Logging-based prediction of organic geochemical parameters in oil shale during thermal evolution using the XGBoost algorithm

TOC prediction using a gradient boosting decision tree method: A case study of shale reservoirs in Qinshui Basin

Evaluation of a source rock can use several parameters, one of which is the determination of Total Organic Carbon (TOC). The determination of TOC is a method that relies on expensive laboratory testing and is limited by the availability of rock samples. TOC prediction using well log data can be performed on most oil and gas wells, which can provide information regarding organic content and continuous data recording. So, the prediction method using well log data is an ideal method to determine TOC in source rock units. The purpose of this study is to predict the TOC value using a well log by applying the machine learning method with the Multi-Layer Perceptron Artificial Neural Network (ANN) technique. Eighteen data samples from the Talang Akar Formation were used for training and testing the MLP-ANN model. The well log data used to predict TOC are density log (RHOB), transit time (DT), deep resistivity (ILD), gamma-rays (GR), and neutron porosity (NPHI), and produce a high correlation (R2 0.87 and the mean absolute percentage error (AAPE) 10%) against the resulting MLP-ANN model. The TOC prediction technique carried out will help a geophysicist (geophysicist and reservoir geology) to evaluate the source rock in an oil and gas field without the need to have a large number of source rock sample data.

Summary As global energy demand continues to rise, the need to exploit unconventional resources like shale oil and gas has become progressively urgent. Total organic carbon (TOC) functions as an essential index for evaluating sweet spot and reservoir production in shale oil and gas exploration. To address the shortcomings of existing TOC prediction approaches, we establish a novel TOC prediction model using the extreme gradient boosting (XGBoost) algorithm optimized by grid search (GS) and particle swarm optimization (PSO). Initially, the GS method is utilized to ascertain the optimal values for the three integer hyperparameters, along with the optimal value ranges for four decimal hyperparameters within the XGBoost model. Subsequently, the PSO method is capable of swiftly identifying the optimal values for the four decimal hyperparameters based on the preceding work. Thus, a GS-PSO-XGBoost model with seven optimal hyperparameters is formulated for TOC prediction utilizing conventional well logs. Meanwhile, Shapley additive explanation (SHAP) is used to enhance the interpretability of the model. When compared with extreme learning machine (ELM), support vector regression (SVR), random forest (RF), XGBoost, and GS-XGBoost models, the GS-PSO-XGBoost method demonstrates superior performance for TOC prediction. The GS-PSO-XGBoost method effectively addresses issues previously encountered in TOC prediction studies, such as slow calculating speed, overfitting, and convergence to local minima, thereby significantly enhancing prediction accuracy. This study deepens the use of machine learning (ML) within petroleum engineering, offering a dependable technical reference for the further analysis of unconventional oil-gas resources exploration.

The prediction of total organic carbon (TOC) content and grading evaluation of shale formation are very much significant and essential for reservoir description of rolling exploration and development in the new shale exploration area (Shuangcheng) in Songliao basin, China. In order to improve exploration efficiency and obtain continuous TOC content curve of wells, the variable coefficient △logR technique was developed for TOC estimating which is based on the two of acoustic time difference and deep lateral resistivity logging curve and the variable scale coefficient (K) between them as well as another scale coefficient (A) between TOC and △logR. A prediction model of TOC was established for the well which TOC is measured by evaluation of side wall cores, then apply it to other wells to verify the reliability of the model. The application result of eleven exploration Wells in Shuangcheng area show that the TOC of shale is linearly correlated with △logR, and the maximum prediction accuracy k value varies with wells, so it is necessary to determine the undetermined coefficient k according to a single well, but the A value having no big change from one well to another in similar sedimentary facies and thermal evolution degree of shale. The average relative error of TOC between prediction model and core measurement is 10.6% which verifies the accuracy of this method. On this basis of TOC prediction, we establish shale grading evaluation criteria for the study area. In the establishment process, not only the relationship between TOC and S1, but also vitrinite reflectance (Ro) are considered. The shale in Shuangcheng area can be divided into three types (Class I: TOC > 3.5% and Ro > 0.9%; Class II: TOC 2%–3.5% and Ro > 0.9; Class III: TOC < 2% or Ro < 0.9%), and achieved shale classification on the well profile with TOC and Ro which are easy to predict and reliable. According to the relationship between the thickness of shale of disparate classes and the total thickness of shale in different zones, the thickness of shale of disparate classes in each well is predicted.

