To improve the accuracy and portability of secondary circuit fault location in smart substations, a graph neural network-based smart substation secondary circuit fault location method is proposed. From the substation configuration description file, a graph database is established to obtain the connection relationship between the secondary equipment. Combined with the alarm signal representation of the secondary equipment, a kind of information representation of the secondary equipment node is proposed as the input of the graph neural network. Graph neural network theory is used to establish the fault location model. Taking the secondary circuits of a 220 kV smart substation as a benchmark, using the fault generation model to change the networking mode, a subscription relationship and network configuration is used to generate different secondary circuits faults. The experiment compares the accuracy of the proposed fault location model with other models and the experimental results show that this method has higher fault location accuracy and better robustness. This work is supported by the Science and Technology Project of State Grid Corporation of China (No. 5108-202018037A-0-0-00).