CN-121981007-A - Uncertainty constraint-based fluid migration prediction method and device

CN121981007ACN 121981007 ACN121981007 ACN 121981007ACN-121981007-A

Abstract

The application discloses a fluid migration prediction method and device based on uncertainty constraint, relates to the technical field of data processing, and mainly aims to solve the problems of poor accuracy and effectiveness of fluid migration prediction of the existing uncertainty constraint. The method comprises the steps of obtaining a noisy data set of a medium for carrying out underground migration of uranium-containing solution, constructing a convection diffusion physical control model for representing a convection diffusion process of the uranium-containing solution, constructing a joint posterior probability model based on joint posterior distribution of the noisy data set, physical consistency likelihood items of the convection diffusion physical control model and physical residual items of a physical information neural network model, solving the joint posterior probability model based on variation gradient descent to obtain a particle set of posterior distribution, and predicting the noisy data set according to the particle set and the physical information neural network model to obtain a fluid migration prediction result.

Inventors

ZHANG ZHEAN
LIU LONGCHENG
JIANG GUOPING
BAI YUNLONG
WANG YAAN
CHEN XI
XU LUYAN

Assignees

核工业北京化工冶金研究院

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (10)

1. A method of fluid migration prediction based on uncertainty constraints, comprising: Acquiring a noisy data set of a medium for the uranium-containing solution to migrate underground, and constructing a convection diffusion physical control model for representing a convection diffusion process of the uranium-containing solution, wherein the noisy data set comprises stratum physical parameters with noise data and uncertainty data and fluid velocity field data; constructing a joint posterior probability model based on the joint posterior distribution of the noisy data set, the physical consistency likelihood term of the convection diffusion physical control model and the physical residual term of the physical information neural network model, wherein the joint posterior probability model is used for describing the model parameter uncertainty distribution of the physical information neural network model; And carrying out inference solving on the combined posterior probability model to obtain a particle set with posterior distribution, and predicting the noisy data set according to the particle set and the physical information neural network model to obtain a fluid migration prediction result, wherein the fluid migration prediction result comprises uranium-containing solution concentration data with different time and space positions and corresponding probability distribution.
2. The method of claim 1, wherein prior to constructing the joint posterior probability model based on the joint posterior distribution of the noisy dataset, the physical consistency likelihood terms of the convective diffusion physical control model, and the physical residual terms of the physical information neural network model, the method further comprises: Constructing a joint posterior distribution of the noisy data set containing model parameter prior distribution, physical parameter prior distribution and data likelihood items based on Bayes; Determining a physical consistency likelihood term of the convection diffusion physical control model based on the residual error of the convection diffusion physical control model and Gaussian noise; and constructing a physical residual term based on the output differential term of the physical information neural network model.
3. The method of claim 2, wherein said performing an inference solution to the joint posterior probability model to obtain a set of particles of posterior distribution comprises: Carrying out iterative evolution on parameter particles of the physical information neural network model through the joint posterior probability model, and carrying out iterative solution based on variation gradient descent in the iterative evolution process, wherein each pair of parameter particles corresponds to a group of model parameters and physical parameters; and when the parameter particles are matched with a preset convergence condition in the iterative solving process, determining a particle set of posterior distribution.
4. The method of claim 1, wherein predicting the noisy data set from the collection of particles, the physical information neural network model, the obtaining a fluid migration prediction result comprises: in the process of predicting the noisy data set by utilizing the physical information neural network model, forward predicting the physical information neural network model based on the particle set to obtain a plurality of groups of prediction results; Carrying out mean statistics on the multiple groups of predicted results to obtain deterministic predicted results, and carrying out discrete distribution statistics on the multiple groups of predicted results to obtain the credibility of the deterministic predicted results; generating the fluid migration prediction result based on the deterministic prediction result and the credibility.
5. The method according to claim 1, wherein the method further comprises: Constructing a physical information neural network model, wherein the physical information neural network model comprises an input layer, an output layer and at least one hidden layer; And calculating differential derivatives of the output parameters of the output layer and the input parameters of the input layer, and constructing a physical residual error item of the physical information neural network model based on the differential derivatives and the convection diffusion physical control model.
6. The method of claim 1, wherein the acquiring the noisy dataset of the medium in which the uranium-containing solution is traveling in the subsurface comprises: determining the noise data and the uncertainty data based on random variables matching a preset statistical distribution; The noise data and the uncertainty data are configured in random noise, physical parameter measurement errors, and disturbance parameters in the fluid velocity field data to form the noisy dataset.
7. The method of claim 1, wherein the constructing a convective diffusion physical control model characterizing the uranium containing solution convective diffusion process comprises: Defining initial physical conditions and boundary conditions; Based on the time derivative term, the convective term and the diffusion or dispersive term, a convective diffusion physical control model under the initial physical condition and the boundary condition is constructed.
8. A fluid migration prediction apparatus based on uncertainty constraints, comprising: The acquisition module is used for acquiring a noisy data set of a medium for the uranium-containing solution to migrate underground, and constructing a convection diffusion physical control model for representing a convection diffusion process of the uranium-containing solution, wherein the noisy data set comprises stratum physical parameters with noise data and uncertainty data and fluid velocity field data; The construction module is used for constructing a joint posterior probability model based on the joint posterior distribution of the noisy data set, the physical consistency likelihood item of the convection diffusion physical control model and the physical residual error item of the physical information neural network model, and the joint posterior probability model is used for describing the model parameter uncertainty distribution of the physical information neural network model; The prediction module is used for carrying out inference solving on the combined posterior probability model to obtain a particle set with posterior distribution, and predicting the noisy data set according to the particle set and the physical information neural network model to obtain a fluid migration prediction result, wherein the fluid migration prediction result comprises uranium-containing solution concentration data with different time and space positions and corresponding probability distribution.
9. A computer readable storage medium having stored thereon a computer program/instruction, which when executed by a processor, implements the steps of the method according to any of claims 1-7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to carry out the steps of the method according to any one of claims 1-7.