Total organic carbon (TOC) is important geochemical data for evaluating the hydrocarbon generation potential of source rocks. TOC is commonly measured experimentally using cutting and core samples. The coring process and experimentation are always expensive and time-consuming. In this study, we evaluated the use of three machine learning (ML) models and two multiple regression models to predict TOC based on well logs. The well logs involved gamma rays (GR), deep resistivity (RT), density (DEN), acoustic waves (AC), and neutrons (CN). The ML models were developed based on random forest (RF), extreme learning machine (ELM), and back propagation neural network (BPNN). The source rock of Paleocene Yueguifeng Formation in Lishui–Jiaojiang Sag was taken as a case study. The number of TOC measurements used for training and testing were 50 and 27. All well logs and selected well logs (including AC, CN, and DEN) were used as inputs, respectively, for comparison. The performance of each model has been evaluated using different factors, including R2, MAE, MSE, and RMSE. The results suggest that using all well logs as input improved the TOC prediction accuracy, and the error was reduced by more than 30%. The accuracy comparison of ML and multiple regression models indicated the BPNN was the best, followed by RF and then multiple regression. The worst performance was observed in the ELM models. Considering the running time, the BPNN model has higher prediction accuracy but longer running time in small-sample regression prediction. The RF model can run faster while ensuring a certain prediction accuracy. This study confirmed the ability of ML models for estimating TOC using well logs data in the study area.

Predicting the maturity and organic richness using artificial neural networks (ANNs): A case study of Montney Formation, NE British Columbia, Canada

Fully connected deep network: An improved method to predict TOC of shale reservoirs from well logs

Enrichment and distribution of shale oil in the Cretaceous Qingshankou Formation, Songliao Basin, Northeast China

Rock physics and seismic reflectivity parameterization and amplitude variation with offsets inversion in terms of total organic carbon indicator

Thermal maturity: The controlling factor of wettability, pore structure, and oil content in the lacustrine Qingshankou shale, Songliao Basin

Total organic carbon (TOC) is an important parameter for characterizing shale gas and oil reservoirs. Estimation of TOC from well logs has previously been achieved by an empirical model. The radial basis function (RBF) neural network is a new quantitative method that can generate a smooth and continuous function of several input variables to approximate the unknown forward model. We investigated the basic principles of the RBF including network structure, basis function, network training method, and its application in the TOC prediction. The nearest neighbor algorithm was selected for the network training. Then, the Gaussian width was investigated to improve the TOC prediction accuracy through leave-one-out cross-validation. Finally, field cases of organic shale were studied for the TOC prediction, and the prediction results using the RBF method were compared with those of the [Formula: see text] method. Furthermore, according to sensitive attribute ranking, the impacts of different input logs on the predicted results were also investigated through various experiments, and the best network model was finally chosen. The error analysis between the prediction results and lab-measured TOC in some examples indicated that the new approach is more accurate than a single empirical regression method and more flexible than the [Formula: see text] method.

Organic matter evolution in pyrolysis experiments of oil shale under high pressure: Guidance for in situ conversion of oil shale in the Songliao Basin

The Cretaceous Era has always been a focus of geologic and palaeoenvironmental studies. Previous researchers believed that the impact of the global carbon cycle represents significant short-term global biogeochemical fluctuations, leading to the formation of a large number of organic rich sediments in the marine environment. During the Turonian, a large number of organic-rich oil shales were deposited in the lakes of the Songliao Basin in the Qingshankou Formation. How the depositional environment affected the formation of oil shales in continental lakes and the characteristics of these oil shales remain controversial. In this paper, through sampling of Qingshankou Formation strata, various testing methods are used to provide a variety of new data to study the characteristics of oil shales and palaeoenvironment evolution history in the Songliao Basin. The research of the sediments in the Qingshankou Formation in the Fuyu oilfield, Songliao Basin, via result analysis revealed that the oil shales possess an excellent oil-generation potential with moderate-high total organic carbon (TOC) levels (0.58–9.43%), high hydrogen index (HI) values (265–959 mg hydrocarbons (HC)/g TOC), high extractable organic matter (EOM) levels (2.50–6.96 mg/g TOC) and high hydrocarbon fractions (48–89%). The sources of the organic matter were mainly zooplankton, red algae and higher plants (including marine organisms). The aqueous palaeoenvironment of the Qingshankou Formation was a saline water environment with a high sulfate concentration, which promoted an increase in nutrients and stratification of the water density in the lake basin. Oxygen consumption in the bottom water layer promoted the accumulation and burial of high-abundance organic matter, thus forming the high-quality oil shales in the Qingshankou Formation. The global carbon cycle, warm-humid palaeoclimate, dynamic local biogeochemical cycling and relative passive tectonism were the most likely reasons for the TOC increase and negative δ13Corg deviation.

