Productivity Prediction in Tight Oil Reservoirs: A Stacking Ensemble Approach with Hybrid Feature Selection

Controls on reservoir heterogeneity of tight sand oil reservoirs in Upper Triassic Yanchang Formation in Longdong Area, southwest Ordos Basin, China: Implications for reservoir quality prediction and oil accumulation

Background Type-II Diabetes Mellitus (T2DM) is a chronic metabolic disorder characterized by elevated blood glucose levels, posing a critical global health challenge. It is largely attributed to lifestyle changes, unhealthy dietary habits, and lack of awareness. If not diagnosed early, T2DM can lead to severe complications, including damage to vital organs like the kidneys, heart, and nerves. While timely and accurate diagnosis is crucial, current diagnostic procedures are often costly and time-intensive, necessitating innovative approaches to improve early detection. Objective This study aimed to enhance the early prediction of T2DM by leveraging advanced hybrid metaheuristic optimization algorithms to improve model efficiency, accuracy, and computational time. Method The methodology employed in this study involved three key steps: feature selection and refinement, model optimization, and evaluation. For feature selection, SHAP (SHapley Additive exPlanations) was integrated with Support Vector Machines (SVMs) to identify the most significant predictive features. This was followed by Particle Swarm Optimization (PSO), which was utilized for feature refinement, ensuring a concise yet highly informative feature set. In the model optimization phase, Genetic Algorithms (GAs) were applied to optimize the hyperparameters of machine learning models, including Artificial Neural Networks (ANNs), Random Forest (RF), and SVM. Bayesian Optimization (BO) was then employed to further refine these hyperparameters, enhancing overall model performance. Finally, the models were evaluated using key classification metrics, such as accuracy, Receiver Operating Characteristic (ROC) curves, and F1 scores, to ensure the robustness and reliability of the proposed approach. Result The hybrid metaheuristic optimization approach, which integrated Random Forest with SHAP, PSO, GA, and Bayesian Optimization, delivered the best performance among all evaluated methods. It achieved an impressive accuracy of 99.0%, an F1-score of 94.8%, and the largest Area Under the Curve (AUC) compared to other approaches. Furthermore, this method demonstrated a significant reduction in computational time while maintaining exceptional reliability and precision. Conclusion The innovative hybrid algorithm demonstrated superior efficiency and reliability, making it a promising tool for early T2DM diagnosis. By integrating metaheuristic optimization techniques with robust machine learning models, the study establishes a framework for improving diagnostic accuracy and computational efficiency in medical support systems. This research highlights the transformative potential of hybrid optimization in advancing healthcare diagnostics.

Horizontal wells and hydraulic fracturing are the key technologies that allow commercial production from tight oil and gas reservoirs. However, rigorous analysis of production data from these reservoirs requires incorporation of the impacts of stress-dependent permeability and multi-phase flow. Changes in the stress state of the system during production may reduce the absolute permeability. Furthermore, gas phase formation and flow in presence of supersaturated oil phase affects fluid dynamics in tight oil reservoirs. This study provides a rigorous methodology for incorporation of the effects of non-static permeability and multi-phase flow in rate transient analysis (RTA) of tight oil and gas reservoirs producing at variable rate/flowing pressures during transient linear flow period. Analytical solutions for the approximate linearized form of the flow equation have been widely used as the basis for RTA tools for conventional reservoirs during transient flow period. However, they lead to considerable error when applied to tight oil and gas reservoirs. In particular, during the transient linear flow period, the slope of the square -root-of-time plot obtained from numerical solution differs from the slope calculated by analytical methods. Efforts have been made by some researchers to obtain a correction factor from the numerical solution of the flow equation to correct the slope of the square-root-of-time plot for single phase flow of gas during transient linear flow period. In this study, an iterative method is used for evaluation of the slope correction factor in the presence of multi-phase flow and non-static permeability for constant-pressure production during transient linear flow period. Further, the correction factor is used for analysis of production data from tight oil and gas reservoirs producing at variable rate/flowing pressures. The correction factor is used in the analysis of different sets of synthetic production data for tight oil and gas reservoirs. The results show that the correction factor can reduce/eliminate the considerable errors associated with the conventional analytical methods in initial permeability estimation. For multi-phase flow cases, the producing gas -oil ratio (GOR) is used to estimate the oil saturation-pressure relationship in the reservoir, which is required for calculation pseudo-pressure and the correction factor. The method developed in this study alleviates the need for using numerical simulation models to generate empirical correlations for the correction factor for the square -root-of-time plot. The easy-to-implement iterative procedure of this method only requires the pressure dependencies of the constituent elements of the hydraulic diffusivity. Therefore, this method is applicable for analysis of production profiles for variety of reservoirs with nonlinear flow equations.

