Pulsed-neutron logs are a staple of time-lapse monitoring programs for saline aquifer carbon capture and sequestration (CCS) projects and are unsurprisingly the most frequently run wireline logs in both injection and monitoring wells. While the emphasis imposed by government regulators and the focus of operators to date has been on the verification of CO2 containment, it is envisioned that a savvy interpretation of the multiple independent measurements should be able to unlock much greater value for the project than merely detecting the location of stored CO2. Recently introduced capabilities for novel measurements and improved environmental compensation should further increase the repeatability, interpretability, and value of these logs. We reviewed more than 30 time-lapse runs of pulsed-neutron logs acquired over a period of 15 years on three mature CCS projects using both previous- and new-generation pulsed-neutron tools, including measurements of formation sigma, hydrogen index, and fast neutron cross section. Special attention in processing is required when changes occur to the wellbore environment between runs, although this is mitigated by the improved environmental compensation scheme of the newer tool. We performed both standalone estimates of CO2 saturation from single-physics time-lapse measurements and simultaneous interpretation of multiple independent time-lapse measurements and studied the results side-by-side with openhole log interpretation, core analysis, and well test results from the evaluation phase. The apparent changes in saturation were framed within the context of the injection history and important events in the life of the wells. A first finding is that differences in apparent CO2 saturation between the various independent measurement physics of the pulsed-neutron tool are often reconcilable and may carry additional information about the state of the well or reservoir. With respect to verification of containment, depending on the well configuration, it may be possible to differentiate between CO2 in the formation and CO2 in the annulus. The interpreted CO2 saturation itself can have different significance depending on the timing of acquisition and the type of well. Measured at the right time, it is a direct in-situ measurement of formation CO2 storage efficiency. In other cases, the interpretation reveals formation dryout in the near-wellbore region of injection wells, a condition that may presage loss of injectivity. We now understand that it is important for operators to plan the timing and frequency of pulsed-neutron runs according to what they want to measure and not based solely on regulatory obligations. In a CCS project, time-lapse pulsed-neutron logs should be thought of as much more than simple indicators of the presence and migration of CO2. They give important information about migration pathways. They can also help to quantify essential uncertainties on reservoir performance that are difficult to ascertain d
{"title":"Time-Lapse Pulsed-Neutron Logs for Carbon Capture and Sequestration: Practical Learnings and Key Insights","authors":"Robert Laronga, Lee Swager, Ulises Bustos","doi":"10.30632/pjv64n5-2023a5","DOIUrl":"https://doi.org/10.30632/pjv64n5-2023a5","url":null,"abstract":"Pulsed-neutron logs are a staple of time-lapse monitoring programs for saline aquifer carbon capture and sequestration (CCS) projects and are unsurprisingly the most frequently run wireline logs in both injection and monitoring wells. While the emphasis imposed by government regulators and the focus of operators to date has been on the verification of CO2 containment, it is envisioned that a savvy interpretation of the multiple independent measurements should be able to unlock much greater value for the project than merely detecting the location of stored CO2. Recently introduced capabilities for novel measurements and improved environmental compensation should further increase the repeatability, interpretability, and value of these logs. We reviewed more than 30 time-lapse runs of pulsed-neutron logs acquired over a period of 15 years on three mature CCS projects using both previous- and new-generation pulsed-neutron tools, including measurements of formation sigma, hydrogen index, and fast neutron cross section. Special attention in processing is required when changes occur to the wellbore environment between runs, although this is mitigated by the improved environmental compensation scheme of the newer tool. We performed both standalone estimates of CO2 saturation from single-physics time-lapse measurements and simultaneous interpretation of multiple independent time-lapse measurements and studied the results side-by-side with openhole log interpretation, core analysis, and well test results from the evaluation phase. The apparent changes in saturation were framed within the context of the injection history and important events in the life of the wells. A first finding is that differences in apparent CO2 saturation between the various independent measurement physics of the pulsed-neutron tool are often reconcilable and may carry additional information about the state of the well or reservoir. With respect to verification of containment, depending on the well configuration, it may be possible to differentiate between CO2 in the formation and CO2 in the annulus. The interpreted CO2 saturation itself can have different significance depending on the timing of acquisition and the type of well. Measured at the right time, it is a direct in-situ measurement of formation CO2 storage efficiency. In other cases, the interpretation reveals formation dryout in the near-wellbore region of injection wells, a condition that may presage loss of injectivity. We now understand that it is important for operators to plan the timing and frequency of pulsed-neutron runs according to what they want to measure and not based solely on regulatory obligations. In a CCS project, time-lapse pulsed-neutron logs should be thought of as much more than simple indicators of the presence and migration of CO2. They give important information about migration pathways. They can also help to quantify essential uncertainties on reservoir performance that are difficult to ascertain d","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"153 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134931123","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
