Scholarlink

<p>The prediction of shear wave velocity in highly heterogeneous reservoirs presents a significant challenge within the geosciences. Accurate analysis of shear waves is crucial for a comprehensive understanding of subsurface heterogeneity and improved reservoir characterization. The study presents a framework for model selection to predict shear wave velocity using heterogeneous subsurface geoparameters. Various machine learning models, including multi-layer perceptron (MLP), support vector machines (SVM), gradient boosting (GB), extreme gradient boosting (XGB), random forests (RF), and bagging random forests (BRF), were evaluated. Furthermore, hyperparameter optimization was employed, along with feature importance analysis, sensitivity analysis, and uncertainty quantification, to identify the most suitable predictive approach. To build high-performance models, the datasets underwent a preprocessing stage to enhance the quality of data by eliminating outliers and standardizing the raw data. Furthermore, the data were split into a training set (70%) and a testing set (30%), and over 2000 blind datasets were employed to validate the reliability of the models. Additionally, the hyperparameters were optimized using grid search cross-validation to ensure the models’ architecture. The findings indicate that the MLP and BRF models obtained superior scores, achieving <i>R</i><sup>2</sup> values of 0.82 and 0.81, respectively. The remaining models—SVM, GB, XGB, and RF—yielded <i>R</i><sup>2</sup> values of 0.76, 0.77, 0.78, and 0.81, respectively. Furthermore, the finding of sensitive methods indicated that sonic log and neutron porosity log are the most influential features in shear wave velocity prediction.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploring Shear Waves Predictions in Heterogeneous Reservoirs Using Multi-machine Learning Models Based on Conventional Wireline Logs: A Case Study from the Nubian Reservoir, Sirt Basin, Libya

  • 期刊论文

  • 发表日期：

    2025年3月13日

查看全文
  • Mohammed A. Amir,
  • Hamzah S. Amir

摘要

The prediction of shear wave velocity in highly heterogeneous reservoirs presents a significant challenge within the geosciences. Accurate analysis of shear waves is crucial for a comprehensive understanding of subsurface heterogeneity and improved reservoir characterization. The study presents a framework for model selection to predict shear wave velocity using heterogeneous subsurface geoparameters. Various machine learning models, including multi-layer perceptron (MLP), support vector machines (SVM), gradient boosting (GB), extreme gradient boosting (XGB), random forests (RF), and bagging random forests (BRF), were evaluated. Furthermore, hyperparameter optimization was employed, along with feature importance analysis, sensitivity analysis, and uncertainty quantification, to identify the most suitable predictive approach. To build high-performance models, the datasets underwent a preprocessing stage to enhance the quality of data by eliminating outliers and standardizing the raw data. Furthermore, the data were split into a training set (70%) and a testing set (30%), and over 2000 blind datasets were employed to validate the reliability of the models. Additionally, the hyperparameters were optimized using grid search cross-validation to ensure the models’ architecture. The findings indicate that the MLP and BRF models obtained superior scores, achieving R2 values of 0.82 and 0.81, respectively. The remaining models—SVM, GB, XGB, and RF—yielded R2 values of 0.76, 0.77, 0.78, and 0.81, respectively. Furthermore, the finding of sensitive methods indicated that sonic log and neutron porosity log are the most influential features in shear wave velocity prediction.