A New Differentiable Parameterization Based on Principal Component Analysis for the Low-Dimensional Representation of Complex Geological Models

A new approach based on principal component analysis (PCA) for the representation of complex geological models in terms of a small number of parameters is presented. The basis matrix required by the method is constructed from a set of prior geological realizations generated using a geostatistical algorithm. Unlike standard PCA-based methods, in which the high-dimensional model is constructed from a (small) set of parameters by simply performing a multiplication using the basis matrix, in this method the mapping is formulated as an optimization problem. This enables the inclusion of bound constraints and regularization, which are shown to be useful for capturing highly connected geological features and binary/bimodal (rather than Gaussian) property distributions. The approach, referred to as optimization-based PCA (O-PCA), is applied here mainly for binary-facies systems, in which case the requisite optimization problem is separable and convex. The analytical solution of the optimization problem, as well as the derivative of the model with respect to the parameters, is obtained analytically. It is shown that the O-PCA mapping can also be viewed as a post-processing of the standard PCA model. The O-PCA procedure is applied both to generate new (random) realizations and for gradient-based history matching. For the latter, two- and three-dimensional systems, involving channelized and deltaic-fan geological models, are considered. The O-PCA method is shown to perform very well for these history matching problems, and to provide models that capture the key sand–sand and sand–shale connectivities evident in the true model. Finally, the approach is extended to generate bimodal systems in which the properties of both facies are characterized by Gaussian distributions. MATLAB code with the O-PCA implementation, and examples demonstrating its use are provided online as Supplementary Materials.     

Kernel Principal Component Analysis for Efficient,Differentiable Parameterization of Multipoint Geostatistics

This paper describes a novel approach for creating an efficient, general, and differentiable parameterization of large-scale non-Gaussian, non-stationary random fields (represented by multipoint geostatistics) that is capable of reproducing complex geological structures such as channels. Such parameterizations are appropriate for use with gradient-based algorithms applied to, for example, history-matching or uncertainty propagation. It is known that the standard Karhunen–Loeve (K–L) expansion, also called linear principal component analysis or PCA, can be used as a differentiable parameterization of input random fields defining the geological model. The standard K–L model is, however, limited in two respects. It requires an eigen-decomposition of the covariance matrix of the random field, which is prohibitively expensive for large models. In addition, it preserves only the two-point statistics of a random field, which is insufficient for reproducing complex structures. In this work, kernel PCA is applied to address the limitations associated with the standard K–L expansion. Although widely used in machine learning applications, it does not appear to have found any application for geological model parameterization. With kernel PCA, an eigen-decomposition of a small matrix called the kernel matrix is performed instead of the full covariance matrix. The method is much more efficient than the standard K–L procedure. Through use of higher order polynomial kernels, which implicitly define a high-dimensionality feature space, kernel PCA further enables the preservation of high-order statistics of the random field, instead of just two-point statistics as in the K–L method. The kernel PCA eigen-decomposition proceeds using a set of realizations created by geostatistical simulation (honoring two-point or multipoint statistics) rather than the analytical covariance function. We demonstrate that kernel PCA is capable of generating differentiable parameterizations that reproduce the essential features of complex geological structures represented by multipoint geostatistics. The kernel PCA representation is then applied to history match a water flooding problem. This example demonstrates that kernel PCA can be used with gradient-based history matching to provide models that match production history while maintaining multipoint geostatistics consistent with the underlying training image.     

基于深度学习的地质找矿大数据挖掘与集成的挑战

基于深度学习的地质找矿信息挖掘与集成已经成为数学地球科学的前沿领域.深度学习作为一种具有多级非线性变换的层级机器学习算法,在地质找矿大数据挖掘与集成中仍处于探索阶段,还有一系列问题亟需解决.以卷积神经网络为例,探讨了基于深度学习的地质找矿大数据挖掘与集成过程中两大挑战:训练样本不足和深度学习网络模型构建困难,重点分析了...     

基于深度学习的岩体遥感智能解译模型研究——以苇河镇、亚布力镇、绥阳镇地区为例

在东北地区选取试验区,对比多种分类模型,提出一种基于多源多模态数据和多流CNN的岩体分类模型. 其中包括两个子模型:一是基于大尺度邻域和深度卷积神经网络的岩体提取模型;二是基于波段组合和多模态数据的多流CNN融合模型. 研究结果表明,预测结果图整体区域预测分布正确,总体精度评价指标达到84.4%,具有智能化程度高、客观性强的特点,能够为地质工作者提供辅助决策依据. 此外,还采用迁移学习策略对样本数量进行扩容,解决了CNN模型小样本问题.     

