Turbulence simulation efficient partial differential equation solving device based on diffusion model

The invention provides a turbulence simulation efficient partial differential equation solving device based on a diffusion model. Comprising a data noise adding module, a data noise reduction module and a physical embedding module, the data noise adding module is used for uniformly sampling from a dynamically changing turbulent flow physical field, taking physical field information at different moments as input data, carrying out data preprocessing, designing a forward diffusion process of a diffusion model, and creating a series of diffusion samples containing noise data; the diffusion sample is sent to a data noise reduction module, and the data noise reduction module receives physical prior information provided by a physical embedding module and embeds the physical prior information into a diffusion model to ensure that a solution of a generated physical field conforms to actual constraints of a physical system; on the basis, Gaussian noise estimation is carried out through a reverse stochastic differentia