Explore the vast range of titles available.

Geophysical well log data are widely used in the field of structural geology, sedimentary geology and petroleum geology. Gaps and misunderstandings are still existing in the scientific interpretation of geophysical well logs. Logging environments and log curves need correction and standardization before interpretation, additionally, there are some special geological phenomena that will mislead the well log interpretation. This review critically highlights the typical misunderstandings existing in the well log data interpretation, and proposes countermeasures as well as scientific interpretation of well logs when encounter these misunderstandings. The factors that affect the well log data acquisition are summarized in terms of types of drilling muds, borehole stability and logging instrument rotation. The vertical resolution of various log series spans a wide range from 5 mm to about 10 m. In the field of structural geology, well logs can be used for determination of stratum attitude, fault recognition, fracture and in situ stress characterization as well as unconformity identification. Lithology and depositional facies can be interpreted using well logs. Well logs aim at finding hydrocarbons, and are used for source rock characterization and logging reservoir evaluation in the petroleum geology field. Then the typical misunderstandings and countermeasures in solving geological issues using geophysical well logs are reviewed from published papers as well as from the authors’ personal experiences. This review will provide insights into the scientific interpretation of geophysical well log data, and help solving geological issues for the petrophysicist and geologist.

Most carbonate reservoirs exhibit heterogeneous pore distribution, whereby the matrix displays low permeability, thus impeding the flow of oil. On the other hand, highly permeable fractures function as the main flow conduits within such reservoirs. Permeability measurements are obtained from core and well test analysis, which are too expensive and not available for many wells. Therefore, accurate permeability prediction is a vital step in developing an efficient field development plan, as it plays a pivotal role in the accurate distribution of 3D petrophysical properties throughout a reservoir. Machine learning (ML) algorithms are now widely applied to predict core permeability using conventional well logs to build a model for permeability prediction in uncored wells. This review considers the performance of six ML algorithms (LightGBM, CATBoost, XGBoost, Adaboost, random forest and gradient boosting) for permeability prediction from a high-quality dataset. The dataset incorporates multiple well-log inputs (gamma ray, caliper, density, neutron porosity, shallow and deep resistivity, total porosity, spontaneous potential, water saturation, depth, and facies) in addition to direct core permeability and porosity measurements. Data pre-processing techniques applied include missing data imputation, scale correction, normalization with three different transformations (log, Box-Cox, and NST) and outlier detection. To enhance the ML performance, two search algorithms (random search and Bayesian optimization) are compared in their ability to tune the ML hyperparameters. There is a need to identify a suitable parameter space, especially when the target variable range is changing. ML performance was evaluated with four evaluation metrics (RMSE, MAE, R 2 , and Adjusted R 2 ). Results showed that the XGBoost algorithm with configuration of (RS as search algorithm, Box Cox as the normalization method, Z-score for outlier detection, without scale correction, old parameter space) delivered the best prediction performance for permeability with RMSE values of 6.9 md and 9.78 md for training and testing, respectively.

Precise estimation of total organic carbon (TOC) is extremely important for the successful characterization of an unconventional shale reservoir. Indirect traditional continuous TOC prediction methods from well-logs fail to provide accurate TOC in complex and heterogeneous shale reservoirs. A workflow is proposed to predict a continuous TOC profile from well-logs through various ensemble learning regression models in the Goldwyer shale formation of the Canning Basin, WA. A total of 283 TOC data points from ten wells is available from the Rock-Eval analysis of the core specimen where each sample point contains three to five petrophysical logs. The core TOC varies largely, ranging from 0.16 wt % to 4.47 wt % with an average of 1.20 wt %. In addition to the conventional MLR method, four supervised machine learning methods, i.e., ANN, RF, SVM, and GB are trained, validated, and tested for continuous TOC prediction using the ensemble learning approach. To ensure robust TOC prediction, an aggregated model predictor is designed by combining the four ensemble-based models. The model achieved estimation accuracy with R2 value of 87%. Careful data preparation and feature selection, reconstruction of corrupted or missing logs, and the ensemble learning implementation and optimization have improved TOC prediction accuracy significantly compared to a single model approach.

Well logging helps geologists find hidden oil, natural gas and other resources. However, well log data are systematically insufficient because they can only be obtained by drilling, which involves costly and time-consuming field trials. Additionally, missing or distorted well log data are common in old oilfields owing to shutdowns, poor borehole conditions, damaged instruments and so on. As a workaround, pseudo-data can be generated from actual field data. In this study, we propose a spatio-temporal neural network (STNN) algorithm, which is built by leveraging the combined strengths of a convolutional neural network (CNN) and a long short-term memory network (LSTM). The STNN exploits the ability of the CNN to effectively extract features related to pseudo-well log data and the ability of the LSTM to extract the key features from well log data along the depth direction. The STNN method allows full consideration of the well log data trend with depth, the correlation across different log series and the actual depth accumulation effect. The method proved successful in predicting acoustic sonic log data from gamma-ray, density, compensated neutron, formation resistivity and borehole diameter logs. Results show that the proposed method achieves higher prediction accuracy because it takes into account the spatio-temporal information of well logs.

