Jingsheng Ma

Stochastic Pore Network Generation from 3D Rock Images

Pore networks can be extracted from 3D rock images to accurately predict multi-phase flow properties of rocks by network flow simulation. However, the predicted flow properties may be sensitive to the extracted pore network if it is small, even though its underlying characteristics are representative. Therefore, it is a challenge to investigate the effects on flow properties of microscopic rock features individually and collectively based on small samples. In this article, a new approach is introduced to generate from an initial network a stochastic network of arbitrary size that has the same flow properties as the parent network. Firstly, we characterise the realistic parent network in terms of distributions of the geometrical pore properties and correlations between these properties, as well as the connectivity function describing the detailed network topology. Secondly, to create a stochastic network of arbitrary size, we generate the required number of nodes and bonds with the correlated properties of the original network. The nodes are randomly located in the given network domain and connected by bonds according to the strongest correlation between node and bond properties, while honouring the connectivity function. Thirdly, using a state-of-the-art two-phase flow network model, we demonstrate for two samples that the rock flow properties (capillary pressure, absolute and relative permeability) are preserved in the stochastic networks, in particular, if the latter are larger than the original, or the method reveals that the size of the original sample is not representative. We also show the information that is necessary to reproduce the realistic networks correctly, in particular the connectivity function. This approach forms the basis for the stochastic generation of networks from multiple rock images at different resolutions by combining the relevant statistics from the corresponding networks, which will be presented in a future publication.

Providing spatial statistical data analysis functionality for the GIS user: the SAGE project

Geographical Information Systems (GIS) are being used in a growing number of application areas. As a consequence there have been frequent calls to expand the range of spatial analysis tools available to users of GIS but a reluctance on the part of GIS software vendors to include such tools in standard software packages. An alternative approach is to link extra tools to GIS packages which raises a series of issues, such as, What sort of tools should be included? How should the linkage be done? To what extent can the functionality of the GIS be used? This paper draws on the results of a project in which software for statistical spatial data analysis (SSDA) was linked to ARC/INFO to produce a software system called SAGE. The statistical tools implemented included those which were felt to be useful to the general GIS user (as opposed to the specialist spatial statistician or econometrician), and they were linked to ARC/INFO using a client server architecture. The GIS was used within the context of SSDA for map drawing, spatial queries and operations on the topology of the spatial data, although it was found that the map drawing facilities of ARC/INFO were not well suited to the needs of this application. One of the conclusions of the project was that many of the techniques of exploratory spatial data analysis, such as providing graphical data summaries and linking these to cartographic views of the data could be easily integrated into existing GIS packages, providing a useful addition to their functionality for many GIS users. Many of the other SSDA facilities are probably still best provided in specialist software, but there is a need for a robust and standardised means for such software to extract information about the topology of spatial data from within GIS packages.

Locally informed gravitational search algorithm

Gravitational search algorithm (GSA) has been successfully applied to many scientific and engineering applications in the past few years. In the original GSA and most of its variants, every agent learns from all the agents stored in the same elite group, namely Kbest. This type of learning strategy is in nature a fully-informed learning strategy, in which every agent has exactly the same global neighborhood topology structure. Obviously, the learning strategy overlooks the impact of environmental heterogeneity on individual behavior, which easily resulting in premature convergence and high runtime consuming. To tackle these problems, we take individual heterogeneity into account and propose a locally informed GSA (LIGSA) in this paper. To be specific, in LIGSA, each agent learns from its unique neighborhood formed by k local neighbors and the historically global best agent rather than from just the single Kbest elite group. Learning from the k local neighbors promotes LIGSA fully and quickly explores the search space as well as effectively prevents premature convergence while the guidance of global best agent can accelerate the convergence speed of LIGSA. The proposed LIGSA has been extensively evaluated on 30 CEC2014 benchmark functions with different dimensions. Experimental results reveal that LIGSA remarkably outperforms the compared algorithms in solution quality and convergence speed in general.

Pore-Scale Characterization of Gas Flow Properties in Shale by Digital Core Analysis

Abstract The characterization of flow properties for shale gas reservoirs is of importance but is more complex than for conventional petroleum reservoirs. The complications arise from the tightness of pore space and the diverse chemical compositions of the matrix as well as nonDarcy flow processes taking place within pore space. All these pose real challenges to develop reliable and robust laboratory core measurement techniques. Digital core analysis is shown to have potential to compliment laboratory measurement techniques. This chapter summarizes some of the work on developing digital core analysis technology for flow characterization of gas shale and on assessing the impact of slip flow, Knudsen diffusion, and gas adsorption on the permeability of free gas.

Pore-scale Modelling of Shale Gas Permeability Considering Shale Gas Adsorption

Shale gas permeability needs to be estimated in order to predict the quality of shale gas reservoirs and to develop shale gas production strategies. With advances in high-resolution imaging technology, one can characterise the pore space of a gas shale sample, which typically contains pores ranging from micrometers to nanometers, and to construct a pore-space model to simulate the gas flow numerically and to calculate the permeability. Gas flow has long been known to behave differently in such a confined space, and the smaller the pores the larger discrepancy is generally expected between gas and liquid (e.g. water) permeability. Since shale gas molecules stored mainly in nano-metre pores in kerogens by gas adsorption, adsorbed gas molecules, of half-nanometres in diameter, could reduce the pore size for free gas flow substantially and so alter the gas permeability significantly. In this work, we extended a model for modelling shale gas flow to account for the gas adsorption effect. We adopted the Langmuir single-layer adsorption model to the multiple layers. We analysed the gas adsorption impact on the permeability on a cylindrical pore analytically, and on a shale sample whose pore space are represented as a node-and–bond pore network, using our network flow model (Ma et al., 2014). The results revealed that the adsorption effect depends strongly on the gas pressure and the radii of pores. Given that low gas pressure increases gas slippage at pore surfaces and decreases the thickness of the adsorption layers then, consequently, enhances the permeability, undesirable operation conditions could lead to an earlier decline of gas production.

