Search

Abstract: Unconventional shale reservoirs as the result of extremely low matrix permeability, higher potential gas productivity requires not only sufficient gas-in-place, but also a high concentration of brittle minerals (silica and/or carbonate) that is amenable to hydraulic fracturing. Shale lithofacies is primarily defined by mineral composition and organic matter richness, and its representation as a 3-D model has advantages in recognizing productive zones of shale-gas reservoirs, designing horizontal wells and stimulation strategy, and aiding in understanding depositional process of organic-rich shale. A challenging and key step is to effectively recognize shale lithofacies from well conventional logs, where the relationship is very complex and nonlinear. In the recognition of shale lithofacies, the application of support vector machine (SVM), which underlies statistical learning theory and structural risk minimization principle, is superior to the traditional empirical risk minimization principle employed by artificial neural network (ANN). We propose SVM classifier combined with learning algorithms, such as grid searching, genetic algorithm and particle swarm optimization, and various kernel functions the approach to identify Marcellus Shale lithofacies. Compared with ANN classifiers, the experimental results of SVM classifiers showed higher cross-validation accuracy, better stability and less computational time cost. The SVM classifier with radius basis function as kernel worked best as it is trained by particle swarm optimization. The lithofacies predicted using the SVM classifier are used to build a 3-D Marcellus Shale lithofacies model, which assists in identifying higher productive zones, especially with thermal maturity and natural fractures. [Copyright &y& Elsevier]

Abstract: The success of shale gas in North America has attracted increased interest in “unconventional” reservoirs. Two critical factors for shale-gas reservoirs are units amenable to hydrologic fracture stimulation and sufficient natural gas content. The effectiveness of hydrologic fracture stimulation is influenced by rock geomechanical properties, which are related to rock mineralogy. The natural gas content in shale reservoirs has a strong relationship with organic matter, which is measured by total organic carbon (TOC). A 3D shale lithofacies model constructed using mineral composition, rock geomechanical properties and TOC content can be applied to optimize the design of horizontal well trajectories and stimulation strategies. Core analysis data, log data and seismic data were used to build a 3D shale lithofacies from core to well and finally to regional scale. Core, advanced and common logs were utilized as inputs to petrophysical analysis, and various pattern recognition methods, such as discriminant analysis, fuzzy logic, neural network and support vector machine. A limited set of eight derived parameters from common logs were determined as critical inputs for pattern recognition methods. Advanced logs, such as pulsed neutron spectroscopy, are used to determine mineral composition and TOC data improve and confirm the quantitative relationship between conventional logs and lithofacies. Seismic data, interpreted sequence stratigraphy and depositional environments were used as constraints to build deterministic and stochastic 3D lithofacies models and to extrapolate lithofacies from well scale to regional scale. [Copyright &y& Elsevier]

The generation and primary migration of hydrocarbons in organic-rich shale leaves void space in organic matter, which is the porosity associated with organic matter commonly observed under scanning electron microscope (SEM). In this study, Middle Devonian black shale core samples were collected from three wells penetrating the organic-rich Marcellus Shale and the organic-lean Mahantango Formation in Pennsylvania and West Virginia. Pyrolysis, ion milled SEM and low-pressure nitrogen adsorption analysis were conducted to investigate the organic richness and the properties of the pore system. Vitrinite reflectance (R o) values in the range of 1.36%–2.89% represent a maturity spectrum covering the wet-gas to post-mature zones. In general, the pore system is composed of organic matter-hosted pores and mineral-hosted pores. However, the dominant pore types and pore sizes vary stratigraphically across lithology and abundance of organic matter. All the organic matter observed in this study shows an amorphous occurrence. Pore space between mineral grains (both silt-size and clay-size) can be filled by organic matter, which contains secondary porosity generated by thermal cracking of kerogen. Mineral-hosted pores are concentrated in organic-lean samples in which secondary organic matter could not fill most of the primary pore space. The destruction of primary mineral-hosted pores and the generation of secondary organic matter-hosted pores were observed. TOC values show positive correlations with the porosity, specific surface area, and the abundance of micropores. Increasing thermal maturity correlates with a significant decrease of pore volume and surface area, primarily through diminishing or vanishing of micropores. The richness and thermal maturity of organic matter in organic-rich Devonian shale can be effective parameters for evaluation of reservoir quality and upscaling the appraisal. • The pore structure of the Marcellus Shale and Mahantango Formation were characterized. • The evolution of the pore system is driven by compaction, cementation, and organic matter maturation. • A stratigraphic distribution of pore networks is identified and classified into four pore facies. • The mechanism of generation and preservation of organic matter hosted porosity is discussed. • The storage capacity also shows a stratigraphic distribution with respect to the pore system. [ABSTRACT FROM AUTHOR]