Well log analysis is significant for hydrocarbon exploration, providing detailed insights into subsurface geological formations. However, gaps and inaccuracies in well log data, often due to equipment limitations, operational challenges, and harsh subsurface conditions, can introduce significant uncertainties in reservoir evaluation. Addressing these challenges requires effective methods for both synthetic data generation and precise imputation of missing data, ensuring data completeness and reliability. This study introduces a novel framework utilizing sequence-based generative adversarial networks (GANs) specifically designed for well log data generation and imputation. The framework integrates two distinct sequence-based GAN models: time series GAN (TSGAN) for generating synthetic well log data and sequence GAN (SeqGAN) for imputing missing data. Both models were tested on a dataset from the North Sea, Netherlands region. For the imputation task, the input comprises logs with missing values and the output is the corresponding imputed logs; for the synthetic data generation task, the input is complete real logs and the output is synthetic logs that mimic the statistical properties of the original data. All log measurements are normalized to a 0-1 range using min-max scaling, and error metrics are reported in these normalized units. Different sections of 5, 10, and 50 data points were used. Experimental results demonstrate that this approach achieves superior accuracy in filling data gaps compared to other deep learning models for spatial series analysis. The imputation method yielded values of 0.92, 0.86, and 0.57, with corresponding mean absolute percentage error (MAPE) values of 8.320, 0.005, and 166.6, and mean absolute error (MAE) values of 0.012, 0.002, and 0.03, respectively. The synthetic generation yielded of 0.92, MAE, of 0.35, and MRLE of 0.01. These results set a new benchmark for data integrity and utility in geosciences, particularly in well log data analysis.

This paper presents a case study on the application of simultaneous inversion (SI) technique and well log analysis to accurately seismic-geomechanical characterization of Asmari formation in an Iranian oil field. To achieve this goal, a one-dimensional (1D) prediction of geomechanical properties such as Young’s modulus (E), bulk modulus (K), shear modulus ( ), Poisson’s ratio (PR), Vp/Vs ratio and brittleness (BRI) was first generated from the analyzing the well log data. In the next step, the spatial distribution of the geomechanical parameters in the reservoir area was predicted on the basis of the results of the simultaneous inversion. Lithological facies discrimination and fluid detection of the Asmari formation were performed using simultaneous inversion of pre-stack seismic data and conventional cross-plotting analysis of well data. To accomplish this objective, elastic and geomechanical parameters were cross-plotted against conventional petrophysical logs along with BRI, water saturation (Sw), gamma ray (GR), acoustic impedance (Zp), , and lithology logs such as quartz and calcite volumes. As a result of this work, two reservoirs were identified in the Asmari formation: sandstone and carbonate. The carbonate reservoir consists predominantly of calcite and dolomite, with minimal shale content, while the sandstone reservoir is mainly composed of shale and contains less quartz. In addition, the carbonate section within the Asmari formation exhibits better reservoir quality due to lower water saturation and higher porosity compared to the sandstone zone. To determine the type of fluid in the Asmari formation of the studied oil field, LMR (lambda-mu-rho) scatterplots were employed in both well and seismic domains. The findings reveal that the sandstone reservoir in the Asmari formation is water-saturated, while the carbonate reservoir is oil-saturated. Furthermore, a gas cap is present at the top of the Asmari formation.

The well logging technique is used to determine the petrophysical properties like porosity, permeability and fluid saturations of subsurface formations. However, the conventional way of log evaluation is very expensive and tiresome. The post-acquisition processing and inversion provides an alternative to determine the properties of drilled formations. This study proposes a novel approach to predict sonic log, adopting a regression method using a supervised machine learning (ML) algorithm, along with the determination of lithology employing clustering and a neural network approach grounded on the basis of gamma-ray log values and hence creating a correlation between the two. The scarce acoustic data obtained upon the traditional well logging procedure often pose a barrier in further determining the rock physics. Regression analysis, a predictive modeling technique, uses other petrophysical data to predict the sonic wave travel time (shear and compressional) by estimating a relationship between the two variables. The model is trained on a set of 10,000 points with 80% training points giving an 86.314% accuracy result. RMSE and R2 scores for training points and testing points came out to be 2.622 and 0.95, and 2.55 and 0.96, respectively, which helps in the validation of the model. Effective lithology determination is a crucial step of reservoir characterization. Traditional methods of core sample inspection and using well logs, however, cannot meet the needs of real-time due to complex sediment environment and reservoir heterogeneity. To deal with the problem, an unsupervised ML model, K-means clustering, a method of vector quantization grouping unlabeled data into arbitrary clusters based on similarities with respect to distance from the center is used. The model gave the optimum number of clusters as 5 and showed a presence of siltstone, coal and sandstone separated between these clusters. The Silhouette score which tests the accuracy came out to be 0.5840 along with a CH score of 27,192.