Description

Uncertainty constraint-based fluid migration prediction method and device Technical Field The present application relates to the field of data processing technologies, and in particular, to a fluid migration prediction method and apparatus based on uncertainty constraint. Background Porous media flow is a phenomenon of fluid seepage in tiny pores, which is widely found in the natural and engineering fields, especially in uranium-containing solution transport processes in the nuclear industry field. In order to predict and grasp the flow condition of the oil-containing solution, a simulation mode is generally adopted to predict the migration process of the uranium-containing solution in the porous medium in an unknown environment. Currently, existing fluid prediction generally adopts numerical simulation, such as solving a flow-diffusion equation based on a finite difference or finite element numerical method, and at this time, it is generally assumed that parameters such as a velocity field, a dispersion coefficient and the like are determined values to fulfill the purpose of prediction. However, in practical engineering, the parameters often originate from limited monitoring data or inversion results, measurement errors and model uncertainty exist, and when the uncertainty parameters are directly substituted into a discrete numerical model, errors are accumulated continuously in the time advancing process, so that the position deviation, the diffusion intensity distortion, and even the numerical oscillation or the instability phenomenon of the front edge of the contaminated plume of the uranium-containing solution are easily caused, and therefore, a fluid migration prediction method based on uncertainty constraint is needed to solve the problems. Disclosure of Invention In view of the above, the application provides a fluid migration prediction method and device based on uncertainty constraint, and a method and device thereof, which mainly aim to solve the problems of poor fluid migration prediction accuracy and poor effectiveness of the existing uncertainty constraint. According to one aspect of the present application, there is provided a fluid migration prediction method based on uncertainty constraint, comprising: acquiring a medium noisy data set of underground migration of uranium-containing solution, and constructing a convection diffusion physical control model representing a convection diffusion process of the uranium-containing solution, wherein the noisy data set comprises stratum physical parameters with noise data and uncertainty data and fluid velocity field data; constructing a joint posterior probability model based on the joint posterior distribution of the noisy data set, the physical consistency likelihood term of the convection diffusion physical control model and the physical residual term of the physical information neural network model, wherein the joint posterior probability model is used for describing the model parameter uncertainty distribution of the physical information neural network model; And carrying out inference solving on the combined posterior probability model to obtain a particle set with posterior distribution, and predicting the noisy data set according to the particle set and the physical information neural network model to obtain a fluid migration prediction result, wherein the fluid migration prediction result comprises uranium-containing solution concentration data with different time and space positions and corresponding probability distribution. Further, before the constructing the joint posterior probability model based on the joint posterior distribution of the noisy dataset, the physical consistency likelihood term of the convective diffusion physical control model, and the physical residual term of the physical information neural network model, the method further includes: Constructing a joint posterior distribution of the noisy data set containing model parameter prior distribution, physical parameter prior distribution and data likelihood items based on Bayes; Determining a physical consistency likelihood term of the convection diffusion physical control model based on the residual error of the convection diffusion physical control model and Gaussian noise; and constructing a physical residual term based on the output differential term of the physical information neural network model. Further, the performing inference solving on the joint posterior probability model to obtain a particle set of posterior distribution includes: Carrying out iterative evolution on parameter particles of the physical information neural network model through the joint posterior probability model, and carrying out iterative solution based on variation gradient descent in the iterative evolution process, wherein each pair of parameter particles corresponds to a group of model parameters and physical parameters; and when the parameter particles are matched with a preset converge