Abstract: This paper proposes a hybrid variable step method to address the low efficiency of using fixed time steps in the current transient temperature rise simulation of oil immersed power transformer windings. First, an initial solution optimization algorithm is adopted to effectively reduce the number of iterations of the equation during the calculation process. Secondly, combined with the proper orthogonal decomposition (POD) method, the problems of high order and condition number of large-scale finite element equations are improved, and the solving efficiency and numerical stability of equations are also improved. Thirdly, an adaptive time stepping (ATS) - heuristic time stepping (HTS) hybrid variable step size mehod is proposed, which effectively solves the contradiction between computational efficiency and accuracy in transient calculations through adaptive and heuristic adjustments to the time step size. Finally, this article establishes a two-dimensional single zone and turn separation of oil immersed power transformer windings to verify the correctness and efficiency of the proposed mehod. The numerical calculation results show that in the flow field, compared with the fixed step size calculation results, the error of the hybrid variable step size method is less than 0.46%, and the computational efficiency is improved by 18.45 times. In the temperature field, compared with the calculation results with a fixed step size, the error of the method proposed in this paper is less than 0.04%, and the computational efficiency is improved by 6 times. Meanwhile, by comparing the calculation results with traditional variable step size method, it is demonstrated that the proposed hybrid variable step size method has certain advantages in terms of calculation accuracy, efficiency, and variable step size effect. This article also explores the impact of different parameter settings on the transient calculation results and state change process in hybrid variable step calculation, laying a certain foundation for its engineering application.