German Merletti, Michael Rabinovich, Salim Al Hajri, William Dawson, Russell Farmer, Joaquin Ambia, Carlos Torres-Verdín
A new iterative modeling workflow has been designed to reduce the uncertainty of water saturation (Sw) calculations in the tight Barik sandstone in the Sultanate of Oman. Results from this case study indicate that Sw can be overestimated by up to 20 s.u. if the as-acquired deep resistivity is used in volumetric calculations. Overbalanced drilling causes deep invasion of water-based mud (WBM) filtrate into porous and permeable rocks, leading to the radial displacement of in-situ saturating fluids away from the wellbore. In low-porosity reservoirs drilled with WBM, the inability of the filtration process to quickly build impermeable mudcake translates into long radial transition zones. Under certain reservoir and drilling conditions, deep resistivity logs cannot reliably measure true formation resistivity and are, therefore, unable to provide an accurate assessment of hydrocarbon saturation. The effect of mud-filtrate invasion on resistivity logs has been extensively documented. Processing techniques use resistivity inversion and tool-specific forward modeling to provide uninvaded formation resistivity logs, which are much better suited for in-place resource volume assessment. However, sensitivity analysis shows that the accuracy of invasion-corrected logs dramatically decreases as the depth of invasion increases, whereby the inversion process needs to be further constrained. The new workflow is designed to reduce the non-uniqueness of true formation resistivity models so that they honor multiple and independent petrophysical data. The inversion routine utilizes a Bayesian algorithm coupled with Markov-Chain Monte Carlo (MCMC) sampling. Inversion results are iteratively modified based on two rock property models : one derived from rock-core data (helium expansion porosity and Dean-Stark saturations) and the other using an equivalent log interpretation of thick reservoir intervals from oil-based mud (OBM) wells. Simulated borehole resistivity is compared to field logs after each validation loop against rock property models. The new inversion-based workflow is extensively tested in the unconventional tight Barik Formation across water-free hydrocarbon and perched water intervals, and inversion-derived Sw models are independently validated by capillary-pressure-derived saturation-height models and fluid inflow rate from production logs.
{"title":"New Iterative Resistivity Modeling Workflow Reduces Uncertainty in the Assessment of Water Saturation in Deeply Invaded Reservoirs","authors":"German Merletti, Michael Rabinovich, Salim Al Hajri, William Dawson, Russell Farmer, Joaquin Ambia, Carlos Torres-Verdín","doi":"10.30632/pjv64n4-2023a5","DOIUrl":"https://doi.org/10.30632/pjv64n4-2023a5","url":null,"abstract":"A new iterative modeling workflow has been designed to reduce the uncertainty of water saturation (Sw) calculations in the tight Barik sandstone in the Sultanate of Oman. Results from this case study indicate that Sw can be overestimated by up to 20 s.u. if the as-acquired deep resistivity is used in volumetric calculations. Overbalanced drilling causes deep invasion of water-based mud (WBM) filtrate into porous and permeable rocks, leading to the radial displacement of in-situ saturating fluids away from the wellbore. In low-porosity reservoirs drilled with WBM, the inability of the filtration process to quickly build impermeable mudcake translates into long radial transition zones. Under certain reservoir and drilling conditions, deep resistivity logs cannot reliably measure true formation resistivity and are, therefore, unable to provide an accurate assessment of hydrocarbon saturation. The effect of mud-filtrate invasion on resistivity logs has been extensively documented. Processing techniques use resistivity inversion and tool-specific forward modeling to provide uninvaded formation resistivity logs, which are much better suited for in-place resource volume assessment. However, sensitivity analysis shows that the accuracy of invasion-corrected logs dramatically decreases as the depth of invasion increases, whereby the inversion process needs to be further constrained. The new workflow is designed to reduce the non-uniqueness of true formation resistivity models so that they honor multiple and independent petrophysical data. The inversion routine utilizes a Bayesian algorithm coupled with Markov-Chain Monte Carlo (MCMC) sampling. Inversion results are iteratively modified based on two rock property models : one derived from rock-core data (helium expansion porosity and Dean-Stark saturations) and the other using an equivalent log interpretation of thick reservoir intervals from oil-based mud (OBM) wells. Simulated borehole resistivity is compared to field logs after each validation loop against rock property models. The new inversion-based workflow is extensively tested in the unconventional tight Barik Formation across water-free hydrocarbon and perched water intervals, and inversion-derived Sw models are independently validated by capillary-pressure-derived saturation-height models and fluid inflow rate from production logs.","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"135872899","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
