To address the problem that multi‑source heterogeneous data is difficult to be used uniformly and effectively in fault diagnosis of secondary equipment in intelligent substations

a fault diagnosis method integrating a knowledge graph and a graph attention network (GAT) is proposed. First

secondary equipment

their connection relationship

equipment inventory information

and historical equipment defect records are stored in the static knowledge graph

while real‑time alarm information and self‑diagnostic information generated during faults are incorporated into the dynamic knowledge graph through continuous updates. Second

a deep learning model based on GAT is constructed

and the knowledge graph is used as the input. A multi‑head attention mechanism is introduced to adaptively aggregate the node features of the graph and output the fault diagnosis results. Finally

the fault diagnosis results are returned to the knowledge graph for auxiliary decision‑making. Case studies demonstrate that the proposed dynamic‑static integrated knowledge graph representation effectively establishes association among multi‑source heterogeneous data of secondary equipment. The GAT model enables adaptive feature aggregation of multi‑source heterogeneous data in graph representation and achieves higher fault diagnosis accuracy than other deep learning models. Moreover

it exhibits good robustness in noisy environments involving false alarms and missed alarms. By integrating the knowledge graph with deep learning

a complete closed loop from fault perception to auxiliary decision‑making is realized.