Layer | Fill | Outline |
---|
Map layers
Theme | Visible | Selectable | Appearance | Zoom Range (now: 0) |
---|
Fill | Stroke |
---|---|
Collaborating Authors
Results
Abstract Improving data quality during log preprocessing is an important task that can consume most of the time of the petrophysicist, with a high impact on the final interpretation. As part of the initiative to increase automation and homogeneity in the data completeness of logs in a field, we organized a systematic comparison of multiple regression models that provided successful predictions of wellbore logs. These approaches can be potentially valuable when extrapolating measurements available on a few wells to a more extensive set of wellbores, predicting low-quality data intervals, and increasing the availability of complete data sets. This study aims to compare the performance of three promising machine-learning (ML) methods when predicting one of the following curves: density, neutron porosity, and compressional slowness curves. We view the need to evaluate models that could provide answers even in the presence of multiple missing logs or logs with alteration, which is a common scenario in petrophysics. Because of that, we built a comparison based on three ML methods that can handle those issues: window-based convolutional neural network autoencoder (WAE), pointwise fully connected autoencoder (PAE), and eXtreme Gradient Boosting (XGBoost). We developed the PAE and WAE methods to handle challenging scenarios of interest, and we used the original implementation of XGBoost, which is already built to handle missing values. We compare the computational complexity, prediction errors [root mean square error (RMSE) and mean absolute error (MAE)], Pearson’s correlation, peak signal-to-noise ratio (PSNR), and the visual analysis of both high- and low-scale feature reconstruction, conducting the comparison in two field data sets. We also used the same methods to predict photoelectric factors and interpreted formation properties such as total organic content in multiple field data sets.
- Europe (0.93)
- Asia (0.68)
- North America > United States > California (0.46)
- North America > United States > Texas (0.28)
- Geophysics > Seismic Surveying (1.00)
- Geophysics > Borehole Geophysics (1.00)
- Europe > Netherlands > Groningen > Southern North Sea - Anglo Dutch Basin > Groningen License > Groningen Field > Upper Rotliegend Formation (0.90)
- Europe > Netherlands > Groningen > Southern North Sea - Anglo Dutch Basin > Groningen License > Groningen Field > Limburg Formation (0.90)
- North America > United States > South Dakota > Williston Basin > Bakken Shale Formation (0.89)
- (2 more...)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Open hole/cased hole log analysis (1.00)
- Data Science & Engineering Analytics > Information Management and Systems > Neural networks (1.00)
- Data Science & Engineering Analytics > Information Management and Systems > Artificial intelligence (1.00)
Abstract Supervised learning algorithms can be employed for the automation of time-intensive tasks, such as image-based rock classification. However, labeled data are not always available. Alternatively, unsupervised learning algorithms, which do not require labeled data, can be employed. Using either of these methods depends on the evaluated formations and the available training/input data sets. Therefore, further investigation is needed to compare the performance of both approaches. The objectives of this paper are (a) to train two supervised learning models for image-based rock classification employing image-based features from computerized tomography (CT) scan images and core photos, (b) to conduct image-based rock classification using the trained model, (c) to compare the results obtained using supervised learning models against an unsupervised learning-based workflow for rock classification, and (d) to derive class-based petrophysical models for improved estimation of petrophysical properties. First, we removed non-formation visual elements from the core image data, such as induced fractures, the core barrel, and the seal peel tag on core photos. Then, we computed image-based features such as grayscale, color, and textural features from core image data and conducted feature selection. Then, we employed the extracted features for model training. Finally, we used the trained model to conduct rock classification and compared the obtained rock classes against the results obtained from an unsupervised image-based rock classification workflow. This workflow uses image-based rock fabric features coupled with a physics-based cost function for the optimization of rock classes. We applied the workflow to one well intersecting three formations with rapid spatial variation in rock fabric. We used 60% of the data to train a random forest and a support vector machines classifier using a 5-fold cross-validation approach. The remaining 40% of the data was used to test the accuracy of the supervised models. We established a base case of unsupervised learning rock classification and four different cases of supervised learning rock classification. The highest accuracy obtained for supervised rock classification was 97.4%. The accuracy obtained in the unsupervised learning rock classification approach was 82.7% when compared against expert-derived lithofacies. Class-based permeability estimates decreased the mean relative error by 34% and 35% when compared with formation-based permeability estimates, for the supervised and unsupervised approaches, respectively. The highest accuracies for the supervised and unsupervised models were obtained when integrating features from CT-scan images and core photos, highlighting the importance of feature selection for machine-learning workflows. A comparison of the two approaches for rock classification showed higher accuracy obtained from the supervised learning approach. However, the unsupervised method provided reasonable accuracy as well as a more general and faster approach for rock classification and enhanced formation evaluation.
