Scholarlink

<p>The occurrence and distribution of hydrocarbon in subsurface reservoirs are governed by a combination of geological factors, including lithological heterogeneity, burial depth, depositional environment, fluid migration pathways, and the extent of diagenetic alteration. Accurate characterization of reservoir properties, such as porosity and fluid saturation, is therefore essential for reliable estimation of hydrocarbon reserves and optimizing field development strategies. In this study, high-quality three-dimensional (3D) seismic reflection data, integrated with acoustic impedance (AI) inversion volume and borehole log data are used to estimate spatial distribution of fluid saturation within a sandstone-dominated Miocene reservoir interval of the Chandmari prospect in the North Assam Shelf, NE India. An ensemble-based random forest machine learning approach was implemented to predict fluid saturation within the target interval, leveraging the nonlinear relationships between seismic attributes and petrophysical properties. The model achieved a correlation coefficient of ~ 0.85 between predicted and measured saturation values, indicating good agreement with the available well data. The predicted water saturation (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:{S}_{w}\)</EquationSource> </InlineEquation>) within the reservoir ranges from 0.12 to 0.25, revealing significant spatial variability across the study area. It is observed that the south-eastern part of the reservoir exhibits comparatively low water saturation, suggesting favourable hydrocarbon accumulation. Furthermore, integration of seismic inversion results with well-log analysis confirms the presence of hydrocarbon within the sandstone reservoir. The methodology presented in this study highlights the broader applicability of machine learning–assisted seismic interpretation for reducing exploration risk and enhancing reservoir characterization in structurally complex sedimentary basins.</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

Random forest assisted machine learning approach for delimiting fluid saturation: a study from sandstone reservoir, North Assam shelf

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Jitender Kumar,
  • Priyadarshi Chinmoy Kumar,
  • Kalachand Sain

摘要

The occurrence and distribution of hydrocarbon in subsurface reservoirs are governed by a combination of geological factors, including lithological heterogeneity, burial depth, depositional environment, fluid migration pathways, and the extent of diagenetic alteration. Accurate characterization of reservoir properties, such as porosity and fluid saturation, is therefore essential for reliable estimation of hydrocarbon reserves and optimizing field development strategies. In this study, high-quality three-dimensional (3D) seismic reflection data, integrated with acoustic impedance (AI) inversion volume and borehole log data are used to estimate spatial distribution of fluid saturation within a sandstone-dominated Miocene reservoir interval of the Chandmari prospect in the North Assam Shelf, NE India. An ensemble-based random forest machine learning approach was implemented to predict fluid saturation within the target interval, leveraging the nonlinear relationships between seismic attributes and petrophysical properties. The model achieved a correlation coefficient of ~ 0.85 between predicted and measured saturation values, indicating good agreement with the available well data. The predicted water saturation ( \(\:{S}_{w}\) ) within the reservoir ranges from 0.12 to 0.25, revealing significant spatial variability across the study area. It is observed that the south-eastern part of the reservoir exhibits comparatively low water saturation, suggesting favourable hydrocarbon accumulation. Furthermore, integration of seismic inversion results with well-log analysis confirms the presence of hydrocarbon within the sandstone reservoir. The methodology presented in this study highlights the broader applicability of machine learning–assisted seismic interpretation for reducing exploration risk and enhancing reservoir characterization in structurally complex sedimentary basins.