Characteristics and controlling factors of lacustrine tight oil reservoirs of the Triassic Yanchang Formation Chang 7 in the Ordos Basin, China

A new approach to calculate permeability stress sensitivity in tight sandstone oil reservoirs considering micro-pore-throat structure

The most commonly used technology for development of unconventional liquid-rich and light oil reservoirs is horizontal wells combined with large multi-stage hydraulic fracture treatments. However, even with these technological advancements, primary recovery factors are generally less than 10% (Shoaib and Hoffman, 2009) of the original oil in place (OOIP). Logically, operators have investigated the use of waterflooding to improve recovery in some tight oil reservoirs, but the success has been mixed. Low matrix permeability in some unconventional (tight) oil reservoirs will not allow effective displacement or movement of water through the reservoir. In some cases, even flooding with a gas will be a challenge, if matrix permeabilities are too low. This study investigates the feasibility of enhanced oil recovery (EOR) in a prominent tight oil reservoir in North America using cyclic solvent injection (CSI, sometimes referred to as "huff-n-puff") with carbon dioxide (CO2) as the solvent. CSI is a single well process, with the solvent remaining in the vicinity of the wellbore, as flow of the solvent through the reservoir to another well is not necessary. This type of process may be attractive from a capital cost point-of-view, as large expenditures on specialized facilities, in-field pipelines and well conversions are unnecessary. In this study, the success and profitability of huff-n-puff is evaluated for the Bakken tight oil reservoir. Knowledge gained from a parallel study (Kanfar and Clarkson, 2017) served to provide guidelines for optimizing the huff-n-puff process. Importantly, a genetic algorithm (GA) is utilized to find the optimum huff-n-puff program that maximizes net present value (NPV). Optimized parameters include: the number of cycles; duration of injection, soaking and production periods; and the start time of huff-n-puff operations. The target reservoir for evaluation is the US Bakken deep tight oil reservoir in North Dakota. The huff-n-puff EOR scheme was found to be successful, but only after the aforementioned operational parameters are optimized with GA. In particular, it is important to delay huff-n-puff until production rates decline and boundary-dominated flow (after fracture interference) is reached. Importantly, as with the parallel study (Kanfar and Clarkson 2017), the gridding scheme used in the simulation is found to have a profound impact on results of huff-n-puff.

