Underground Stratigraphic Surface Reconstruction Under Uncertain Conditions

Point cloud surface reconstruction is the construction of a corresponding mathemati-cal model by using geometric information of the 3D scatters for the purpose of describing and drawing spatial surfaces.It is used in many fields such as product design,heritage con-servation,and oil & gas exploration.Underground stratigraphic surface reconstruction,the foundation and support for the establishment of underground stratigraphic sequence models and reservoir models,construct stratigraphic surfaces such as horizons and faults on basis of the stratigraphic scatter information obtained from seismic surveys.Involved with a large number of information processing techniques,geological rules and regional geological knowledge,underground stratigraphic surface reconstruction is one of the fron-tier issues of interest to both the information processing field and the oil & gas exploration field.Point cloud data for underground stratigraphic surface reconstruction is characterized by inhomogeneity,sparseness and low accuracy.These characteristics lead to uncertain-ties in the spatial location,intersected relationships and morphological characteristics of stratigraphic surfaces.They also lead to deficiencies of the traditional point cloud recon-struction method in three aspects: 1.effectively characterize the uncertainty of the spa-tial position of the underground stratigraphic surface? 2.maintain the correctness of the intersected relationship between stratigraphic surfaces? 3.accurately portrait the morpho-logical characteristics of stratigraphic surfaces.This dissertation provides a new technical solution for underground stratigraphic surface reconstruction under uncertain conditions by carrying out innovative research on the above deficiencies,promotes the development of research in related fields,and has certain theoretical value and practical significance.The specific research results are as follows:(1)In response to the uncertainty characterization of the spatial location of under-ground stratigraphic surface,an uncertainty characterization method based on support vector interval regression is proposed in this dissertation.The method abstracts the uncer-tainty characterization of the stratigraphic surface into an optimization model by taking the stratigraphic sedimentation rule,the spatial relationship between the stratigraphic sur-face and the well paths,and the sampling value confidence coefficient of the sample point as constraint conditions.Then twin support vector regression is introduced to decompose the uncertainty characterization problem into two smaller sub-problems,which improves the time efficiency of uncertainty characterization.Studies have shown that the proposed method can effectively characterize the spatial location uncertainty of the stratigraphic surface.The generated envelopes can provide spatial position constraints for subsequent stratigraphic surface reconstruction.Besides,by integrating well data,this method can partially reduce the uncertainty of the spatial position of the stratigraphic surface.(2)To solve the problem that makes it difficult for the traditional point cloud re-construction method to correctly describe the intersected relationship between the strati-graphic surfaces,a stratigraphic surface reconstruction method constrained by the inter-sected relationship is proposed in this dissertation.By using Bayesian inference tech-nology in combination with surface inclination,geological rules and regional geological knowledge,this method achieves automatic and efficient determination of the intersec-tion of the stratigraphic surface.Then,the stratigraphic surface reconstruction model constrained by the intersected relationship is established with the envelope as the spa-tial position constraint and the intersection point as the high-confidence sample point,achieving automatic stitching of the intersecting stratigraphic surfaces.Studies show that this method avoids the additional adjustment operations of the traditional method and improves the efficiency of the reconstruction of the stratigraphic surface.Besides,this method effectively guarantees the correctness of the intersected relationship between the stratigraphic surfaces,thereby reducing the uncertainty of the intersected relationship be-tween the stratigraphic surfaces.(3)It is difficult for the traditional stratigraphic surface reconstruction method to maintain the geometric features of the surface.To solve this problem,an underground stratigraphic surface reconstruction method constrained by the surface morphology is pro-posed in this dissertation.This method extracts the ridge(valley)lines that describe the morphological features of the stratigraphic surface by using the ridge(valley)feature infor-mation contained in stratum scatter points,self-paced ensemble classification technology,union-find algorithm and support vector regression.Then,on basis of the spatial posi-tion limitation provided by the envelope,a stratigraphic surface reconstruction model is constructed with the ridge(valley)line as the constraint,and the stratigraphic surface re-construction is achieved while maintaining the morphological features.Studies show that this method can quickly extract ridge(valley)lines,effectively maintain the inherent geo-metric features of the stratigraphic surface,and reduce the uncertainty of the morphology of the stratigraphic surface.(4)The envelope constraint,the intersection constraint and the ridge(valley)line constraint ensure the authenticity and rationality of the reconstructed stratigraphic sur-face.However,a large number of constraints will lead to high time costs for the model solution and even solution algorithm failure.To solve this problem,this dissertation pro-poses a method for determining the optimal stratigraphic surface based on disturbance.By determining the optimal fault surface from stochastic disturbance and dip differences,and deciding the optimal horizon surface through the use of the uncertain reconstruction model of stratigraphic surface based on chance-constrained programming proposed in this dissertation and feature similarity,this method avoids the problem of difficult model solv-ing caused by a large number of constraints.Studies show that this method can maintain the morphological features of the stratigraphic surface while ensuring its intersected re-lationship correctness,thus laying a solid foundation for the industrial application of the underground stratigraphic surface reconstruction method under uncertain conditions.