Abstract Predicting gas production from stimulated unconventional reservoirs has always been a challenge to oil and gas companies because the physics of fluid flow through such reservoirs is not well understood. On the other hand, modeling gas production from shale reservoirs is a complex multi-variate problem that requires proper integration of data from multiple disciplines, such as geology, petrophysics, geomechanics, completions, and surface-based measurements, etc. This study demonstrates the application of data-driven machine learning algorithms, integrating geoscientific, distributed acoustic sensing (DAS), distributed temperature sensing (DTS) fiber-optic, completions, flow scanner production log, and surface data to model daily gas production from a 28-stage stimulated horizontal well drilled in the Marcellus Shale of the Appalachian basin in North America. In addition, this study aims to utilize the data from a fiber-optic monitoring system, such as DAS and DTS to evaluate the well performance. We build supervised data-driven machine learning models using Random Forest (RF), Artificial Neural Network (ANN), and Support Vector Machine (SVM) algorithms to understand the well performance and forecast the daily gas production. The study compares the forecasted results, tests the level of accuracy, and addresses different issues with these machine learning algorithms. A spatio-temporal database is constructed and used to perform sensitivity analyses to identify the key drivers for gas production. The results show that RF and ANN algorithms can be used to predict daily gas production with significantly high accuracy; however, RF algorithm is the best predictor in terms of highest accuracy (96%), less computational time and cost. Based on sensitivity analysis, it appears that Poisson's ratio, minimum horizontal stress, DAS, casing pressure, gamma log are the most important parameters to predict gas production. Highlights • Application of machine learning algorithms to forecast daily gas production from unconventional reservoirs using large multi-sensor dataset. • Random Forest and Artificial Neural Network can model daily gas production from each stage efficiently. [ABSTRACT FROM AUTHOR]

Abstract Porosity and pore size distribution (PSD) are critical reservoir parameters. Pore surface area, pore volume, PSD, and porosity were measured using subcritical nitrogen (N 2) adsorption, and helium porosimetry. A suite of 17 samples were collected from 4 wells in Pennsylvania and West Virginia to analyze the evolution of porosity with increasing thermal maturity in Middle Devonian shales of the Appalachian Basin. The thermal maturity of the tested samples covers a wide range in the hydrocarbon generation sequence from wet gas/condensate zone (vitrinite reflectance (R o) = 1.16%) to post-mature zone (R o = 2.79%). Shale samples from the Marcellus Shale and Mahantango Formation used in this study have total organic carbon contents from 0.41 to 7.88 wt%. Results indicate that total organic carbon (TOC) has the strongest effect on porosity and pore structure. The presence of organic matter in shale strongly enhances the storage capacity by increasing the specific surface area and pore volume, which represents sorption storage capacity and free-gas storage capacity. Differences in porosity and pore structure have a complex relationship to thermal maturity, micro texture, mineralogy, clay content, and TOC. [ABSTRACT FROM AUTHOR]

Abstract: The organic-rich Marcellus Shale was deposited in a foreland basin during Middle Devonian. In terms of mineral composition and organic matter richness, we define seven mudrock lithofacies: three organic-rich lithofacies and four organic-poor lithofacies. The 3D lithofacies model is very helpful to determine geologic and engineering sweet spots, and consequently useful for designing horizontal well trajectories and stimulation strategies. The NeuroEvolution of Augmenting Topologies (NEAT) is relatively new idea in the design of neural networks, and shed light on classification (i.e., Marcellus Shale lithofacies prediction). We have successfully enhanced the capability and efficiency of NEAT in three aspects. First, we introduced two new attributes of node gene, the node location and recurrent connection (RCC), to increase the calculation efficiency. Second, we evolved the population size from an initial small value to big, instead of using the constant value, which saves time and computer memory, especially for complex learning tasks. Third, in multiclass pattern recognition problems, we combined feature selection of input variables and modular neural network to automatically select input variables and optimize network topology for each binary classifier. These improvements were tested and verified by true if an odd number of its arguments are true and false otherwise (XOR) experiments, and were powerful for classification. [Copyright &y& Elsevier]