Ensemble level upscaling for compositional flow simulation

Uncertainty quantification is typically accomplished by simulating multiple geological realizations, which can be very expensive computationally if the flow process is complicated and the models are highly resolved. Upscaling procedures can be applied to reduce computational demands, though it is essential that the resulting coarse-model predictions correspond to reference fine-scale solutions. In this work, we develop an ensemble level upscaling (EnLU) procedure for compositional systems, which enables the efficient generation of multiple coarse models for use in uncertainty quantification. We apply a newly developed global compositional upscaling method to provide coarse-scale parameters and functions for selected realizations. This global upscaling entails transmissibility and relative permeability upscaling, along with the computation of a-factors to capture component fluxes. Additional features include near-well upscaling for all coarse parameters and functions, and iteration on the a-factors, which is shown to improve accuracy. In the EnLU framework, this global upscaling is applied for only a few selected realizations. For 90 % or more of the realizations, upscaled functions are assigned statistically based on quickly computed flow and permeability attributes. A sequential Gaussian co-simulation procedure is incorporated to provide coarse models that honor the spatial correlation structure of the upscaled properties. The resulting EnLU procedure is applied for multiple realizations of two-dimensional models, for both Gaussian and channelized permeability fields. Results demonstrate that EnLU provides P10, P50, and P90 results for phase and component production rates that are in close agreement with reference fine-scale results. Less accuracy is observed in realization-by-realization comparisons, though the models are still much more accurate than those generated using standard coarsening procedures.     

Deep learning based classification of rock structure of tunnel face

The automated interpretation of rock structure can improve the efficiency,accuracy,and consistency of the geological risk assessment of tunnel face.Because of the high uncertainties in the geological images as a result of different regional rock types,as well as in-situ conditions(e.g.,temperature,humidity,and construction procedure),previous automated methods have limited performance in classification of rock structure of tunnel face during construction.This paper presents a framework for classifying multiple rock structures based on the geological images of tunnel face using convolutional neural networks(CNN),namely Inception-ResNet-V2(IRV2).A prototype recognition system is implemented to classify 5 types of rock structures including mosaic,granular,layered,block,and fragmentation structures.The proposed IRV2 network is trained by over 35,000 out of 42,400 images extracted from over 150 sections of tunnel faces and tested by the remaining 7400 images.Furthermore,different hyperparameters of the CNN model are introduced to optimize the most efficient algorithm parameter.Among all the discussed models,i.e.,ResNet-50,ResNet-101,and Inception-v4,Inception-ResNet-V2 exhibits the best performance in terms of various indicators,such as precision,recall,F-score,and testing time per image.Meanwhile,the model trained by a large database can obtain the object features more comprehensively,leading to higher accuracy.Compared with the original image classification method,the sub-image method is closer to the reality considering both the accuracy and the perspective of error divergence.The experimental results reveal that the proposed method is optimal and efficient for automated classification of rock structure using the geological images of the tunnel face.     

Conditioning Surface-Based Geological Models to Well and Thickness Data

Geostatistical simulation methods aim to represent spatial uncertainty through realizations that reflect a certain geological concept by means of a spatial continuity model. Most common spatial continuity models are either variogram, training image, or Boolean based. In this paper, a more recent spatial model of geological continuity is developed, termed the event, or surface-based model, which is specifically applicable to modeling cases with complex stratigraphy, such as in sedimentary systems. These methods rely on a rule-based stacking of events, which are mathematically represented by two-dimensional thickness variations over the domain, where positive thickness is associated with deposition and negative thickness with erosion. Although it has been demonstrated that the surface-based models accurately represent the geological variation present in complex layered systems, they are more difficult to constrain to hard and soft data as is typically required of practical geostatistical techniques. In this paper, we develop a practical methodology for constraining such models to hard data from wells and thickness data interpreted from geophysics, such as seismic data. Our iterative methodology relies on a decomposition of the parameter optimization problem into smaller, manageable problems that are solved sequentially. We demonstrate this method on a real case study of a turbidite sedimentary basin.     