Producing low-mature shale oil resource is becoming economically available as in situ organic matter or heavy oil conversion technologies are introduced to shale oil production. However, we still lack suitable evaluation parameters and criteria for characterizing oil content and mobility in low-mature shales, which are very important for low-mature shale oil economic production. With the aim to better understand the influence factors for low-mature shale oil content and mobility, 16 shale core samples were collected from low-mature Upper Cretaceous Nenjiang Formation (K2n), Gulong Sag, in the Songliao Basin, to investigate their mineral composition and organic geochemical and pore characteristics. The results show the following: (1) Four shale lithofacies were identified, such as mudstone, carbonate lamina shale, bioclastic shale, and mixed lamina shale. (2) Shale oil mobility [represented by the oil saturation index (OSI)] was mainly controlled by large-sized pore distribution. (3) Total organic carbon (TOC) content and lithofacies had great influence on the distribution of oil mobility in the shale. As for mudstone and bioclastic shale, TOC played the most important role in controlling shale oil mobility, whereas for carbonate lamina shale and mixed lamina shale, TOC and mineral variation both have impacts on oil mobility. (4) The average contents of TOC, S1, and OSI in K2n were higher than those in the deeper Qingshankou shale. The oil mobility controlling factors in K2n and K2qn were both TOC and preservation conditions. (5) A new index was proposed to weigh the content of shale oil, in situ OSI, and multiple isothermal pyrolysis parameter S2-2. The in situ OSI shows that K2n has high oil retention capacity, and the in situ oil mobility is 2–3 times the movable oil content.

The exploration of continental shale oil in China has made a breakthrough in many basins, but the pure shale type has only been found in the Qingshankou Formation, Gulong Sag, Songliao Basin, and the evaluation of shale oil occurrence and sweet spot faces great challenges. Using information about the total organic carbon (TOC), Rock-Eval pyrolysis, vitrinite reflectance (Ro), kerogen elemental composition, carbon isotopes, gas chromatography (GC), bitumen extraction, and component separation, this paper systematically studies the organic geochemical characteristics and shale oil occurrence at the Qingshankou Formation. The G1 well, which was cored through the entire section of the Qingshankou Formation in the Gulong Sag, was the object of this study. On this basis, the favorable sweet spots for shale oil exploration are predicted. It is concluded that the shale of the Qingshankou Formation has high organic heterogeneity in terms of organic matter features. The TOC content of the source rocks in the Qingshankou Formation is enhanced with the increase in the burial depth, and the corresponding organic matter types gradually changed from Ⅱ2 and Ⅱ1 types to the Ⅰ type. The distribution of Ro ranges from 1.09% to 1.67%, and it is the mature to high-mature evolution stage that generates a large amount of normal crude oil and gas condensate. The high-quality source rocks of good to excellent grade are mainly distributed in the Qing 1 member and the lower part of the Qing 2 member. After the recovery of light hydrocarbons and the correction of pyrolytic heavy soluble hydrocarbons, it is concluded that the occurrence state of shale oil in the Qingshankou Formation is mainly the free-state form, with an average value of 6.9 mg/g, and there is four times as much free oil as adsorbed oil. The oil saturation index (OSI), mobile hydrocarbon content, Ro, and TOC were selected to establish the geochemical evaluation criteria for shale oil sweet spots in the Qingshankou Formation. The evaluation results show that interval 3 and interval 5 of the Qingshankou Formation in the G1 well are the most favorable sections for shale oil exploration.