Abstract: The traditional acoustic finite element method (FEM) is difficult to accurately characterize the change of air medium parameters in the large space of the main transformer room caused by the temperature rise effect, resulting in the large calculation error of the noise field in the main transformer room of the substation under the coupling effect of temperature field and sound field. Based on the acoustic FEM algorithm, computational fluid dynamics (CFD) is innovatively introduced to extract the complex spatial medium parameters of the large space main transformer room, and the wave integral equation is improved, and a noise field solution algorithm in the main transformer room based on the improved acoustic FEM is proposed. Firstly, the rheological model of the main transformer chamber under the influence of temperature field is established, and the discrete space medium parameters of the large space temperature field of the main transformer chamber are characterized by CFD; Then, based on the fluid-acoustic grid mapping theory, the discrete space medium parameters of the temperature field are mapped to the acoustic grid, and the acoustic FEM integral equation is established after correcting the air medium parameters in large space; Finally, based on the conventional Gauss numerical integration method and the introduction of Kirchhoff-Helmholtz equation, the modified acoustic FEM integral equation is jointly solved. The algorithm has been successfully applied to the solution and analysis of the noise field in the No.1 main transformer room of Xi'an 110 kV Changming Substation, with an error of 2.168% from the measured value.