German Merletti, Salim Al Hajri, Michael Rabinovich, Russell Farmer, Mohamed Bennis, Carlos Torres-Verdin
The process of mud-filtrate invasion involves immiscible fluid displacement and salt mixing between mud filtrate and formation fluids in porous and permeable rocks. Consequently, the post-invasion spatial distribution of fluids and electrolyte concentration around the borehole affects resistivity measurements with different depths of investigation (DOI). In the presence of deep mud-filtrate invasion, the assessment of water saturation in the uninvaded zone based on the deep resistivity log can be inaccurate. Deep and electrically conductive filtrate invasion coupled with shoulder-bed effects can artificially increase water saturation (Sw) estimations by 20 saturation units (s.u.) in the Barik reservoir, resulting in pessimistic estimates of hydrocarbon pore volume if no corrections are applied. The Barik sandstone reservoir, which is characterized by low porosity ( up to 14%), low-to-medium permeability (up to 40 md), and high residual gas saturation (40 to 50%), exhibits low storage capacity to admit the critical filtrate volume necessary for building an impermeable mudcake. Combined with multiple days of overbalanced exposure to saline water-based mud (WBM), mud-filtrate invasion results in deep and smooth radial transition zones where the uninvaded formation is far beyond the depth of investigation of laterolog tools. Deep resistivity values are, therefore, lower than the true formation resistivity. Additionally, numerical simulations of resistivity logs show that the resistivity reduction by conductive invasion is further aggravated by shoulder-bed effects when individual reservoir thickness falls below 2.5 m. This paper describes the implementation of a compositional fluid-flow simulator to numerically model WBM-filtrate invasion and mudcake buildup in vertical boreholes. The algorithm allows the simulation of physical dispersion and fluid displacement around the borehole in a multilayer model. Time-dependent radial profiles of Sw and salinity are combined with core-calibrated porosity and electrical properties to compute electrical resistivity via Archie’s formulation. Subsequently, numerically simulated logs are generated using vendor-specific forward model processing and compared against field measurements. This workflow was extensively tested in various reservoir intervals with a wide range of petrophysical rock types and drilling conditions. Results show that the deep laterolog exhibits low sensitivity to conductive filtrate invasion when reservoirs’ porosities are lower than 8%. Above that threshold value, invasion length is a nontrivial process involving multiple variables. Even though exposure time to openhole conditions is a key factor leading to deep invasion, certain reservoir characteristics can lead to deeper invasion at short exposure times and significantly increase uncorrected Sw estimates.
{"title":"Assessment of True Formation Resistivity and Water Saturation in Deeply Invaded Tight-Gas Sandstones Based on the Combined Numerical Simulation of Mud-Filtrate Invasion and Resistivity Logs","authors":"German Merletti, Salim Al Hajri, Michael Rabinovich, Russell Farmer, Mohamed Bennis, Carlos Torres-Verdin","doi":"10.30632/pjv64n4-2023a2","DOIUrl":"https://doi.org/10.30632/pjv64n4-2023a2","url":null,"abstract":"The process of mud-filtrate invasion involves immiscible fluid displacement and salt mixing between mud filtrate and formation fluids in porous and permeable rocks. Consequently, the post-invasion spatial distribution of fluids and electrolyte concentration around the borehole affects resistivity measurements with different depths of investigation (DOI). In the presence of deep mud-filtrate invasion, the assessment of water saturation in the uninvaded zone based on the deep resistivity log can be inaccurate. Deep and electrically conductive filtrate invasion coupled with shoulder-bed effects can artificially increase water saturation (Sw) estimations by 20 saturation units (s.u.) in the Barik reservoir, resulting in pessimistic estimates of hydrocarbon pore volume if no corrections are applied. The Barik sandstone reservoir, which is characterized by low porosity ( up to 14%), low-to-medium permeability (up to 40 md), and high residual gas saturation (40 to 50%), exhibits low storage capacity to admit the critical filtrate volume necessary for building an impermeable mudcake. Combined with multiple days of overbalanced exposure to saline water-based mud (WBM), mud-filtrate invasion results in deep and smooth radial transition zones where the uninvaded formation is far beyond the depth of investigation of laterolog tools. Deep resistivity values are, therefore, lower than the true formation resistivity. Additionally, numerical simulations of resistivity logs show that the resistivity reduction by conductive invasion is further aggravated by shoulder-bed effects when individual reservoir thickness falls below 2.5 m. This paper describes the implementation of a compositional fluid-flow simulator to numerically model WBM-filtrate invasion and mudcake buildup in vertical boreholes. The algorithm allows the simulation of physical dispersion and fluid displacement around the borehole in a multilayer model. Time-dependent radial profiles of Sw and salinity are combined with core-calibrated porosity and electrical properties to compute electrical resistivity via Archie’s formulation. Subsequently, numerically simulated logs are generated using vendor-specific forward model processing and compared against field measurements. This workflow was extensively tested in