ObjectiveThe aim of this study was to establish an ensemble learning model based on clinicopathological parameter and ultrasound radomics for assessing the risk of lateral cervical lymph node with short diameter less than 8 mm (small lymph nodes were used instead) metastasis in patients with papillary thyroid cancer (PTC), thereby guiding the selection of surgical methods.MethodsThis retrospective analysis was conducted on 454 patients diagnosed with papillary thyroid carcinoma who underwent total thyroidectomy and lateral neck lymph node dissection or lymph node intraoperative frozen section biopsy at the First Hospital of China Medical University between January 2015 and April 2022. In a ratio of 8:2, 362(80%) patients were assigned to the training set and 92(20%) patients were assigned to the test set. Clinical pathological features and radomics features related to ultrasound imaging were extracted, followed by feature selection using recursive feature elimination (RFE). Based on distinct feature sets, we constructed ensemble learning models comprising random forest (RF), extreme gradient boosting (XGBoost), categorical boosting (CatBoost), gradient boosting decision tree (GBDT), and light gradient boosting machine (Lightgbm) to develop clinical models, radiomics models, and clinical-radiomic models. Through the comparison of performance metrics such as area under curve (AUC), accuracy (ACC), specificity (SPE), precision (PRE), recall rate, F1 score, mean squared error (MSE) etc., we identified the optimal model and visualized its results using shapley additive exPlanations (SHAP).ResultsIn this study, a total of 454 patients were included, among whom 342 PTC patients had small lymph node metastasis in the lateral neck region, while 112 did not have any metastasis. A total of 1035 features were initially considered for inclusion in this study, which were then narrowed down to 10 clinical features, 8 radiomics features, and 17 combined clinical-omics features. Based on these three feature sets, a total of fifteen ensemble learning models were established. In the test set, RF model in the clinical model is outperforms other models (AUC = 0.72, F1 = 0.75, Jaccard = 0.60 and Recall = 0.84), while CatBoost model in the radiomics model is superior to other models (AUC = 0.91, BA = 0.83 and SPE = 0.76). Among the clinical-radiomic models, Catboost exhibits optimal performance (AUC = 0.93, ACC = 0.88, BA = 0.87, F1 = 0.91, SPE = 0.83, PRE = 0.88, Jaccard = 0.83 and Recall = 0.92). Using the SHAP algorithm to visualize the operation process of the clinical-omics CatBoost model, we found that clinical omics features such as central lymph node metastasis (CLNM), Origin_Shape_Sphericity (o_shap_sphericity), LoG-sigma3_first order_ Skewness (log-3_fo_skewness), wavelet-HH_first order_Skewness (w-HH_fo_skewness) and wavelet-HH_first order_Skewness (sqr_gldm_DNUN) had the greatest impact on predicting the presence of lateral cervical small lymph node metastasis in PTC patients.Conclusions(1) In this study, among the ensemble learning models established based on clinicopathological features and radiomics features for predicting PTC lateral small lymph node metastasis, the clinical-radiomic CatBoost model has the best performance. (2) SHAP can visualize how the clinical and radiomics features affect the results and realize the interpretation of the model. (3) The combined CatBoost model can improve the diagnostic accuracy of suspicious lymph nodes with short diameter < 8 mm that are difficult to obtain accurate puncture results. The combined application of radiomics features is more accurate and reasonable than the prediction of clinical data alone, which helps to accurately evaluate the surgical scope and provide support for individual clinical decision making.

Evaluation criteria, major types, characteristics and resource prospects of tight oil in China

Remote sensing estimation of rice chlorophyll content based on UAV image feature selection and PSO-optimized ensemble learning

The quality of an airline's services cannot be measured from the company's point of view, but must be seen from the point of view of customer satisfaction. Data mining techniques make it possible to predict airline customer satisfaction with a classification model. The Naïve Bayes algorithm has demonstrated outstanding classification accuracy, but currently independent assumptions are rarely discussed. Some literature suggests the use of attribute weighting to reduce independent assumptions, which can be done using particle swarm optimization (PSO) and genetic algorithm (GA) through feature selection. This study conducted a comparison of PSO and GA optimization on Naïve Bayes for the classification of Airline Passenger Satisfaction data taken from www.kaggle.com. After testing, the best performance is obtained from the model formed, namely the classification of Airline Passenger Satisfaction data using the Naïve Bayes algorithm with PSO optimization, where the accuracy value is 86.13%, the precision value is 87.90%, the recall value is 87.29%, and the value is AUC of 0.923.

Continental tight sandstone oil reservoirs have strong heterogeneity, and staged fracturing technology of horizontal well is a crucial measure for successful development of oil and gas. In this study, the fracturing effect of horizontal wells in tight oil reservoirs of Yanchang Formation in the western Ordos Basin was systematically studied using the rock mechanics, array acoustic and microseismic testing data and the staged fracturing technology. The hydraulic fracturing method was used to calculate the horizontal principal stress difference (σH-σh). It showed that as the buried depth increases, σH-σh tends to decrease first and then increase. Small-scale fracturing should be used for areas with smaller σH-σh values. Fracturing construction parameters have an impact on oil production capacity, which is mainly manifested in that the usage of prepad fluid, sand-carrying fluid and proppant is proportional to productivity. Excessive displacement and construction scale should not be used in the fracturing process, and the fracture height of the target layer should be strictly controlled within the range of 26 m. The analysis of the “rupture points” in the fracturing curves shows that wells with relatively obvious rupture points usually have a higher oil production capacity. These wells have a good fracturing effect and an effective fracture network was formed in the tight oil reservoir. The optimization simulation results of the horizontal well pattern form show that the seven-point combined well pattern is the best well pattern, which is suitable for the development of tight oil sandstone in the Yanchang Formation.