- North America > United States > Texas (0.69)
- Asia (0.67)
- Geophysics > Borehole Geophysics (1.00)
- Geophysics > Seismic Surveying (0.93)
- North America > United States > Texas > Permian Basin > Yeso Formation (0.99)
- North America > United States > Texas > Permian Basin > Yates Formation (0.99)
- North America > United States > Texas > Permian Basin > Wolfcamp Formation (0.99)
- (22 more...)
Abstract Due to the complexity of lithologies and pore types, the permeability calculation of complex carbonate reservoirs has always been a difficult problem. To accurately calculate the permeability of complex carbonate reservoirs, a data mining technique is introduced. The technical process of data mining is established and divided into seven steps: data warehousing, data preprocessing, classification of reservoir types, selection of sensitive parameters, establishment of the classification model, evaluation of classification model, and application of classification model. The data-driven method can find effective knowledge that conventional reservoir evaluation methods cannot recognize and that are still contained in oil and gas data. Since the data-driven method may acquire a large amount of invalid knowledge while obtaining effective knowledge, the domain knowledge needs to be introduced to participate in the data mining process. The domain-knowledge-driven method can extract the most valuable and effective information from oil and gas data. The combination of data-driven and domain knowledge-driven methods is possible to avoid subdividing lithologies and pore types of complex carbonate reservoirs. As a result, the permeability of complex carbonate reservoirs can be accurately calculated based on the combination of data-driven and domain-knowledge-driven methods. Compared with the permeability calculation result by the previous method, the accuracy of the permeability calculation result by the data mining technique is improved by 18.39%. The combination of data-driven and domain-knowledge- driven methods can solve the difficult problem that traditional reservoir evaluation methods cannot overcome. Additionally, they can also provide new theories and techniques for reservoir evaluation. The permeability calculation result proves the feasibility and correctness of the method.
- Asia > China (0.29)
- Asia > Middle East (0.28)
- North America > United States > Texas (0.28)
- North America > United States > Texas > Permian Basin > Yeso Formation (0.99)
- North America > United States > Texas > Permian Basin > Yates Formation (0.99)
- North America > United States > Texas > Permian Basin > Wolfcamp Formation (0.99)
- (21 more...)
