Abstract: Transformer operation and maintenance data comes from a wide range of sources, and the diversity, heterogeneity, and complexity of the data make its fusion difficult. To this end, design a multi-source data fusion method for the operation and maintenance of high-voltage transmission and transformation transformers. Based on the Dempster Shafer(DS) evidence theory, the principle of multi-source data fusion is introduced. Introducing deep learning networks(DLN), the input of the DLN fusion model includes multi-source data such as oil color spectrum indicators, oil chemical test indicators, and electrical test indicators in the operation and maintenance of high-voltage transmission and transformation transformers. Automatically extract key features from multi-source data through DLN fusion model. Based on this, combined with the DLN model and DS evidence theory, the multi-source data fusion of high-voltage transmission and transformation transformer operation and maintenance is completed. Through case application analysis, the feasibility and effectiveness of the multi-source data fusion method proposed in this paper in practical operation and maintenance have been verified, providing strong support for the intelligent operation and maintenance of high-voltage transmission and transformation transformers.