One of the major challenges associated with the exploitation of unconventional hydrocarbon resources is determining the optimal stimulation design. In this sense, it is necessary to understand how the parameters and variables involved in the completion process impact on production performance; the purpose is to act on such controllable variables and, consequently, maximize production and field development efficiency. Whereas physical driven tools frequently used in the oil industry are very helpful, they always imply a set of assumptions and simplifications regarding the system or phenomenon they try to model; they also require a large amount of unavailable or expensive data to calibrate them. Generally, different combinations of model parameters could explain well production behavior and for each of these solutions the way to optimize completion and development may be different.
Because of these drawbacks, and the big number of unconventional wells available, data-driven workflows have gained popularity in the last years. These models represent an excellent complement to physical driven tools in the attempt to optimize the completion and development strategy in shale plays. Several publications used both parametrical and non-parametrical models in the search of the Holy Grail: a statistical model capable of predicting how stimulation design affects productivity. The aim of this paper is to develop a novel methodology to understand the relation between formation parameters, completion design variables and production performance. An artificial neural network model (ANN) was chosen for this study.
Public production and stimulation data was merged with geological and petrophysical properties maps for almost 13.000 horizontal wells landed in Eagle Ford formation. A back propagation ANN algorithm was trained with this data-set and a cross-validation criterion was used for hyper-parameters optimization. Once the optimal model was selected, a bootstrap algorithm was run to assess for uncertainty in model prediction; these models were trained to determine which part of the input space presented enough data to get a clear signal and in which part the amount of data was not enough to differentiate signal from noise.
ANN models proved to be a fine method for this purpose obtaining R-Squared values between 0.5 and 0.7 for cross-validation sets. Significant relations were observed between production performance and lateral length, true vertical depth, porosity and fracture fluid intensity.
The methodology presented in this paper introduces a novel feature in comparison to previous publications regarding model uncertainty assessment. The coupling of the ANN model with the bootstrap re-sampling technique allowed to better understand which conclusions were statistically significant and which not, a fact that proved to be vital to correctly interpret results. It was demonstrated that such methodology is a good complement to physical-driven models in the aim to comprehend the relation between formation parameters, completion design variables and production performance.
The objective of our research is to reconcile the differences, in both age and relative stratigraphic position, between the Woodbine and Eagle Ford Groups in the outcrop and subsurface of the East Texas Basin. In the outcrop belt, organic- and carbonate-rich Middle Cenomanian mudstones are mapped within the Eagle Ford Group, where they overlie, and are separated by a regional unconformity from Early Cenomanian, organic-poor, and clay-rich mudstones of the Woodbine Group (Pepper Shale). In southern portions of the East Texas Basin, however, these same organic- and carbonate-rich Middle Cenomanian mudstones are mapped within the Maness Shale, which in turn, is overlain by Late Cenomanian to Turonian-aged mudstones (Pepper Shale) and sandstones (Dexter Formation) mapped as the Woodbine Group. Our approach to reconcile the lithostratigraphic juxtaposition between the two regions was to use chemo-stratigraphic and petrophysical data collected from the outcrops, as well as an adjacent shallow research borehole, in order to define key sequence stratigraphic units/surfaces, and then correlate the key units/surfaces from the outcrop belt into the subsurface.
Our research indicates that the Woodbine Group, is an older unconformity-bounded depositional sequence which is Early Cenomanian, whereas the Eagle Ford Group, is an overlying (younger) unconformity-bounded depositional sequence, which is Middle Cenomanian to Late Turonian. The unconformities that bound these units can be mapped from the outcrop belt into the subsurface of the East Texas Basin, to define coeval depositional sequences. As defined in this study, marine mudstones of the Woodbine Group, are clay- & silica-rich, TOC-poor, and characterized by low resistivity on geophysical logs. In general, the Woodbine Group thins, as well as transitions to more mudstone-prone facies, from northeast to southwest within the basin. While beyond the scope of this study, the Woodbine Group contains numerous higher-frequency sequences, which are stacked in an overall progradational (highstand) sequence set. The depositional profile of the unconformity which forms the top of this progradational succession sets up the relict physiographic (depositional shelf/slope/basin) profile for the overlying Eagle Ford Group.