Mechanical rock properties are often determined using sonic log data—compressional velocity (Vp) and shear velocity (Vs). However, a sonic well log is not always acquired due to deteriorated hole condition (i.e., hole washout), sonic tool failures, especially in high-pressure, high-temperature (HPHT) wells, and relatively high cost. This paper introduces two data-driven models, namely artificial neural network (ANN) and random forest (RF), to estimate Vp and Vs across different formations that are characterized by deep burial depth and strong heterogeneity. Two types of actual field data were used to develop the models: (i) drilling surface parameters, which include flow rate, standpipe pressure, rotary speed, and surface torque, and (ii) acoustic velocities Vp and Vs, which were acquired by a conventional sonic log. Well-1 and Well-2 with data points of 6,846 were used to develop the models, while Well-3 with 1,016 data points was used to evaluate the capability of the developed models to generalize on an unseen data set with different statistical behavior. Furthermore, Well-3 was used to compare the accuracy of the developed models with the earliest published correlations in estimating the Vs. The results showed that the RF outperformed the optimized ANN in estimating Vp and Vs in Well-3. The RF predicted the Vp with a low average absolute percentage error (AAPE) of 0.9% and correlation of coefficient (R) of 0.87, while the AAPE and R were 6.7 % and 0.45 in the case of ANN. Similarly, the RF estimated the Vs with an AAPE of 1.1% and R of 0.85, whereas the ANN predicted the Vs with an AAPE of 9.5% and R of 0.40. Furthermore, the RF was the most accurate in determining Vs in Well-3 compared to the earliest published correlations.
- North America > United States (0.68)
- Europe (0.67)
- Asia > Middle East > Saudi Arabia > Eastern Province (0.46)
- Overview (0.67)
- Research Report > New Finding (0.66)
- Geology > Geological Subdiscipline > Geomechanics (1.00)
- Geology > Rock Type > Sedimentary Rock > Clastic Rock > Mudrock (0.46)
- Geophysics > Borehole Geophysics (1.00)
- Geophysics > Seismic Surveying > Seismic Processing (0.67)
- Geophysics > Seismic Surveying > Seismic Modeling > Velocity Modeling (0.35)
- Oceania > Australia > Western Australia > Carnarvon Basin (0.99)
- Asia > Middle East > Saudi Arabia > Eastern Province > Al-Ahsa Governorate > Arabian Basin > Widyan Basin > Ghawar Field > Lower Fadhili Formation (0.99)
- Asia > Middle East > Saudi Arabia > Eastern Province > Al-Ahsa Governorate > Arabian Basin > Widyan Basin > Ghawar Field > Khuff D Formation (0.99)
- (9 more...)
A Comparative Study of Three Supervised Machine-Learning Algorithms for Classifying Carbonate Vuggy Facies in the Kansas Arbuckle Formation
Deng, Tianqi (The University of Texas at Austin) | Xu, Chicheng (Aramco Services Company) | Jobe, Dawn (Consultant) | Xu, Rui (The University of Texas at Austin)
Diagenetic features, such as vugs, fractures and dolomite bodies can have significant impacts on carbonate reservoir quality. Challenges remain in characterizing these diagenetic features from well logs, as they are often mixed with changes in mineral and fluid concentrations. In this paper, a data-driven approach is developed to classify vuggy facies based on core and well logs from a key well penetrating the Arbuckle formation in Kansas. Three supervised machine-learning methods, namely artificial neural networks (ANN), support vector machines (SVM), and random forests (RF), are compared for their accuracy, stability, and computational efficiency. Hyperparameters are tuned using cross-validation and Bayesian optimization. Different feature selection methods and data labeling schemes are also evaluated to optimize the prediction. Results indicate predicting a binary classification (vuggy/nonvuggy) presents an ~80% accuracy, compared to a ~65% accuracy using a five-class vug-size-based classification label. A direct input of well logs as training features is recommended instead of using derived Petrophysical properties. Among the three machine-learning algorithms, ANN outperforms the other two methods for vug/nonvug detection, whereas for vug-size classification, RF is the best algorithm to apply. This work also suggests RF shows the least sensitivity to hyperparameters (i.e., maximum number of splits and minimum leaf sizes) according to the response surfaces constructed via Bayesian optimization. For the dataset used in this study, SVM is the most computationally efficient algorithm.
- Asia (1.00)
- North America > United States > Kansas (0.62)
- North America > United States > Texas > Harris County > Houston (0.46)
- Geology > Geological Subdiscipline (1.00)
- Geology > Rock Type > Sedimentary Rock > Carbonate Rock > Dolomite (0.48)
- North America > United States > Texas > Permian Basin > Yeso Formation (0.99)
- North America > United States > Texas > Permian Basin > Yates Formation (0.99)
- North America > United States > Texas > Permian Basin > Wolfcamp Formation (0.99)
- (25 more...)
