To address the problems of low calculation efficiency of transient temperature rise in oil-immersed power transformer windings and significant computational error of existing rapid calculation methods

a reduced-order fast prediction model based on radial basis function-extended dynamic mode decomposition (RBF-EDMD) is proposed in this paper. The model uses the historical simulation data of transformers to obtain the dynamic mode and mode time coefficient to predict the transient temperature field. The adaptive scale selection strategy for the snapshot matrix is introduced to reduce the data cost in model construction. At the same time

in order to improve the calculation accuracy

the radial basis function is used as the dictionary function to upgrade the data dimension

which mitigates the problem that the computational error of dynamic mode decomposition (DMD) increases significantly in long-term highly nonlinear prediction. By selecting the main modes

the full-field temperature is calculated quickly. Finally

an eight-zone winding model and a temperature rise test platform are established based on the fundamental winding structure of the product-grade 110 kV oil-immersed power transformer to validate the proposed model. The results show that the computational accuracy of the proposed model is significantly improved compared to that of DMD

and is within the acceptable range in engineering. For the temperature rise experiment

the average computational time per time step is only 0.018 s. The conclusion of the study proves that the proposed model can ensure the high fidelity of the transient temperature rise’s reduced-order calculation and the real-time application of the digital transformation of electric power equipment.