To address the limitations of conventional fault diagnosis and evaluation techniques for converter transformers under conditions of unclear fault mechanisms and insufficient fault data, this paper proposes an interpretable fault diagnosis method for converter transformers based on knowledge reasoning. Firstly, grounded theory is employed to establish a fault characteristic framework, thereby recon-structing textual information from historical fault cases into structured data. Subsequently, grey relational analysis is utilized to optimize the selection of key fault characteristics, and a causal feature separation-based Apriori algorithm is applied to extract fault diagnosis knowledge rules. A fault diagnosis knowledge graph is constructed to enhance the interpretability of the fault diagnosis mechanism and process. The effectiveness of the proposed method is validated using actual converter transformer fault data. The results demonstrate that the method achieves a diagnostic accuracy of 78.26%, providing accurate, reliable, and interpretable decision support for converter transformer fault diagnosis and condition assessment. This approach holds significant practical value for guiding field testing and maintenance of converter transformers.