Within the Lower Eagle Ford Formation, two high-frequency sequences, defined as the Lower and Upper Members, were defined. Within the Upper Eagle Ford Formation, three high-frequency sequences, defined as the Lower, Middle, and Upper Members, were defined. The Lower and Upper Members of the Lower Eagle Ford Formation, as well as the Lower Member of the Upper Eagle Ford Formation range from Middle Cenomanian to Early Turonian. These three high-frequency sequences contain marine mudstones that are carbonate- & TOC-rich, as well as clay- and quartz-poor, and are characterized by high resistivity values on geophysical logs. Furthermore, they are interpreted as a transgressive sequence set, with sequences that sequentially onlap, from older to younger, the inherited relict physiographic (depositional shelf/slope/basin) profile of the underlying Woodbine Group. In stark contrast, mudstones within the Middle and Upper Members of the Upper Eagle Ford Formation, which are Middle to Late Turonian, are clay-rich, TOC-poor, and characterized by low resistivity on geophysical logs. These two sequences, which are interpreted as a highstand sequence set, are sandstone-prone, and contain petroleum reservoirs that previously were incorrectly included within the Woodbine Group. Based on these correlations, updated sequence-based paleogeographic maps can be constructed for the first time across the East Texas Basin. These maps can in turn be used to define a robust portfolio of conventional, as well as unconventional tight-rock and source-rock, plays and play fairways, which are now based on a modern sequence stratigraphic, versus the traditional archaic lithostratigraphic framework.
Geochemical data measured on oil samples produced from wells landed in the Austin Chalk, the Eagle Ford Formation, and the Buda Formation and on petroleum samples sequentially extracted from Upper Eagle Ford and Lower Eagle Ford marl and calcareous shale core pucks using several solvents were used to estimate the amount and properties of producible oil, immobile adsorbed/dissolved oil, and non-producible bitumen in those core samples. Crushed core samples obtained from two monitor wells located on the San Marcos Arch where Eagle Ford source-rock beds have reached different levels of maturity were sequentially extracted using a weak solvent (cyclohexane; CH), two stronger solvents (toluene and DCM), and a very strong solvent (chloroform-methanol; CM). Similar geochemical data were measured on the core extracts (after heating them to evaporate the solvents), and on native and topped oil samples. The CH extracts exhibit n-alkane profiles characteristic of crude oil, but extracts obtained using stronger solvent do not resemble oil. C15-C35 HC compounds present in produced oils are more abundant in CH extracts (which principally contain producible oil and adsorbed/dissolved oil) than in extracts obtained using stronger solvents (which principally contain bitumen). The SARA composition of topped oil samples also is more similar to the composition of core extracts obtained using CH than extracts obtained using stronger solvents (which contain significantly more resins and asphaltenes). The extract obtained from lower-maturity marl core pucks using CH contains much more sulfur (≈4.4 wt%) than the CH extract obtained from more thermally mature marl core pucks (≈2.0 wt%). Calibrations between the API gravity, C7 temperature, and sulfur content of native and topped oil samples were used to estimate the gravity and sulfur content of core extracts obtained using different solvents. The amount of resin-rich immobile oil in the core extracts was estimated using reasonable assumptions about the composition of that component. The Lower Eagle Ford marl at the higher-maturity monitor well contains ≈0.35 wt% of ≈30-31°API producible oil and ≈0.27 wt% of non-producible bitumen. That reservoir contains only ≈0.12 wt% of ≈27°API producible oil and ≈0.38 wt% of non-producible bitumen at the lower-maturity monitor well. The LEF calcareous shale contains approximately the same amount of producible oil as the overlying marl at the more mature monitor well, but it contains much less non-producible bitumen (≈0.12 wt%).
This study introduces a methodology for estimating uncertainty in production of new shale wells. The methodology combines geostatistical modeling and machine learning and accounts for geological uncertainty. Our approach improves uncertainty quantification by merging local and global trends.
A functional random forest regression model is trained, connecting completion parameters and geological parameters with production profiles. The geological parameters are simulated on a two-dimensional areal grid using Sequential Gaussian Simulation with information from pilot wells. Production profiles are generated for each cell in the simulation grid based on optimal completion and simulated geological parameters. These machine learning-based realizations account for the global trends. We account for local effects by using cokriging to merge total production from nearby wells with the production from machine learning-based realizations.
We test the methodology on a dataset from the Eagle Ford formation. The result of the study is a map of the play highlighting the most probable production (P50) for different areas and associated risk (P90–P10). The resulting map allows us to rank locations for new wells for drilling. The proposed methodology provides a first estimate, and a more detailed data investigation is required to sanction a new well in a particular location.
Dutta, Sandipan (Cairn Oil & Gas, Vedanta Ltd.) | Kuila, Utpalendu (Cairn Oil & Gas, Vedanta Ltd.) | Naidu, Bodapati (Cairn Oil & Gas, Vedanta Ltd.) | Yadav, Raj (Cairn Oil & Gas, Vedanta Ltd.) | Dolson, John (DSP Geosciences and Associates LLC) | Mandal, Arpita (Cairn Oil & Gas, Vedanta Ltd.) | Dasgupta, Soumen (Cairn Oil & Gas, Vedanta Ltd.) | Mishra, Premanand (Cairn Oil & Gas, Vedanta Ltd.) | Mohapatra, Pinakadhar (Cairn Oil & Gas, Vedanta Ltd.)
