Scholarlink

<p>Accurate reservoir characterization in structurally complex fields is essential for optimizing hydrocarbon exploration and production. This study presents a detailed analysis of the AEB-3E reservoir within the Berenice Oil Field. It integrates well log and 3D seismic data using the EMBER (Ensemble Machine Learning for Better Estimation of Reservoir properties) workflow. Four wells provided composite logs, including gamma ray, density, neutron porosity, resistivity, and sonic measurements. These logs were quality-controlled, depth-matched, corrected, and upscaled to a 3D geological grid. Seismic-derived attributes were used to capture lateral heterogeneity and structural trends. They also enabled improved interwell property prediction in areas with sparse data. Petrophysical properties (shale volume, effective porosity, and water saturation) were modeled using EMBER. The method combines ensemble decision tree regression with embedded geostatistical features. This allows spatial continuity and stratigraphic relationships to be honored. Deterministic results indicate that the AEB-3E interval is predominantly clean. Average shale volume ranges from 0.11 to 0.18, effective porosity from 0.11 to 0.147, and water saturation from 0.28 to 0.45. Cross-sectional analysis from eastern and western parts of the field confirms lateral consistency. It also highlights localized variations controlled by structural features. Stochastic simulations were used to quantify uncertainty. They reveal low to moderate variability and indicate higher uncertainty near faults. The EMBER workflow significantly reduces modeling time and manual effort. At the same time, it delivers reliable and interpretable reservoir property predictions. Despite the limitation of only four wells, the methodology shows potential for application in similar geological settings. Overall, integrating machine learning with seismic and well data shows promise as an effective framework for reservoir evaluation, offering valuable insights to support informed decision-making and risk assessment in hydrocarbon development. However, the limited dataset warrants cautious interpretation, and further validation is recommended for broader application.</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

Machine learning driven reservoir property modeling of the AEB-3E reservoir in the Berenice field Egypt

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月25日

查看全文
  • Amr M. Eid,
  • Walid M. Mabrouk,
  • Mohammed Amer,
  • Ahmed Metwally

摘要

Accurate reservoir characterization in structurally complex fields is essential for optimizing hydrocarbon exploration and production. This study presents a detailed analysis of the AEB-3E reservoir within the Berenice Oil Field. It integrates well log and 3D seismic data using the EMBER (Ensemble Machine Learning for Better Estimation of Reservoir properties) workflow. Four wells provided composite logs, including gamma ray, density, neutron porosity, resistivity, and sonic measurements. These logs were quality-controlled, depth-matched, corrected, and upscaled to a 3D geological grid. Seismic-derived attributes were used to capture lateral heterogeneity and structural trends. They also enabled improved interwell property prediction in areas with sparse data. Petrophysical properties (shale volume, effective porosity, and water saturation) were modeled using EMBER. The method combines ensemble decision tree regression with embedded geostatistical features. This allows spatial continuity and stratigraphic relationships to be honored. Deterministic results indicate that the AEB-3E interval is predominantly clean. Average shale volume ranges from 0.11 to 0.18, effective porosity from 0.11 to 0.147, and water saturation from 0.28 to 0.45. Cross-sectional analysis from eastern and western parts of the field confirms lateral consistency. It also highlights localized variations controlled by structural features. Stochastic simulations were used to quantify uncertainty. They reveal low to moderate variability and indicate higher uncertainty near faults. The EMBER workflow significantly reduces modeling time and manual effort. At the same time, it delivers reliable and interpretable reservoir property predictions. Despite the limitation of only four wells, the methodology shows potential for application in similar geological settings. Overall, integrating machine learning with seismic and well data shows promise as an effective framework for reservoir evaluation, offering valuable insights to support informed decision-making and risk assessment in hydrocarbon development. However, the limited dataset warrants cautious interpretation, and further validation is recommended for broader application.