ABSTRACT Driller's depth has always been the reference measurement used when logging while drilling (LWD), calculated using the sum of pipe-strap measured while the pipe is on surface. However, environmental corrections must be applied to the driller's depth to account for the dynamic changes the pipe undergoes while in the borehole. These include dynamic mechanical changes due to drilling activities, changes in the wellbore profile, torque, drag, friction factor, and temperature. They all result in LWD depth being shallower than the absolute depth. Understanding the position objectives, assigning the depth uncertainty and environmental modeling to predict the magnitude of depth correction should be applied in advance, along with the surveying technique to be used. These are critical components in the prejob analysis to ensure the position objectives can be met prior to drilling. Over the years, depth correction has successfully been run on several projects around the world. Environmental corrections were applied in various applications to improve the accuracy of the depth and demonstrate the significance of the correction for reservoir development. The intention of this paper is to raise awareness of the impact of LWD depth errors and existing approaches for correcting them. Case studies are presented to demonstrate the benefits derivable from applying depth corrections. In one case, the placement of the pressure and sample points provided the most accurate TVD possible. In another case, the corrected measurement enabled determination of where to set the casing depth to within the expected rat hole. Applying depth correction allows for accurate mapping of the geological markers, reservoir tops, sand continuity, and fluid contacts as well as for setting casing and other drilling applications in offshore deepwater and extended-reach drilling worldwide.
- Europe (0.69)
- North America > United States > Texas > Harris County > Houston (0.28)
Calculating the Total Porosity of Shale Reservoirs by Combining Conventional Logging and Elemental Logging to Eliminate the Effects of Gas Saturation
Zhu, Linqi (Yangtze University) | Zhang, Chong (Yangtze University) | Guo, Cong (Yangtze University) | Jiao, Yifeng (Yangtze University) | Chen, Lie (Yangtze University) | Zhou, Xueqing (Yangtze University) | Zhang, Chaomo (Yangtze University) | Zhang, Zhansong (Yangtze University)
Abstract Shale reservoir exploration technology has attracted increasing attention, and total porosity is a parameter that characterizes the shale storage. Due to the complexity of mineral components and the large variety of pore types, the evaluation accuracy of total porosity of shale reservoirs is not satisfactory, at present. To address this problem, this paper proposes an evaluation method for shale reservoir total porosity based on a shale petrophysical model. We first established the petrophysical model for the calculation of total porosity and then eliminated the effect of gas saturation in the petrophysical model by combining density and neutron-porosity logging. After that, evaluations of matrix density, matrix neutron porosity, and organic matter were conducted using a combined method of elemental logging and conventional logging. Finally, the total porosity of the shale reservoir was calculated. The calculation results showed that by using the elemental logging method and based on actual conditions in the research area, the shale mineral composition could be obtained, and an accurate evaluation of matrix neutron porosity and matrix density could be realized. The total organic carbon (TOC) and organic matter (OM) in the shale reservoir can be accurately calculated according to conventional logging data. The evaluation accuracy of total porosity by this method was high, wherein the predicted relative error was only 0.4. Moreover, based on theoretical deduction, it is known that the proposed method has high applicability for shale reservoirs. If the inversion effect of matrix minerals can be guaranteed, an accurate calculation of shale total porosity can be obtained. In summary, the proposed method can accurately calculate the total porosity of shale reservoirs, which provides a reference for the exploration and exploitation of shale reservoirs.
- Europe (0.93)
- North America > United States > Texas (0.46)
- Asia > China > Sichuan Province (0.29)
- (2 more...)