分块区域三维地质建模方法

区域三维地质建模是区域三维地质调查的关键。在大量反复实践的基础上,提出了分块三维地质建模方法,该方法的主要步骤为：以断裂、岩体边界和不整合等为边界,把复杂的三维地质建模区域分解为内部构造相对简单的建模块或建模地质单元;分别对各建模块进行地质地球物理综合研究,编制深部地质剖面,揭示深部地质结构;采用基于剖面的建模方法按照全区统一的坐标系统构建三维地质模型;在完成全区所有建模块三维地质模型构建之后,把各个建模块的三维地质模型集成在统一的三维空间框架下,形成全区的三维地质模型。该方法具有能够简化三维地质建模过程、易于修改完善模型、易于集成模型等优势,能够克服常规的基于剖面的三维地质建模方法中存在的问题,突破了大规模区域三维地质建模的瓶颈。在本溪-临江深部地质调查中的应用表明,采用该方法可以有效构建研究区的三维地质模型,并能充分表达复杂的地下深部地质结构,为开展复杂地区大规模区域三维地质建模提供了一个重要途径。     

基于深度学习的山东大尹格庄金矿床深部三维预测模型

在隐伏矿体三维预测中,预测模型的准确性在很大程度上取决于找矿指标对矿化富集部位的指示性。然而,找矿指标容易受到找矿概念模型可靠性和成矿信息提取有效性限制,从而影响预测的准确性。论文以山东大尹格庄金矿隐伏矿体三维预测为例,基于深度学习方法,构建矿床深部隐伏矿体三维预测模型,旨在利用深度网络模型,学习获得对矿化具有显著指示性的找矿指标,提升三维预测的准确性。该方法将三维地质模型及其形态特征转换为适合卷积网络二维图像,采用卷积神经网络实现找矿指标的自动提取,并构建三维地质模型到矿化富集地段的定量关联。利用该方法建立了大尹格庄金矿的三维预测模型,经与几种人工建立找矿指标预测模型的对比分析,表明基于深度学习的预测模型较大地提升了预测准确性。     

Hierarchical object-based stochastic modeling of fluvial reservoirs

This paper describes a novel approach to modeling braided stream fluvial reservoirs. The approach is based on a hierarchical set of coordinate transformations involving relative straingraphic coordinates, translations, rotations, and straightening functions. The emphasis is placed on geologically sound geometric concepts and realistically-attainable conditioning statistics including areal and vertical facies proportions. Modeling proceeds in a hierarchical fashion, that is (1) a stratigraphic coordinate system is established for each reservoir layer, (2) a number of channel complexes are positioned within each layer, and then (3) channels are positioned within each channel complex. The geometric specification of each sand-filled channel within the background of floodplain shales is a marked point process. Each channel is marked with a starting location, size parameters, and sinuosity parameters. We present the hierarchy of eight coordinate transformations, introduce an analytical expression for the channel cross-section shape, describe the simulation algorithm, and demonstrate how the realizations are made to honor local conditioning data from wells and global conditioning data such as areal and vertical proportions.     

Conditioning generative adversarial networks on nonlinear data for subsurface flow model calibration and uncertainty quantification

Conditioning complex subsurface flow models on nonlinear data is complicated by the need to preserve the expected geological connectivity patterns to maintain solution plausibility. Generative adversarial networks (GANs) have recently been proposed as a promising approach for low-dimensional representation of complex high-dimensional images. The method has also been adopted for low-rank parameterization of complex geologic models to facilitate uncertainty quantification workflows. A difficulty in adopting these methods for subsurface flow modeling is the complexity associated with nonlinear flow data conditioning. While conditional GAN (CGAN) can condition simulated images on labels, application to subsurface problems requires efficient conditioning workflows for nonlinear data, which is far more complex. We present two approaches for generating flow-conditioned models with complex spatial patterns using GAN. The first method is through conditional GAN, whereby a production response label is used as an auxiliary input during the training stage of GAN. The production label is derived from clustering of the flow responses of the prior model realizations (i.e., training data). The underlying assumption of this approach is that GAN can learn the association between the spatial features corresponding to the production responses within each cluster. An alternative method is to use a subset of samples from the training data that are within a certain distance from the observed flow responses and use them as training data within GAN to generate new model realizations. In this case, GAN is not required to learn the nonlinear relation between production responses and spatial patterns. Instead, it is tasked to learn the patterns in the selected realizations that provide a close match to the observed data. The conditional low-dimensional parameterization for complex geologic models with diverse spatial features (i.e., when multiple geologic scenarios are plausible) performed by GAN allows for exploring the spatial variability in the conditional realizations, which can be critical for decision-making. We present and discuss the important properties of GAN for data conditioning using several examples with increasing complexity.