various reservoir intervals with a wide range of petrophysical rock types and drilling conditions. Results show that the deep laterolog exhibits low sensitivity to conductive filtrate invasion when reservoirs’ porosities are lower than 8%. Above that threshold value, invasion length is a nontrivial process involving multiple variables. Even though exposure time to openhole conditions is a key factor leading to deep invasion, certain reservoir characteristics can lead to deeper invasion at short exposure times and significantly increase uncorrected Sw estimates.","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134950193","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Complex lithology petrophysical interpretation with multiphysics logging tools has been and continues to be a major challenge in formation evaluation. Many currently used data-driven approaches, such as a neural network (NN), deliver predicted results in numerical quantities rather than analytical equations. It is more challenging if multiphysics logging measurements are collectively used to estimate a petrophysical parameter. To overcome these problems, a physics-guided, artificial intelligence (AI) machine-learning (ML) method for petrophysical interpretation model development is described. The workflow consists of the following five constituents: (1) statistical tools such as correlation heatmaps are employed to select the best candidate input variables for the target petrophysical equations; (2) a genetic programming-based symbolic regression approach is used to fuse multiphysics measurements data for training the petrophysical prediction equations; (3) an optional ensemble modeling procedure is applied for maximally utilizing all available training data by integrating multiple instances of prediction equations objectively, which is especially useful for a small training data set; (4) a means of obtaining conditional branching in prediction equations is enabled in symbolic regression to handle certain formation heterogeneity; and (5) a model discrimination framework is introduced to finalize the model selection based on mathematical complexity, physics complexity, and model performance. The efficacy of the five-constituents petrophysical interpretation development process is demonstrated on a data set collected from six wells with the goal of obtaining formation resistivity factor (F) and permeability (k) equations for heterogeneous carbonate reservoirs. We show quantitatively how individual constituents of the workflow improve the model performance with two error metrics. A comparison of NN-method-predicted permeability values vs. SR-based-workflow-predicted permeability equation is included to showcase many advantages of the latter. Beyond the transparency of an analytical form of the prediction equations, the SR method intrinsically has a more relaxed requirement on the training data size, is less prone to overfitting, yet can deliver superior model performance rival to the NN approach.
{"title":"Use of Symbolic Regression for Developing Petrophysical Interpretation Models","authors":"Songhua Chen, Wei Shao, Huiwen Sheng, Hyung Kwak","doi":"10.30632/pjv64n2-2023a3","DOIUrl":"https://doi.org/10.30632/pjv64n2-2023a3","url":null,"abstract":"Complex lithology petrophysical interpretation with multiphysics logging tools has been and continues to be a major challenge in formation evaluation. Many currently used data-driven approaches, such as a neural network (NN), deliver predicted results in numerical quantities rather than analytical equations. It is more challenging if multiphysics logging measurements are collectively used to estimate a petrophysical parameter. To overcome these problems, a physics-guided, artificial intelligence (AI) machine-learning (ML) method for petrophysical interpretation model development is described. The workflow consists of the following five constituents: (1) statistical tools such as correlation heatmaps are employed to select the best candidate input variables for the target petrophysical equations; (2) a genetic programming-based symbolic regression approach is used to fuse multiphysics measurements data for training the petrophysical prediction equations; (3) an optional ensemble modeling procedure is applied for maximally utilizing all available training data by integrating multiple instances of prediction equations objectively, which is especially useful for a small training data set; (4) a means of obtaining conditional branching in prediction equations is enabled in symbolic regression to handle certain formation heterogeneity; and (5) a model discrimination framework is introduced to finalize the model selection based on mathematical complexity, physics complexity, and model performance. The efficacy of the five-constituents petrophysical interpretation development process is demonstrated on a data set collected from six wells with the goal of obtaining formation resistivity factor (F) and permeability (k) equations for heterogeneous carbonate reservoirs. We show quantitatively how individual constituents of the workflow improve the model performance with two error metrics. A comparison of NN-method-predicted permeability values vs. SR-based-workflow-predicted permeability equation is included to showcase many advantages of the latter. Beyond the transparency of an analytical form of the prediction equations, the SR method intrinsically has a more relaxed requirement on the training data size, is less prone to overfitting, yet can deliver superior model performance rival to the NN approach.","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134946665","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Vanessa Simoes, Hiren Maniar, Aria Abubakar, Tao Zhao
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.