The amplitude-variation-with-angle (AVA) inversion for seismic data has been widely used for hydrocarbon detection in exploration seismology. Traditional AVA inversion quantitatively estimates high-resolution elastic parameters, i.e., P-wave velocity, S-wave velocity and density, from migrated seismic gathers by solving either a linear or nonlinear inverse problem. It is commonly an ill-posed problem and the inversion accuracy depends on initial models. Recently, deep learning has been introduced into the AVA inversion by building a complicated nonlinear relation between seismic data and elastic parameters based on the training on a large amount of labeled data. The performance of the deep-learning-based inversion is determined by the diversity of training datasets. Because of sparse well locations, the application of deep-learning-based AVA inversion is limited by well-log label sets in production. To mitigate this problem, we present an intelligent AVA inversion method using a convolutional neural network trained by realistic pseudo-well logs. By considering spatial and inter-parameter correlation of elastic parameters, we first generate a large number of realistic pseudo-well logs based on Monte Carlo simulation. Then, angle-domain common-image gathers are computed by convolving a source wavelet with angle-dependent reflectivity series, which are used to train a convolutional neural network (CNN) to predict elastic parameters. In this study, we introduce two CNN frameworks to investigate the feasibility of the proposed pseudo-well-based CNN AVA inversion method using both synthetic and field data. We also compare the proposed CNN-based AVA inversion method with traditional linear and nonlinear inversion methods constrained by prior knowledge in terms of efficiency and accuracy. The results of synthetic data show that the pseudo-well-based CNN AVA inversion method can accurately and efficiently estimate P-wave velocity, S-wave velocity and density, and has a potential to reduce inter-parameter crosstalk artifacts. In the tests of field data, because of inaccurate background velocity models and noisy angle-domain gathers, the accuracy of CNN prediction results is not as high as in synthetic example. However, the pseudo-well-based CNN AVA inversion method still has better performance to reduce inter-parameter crosstalk artifacts and requires less computing time than traditional AVA inversion method.

Accurate reservoir quality (RQ) assessment is a critical component of hydrocarbon exploration and production, particularly in heterogeneous reservoirs with complex geological characteristics. The current study proposes an integrated reservoir quality index (IRQI) that combines parameters from different disciplines, including petrophysical, geomechanical and elastic attributes derived from well-log data. To ensure consistency among inputs, all parameters were normalized to a common range between 0 and 1, and no predefined or arbitrary weighting factors were applied. This approach allows the relative influence of each parameter to be governed by reservoir-specific geological conditions rather than subjective weighting assumptions. The proposed IRQI is implemented within a well–seismic integrated workflow, enabling reservoir quality evaluation at both well locations and the whole reservoir. At the well domain, IRQI was calculated using log-derived parameters and independently validated using laboratory core data. At the reservoir domain, IRQI was propagated into inter-well spaces using acoustic impedance inversion and seismic attributes, allowing continuous spatial characterization of reservoir quality. The methodology was applied to the Asmari Formation in two oilfields with diverse geological complexities and reservoir heterogeneities. One field exhibits a moderate geological complexity and heterogeneity, while the other is characterized by a high geological complexity and a pronounced reservoir heterogeneity. Unlike most previous studies that primarily rely on petrophysical parameters alone, this research integrates petrophysical, elastic and geomechanical properties into a single quantitative index. The performance of IRQI was evaluated at well locations and across the seismic domain in both fields, allowing consistent identification of high-quality reservoir intervals. The novelty of this study relies on the multi-domain integration of well logs-derived parameters, validated using core data, and field-scale application through well–seismic integration across multiple oilfields. The results highlight the robustness and applicability of IRQI for reservoir characterization in geologically heterogeneous reservoirs.

Derivative and volatility attributes calculated for well-log versus depth sequences extract characteristics that can be usefully exploited by automated machine-learning (ML) lithofacies classification models. That information is valuable for wellbores that have a restricted suite of recorded well logs and no cores recovered, limiting the detailed geological information available. In this study, a ten-well dataset, calibrated with core information, from the large Panoma gas Field (Kansas, U.S.A), with a suite of five well logs through a complex Lower Permian carbonate and siliciclastic lithofacies sequence is evaluated with seven ML models. Detailed cross-validation. feature selection analysis identifies that the volatility of the neutron-density porosity log added to the five recorded well logs improves lithofacies classification performance with this dataset without recourse to geological input. The support vector classifier (SVC) provides the most accurate facies class prediction of the ML models tested, achieving prediction accuracy of 61.2% with this six feature dataset. However, the addition of calculated facies boundary influence attributes further improves the SVC’s facies class prediction accuracy to 77.2% (weighted average F1 score of 0,7620) for data unseen by the trained model. Training, validation and unseen data testing of the ML models reveals that the SVC model is less prone to overfitting than the other ML models evaluated with the attribute enhanced datasets. In the absence of detailed geological inputs, such attribute-enhanced well-log datasets can be used to reliably locate target reservoir lithofacies, such as high productivity zones associated with high porosity and permeability, with automated ML models.