Abstract: Objective Addressing the limitation issue of traditional traction transformer relying on single-sensor data, which hinders comprehensive state perception, an enhanced perception data-based health assessment method is proposed. It is aimed to achieve holistic and precise condition evaluation through multi-source heterogeneous data fusion. Method Multi-modal sensing layer and data fusion architecture are designed to integrate electrical parameters, non-electrical parameters, and other multi-source data. Combining the AHP（analytic hierarchy process） and G1 weighting methods, indicator weights and quantification approaches are determined to construct a health assessment model. Using on-site application of metro transformers as object, the enhanced perception system is deployed and health assessment is conducted. Result ＆ Conclusion The health assessment method based on enhanced perception data not only reliably determines the overall health status of equipment but also reveals subtle state differentiation among individual units and their potential failure modes. The evaluation results align with actual operational performance and physical degradation mechanisms of the equipment.