{"title":"Comparative Study of Machine-Learning-Based Methods for Log Prediction","authors":"Vanessa Simoes, Hiren Maniar, Aria Abubakar, Tao Zhao","doi":"10.30632/pjv64n2-2023a4","DOIUrl":"https://doi.org/10.30632/pjv64n2-2023a4","url":null,"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.","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134946803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
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
{"title":"Data-Driven Algorithms for Image-Based Rock Classification and Formation Evaluation in Formations With Rapid Spatial Variation in Rock Fabric","authors":"Andres Gonzalez, Zoya Heidari, Oliver Lopez","doi":"10.30632/pjv64n2-2023a2","DOIUrl":"https://doi.org/10.30632/pjv64n2-2023a2","url":null,"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 ","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134946804","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
C. Cavalleri, Schlumberger, G. Brouwer, Dimas Kodri, D. Rose, J. Brinks, B V Nederlandse Aardolie Maatschappij
Casedhole logging for formation evaluation and input to determine the redevelopment potential of an oil producer with a challenging production history was conducted. This included an intelligent assessment of formation gas pressure through casing, which was later confirmed by perforating. The target reservoirs are Triassic sandstones drilled as a gas exploration prospect. Based on openhole log data, the prospect appeared to be oil bearing. The well has been producing oil for several years and is now a candidate for a gas cap blowdown. The presence of heterogeneous layers with varied rock quality and producibility indexes coupled to complexity in fluids distribution and zonal isolation issues complicates the development process and ability to optimize recovery from any contributing level. Recently, a new-generation casedhole formation evaluation tool that provides multiple independent formation property measurements was deployed to enhance knowledge of the formation parameters while describing the current gas and oil volumes. Sigma, neutron porosity, fast-neutron cross section (FNXS), and elemental concentrations, including total organic carbon from inelastic and capture spectroscopy, were simultaneously recorded. Because the well is highly deviated in the zones of interest, the tool was efficiently conveyed on wireline using tractor technology. The evaluation techniques used to study this rich set of data reveal several pieces of information that are essential to the petrophysicists and geologists, and to the reservoir and production engineers. A multimineral solver analysis guided by the prior knowledge of the rocks using cores from offset wells was conducted to quantify the porosity and gas-oil contact levels while giving access to detailed knowledge of matrix and rock composition for refining the reservoir models. Additionally, a novel method to determine gas pressure at the current time from the casedhole log measurements was applied to support reservoir management. The highly sensitive sigma, neutron porosity, and FNXS gas properties can be parameterized as a function of pressure and temperature if the formation and fluid properties are known. This is a well-established principle that can finally be applied independently and directly to multiple measurements. The computations are done independently and checked against each other for consistency and to support optimal parameter setting, in an iterative manner. This is particularly important in this scenario where the complexity of the wellbore environment and history of the well could have complicated the ability to achieve enough precision on the estimated pressures when coming from a single method or if modeling is required. The log results, with validation, and implications for the well redevelopment are presented together with a general discussion on the methodology and applicability based on this well experience. The importance of meticulous job preparation, prejob modeling, and data qua
{"title":"Maximizing the Value of Pulsed-Neutron Logs: A Complex Case Study of Gas Pressure Assessment Through Casing","authors":"C. Cavalleri, Schlumberger, G. Brouwer, Dimas Kodri, D. Rose, J. Brinks, B V Nederlandse Aardolie Maatschappij","doi":"10.30632/pjv61n6-2020a6","DOIUrl":"https://doi.org/10.30632/pjv61n6-2020a6","url":null,"abstract":"Casedhole logging for formation evaluation and input to determine the redevelopment potential of an oil producer with a challenging production history was conducted. This included an intelligent assessment of formation gas pressure through casing, which was later confirmed by perforating. The target reservoirs are Triassic sandstones drilled as a gas exploration prospect. Based on openhole log data, the prospect appeared to be oil bearing. The well has been producing oil for several years and is now a candidate for a gas cap blowdown. The presence of heterogeneous layers with varied rock quality and producibility indexes coupled to complexity in fluids distribution and zonal isolation issues complicates the development process and ability to optimize recovery from any contributing level. Recently, a new-generation