The Eocene Lower Barmer Hill (LBH) Formation is the major regional source rock in the Barmer Basin rift, located in Rajasthan, India, and has substantial unconventional shale potential. The basin is almost completely covered with 3D seismic, providing an opportunity for more surgical mapping of the rapid structural and stratigraphic changes typical with any syn-rift deposit. Thick sections of organic-rich black shales reaching 400 meters thickness with TOC up to 14 wt. %, were deposited during a period of widespread basin deepening. Algal-rich type I oil prone kerogens dominate in north and generate oil, with very little gas. These shales mature at much lower temperatures than the mixed type I and III kerogens in the south, which also generate much larger amounts of gas and oil, and at higher threshold temperatures. The variable kinetics, as well as rapid facies variations typical of rifts, provide challenges to high-grading and testing unconventional shale plays.
Extensive Rock Eval pyrolysis and source rock kinetic databases were combined with petrophysical analysis to determine log-based porosity and saturations and productive potential. Modified Passey techniques calibrated to NMR log porosities provide estimates of organic richness as well as maturity and shale oil saturation. Basin modeling using Trinity software provides probabilistic ranges of generated and expelled hydrocarbons to determine storage capacity. The modeled oil window storage capacity varies between 6 to 13 MMBOE/km2, comparable to the values observed in Eagle Ford and Barnett Shale plays, but in a rifted basin and not broad cratonic shelf deposits.
Excess pore pressure was modeled using the kinetics of kerogen-to-oil conversion, and is noted in some of the deeper wells in tight sandstones, but not confirmed in the undrilled grabens. These pressure-gradient maps, along with oil properties (viscosity and oil mass fractions) derived from the geochemical model, are used to compute the producibility index. Composited storage capacity and producibility index maps have high-graded potential pilot areas.
In contrast to cratonic shale plays such as the Bakken or Eagle Ford, rapid and substantial facies variations occur due to local input of clastics and variable turbidite geometries which form potential targets for horizontal drilling. Increasingly more detailed paleogeographic maps are highlighting both the challenge and potential of the rich source rock in this basin.
This paper will cover how geochemical, structural, paleogeographic, petrophysical and other data are being used to derisk unconventional potential in this rich and complex rift system. Learnings from future testing of the Barmer Basin shale plays will be important to understand how to develop shale plays in other lacustrine rift basins.
Hydrocarbon production can be hindered as a result of fluid-induced formation damage caused by shale damage (swelling, sloughing, or fines migrating) or chemical damage (insoluble residue or polymer buildup). The proper selection of completion and stimulation fluid with additives provides the leading approach to mitigate formation damage.
Formation-specific damage mechanisms were determined from formation core, drilled cuttings, and outcrop materials for more than 100 North American resource shale samples. The formation materials were characterized using x-ray diffraction (XRD), cation exchange capacity (CEC), swelling sensitivity time (SST), mechanical stability turbidity (MST), and column flow testing to determine mineralogy, fluid sensitivity, and dominate fluid-induced damage mechanism. Preventing formation damage is predominately achieved with cationic clay stabilization chemicals. The efficacy of numerous chemical additives prepared at the same activity but with varied molecular weights (MWs) from 0.1 to 1,200 kDa was evaluated on ultra-low permeability shale samples based on reducing the swelling, fines generation, and mechanical destabilization tendencies. The same treatment chemicals were evaluated for permanency, compatibility with anionic friction reducer (FR) polymers, and mobility within porous media to determine the ideal North American formation stabilization material.
Fines generation was determined to be the dominant fluid-induced damage mechanism for ultra-low permeability North American hydrocarbon-producing formations. Clay content for these active formations range from 1 to 70 wt%, with an average of 30 wt% and a CEC of 4.5 milli-equivalence (meq)/100 g, indicating that most North American formations have moderate fluid instability. Fluid sensitives found an average swelling damage comparable to a 1 wt% smectite sample (30 seconds) mass loss due to mild mechanical agitation similar to an illite sample (3.2 wt%/hr) and fines generation due to flow approximately half of an illite sample (6.22 mg/PV). Prevention of these fluid-induced formation damage effects was determined for cationic clay stabilization chemicals ranging in MWs from monovalent salt solutions to large polymeric materials. Highly mobile monovalent salts effectively prevent swelling and remain compatible in an anionic FR solution; however, these treatments are temporary and less effective for remediating fines generation. Increasing the MW of cationic treatments improves the performance in terms of preventing swelling, fines, and wash-off; however, if the MW increase is too large, the polymers reduce permeability and become incompatible with anionic FR polymers. There is a range of moderate MW materials that provides permanent protection against swelling and fines damage while remaining compatible with anionic FRs. This range of moderate MW cationic treatments is the optimal chemical additive for most North American formations, offering the most effective protection against prevalent fluid-induced formation damage mechanisms while preventing chemical damage.