- Oceania > Australia > Queensland > Cooper Basin (0.99)
- North America > United States > West Virginia > Appalachian Basin > Marcellus Shale Formation (0.99)
- North America > United States > Virginia > Appalachian Basin > Marcellus Shale Formation (0.99)
- (42 more...)
Abstract In our business, accurate subsurface measurements are crucial. Depth is the most fundamental measurement made, tying together all the various along-hole measurements made and services provided. Logging-while-drilling (LWD) depths are based on driller's depths. Driller's depths have been plagued with accuracy issues, with numerous papers highlighting this. This paper combines a wireline-depth-determination methodology with driller's depth and shows how this can be used to arrive at a calibrated and corrected drillstring depth. The objective is to arrive at "true along-hole (TAH)" depth. Driller's depth measurement is based on drillstring length, typically as identified in the "tally book". Together with the dimensions of the bottomhole assembly (BHA), the measured pipe lengths represent the calibrated drillstring length. Similar to wireline correction, a way-point method is described for correcting the drillstring length for thermal elongation and elastic stretch. This allows a corrected depth to be defined for the bit and the associated LWD sensors. A crucial difference to conventional driller's depth is that the way-point method is applied during pull out of the hole (POOH). Most of the parameters that cause complications in driller's depth correction are mitigated when pulling out of hole. The correction elements of thermal elongation and elastic stretch are the only ones then applicable. The waypoint method described can provide corrections in wells with complex and extended-reach trajectories. The paper discusses how to arrive at an uncertainty so that measured depths are the TAH depth and that these TAH depths can be verified against repeat runs and wireline measured depth. Introduction Along-hole depth is the basis of all well construction, reservoir and field modeling and economic evaluations. Depth impacts all aspects of the oil and gas upstream subsurface activity. Driller's depth is derived from a composite record of the drilling activity as the drillstring moves into the well during the drilling process. The economic need for absolute and relative depth accuracy is not discussed in this paper, but it is clear that there are efficiency gains and asset value improvements with increased certainty and congruence of depth data.
- Europe (0.68)
- North America > United States > Texas (0.28)
- Well Drilling > Well Planning > Trajectory design (1.00)
- Well Drilling > Drilling Operations (1.00)
Abstract Three-dimensional (3D) printing is a unique technology that enables building of 3D pore-network proxies from digital models. Proxies allow us to experimentally test petrophysical properties (e.g., porosity and permeability) that can supplement reservoir rock analysis. In this study, we tested the resolution and accuracy of a polyjet 3D printer for generating rock proxies from a digital model of Berea sandstone. A 20×25-mm (length × diameter) cylindrical sandstone "macroplug" (21.6% porosity) and a smaller 3.5×4.0-mm "microplug" (21.3% porosity) were analyzed with mercury intrusion porosimetry and were scanned with computed tomography at 10 and 4 µm per voxel, respectively. A microplug digital model, with a porosity of 21.3%, a volume of 8 mm, and a modal pore-throat diameter of 18 µm, was extracted from tomographic data and rescaled at 10x magnification to meet the minimum pore resolution of the 3D printer (~132 µm). Proxies and core-plug samples were compared for porosity and pore-throat size distribution using two approaches:mercury porosimetry; and digital measurements from tomographic data. This comparison revealed a decrease in proxy porosity by ~2 percentage points and a decrease in pore-throat diameter by ~56 µm relative to natural samples. These discrepancies could arise due to insufficient magnification of the digital model or due to incomplete removal of the wax support material from the proxy pore space. Development of enhanced cleaning methods for pore space in polyjet proxies is needed to generate more accurate reservoir rock models. Introduction Multiscale reservoir characterization requires an understanding of a rock's mineralogical and textural characteristics as well as physical and chemical properties of fluids occupying its pore space. Pore geometry and topology also control key petrophysical properties such as porosity and permeability (Doyen, 1988; Bera et al., 2011; Peng et al., 2012). These properties determine reservoir quality with respect to the extraction of water and hydrocarbons, or the sequestration of carbon dioxide.