casedhole formation evaluation tool that provides multiple independent formation property measurements was deployed to enhance knowledge of the formation parameters while describing the current gas and oil volumes. Sigma, neutron porosity, fast-neutron cross section (FNXS), and elemental concentrations, including total organic carbon from inelastic and capture spectroscopy, were simultaneously recorded. Because the well is highly deviated in the zones of interest, the tool was efficiently conveyed on wireline using tractor technology. The evaluation techniques used to study this rich set of data reveal several pieces of information that are essential to the petrophysicists and geologists, and to the reservoir and production engineers. A multimineral solver analysis guided by the prior knowledge of the rocks using cores from offset wells was conducted to quantify the porosity and gas-oil contact levels while giving access to detailed knowledge of matrix and rock composition for refining the reservoir models. Additionally, a novel method to determine gas pressure at the current time from the casedhole log measurements was applied to support reservoir management. The highly sensitive sigma, neutron porosity, and FNXS gas properties can be parameterized as a function of pressure and temperature if the formation and fluid properties are known. This is a well-established principle that can finally be applied independently and directly to multiple measurements. The computations are done independently and checked against each other for consistency and to support optimal parameter setting, in an iterative manner. This is particularly important in this scenario where the complexity of the wellbore environment and history of the well could have complicated the ability to achieve enough precision on the estimated pressures when coming from a single method or if modeling is required. The log results, with validation, and implications for the well redevelopment are presented together with a general discussion on the methodology and applicability based on this well experience. The importance of meticulous job preparation, prejob modeling, and data qua","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"61 1","pages":"610-622"},"PeriodicalIF":0.9,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47979348","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
T. Zhou, Schlumberger, D. Rose, J. Miles, J. Gendur, Haijing Wang, M. Sullivan, Chevron Canada Resources
Formation elemental composition and mineralogy measurements, including organic carbon from recently developed spectroscopy tools, provide critical information for formation evaluation in both conventional and unconventional reservoirs. These measurements can be obtained under conditions by using a slim pulsed-neutron tool with two spectroscopy detectors. One primary limitation is that users must manually provide offsets for the elements (silicon (Si), calcium (Ca), and iron (Fe)) present in casing and cement before performing the oxide closure computation to obtain elemental concentrations. This process is time consuming, and the results could be inaccurate and subjective, especially without any local reference. Another limitation is that the formation element signals are smaller in cased hole than in open hole. This increases the noise in the oxide closure-derived environmental yield-to-weight normalization factor (FY2W), which is propagated to all the elemental weight fractions. A self-compensated spectroscopy algorithm was developed to overcome these two limitations. The key breakthrough is the use of raw measurements with very high precision from the two spectroscopy detectors to predict FY2Ws instead of using the oxide closure or inelastic capture (INCP) closure methods. The capture FY2W is mainly determined by the borehole and formation sigma. It can be characterized by using multiple measured apparent sigma values in different timing gates from multiple detectors, which have different sensitivities to borehole and formation sigma. The inelastic FY2W is mainly determined by the borehole and formation geometry and hydrogen index. It can be characterized by using count rate ratios in both burst-on (inelastic) and burst-off (capture) timing gates from multiple detectors. This method reduces the noise in the FY2Ws by an order of magnitude, which improves the precision of all the final elemental weight fractions. Two independent sets of apparent elemental weight fractions can be calculated from the two spectroscopy detectors. The measured elements from the detector with shorter spacing are more sensitive to the borehole environment, including the casing and cement, whereas the ones from the detector farther away are more sensitive to the formation. This enables self-compensation for casing and cement effects. The new processing can be done without user intervention and results in a more accurate, more precise, and less subjective elemental composition and mineralogy. More than 1,600 laboratory measurements in different conditions were used to characterize the algorithm. Several log examples demonstrate the excellent performance of the new compensated spectroscopy measurements.
