The current power grid is undergoing digital transformation and upgrading, and the intelligent health management technology of power transformers is rapidly advancing. However, there are issues of weak information correlation and low decision-making efficiency in the operation and maintenance process and there are few papers on knowledge graph construction specifically related to power transformer maintenance. Additionally, there is limited public data available specifically for power transformer operation and maintenance, making it difficult to effectively construct maintenance knowledge. This paper proposes a method for constructing a knowledge graph for power transformer operation and maintenance based on Roberta-GPliner. Firstly, public literature in the field of power transformers is obtained to enhance the training dataset of power transformer operation and maintenance. Then, we use Roberta as the embedding layer and employ the GPliner joint extraction model to extract knowledge triplets on power transformer operation and maintenance. Roberta-GPliner is compared with other pre-training models, validating that the joint knowledge extraction algorithm based on Roberta-GPliner performs better. Finally, an intelligent operation and maintenance platform is built for power transformers, enabling knowledge retrieval and decision-making support.