A Fast Modeling Method For 3D Porous Media Based On Column Searching Multiple-Point Statistics

With the rapid development of the economy and society,the demand for natural resources,especially oil and gas resources has increased sharply.To efficiently and effectively exploit underground resources,it is necessary to accurately construct geological spatial models to better understand the spatial relationships between geological objects.As an emerging geological modeling method,multiple-point statistics(MPS)uses training images as prior models,which can accurately reproduce the spatial structure of the geological bodies of interest and generate a large and diverse set of geological models.However,large scale,multivariate,and heterogeneous complex modeling scenarios pose challenges to the computational efficiency and modeling accuracy of multiple-point statistics.Creating high-quality geological models quickly in various practical application scenarios has become an urgent problem to be solved.This study proposes a column-based searching simulation(CSSIM)method for efficient and accurate generation of high-quality porous media and geological models for complex underground spatial systems.The method utilizes techniques such as dataset searching,image feature extraction,and machine learning to overcome the limitations in efficiency and accuracy of geological statistical modeling.To save modeling time,this method improves the spatial structure dataset of multiple-point statistics and uses column search technology to quickly find matching spatial structures.To address complex spatial structures,the early stopping strategy is proposed to focus on geometric structures with high correlation and remove redundant operations in the multiple-point statistics searching program.Then,this study focuses on homogeneous geological structures,aiming to understand the spatial evolution rules of underground systems.A method for accurately characterizing pore space evolution is designed to guide the modeling process by identifying key structural features.Finally,the contrastive loss function is introduced into the 3D probability aggregation formula,the current model and target model are compared in real-time,simplifying the setting of multiple-point statistics parameters.In this paper,the proposed methods are applied to model two-dimensional channel,three-dimensional sandstone,multiphase asphalt mixtures,and heterogeneous shale,respectively.To validate the performance of the proposed methods,the several evaluation metrics are used to quantitatively analyze the spatial structure,geometric features,permeability,and electrical properties of the generated models.The experimental results indicate that the proposed method significantly improves the modeling efficiency and generates high-quality porous media models.This provides reliable models for the rational development of natural resources and has important research significance and application value for practical engineering.