{"title":"Self-Compensated Pulsed-Neutron Spectroscopy Measurements","authors":"T. Zhou, Schlumberger, D. Rose, J. Miles, J. Gendur, Haijing Wang, M. Sullivan, Chevron Canada Resources","doi":"10.30632/pjv61n6-2020a3","DOIUrl":"https://doi.org/10.30632/pjv61n6-2020a3","url":null,"abstract":"Formation elemental composition and mineralogy measurements, including organic carbon from recently developed spectroscopy tools, provide critical information for formation evaluation in both conventional and unconventional reservoirs. These measurements can be obtained under conditions by using a slim pulsed-neutron tool with two spectroscopy detectors. One primary limitation is that users must manually provide offsets for the elements (silicon (Si), calcium (Ca), and iron (Fe)) present in casing and cement before performing the oxide closure computation to obtain elemental concentrations. This process is time consuming, and the results could be inaccurate and subjective, especially without any local reference. Another limitation is that the formation element signals are smaller in cased hole than in open hole. This increases the noise in the oxide closure-derived environmental yield-to-weight normalization factor (FY2W), which is propagated to all the elemental weight fractions. A self-compensated spectroscopy algorithm was developed to overcome these two limitations. The key breakthrough is the use of raw measurements with very high precision from the two spectroscopy detectors to predict FY2Ws instead of using the oxide closure or inelastic capture (INCP) closure methods. The capture FY2W is mainly determined by the borehole and formation sigma. It can be characterized by using multiple measured apparent sigma values in different timing gates from multiple detectors, which have different sensitivities to borehole and formation sigma. The inelastic FY2W is mainly determined by the borehole and formation geometry and hydrogen index. It can be characterized by using count rate ratios in both burst-on (inelastic) and burst-off (capture) timing gates from multiple detectors. This method reduces the noise in the FY2Ws by an order of magnitude, which improves the precision of all the final elemental weight fractions. Two independent sets of apparent elemental weight fractions can be calculated from the two spectroscopy detectors. The measured elements from the detector with shorter spacing are more sensitive to the borehole environment, including the casing and cement, whereas the ones from the detector farther away are more sensitive to the formation. This enables self-compensation for casing and cement effects. The new processing can be done without user intervention and results in a more accurate, more precise, and less subjective elemental composition and mineralogy. More than 1,600 laboratory measurements in different conditions were used to characterize the algorithm. Several log examples demonstrate the excellent performance of the new compensated spectroscopy measurements.","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"61 1","pages":"570-584"},"PeriodicalIF":0.9,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45168276","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. A. Wijaya, Rama Aulianagara, Weijun Guo, Fetty Maria Naibaho, Fransiscus Xaverius Asriwan, Usman Amirudin, Pertamina Hulu Sanga-Sanga
In mature fields, pulsed-neutron logging is commonly used to solve for the remaining saturation behind the casing. For years, sigma-based saturation has been used to calculate gas saturation behind casing; however, the high dependency of sigma-to-water salinity of the formation, especially the low-dynamic range at porosity near 12 p.u., has proven to be challenging in low-porosity gas rock. A new measurement from the third detector from a multidetector pulsed-neutron tool (MDPNT) is proposed to provide a better estimation of the gas saturation in a low-porosity reservoir. Two sets of independently measured sigma and the third detector were taken in a casedhole well, with a dual-tubing system of a long string and short string. For the third-detector measurement, the measurement was based on the ratio of the slow capture gate and inelastic gate component from the decay curve created by the long detector. This ratio can be used to detect gas in a tight reservoir with a minimum salinity and lithology effect. This data will then be used to calculate the gas saturation from the third detector, and the result is compared to sigma-based gas saturation. At an interval where the porosity is above 12 p.u., the sigma-based gas saturation and MDPNT-based gas saturation are very much in agreement. However, in a low-porosity reservoir near 12 p.u. or below, the sigma-based measurement starts to show its limitation. Meanwhile, the MDPNT-based gas saturation clearly shows the remaining gas saturation where sigma-based measurements failed to detect it. The subsequent decision was made based on the log analysis result, and perforation was done at a potential interval based on the MDPNT result. The results from the production test confirm the MDPNT-based gas saturation with 700-Mscf/d gas production added. This study showcases a new technology to solve a low-porosity gas reservoir issue where a sigma-based measurement underestimates the remaining gas saturation. Using two different measurements in the same well, the results from the MDPNT measurement demonstrated a better result compared to the sigma-based measurement in low-porosity rock