Gravitation-Based Edge Detection in Hyperspectral Images

Edge detection is one of the key issues in the field of computer vision and remote sensing image analysis. Although many different edge-detection methods have been proposed for gray-scale, color, and multispectral images, they still face difficulties when extracting edge features from hyperspectral images (HSIs) that contain a large number of bands with very narrow gap in the spectral domain. Inspired by the clustering characteristic of the gravitational theory, a novel edge-detection algorithm for HSIs is presented in this paper. In the proposed method, we first construct a joint feature space by combining the spatial and spectral features. Each pixel of HSI is assumed to be a celestial object in the joint feature space, which exerts gravitational force to each of its neighboring pixel. Accordingly, each object travels in the joint feature space until it reaches a stable equilibrium. At the equilibrium, the image is smoothed and the edges are enhanced, where the edge pixels can be easily distinguished by calculating the gravitational potential energy. The proposed edge-detection method is tested on several benchmark HSIs and the obtained results were compared with those of four state-of-the-art approaches. The experimental results confirm the efficacy of the proposed method.

Construction of guiding grids for flow modelling in fault damage zones with through-going regions of connected matrix

A natural fault damage zone (FDZ) often contains many small, thin faults that form complex arrays. Because the small faults reduce the permeability in most siliciclastic rock types, FDZs can have a major impact on fluid flow behaviours at reservoir or aquifer scales. Numerical flow modelling techniques, which are often used to predict the impact of such complexity, face a challenge: how to accurately capture the effects of the fault connectivity in coarse-scale domain- and numerical-discretisation. If thin faults are discretised explicitly along with the matrix, this leads to grids that have many cells, making it difficult to solve the flow equations efficiently. Some discretisation schemes lead to a significant reduction in the number of cells by means of either indirect or direct simplification to faults. For a FDZ model that contains low-permeability faults and through-going regions (TGRs)-matrix material that link fluid inlets and outlets, simplification must not lead to any change to the fault connectivity that could result in mis-representations of flow-influential TGRs as non-TGRs, or vice versa, in the discretisation. In this paper, we describe a scheme that is capable of generating a guiding grid that resolves all flow-influential TGRs in a FDZ model. This grid can then be used to guide fault simplification in the subsequent construction of structured and/or unstructured computational grids that retain the essential fault connectivity. This scheme combines the identification of TGRs with simple network-based flow modelling to estimate the flow influences of TGRs.

3-D Modelling and Experimental Comparison of Reactive Flow in Carbonates under Radial Flow Conditions

We use a two-scale continuum model to simulate reactive flow and wormhole formation in carbonate rocks under 3-D radial flow conditions. More specifically, we present a new structure-property relationship based on the fractal geometry theory, to describe the evolution of local permeability, pore radius, and specific area with porosity variation. In the numerical calculation, to improve the convergence rate, the heterogeneous medium in question is extended by adding a thin layer of homogeneous porous medium to its inlet. We compare the simulation results with the available experimental observations and find that they are qualitatively consistent with each other. Additionally, sensitivity analysis of the dissolution process with respect to acid injection rate and rock heterogeneity, including heterogeneity magnitude and correlation length, is presented.

Multi-scale Pore-network Modelling of WAG in Carbonates

Carbonate reservoirs have textural heterogeneities at all length-scales (triple porosity: pore-vug-fracture) and tend to be mixed- to oil-wet. The choice of an enhanced oil recovery process and the prediction of oil recovery require a sound understanding of the fundamental controls on fluid flow in mixed- to oil-wet carbonate rocks, as well as physically robust flow functions, i.e. relative permeability and capillary pressure functions. Obtaining these flow functions is a challenging task, especially when three fluid phases coexist, such as during water-alternating-gas injection (WAG). We have recently developed a method for integration of pore-networks derived from micro CT images at different length-scales, thus capturing pore structures from different types of porosity. The network integration method honours the connectivity between different pore types, including micro-fractures, and their spatial distribution. In this work, we use these multi-scale networks as input for our three-phase flow pore-network model, which comprises a novel thermodynamic criterion for formation and collapse of oil layers that strongly depends on the fluid spreading behaviour and the rock wettability. The criterion affects in particular the oil relative permeability at low oil saturations and the accurate prediction of residual oil saturations. We generate three-phase flow functions for gas injection and WAG from networks with carbonate pore geometries and connectivities and we demonstrate the impact on residual saturations of the different types of porosity and the interaction with different realistic wettability scenarios. We also show that the network generated three-phase flow relative permeabilities are distinctly different from traditional models, such as Stone’s. The flow functions will be used in a heterogeneous carbonate reservoir model and to demonstrate their impact on the sweep efficiency.

FADA: an efficient dimension reduction scheme for image classification

This paper develops a novel and efficient dimension reduction scheme--Fast Adaptive Discriminant Analysis (FADA). FADA can find a good projection with adaptation to different sample distributions and discover the classification in the subspace with naive Bayes classifier. FADA overcomes the high computational cost problem of current Adaptive Discriminant Analysis (ADA) and also alleviates the overfitting problem implicitly caused by ADA. FADA is tested and evaluated using synthetic dataset, COREL dataset and three different face datasets. The experimental results show FADA is more effective and computationally more efficient than ADA for image classification.