- North America > United States > Ohio (0.73)
- North America > United States > West Virginia (0.64)
- North America > United States > Pennsylvania (0.64)
- North America > United States > Kentucky (0.64)
- Geology > Mineral (0.94)
- Geology > Geological Subdiscipline > Geomechanics (0.93)
- Geology > Rock Type > Sedimentary Rock > Clastic Rock > Sandstone (0.86)
- Energy > Oil & Gas > Upstream (1.00)
- Materials > Chemicals > Commodity Chemicals > Petrochemicals (0.46)
New Method to Estimate Porosity More Accurately from NMR Data with Short Relaxation Times
Venkataramanan, Lalitha (Schlumberger Doll Research) | Gruber, Fred K. (GNS Healthcare) | LaVigne, Jack (Schlumberger Doll Research) | Habashy, Tarek M. (Schlumberger Doll Research) | Iglesias, Jorge G. (Schlumberger Doll Research) | Cohorn, Patrick (Bold Energy III LLC) | Anand, Vivek (Schlumberger Doll Research) | Rampurawala, Mansoor A. (Schlumberger Doll Research) | Jain, Vikas (Schlumberger Doll Research) | Heaton, Nick (Schlumberger Doll Research) | Akkurt, Ridvan (Schlumberger Doll Research) | Rylander, Erik (Schlumberger Doll Research) | Lewis, Rick (Schlumberger Doll Research)
Abstract In conventional oilfield applications of low-field nuclear magnetic resonance (NMR), data acquisition and analysis are optimal for T2 relaxation in the center of the spectrum, nominally between several milliseconds and several seconds. However, there are numerous applications where the measured magnetization data have short relaxation components, approaching or even below the time resolution of the downhole and/or laboratory measurement. Examples of these applications include heavy oil, organic– shale reservoirs and hydrocarbon and water in small pores. In these applications, the relaxation spectra of interest are typically a few milliseconds. Because the traditional algorithms used to analyze NMR data to estimate porosity and other petrophysical properties involving short relaxation times can be inaccurate, a new algorithm is proposed to improve the accuracy of these parameters. First, a T2 distribution is estimated from the measured magnetization data using traditional inverse–Laplace–transform (ILT) methods. Second, a porosity sensitivity curve is computed for a given pulse sequence and a set of acquisition and inversion parameters. Third, a correction factor is derived from this sensitivity curve and applied seamlessly as part of the inversion so that the overall porosity sensitivity is more uniform at short relaxation times to obtain a modified T2 distribution. The efficacy of the algorithm is illustrated by Monte Carlo simulations and application on two field examples from unconventional shale reservoirs. Prediction of porosity from NMR measurements is particularly useful in unconventional reservoirs for two reasons. First, NMR measurements provide a direct estimate of effective porosity without requiring detailed knowledge of the complex mineralogy typical of shale formations. Second, the deficit between effective porosity predicted from NMR, and total porosity predicted from nuclear logs can be used to obtain accurate estimates of petrophysical quantities, such as, the kerogen content in shales and the hydrogen index in heavy-oil formations. The two field examples are from reservoirs in the Wolfberry trend in the southwestern United States. Application of the new algorithm to NMR data in the first field example results in an increase of up to 10% in porosity in zones with T2<10 msec. The porosity predictions from the new algorithm show improved correlation with core measurements. In the second field example, the deficit between total porosity and effective porosity predicted from NMR T2 distributions using the new algorithm provides a more accurate estimate of the kerogen content.
- Asia (1.00)
- Europe (0.94)
- North America > United States > Texas > Fort Bend County (0.28)
- North America > United States > Texas > Permian Basin > Yeso Formation (0.99)
- North America > United States > Texas > Permian Basin > Yates Formation (0.99)
- North America > United States > Texas > Permian Basin > Wolfcamp Formation (0.99)
- (29 more...)