{"title":"Multidetector Pulsed-Neutron Tool Application in Low-Porosity Reservoir–A Case Study in Mutiara Field, Indonesia","authors":"A. A. Wijaya, Rama Aulianagara, Weijun Guo, Fetty Maria Naibaho, Fransiscus Xaverius Asriwan, Usman Amirudin, Pertamina Hulu Sanga-Sanga","doi":"10.30632/pjv61n6-2020a7","DOIUrl":"https://doi.org/10.30632/pjv61n6-2020a7","url":null,"abstract":"In mature fields, pulsed-neutron logging is commonly used to solve for the remaining saturation behind the casing. For years, sigma-based saturation has been used to calculate gas saturation behind casing; however, the high dependency of sigma-to-water salinity of the formation, especially the low-dynamic range at porosity near 12 p.u., has proven to be challenging in low-porosity gas rock. A new measurement from the third detector from a multidetector pulsed-neutron tool (MDPNT) is proposed to provide a better estimation of the gas saturation in a low-porosity reservoir. Two sets of independently measured sigma and the third detector were taken in a casedhole well, with a dual-tubing system of a long string and short string. For the third-detector measurement, the measurement was based on the ratio of the slow capture gate and inelastic gate component from the decay curve created by the long detector. This ratio can be used to detect gas in a tight reservoir with a minimum salinity and lithology effect. This data will then be used to calculate the gas saturation from the third detector, and the result is compared to sigma-based gas saturation. At an interval where the porosity is above 12 p.u., the sigma-based gas saturation and MDPNT-based gas saturation are very much in agreement. However, in a low-porosity reservoir near 12 p.u. or below, the sigma-based measurement starts to show its limitation. Meanwhile, the MDPNT-based gas saturation clearly shows the remaining gas saturation where sigma-based measurements failed to detect it. The subsequent decision was made based on the log analysis result, and perforation was done at a potential interval based on the MDPNT result. The results from the production test confirm the MDPNT-based gas saturation with 700-Mscf/d gas production added. This study showcases a new technology to solve a low-porosity gas reservoir issue where a sigma-based measurement underestimates the remaining gas saturation. Using two different measurements in the same well, the results from the MDPNT measurement demonstrated a better result compared to the sigma-based measurement in low-porosity rock","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"61 1","pages":"623-632"},"PeriodicalIF":0.9,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"69376769","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper provides a history of nuclear spectroscopy in well logging from its beginnings in 1939 up until the present day. After the invention and implementation of gamma ray logging, this paper traces the technological development of the pulsed-neutron capture (sigma) log, the spectral gamma ray log, the carbon-oxygen log, tracer identification logs, small-diameter reservoir characterization tools, and finally the geochemical log. The key to the science of nuclear spectroscopy has been the detection of gamma rays, their energies, and the identity of their parent atomic nuclei. From this, the properties of the formation can be better understood. There have been many advances in technology that have led to the current state of nuclear spectroscopy tools. The most notable has been the ability to detect the presence of a gamma ray. After this came numerous advances in scintillator crystal detector technology, the pulsed-neutron generator, the energy digitization of gamma ray pulses, fast-timing electronics, and powerful computers. These advances have made possible the complex, gamma ray-centric logging tools that we have today that have helped petroleum engineers in the energy industry locate and produce hydrocarbon, kerogen, and natural gas reservoirs for the benefit of each individual in the world. This paper discusses the rich history of these historic developments.
{"title":"A History of Nuclear Spectroscopy in Well Logging","authors":"R. Pemper","doi":"10.30632/pjv61n6-2020a1","DOIUrl":"https://doi.org/10.30632/pjv61n6-2020a1","url":null,"abstract":"This paper provides a history of nuclear spectroscopy in well logging from its beginnings in 1939 up until the present day. After the invention and implementation of gamma ray logging, this paper traces the technological development of the pulsed-neutron capture (sigma) log, the spectral gamma ray log, the carbon-oxygen log, tracer identification logs, small-diameter reservoir characterization tools, and finally the geochemical log. The key to the science of nuclear spectroscopy has been the detection of gamma rays, their energies, and the identity of their parent atomic nuclei. From this, the properties of the formation can be better understood. There have been many advances in technology that have led to the current state of nuclear spectroscopy tools. The most notable has been the ability to detect the presence of a gamma ray. After this came numerous advances in scintillator crystal detector technology, the pulsed-neutron generator, the energy digitization of gamma ray pulses, fast-timing electronics, and powerful computers. These advances have made possible the complex, gamma ray-centric logging tools that we have today that have helped petroleum engineers in the energy industry locate and produce hydrocarbon, kerogen, and natural gas reservoirs for the benefit of each individual in the world. This paper discusses the rich history of these historic developments.","PeriodicalId":49703,"journal":{"name":"Petrophysics","volume":"61 1","pages":"523-548"},"PeriodicalIF":0.9,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"43075651","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：