There are many alarm signal texts in the backlog of the monitoring terminal of a substation. Retrieving the hidden fault knowledge in them can help monitoring personnel to quickly grasp the fault situation and improve the efficiency of fault handling. Based on knowledge graph technology, this paper first proposes a method of expressing the fault knowledge of a substation alarm signal. It analyzes the existing classification and analysis content of the signal, and then proposes a method for constructing the fault knowledge graph of alarm signal, i.e. constructing a power grid accident behavior graph with accident and displacement signals, and constructing a device abnormal behavior graph with abnormal signals. Then it uses the graphs to excavate and display the behavioral logic, correlation between alarm signals and important information after the failure, so as to provide a convenient way for on-site monitoring personnel to quickly obtain key failure information and improve the efficiency of failure handling. The feasibility of the method is demonstrated with a case of knowledge graph visualization. Finally, the extended application of knowledge graphs in the fields of signal analysis, fault diagnosis, and fault tracking is analyzed, so that the substation alarm signal data can be better tapped and used in the field of power grid fault processing. This work is supported by the National Natural Science Foundation of China (No. 51877123).