Abstract: Accurately calculating the temperature distribution of oil-immersed transformers and seeking optimization methods are of great significance for equipment operation and maintenance. Firstly, the three-dimensional equivalent model of 35 kV oil-immersed transformer is established. The hysteresis characteristic of transformer core material is simulated by the Jiles-Atherton model, and the loss density distribution of core and windings is obtained by magnetic field calculation. On this basis, the loss density is taken as the heat source, and the temperature field and flow field simulation results of the transformer are obtained by fluid-thermal coupling calculation. According to the calculation results, the temperature distribution in the core and windings and the oil flow velocity distribution in the cross section are extracted, and the maximum temperature and velocity curves of heat sink are drawn. Finally, combined with the central composite design and simulation analysis method, the hot spot temperature of transformer under different radiator structure parameters is obtained, the support vector machine model between hot spot temperature and structure parameters is constructed, and the optimal parameters are obtained by particle swarm optimization. The hot spot temperature of the transformer is 15.12 K lower than that before optimization, and the relative error between simulation and model prediction results is 6.42%, which verifies the effectiveness of the optimization model. The results show that the optimization method can be adopted to significantly reduce the hot spot temperature of the transformer, which provides a new idea and reference for the optimization design of the transformer.