to improve the accuracy of fault section identification and self-healing response speed, and provide a feasible path for intelligent operation and maintenance of distribution networks.