Application of Deep Learning for Reservoir Porosity Prediction and self Organizing Map for Lithofacies Prediction

Journal of Applied Geophysics, Journal Year: 2024, Volume and Issue: 230, P. 105502 - 105502

Published: Aug. 31, 2024

Language: Английский

---

Synthetic Graphic Well Log Generation Using an Enhanced Deep Learning Workflow: Imbalanced Multiclass Data, Sample Size, and Scalability Challenges

SPE Journal, Journal Year: 2023, Volume and Issue: 29(01), P. 1 - 20

Published: Sept. 7, 2023

Summary The present study introduces an enhanced deep learning (DL) workflow based on transfer (TL) for producing high-resolution synthetic graphic well logs (SGWLs). To examine the scalability of proposed workflow, a carbonate reservoir with high geological heterogeneity has been chosen as case study, and developed is evaluated unseen data (i.e., blind well). Data sources include conventional graphical (GWLs) from neighboring wells. During drilling operations, GWLs are standard practice collecting data. GWL provides rapid visual representation subsurface lithofacies to establish correlations. This investigation examines five wells in southwest Iranian oil field. Due heterogeneities, primary challenge this research lies addressing imbalanced facies distribution. traditional artificial intelligence strategies that manage [e.g., modified minority oversampling technique (M-SMOTE) Tomek link (TKL)] mainly designed solve binary problems. However, adapt these methods upcoming multiclass situation, one-vs.-one (OVO) one-vs.-all (OVA) decomposition ad-hoc techniques used. Well-known VGG16-1D ResNet18-1D used adaptive very-deep algorithms. Additionally, highlight robustness efficiency algorithms, shallow approaches support vector machine (SVM) random forest (RF) classification also other main need enough points train very resolved through TL. After identifying well, four wells’ entered model training. average kappa statistic F-measure, appropriate imbalance evaluation metrics, implemented assess workflows’ performance. numerical comparison analysis shows TL performs better set when combined OVA scheme TKL combat tactic. An 86.33% mean F-measure 92.09% demonstrate superiority. Considering prevalence different distributions, scalable can be efficient productive generating SGWL.

Language: Английский

---

A multidisciplinary approach to facies evaluation at regional level using well log analysis, machine learning, and statistical methods

Geomechanics and Geophysics for Geo-Energy and Geo-Resources, Journal Year: 2023, Volume and Issue: 9(1)

Published: Nov. 18, 2023

Abstract Geological facies evaluation is crucial for the exploration and development of hydrocarbon reservoirs. To achieve accurate predictions litho-facies in wells, a multidisciplinary approach using well log analysis, machine learning, statistical methods was proposed Lower Indus Basin. The study utilized five supervised learning techniques, including Random Forest (FR), Support Vector Machine (SVM), Artificial Neural Network (ANN), Extreme Gradient Boosting (XGB), Multilayer Perceptron (MLP), to analyse gamma ray, resistivity, density, neutron porosity, acoustic, photoelectric factor logs. Concentration-Number (C-N) fractal model log–log plots were also used define geothermal features. In on models classifying different rock types Sawan field Southern Basin, it discovered that sand (fine, medium coarse) most accurately classified (87–94%), followed by shale (70–85%) siltstone (65–79%). accuracy assessed various metrics, such as precision, recall, F1 score, ROC curve. found all successfully predicted particular, classified, facies. multilayer perceptron method performed best overall. This has potential save time costs associated with traditional core analysis enhance efficiency development.

Language: Английский

---

A novel machine learning approach for interpolating seismic velocity and electrical resistivity models for early-stage soil-rock assessment

Earth Science Informatics, Journal Year: 2024, Volume and Issue: 17(3), P. 2629 - 2648

Published: April 12, 2024

Language: Английский

---

Classifying lithofacies from well logs using supervised machine learning, cluster, and principal component analysis plus stacking model combinations

Elsevier eBooks, Journal Year: 2025, Volume and Issue: unknown, P. 111 - 150

Published: Jan. 1, 2025

Language: Английский

---