Search

Formation and lithology identification based on well logging curves reflecting geophysical response characteristics is fundamental for drilling planning and reservoir recovery. For the purpose of providing efficient, accurate, and comprehensive insights for drilling operation decisions, the present research evaluates three typical supervised learning algorithms based on machine learning, e.g. Adaboost, decision tree and support vector machine (SVM). By comparing the prediction results from three typical classification algorithms based on performance metrics such as accuracy, precision, recall, F1 score, Adaboost and decision tree are found to present more accurate prediction with relatively higher accuracy, precision, recall and F1 score. The prediction accuracy is positively related to the training data set proportion for all three approaches. Decision tree approach spends less computation time while still provides favorable prediction scores. The accuracy of prediction gradually increases as the number of logging features increases. Accuracy for most parameter combinations beyond four logging parameters can be up to 90%. The neutron porosity and the spontaneous potential are considered as the most influential parameters affecting the prediction accuracy.

Channel fracturing generally refers to a kind of novel hydraulic fracturing treatment that relies on the intermittent pumping of proppant-laden and proppant-free fluid to generate highly conductive channels within the formation. The fracture conductivity within the proppant pillar can increase up to several folds. However, how to effectively evaluate the fracture conductivity in channel fracturing is a tough problem. Unlike conventional fracturing, there are two distinct conductive media within the fracture: proppant pack and free channel. This paper analyzes the special proppant distribution pattern of channel fracturing, and the expression of fracture permeability is derived. On the basis of Hertz contact theory and geometry of proppant embedment, the expression of fracture opening is obtained. Then the effects of proppant distribution density and proppant pillar radius on fracture conductivity are analyzed. In the subsequent chapter, the correlation between proppant pillar permeability and free channel permeability is provided. The results show that the fracture conductivity increases at first and then decreases rapidly with the increase in proppant distribution density. The conductivity factor reaches its maximum value when vertical pillar spacing is equal to horizontal spacing. The more regular the shape of the proppant pillar is, the greater the fracture permeability is. The research in this paper provides a simple method to evaluate the key parameters controlling the fracture conductivity and permeability in channel fracturing.

Hydraulic fracture is the main flow path for gas transport. The proppants are man-made material that filled in the hydraulic fractures to keep them open and allow gas flow through. The permeability change of hydraulic fracture is controlled by the combined effect of compaction and embedment. In this study, we modeled the proppant embedment as a function of effective stress by a transformed Hertz contact model and a proposed power law model which is analogous to the Oliver-Pharr model. The results illustrate that the power law relationship could better fit the experimental data, because the Hertz model becomes invalid when the embedment is large compared to the proppant size. By incorporating the power law correlation into an existing theoretical permeability model as a function of effective stress, a permeability model for the hydraulic fracture filled with proppant packs under combined effect of compaction and embedment is developed. The new model is able to adequately describe the permeability data of proppant packs confined by rock core slices. Although this study puts forward the theoretical basis of the hydraulic permeability modelling under combined effect of compaction and embedment, more fundamental studies are required to investigate the contact behaviour between the proppant packs and the fracture face under various conditions. Therefore, the permeability model could be further improved by introducing the new advanced proppant embedment correlations.

Channel fracturing revolutionarily changes continuous sanding into discontinuous distribution by wrapping proppant particles with a degradable fiber and then injecting these packs or pillars into the formation. Channel fracturing creates open channels between pillars and substantially increases the fracture conductivity. The pulse time of injecting proppants can influence pillar spacing, thereby changing the opening of the fracture. Under different formation conditions, determining a proper pulse time to improve the wellbore productivity is the key problem in channel fracturing. On the basis of Hertz contact theory, we assume that a proppant pillar is a standard cylindrical indenter. We also calculate the penetration depth beneath the original surface. In this manner, we obtain the residual fracture width under a certain confining stress. This residual fracture width serves as our optimization objective. Then, by combining the flow conservation equation, we find that a restriction relationship exists between the pillar spacing and pulse time. By controlling the variables, we adjust the parameters to find the maximum fracture width. The fracture width reaches its maximum when the horizontal spacing equals vertical spacing. A set of simulations are completed and analyzed. The results show that the pulse time, pillar spacing, and formation parameters such as the closure pressure, elasticity modulus and initial crack width, can significantly affect the supporting effectiveness of the proppant pillar. Critical points exist when these parameters increase or decrease to a certain level. For the convenience of field operation, a reference plate is provided. A proper pulse time can be determined on the plate under certain formation conditions, thereby providing theoretical guidance for actual channel fracturing.

Although many coal permeability models have been developed in the past decades to describe the coal permeability behaviour under elastic state, few of them address the coal permeability change under plastic and post-failure state which is often the case within the plastic region adjacent to the excavation face in underground coal mining. In this study, a methodology to model permeability change from elastic to post-failure state is developed by using a modified logistic growth function in conjunction with the classic exponential coal permeability correlation. The proposed coal permeability model is a function of mean effective stress which controls the coal compaction and deviatoric effective stress which controls coal fracturing. The coal permeability may increase by up to several orders of magnitude after failure and then reaches a plateau during triaxial tests. The new model is able to capture this behaviour by matching a set of permeability data in transition from elastic to post-failure state under triaxial stress conditions. This modelling approach may be used to better understand coal permeability changes associated with mining activities, which have applications in the prediction of gas emission, risk assessment of coal and gas outburst, and analysis of gas drainage near mining openings. It is anticipated that the current work may attract more attentions on coal permeability modelling under plastic condition, a critical issue for mining safety.

