Abstract: The digital research of large power equipment is the basis of building a new power system. In order to realize the accurate calculation of transformer winding temperature rise and fall under different steady-state conditions, a method based on discrete orthogonal interpolation is proposed in this paper. First, the least square finite element method and streamline upwind finite element method are used to establish the finite element full order model for the calculation of fluid solid coupling heat transfer in windings. Secondly, based on the basic principle of POD and DEIM algorithm and the characteristics of full order finite element equation, a reduced order model suitable for winding temperature rise calculation is established. Finally, a two-dimensional transformer winding model is constructed, and the model is used to verify and analyze the accuracy, calculation efficiency and robustness of the reduced order algorithm in steady-state heat transfer calculation. The experimental results show that the calculation method of steady-state temperature rise and fall of transformer winding based on POD-DEIM proposed in this paper has quite high calculation accuracy, and the calculation efficiency is about 40 times higher than that of the full-order calculation model. At the same time, for the working conditions outside the sample space, the reduced order model still has high accuracy, and the average relative error of flow field and temperature field is less than 5%, indicating that the reduced order model has excellent robustness, and the algorithm can provide some development ideas for the digital construction of large-scale power equipment.