Abstract: Accurately identifying the location and cause following the occurrence of multiple faults in the power grid is crucial for effective fault management and rapid restoration of power supply. To address this issue, a smart grid fault diagnosis method using knowledge graphs (KGs) as the knowledge representation and abductive reasoning as the reasoning mechanism is proposed. First, fault diagnosis KGs containing primary system topology and secondary protection logic are established, and an automatic construction method for KGs is proposed based on the formal model of KGs, CIM/E files, and SCD files. Subsequently, an abductive reasoning framework is proposed for power grid fault diagnosis, consisting of backward reasoning, forward reasoning, and iterative evidence fusion stages. This framework is capable of effectively handling time constraints and multi-source conflicting evidence while preserving the relative independence of knowledge and reasoning. Finally, the effectiveness of the methodology proposed in this paper is verified by two actual historical fault cases from the actual power grid when single and multiple faults occur in the grid.