Pore-scale analysis of gas huff-n-puff enhanced oil recovery and waterflooding process

Objectives: A hybrid ETLBO-PSO model is developed to improve student performance predictions. It assesses intellectual, social, and economic background of students to increase accuracy of students performance predictions. The model optimizes selecting features, which reduces redundancy and increases efficiency. The efficacy is compared with existing Educational Data Mining techniques. Methods : This study integrates Enhanced Teachers Learners Based Optimization (ETLBO) and Particle Swarm Optimization (PSO) algorithm for optimal feature selection. The suggested technique is utilized as an algorithm for selecting features to identify the most significant elements for predicting student academic performance. The efficacy of the proposed feature selection technique is evaluated using three machine learning classifiers: Extreme Gradient Boosting (XGB), Light Gradient Boosting (LightGB), and Category Gradient Boosting (CatGB) in Student achievement Dataset in secondary education for Mathematics. Findings: The experimental results of ETLBO-PSO provides sustained excellent model performance while reducing accuracy decline. The Meta-Class model of ETLBO-PSO has an F1-score of 82.43%, which makes it an increasingly robust and reliable strategy. Furthermore, an innovative visual and intuitive method is employed to identify the aspects that most significantly impact the score, facilitating the interpretation and comprehension of the complete model. Novelty: ETLBO_PSO is integrated with SHAP (SHapley Additive exPlanations), and Meta-class Model are used to optimize student performance predictions with higher accuracy. Unlike traditional approaches, it continuously refines selecting features throughout training, solving high-dimensional data issues. SHAP's approach assures precise feature attribution, hence improving accessibility and making decisions. Keywords: Feature Selection, Enhance Teacher Learner based Optimization, Particle Swarm Optimization, Academic Student Performance, Classification Algorithm, Optimization Techniques, XGBoost, LGBoost, CATBoost

Nitrous oxide prediction through machine learning and field-based experimentation: A novel strategy for data-driven insights

Over the past decade, multi-stage fractured horizontal wells have been widely applied to develop tight oil and gas reservoirs. Therefore, it has become a standard practice to pump more and more numbers of treatment proppants and fluids with higher and higher pumping rate during hydraulic fracturing treatment in China, but things always did not get the expected results, field test application has displayed that not all wells production have positive correlation with increasingly treatment scales but rather lead to high costs. To investigate this problem, we established a new fracture face damage skin (FFDS) mathematical model based on the classical model of Perkins-Kern-Nordgren(PKN) and Khristianovic-Geertsma-Daneshy(KGD) using fracture mechanics and fluid coupling method. Furthormore, a new evaluation experiment by computed tomography (CT) scanning and rock mechanics and some key factors were revealed. Compared with original model the influence of FFDS on productivity was increase from 5% to 51% using new model when the permeability is 0.1mD. The results indicated the lower of permeability and the greater of impact for a specific low permeability reservoir. A case study was put forward with four different volume of proppants and fluids while different pumping ratel located in the same stratum with a same pay thickness in X tight oil reservoir. The reason was studied that some large-scale fractured well not produce that expected much oil and some small-scale fractured wells even perform better than massive treated ones owing to the fracture face damage skin. What's more, a comprehensive low FFDS treatment was put forward as following: Firstly, utilizing the low damage fracturing fluids such as slick water, low concentration and low molecular weight fluid, etc.; Secondly, optimizing the viscosity value of fracturing fluid with corresponding pumping rate for a certain permeability reservoir;Thirdly, optimizing combination between fracturing fluid and proppant types according new mathematical model; Fourthly, using such as acid-fracturing to minimize fracture face damage skin for an optimized fracture profile. These new observations have significant implications for field development stratedies and hydraulic fracturing design. More than 45 wells have been put into field application using new technology. The performance of post-fracturing is remarkable and it has a great significance in unconventional oil and gas development in China.