     

大数据挖掘与智能预测找矿靶区实验研究——卷积神经网络模型的应用

成矿预测需要通过一定的规则集合将专家观点、地质背景、成矿类型等因素进行综合考虑。但由于受到人类实际计算能力的生物条件限制,影响找矿预测成果的最大因素是找矿者的经验知识。随着大数据时代的到来,成矿预测可充分利用数学计算,即以特定规则对成矿系统进行计算,以概率表示成矿前景。依靠计算机的超级运算能力,结合机器学习的方法技术,可以对地质大数据进行成矿预测特征学习,实现对众多地质变量与矿体相关性之间的验证,从而进行预测。本文以安徽东至兆吉口铅锌矿床为例,示范如何通过机器学习的卷积神经网络方法,学习元素Zn在地表的分布特征与矿体在地下空间就位的耦合关系,并圈定靶区。经过450次训练后,得到了准确率95%,损失率14%的CNN模型,并成功实现智能圈定3块找矿靶区。这种神经网络模型可能表达了矿体在地下就位时元素在地表分布的响应,可以用来进行找矿勘查并圈定靶区。     

A New Data-Space Inversion Procedure for Efficient Uncertainty Quantification in Subsurface Flow Problems

Uncertainty quantification for subsurface flow problems is typically accomplished through model-based inversion procedures in which multiple posterior (history-matched) geological models are generated and used for flow predictions. These procedures can be demanding computationally, however, and it is not always straightforward to maintain geological realism in the resulting history-matched models. In some applications, it is the flow predictions themselves (and the uncertainty associated with these predictions), rather than the posterior geological models, that are of primary interest. This is the motivation for the data-space inversion (DSI) procedure developed in this paper. In the DSI approach, an ensemble of prior model realizations, honoring prior geostatistical information and hard data at wells, are generated and then (flow) simulated. The resulting production data are assembled into data vectors that represent prior ‘realizations’ in the data space. Pattern-based mapping operations and principal component analysis are applied to transform non-Gaussian data variables into lower-dimensional variables that are closer to multivariate Gaussian. The data-space inversion is posed within a Bayesian framework, and a data-space randomized maximum likelihood method is introduced to sample the conditional distribution of data variables given observed data. Extensive numerical results are presented for two example cases involving oil–water flow in a bimodal channelized system and oil–water–gas flow in a Gaussian permeability system. For both cases, DSI results for uncertainty quantification (e.g., P10, P50, P90 posterior predictions) are compared with those obtained from a strict rejection sampling (RS) procedure. Close agreement between the DSI and RS results is consistently achieved, even when the (synthetic) true data to be matched fall near the edge of the prior distribution. Computational savings using DSI are very substantial in that RS requires \(O(10^5\)–\(10^6)\) flow simulations, in contrast to 500 for DSI, for the cases considered.     

Facies Modeling Using a Markov Mesh Model Specification

The spatial continuity of facies is one of the key factors controlling flow in reservoir models. Traditional pixel-based methods such as truncated Gaussian random fields and indicator simulation are based on only two-point statistics, which is insufficient to capture complex facies structures. Current methods for multi-point statistics either lack a consistent statistical model specification or are too computer intensive to be applicable. We propose a Markov mesh model based on generalized linear models for geological facies modeling. The approach defines a consistent statistical model that is facilitated by efficient estimation of model parameters and generation of realizations. Our presentation includes a formulation of the general framework, model specifications in two and three dimensions, and details on how the parameters can be estimated from a training image. We illustrate the method using multiple training images, including binary and trinary images and simulations in two and three dimensions. We also do a thorough comparison to the snesim approach. We find that the current model formulation is applicable for multiple training images and compares favorably to the snesim approach in our test examples. The method is highly memory efficient.     

High-order Statistics of Spatial Random Fields: Exploring Spatial Cumulants for Modeling Complex Non-Gaussian and Non-linear Phenomena

