Thermal optimization research of oil-immersed transformer winding based on the support vector machine response surface

In this paper, the CFD model is established for the low voltage winding region of an oil-immersed transformer according to the design parameters, and the detailed temperature distribution within the region is obtained by numerical simulation. On this basis, the response surface methodology is adopted to optimize the structure parameters with the purpose of minimizing the hot spot temperature. After a sequence of designed experiments, the second-order polynomial response surface and the support vector machine response surface are established, respectively. The analysis of their errors shows that the support vector machine response surface can be better used to fit the approximation. Finally, the particle swarm optimization algorithm is employed to get the optimal structure parameters of the winding based on the support vector machine response surface. The results show that the optimization method can significantly reduce the hot spot temperature of the winding, which provides a guiding direction for the optimal design of the winding structure of transformers.