Abstract: The simulation based on finite element method and other numerical calculation methods has slow calculation speed, which is difficult to meet the real-time requirements of equipment monitoring and analysis in digital twin. Therefore, this paper takes the three-phase compact GIL as the research object and proposes a fast calculation method for temperature field distribution based on data reduction and surrogate model. The input dataset is obtained through electromagnetic-thermal-flow multi physical field coupling simulation. The grid node data of the simulation results are dimensionally reduced and truncated using the EOF method, while a surrogate model is applied to construct a mapping relationship between boundary conditions and truncated data. The impact of the key parameters on the error of fast calculation results is also analyzed. The results indicate that using the data reduction and reconstruction method proposed in this paper can significantly reduce the input dataset for the surrogate model with low truncation errors. The error of the fast calculation method is related to the surrogate model used and the size of the input data. When selecting an adaptive method and parameters, the maximum absolute error can be less than 0.5 ℃, and the simulation time can be shortened to within seconds. The research results of this paper can provide reference and ideas for the fast calculation of multi-physical fields in digital power equipment.