The differences in thermal maturity of migrated, conventional oil vs in situ shale oil can be recognized by standard petroleum geochemical methods. Thermal maturity of the oils was assessed in this study using both gas chromatography (GC) and GC-mass spectrometry (GCMS), often referred to as biomarker analysis.
We present results from a vertical pilot well from the Midland Basin, drilled with water-based mud, which illustrate how to assess the thermal maturity of the oil in rock samples (extracted from cuttings, SWC and conventional core) from the source rock units (notably the Lower Spraberry and Wolfcamp A units here) and the sandstone in the Dean Formation. These results show that the oil in the Dean sandstone is more mature than the in situ shale oil in the Lower Spraberry and Wolfcamp A, and has therefore migrated in to the reservoir at the well location from a deeper, more mature source.
We are also able to recognize the presence of very local oil migration within source rock units into silty layers, which does not result in thermal maturity differences but does cause bulk compositional fractionation, reflected in the SARA data.
We will then use this approach on two produced oils from nearby lateral wells in the Lower Spraberry and Wolfcamp A, to show that they are similar in thermal maturity, whereas the extracted oil from the Lower Spraberry is less mature than the produced oil from the same unit. We infer therefore that the Lower Spraberry produced oil contains a contribution from a more mature source than the in situ oil, and the most likely source for this more mature oil is the underlying Dean sandstone.
Petroleum geochemical methods demonstrate that so-called unconventional source rock, or shale, oils may be a mixture of both conventional migrated oil in sandstone and in situ oil in source rocks, due to the fractures penetrating into the sandstone. This may also help to explain bulk properties, such as API gravity, of the oils.
A significant amount of oil is trapped within organic nanopores of shale that cannot be recovered by primary production from these resources. The main reason for the large unrecovered oil volumes in shale reservoirs is the presence of nanoscale pore sizes, which leads to extremely small permeability values, and trapping of hydrocarbons in the adsorbed state on the surfaces the pores. For these resources, effective enhanced oil recovery (EOR) techniques are required to displace oil from nanoscale shale matrix. Due to small permeability, it is difficult, if not impossible, to conduct water and chemical flooding in these resources. Maintaining a stable flood front in immiscible gas flooding is challenging due to the severe fingering phenomenon cause by the naturally fractured nature of these formations. Gas huff-n-puff becomes the most suitable EOR method in shale reservoir development. For decades, carbon dioxide EOR techniques have been successfully applied in conventional reservoirs to improve oil production. In this work, the physics behind CO2 injection into organic nanopores of shale is investigated using molecular dynamics simulations. A 3D kerogen nanochannel, based on the kerogen unit molecules prepared by Ungerer et al. (2014), is created along with a synthetic oil mixture created based on the experimental study of phase behavior of petroleum mixtures performed by Turek et al. (1984). Supercritical CO2 (sCO2) is then injected into the channel at different pressures and oil recovery factors are computed. Results of this study demonstrates that the C7+ component of the oil sample have higher adsorption tendency than lighter hydrocarbon components. Furthermore, it is shown that sCO2 could potentially produce oil, especially lighter components, from organic matters in shale oil reservoirs. It is observed that as sCO2 injection pressure increases, the required soaking time for maximum process performance increases.
Trapped within organic shale matrix, there is a large amount of unrecovered oil content that cannot be removed through primary depletion due to the nanoscale pore size and extremely low permeability values for unconventional reservoirs. Due to these microscopic pore sizes and low permeability values, effective and efficient enhanced oil recovery (EOR) techniques are vital to increasing the oil recovery factor (Kazemi and Takbiri-Borujeni, 2015). Given the properties of these unconventional shale reservoirs, water and chemical flooding become very difficult to perform. Severe fingering effects also take place throughout the fractures within the matrix, adding to the difficulty of gas flooding procedures. With these challenges in mind, gas huff-n-puff becomes the most effective method of EOR for unconventional shale formations. In conventional reservoirs, CO2 gas injection has been a very successful method for increasing the oil production. Given the great success of CO2 gas injection in conventional reservoirs, recent studies have also shown the effectiveness CO2 injection throughout unconventional reservoirs because CO2 has a very high adsorption affinity to the walls of the porous organic matter found within shale gas reservoirs (Jin et al., 2017). By changing the fluid-fluid, rock-fluid, and gaseous interactions by CO2 injection, larger amounts of hydrocarbons can be produced.