A large amount of experimental data for methane adsorption capacity in shale are available and the Langmuir model is capable to describe most of these adsorption isotherms. Two parameters used in the Langmuir model are the Langmuir pressure constant P L and the Langmuir volume constant V L . However, experimental data also demonstrate that the gas adsorption capacity of shale is greatly affected by temperature and the classic Langmuir model, with temperature independent V L and temperature dependent P L , could not well interpret the experimental data. This is partly attributed to the isosteric heat of adsorption, which makes V L vary with temperature. The change of V L with temperature for gas adsorption in shale and its dependency on shale properties are still not well understood. In this study, the variation characteristics of V L with temperature for gas shale are investigated through a modeling analysis on the published gas sorption data on the shales from the US, China, Canada, and etc. The classic Langmuir model is improved by considering the temperature dependent V L . The results show that the improved Langmuir model can reasonably describe the shale gas sorption data with less fitting parameters required and it allows a better understanding of gas sorption capacity under the combined effect of temperature and pressure. In addition, the model parameter analysis indicates that the adsorption capacity of shale samples with lower TOC is more likely to be affected by temperature change on the basis of the limited data available. More experiments are required to investigate how the shale properties affect the thermal impact on the gas adsorption capacity on gas shale.

As pore size of gas shale ranges widely, various flow types including viscous flow, slip flow and Knudsen flow could coexist during shale gas production. Currently, different models have to be applied to describe the gas transport behaviours in different flow regions and the selection of the models is highly dependent on the flow classification criteria. This also brings difficulty in selecting proper gas flow models near the flow region boundaries, which again depends on the flow classification criteria. To address this problem, this study proposes a unified method to describe the gas flow in the whole spectrum of flow regions by transforming all the gas transport equations for different flow types into a general form. The modelling results by the unified method are comparable with that by the previous models, but less input parameters are required for the unified method. Since the unified method provides a continuous permeability kea for all flow regions, it avoids discontinuity at the boundary of different flow regions caused by using piecewise flow models. The method is able to depict the gas permeability evolution with reservoir pressure change and is applied to interpret the literature permeability data and to predict the permeability variation among all the flow regions with different gas pressures and pore sizes. The results illustrate that the flow types may change within a specific pore (especially for small pores) during the pressure depletion in shale gas production and different gas transport mechanisms may coexist for some shales with typical bimodal pore size distribution under a specific gas pressure condition. This work provides a unified way to analyse gas flow behaviours and gas permeability variation in different regions and can be readily applied to the reservoir simulation models to predict the gas production behaviour especially the long term gas production behaviour.

Although permeability data for different gas shales have been reported previously and attempts have been made to match permeability with empirical correlations, theoretical studies of shale permeability modelling are lacking. In this work, the correlation between fracture permeability and effective stress is established for gas shales through theoretical derivation. This model is able to match the permeability data for different gas shales. The matching results for the gas shale studied show that the model coefficient, fracture compressibility, which decreases as initial shale permeability increases, is strongly affected by the flow directions and varies with the shale’s mineralogical composition. Furthermore, the correlation between fracture permeability and reservoir pressure has also been established. Sensitivity study shows that fracture permeability may decrease significantly with the reservoir pressure drawdown. Moreover, the horizontal fracture permeability drop is found to be significantly affected by the Young’s modulus’ anisotropic ratio ( E h / E v ). The insights gained warrant further theoretical and experimental studies to evaluate shale fracture permeability.

The non-Darcy factor, an indicator for the non-Darcy effect, is dependent on the properties of porous media and pore fluid including permeability, viscosity, density, flow velocity and a coefficient named as β factor. Experimental results show that the β factor can be expressed as a power law of permeability. For conventional gas reservoirs, this β factor can be assumed as a constant as the permeability change is negligible. However, the constant β factor may not be suitable for coal seams with remarkable permeability change and a variable β factor as a function of coal permeability should be an alternative. Moreover, the coal permeability change is complex due to the competing effects of coal cleat compression and sorption induced coal shrinkage/swelling. Few studies have been done previously to incorporate the variable β factor as a function of coal permeability in reservoir simulations. In the present work, both the coal permeability change and the variable β factor are coupled in a dual porosity model to study the non-Darcy flow behavior in coal seams. The simulation results illustrate that the evolution of non-Darcy factor becomes tortuous by using a variable β factor, which differs from the monotonic behavior when constant β factors are applied. Furthermore, increasing the coal cleat compressibility and matrix shrinkage strain tends to intensify the tortuous behavior. The simulation results also indicate that using typical constant β factors, instead of the variable one, may significantly underestimate or overestimate the gas production rate for coalbed methane wells.