Aiming at the potential faults in the smart microgrid that cannot be detected by traditional methods, and the systemic instability of the microgrid caused by changes in the configuration and structure of the microgrid. This paper proposes a fast fault location method based on deep learning, which uses multi-point and multi-state monitoring data at the network end to train a convolutional neural network to obtain an AI network with high precision and fast fault location. On this basis, the network is reconstructed under the fault state of the system, and the power distribution of multiple micro sources is re-coordinated to achieve economic optimization. In this paper, an intelligent microgrid simulation system is constructed and the proposed algorithm is verified. The simulation results show that the proposed method can quickly detect faults and realize economically optimized operation.