Abstract: Aiming at the problem of low troubleshooting efficiency of communication link faults caused by complex links in smart substation, a deep learning based fault location method for intelligent substation communication link of smart substation is proposed. Firstly, based on the network topology of secondary devices in smart substations, a network connectivity matrix is constructed and used as the benchmark, and a fault feature characterization method is proposed for communication link faults. And then, based on the logical relationship between secondary device connection and operation status, a fault sample set of the whole station is constructed. The improved CNN is applied to build the fault location model of the smart substation communication link. Finally, the fault link is accurately located through the fault bay information preliminarily determined by the background information and the model output results. A 220 kV smart substation is taken for case study and some bays of it are taken to construct the fault sample set, and different fault location methods are compared through result analysis. The comparison results show that the proposed location method has higher accuracy.