<p>This paper targets fault tracing of transformer substation devices. Current automated methods have the potential to save vast amount of manpower. However, there are two challenges they face: inadequate representation of the connection topology of substation devices; and the lack of interpretability of the fault tracing process, both of which are highly valued in industrial practices. Thus, this paper constructs a framework to address the problem of fault tracing by preforming graph kernel computations on specially designed event graphs. The constructed event graphs represent the connection topology of substation devices and the alarm signals sent by monitors within a possible fault event. In addition, an Edge-Labeled Weisfeiler-Lehman (EL-WL) graph kernel is proposed, integrating edge information into the tree distance calculation of the graph kernel. In addition, a variant of the unique class association rule mining algorithm performed on subtree histogram vectors is used to mine interpretable fault tracing rules. Simulation results show noticeable performance improvement on real-world transformer substation data.</p>