The spatial distributions of earth science and engineering phenomena under study are currently predicted from finite measurements and second-order geostatistical models. The latter models can be limiting, as geological systems are highly complex, non-Gaussian, and exhibit non-linear patterns of spatial connectivity. Non-linear and non-Gaussian high-order geostatistics based on spatial connectivity measures, namely spatial cumulants, are proposed as a new alternative modeling framework for spatial data. This framework has two parts. The first part is the definition, properties, and inference of spatial cumulants—including understanding the interrelation of cumulant characteristics with the in-situ behavior of geological entities or processes, as examined in this paper. The second part is the research on a random field model for simulation based on its high-order spatial cumulants. Mathematical definitions of non-Gaussian spatial random functions and their high-order spatial statistics are presented herein, stressing the notion of spatial cumulants. The calculation of spatial cumulants with spatial templates follows, including anisotropic experimental cumulants. Several examples of two- and three-dimensional images, including a diamond bearing kimberlite pipe from the Ekati Mine in Canada, are analyzed to assess the relations between cumulants and the spatial behavior of geological processes. Spatial cumulants of orders three to five are shown to capture directional multiple-point periodicity, connectivity including connectivity of extreme values, and spatial architecture. In addition, they provide substantial information on geometric characteristics and anisotropy of geological patterns. It is further shown that effects of complex spatial patterns are seen even if only subsets of all cumulant templates are computed. Compared to second-order statistics, cumulant maps are found to include a wealth of additional information from underlying geological patterns. Further work seeks to integrate this information in the predictive capabilities of a random field model.     

Landslide identification using machine learning

Landslide identification is critical for risk assessment and mitigation.This paper proposes a novel machinelearning and deep-learning method to identify natural-terrain landslides using integrated geodatabases.First,landslide-related data are compiled,including topographic data,geological data and rainfall-related data.Then,three integrated geodatabases are established;namely,Recent Landslide Database(Rec LD),Relict Landslide Database(Rel LD)and Joint Landslide Database(JLD).After that,five machine learning and deep learning algorithms,including logistic regression(LR),support vector machine(SVM),random forest(RF),boosting methods and convolutional neural network(CNN),are utilized and evaluated on each database.A case study in Lantau,Hong Kong,is conducted to demonstrate the application of the proposed method.From the results of the case study,CNN achieves an identification accuracy of 92.5%on Rec LD,and outperforms other algorithms due to its strengths in feature extraction and multi dimensional data processing.Boosting methods come second in terms of accuracy,followed by RF,LR and SVM.By using machine learning and deep learning techniques,the proposed landslide identification method shows outstanding robustness and great potential in tackling the landslide identification problem.     

Validation Techniques for Geological Patterns Simulations Based on Variogram and Multiple-Point Statistics

Traditional simulation methods that are based on some form of kriging are not sensitive to the presence of strings of connectivity of low or high values. They are particularly inappropriate in many earth sciences applications, where the geological structures to be simulated are curvilinear. In such cases, techniques allowing the reproduction of multiple-point statistics are required. The aim of this paper is to point out the advantages of integrating such multiple-statistics in a model in order to allow shape reproduction, as well as heterogeneity structures, of complex geological patterns to emerge. A comparison between a traditional variogram-based simulation algorithm, such as the sequential indicator simulation, and a multiple-point statistics algorithm (e.g., the single normal equation simulation) is presented. In particular, it is shown that the spatial distribution of limestone with meandering channels in Lecce, Italy is better reproduced by using the latter algorithm. The strengths of this study are, first, the use of a training image that is not a fluvial system and, more importantly, the quantitative comparison between the two algorithms. The paper focuses on different metrics that facilitate the comparison of the methods used for limestone spatial distribution simulation: both objective measures of similarity of facies realizations and high-order spatial cumulants based on different third- and fourth-order spatial templates are considered.     

卷积神经网络和视觉注意力语义分割模型在高分辨率遥感影像分类中的性能分析

随着深度学习语义分割的快速发展,基于计算机视觉语义分割模型的高分辨率遥感影像分类方法也大量涌现。为系统定量地研究经典的和先进的视觉语义分割模型在遥感影像分类中的性能,在总结深度学习语义分割进展的基础上,选择9种基于卷积神经网络(CNN)和视觉注意力的语义分割算法,对米级和厘米级2个尺度的遥感数据集进行分析研究。在模型构建上基于计算机视觉通用的语义分割框架,训练时采用红绿蓝3波段遥感图像并基于ImageNet预训练权重进行迁移学习训练。研究结果表明：通用的语义分割模型通过常规训练设置进行训练能取得较好的遥感影像分类效果,部分地物的交并比(IoU)可以达到90%以上;基于视觉注意力的遥感影像分类模型的精度普遍高于基于CNN的模型,且MaskFormer能更有效地提取离散的地物信息;不同类别的精度最高值并不全在总体最优模型中,部分会存在于次优模型中;类似的地物在更高分辨率遥感数据集中可以获得更高的精度。