Scholarlink

<p>This study evaluates porosity systems in the Central Luconia Miocene carbonate reservoirs using systematic literature analysis (2014–<i>2025</i>) and original image-based data. The 317 thin-section images from the X Field are the same dataset reported in our empirical paper, and this review is <i>benchmarked</i> against those results. To estimate porosity distributions, 317 thin-section images were analysed using four deep learning segmentation models UNet, SegNet, PSPNet, and Fully Convolutional Network (FCN); UNet yielded the best performance among the tested architectures. We embed the pore-scale analysis in a <i>validation ladder</i> (manual point counts → <i>NMR T2 rock types</i> → <i>µCT/SEM</i> spot checks) and outline <i>multi-attribute seismic fusion</i> <b>(</b>e.g., <i>MSAT + PNN</i> / ensembles) to upscale pore-scale classes to seismic-scale facies/flow-unit models. The findings suggest that porosity in the reservoir is best characterised by a multiphase model comprising karstification-based, matrix-based, and fracture-based components. The analysis indicates that matrix-based porosity contributes approximately 40–45% of total porosity, followed by karstification features (35–40%) and fracture networks (15–20%). Methodologically, our approach differs from seismic-only or logs-only ML by <i>starting at the pore scale</i> to capture carbonate microfacies heterogeneity and then <i>linking quantitatively</i> to the field scale via the validation and upscaling pipeline; recent case studies in attribute fusion, resolution/bandwidth enhancement, automatic interpretation, diffraction-aware imaging, and AI-based permeability prediction reinforce this route-to-practice. This framework has practical implications for <i>sweet-spot ranking</i>, <i>pattern-aware completions</i>, and risk-aware screening, and it aligns with Sustainable Development Goal 7 (Affordable and Clean Energy) by supporting the more efficient development of underexplored carbonate systems.</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

Porosity estimation and reservoir characterisation in Central Luconia Miocene carbonates: integrating systematic review with deep learning analysis of thin sections

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • Abdulrahman Danlami Isa,
  • Haylay Tsegab Gebretsadik,
  • Abdulrahman Muhammad,
  • Hassan Salisu Mohammed,
  • Mohammed Gali

摘要

This study evaluates porosity systems in the Central Luconia Miocene carbonate reservoirs using systematic literature analysis (2014–2025) and original image-based data. The 317 thin-section images from the X Field are the same dataset reported in our empirical paper, and this review is benchmarked against those results. To estimate porosity distributions, 317 thin-section images were analysed using four deep learning segmentation models UNet, SegNet, PSPNet, and Fully Convolutional Network (FCN); UNet yielded the best performance among the tested architectures. We embed the pore-scale analysis in a validation ladder (manual point counts → NMR T2 rock types → µCT/SEM spot checks) and outline multi-attribute seismic fusion (e.g., MSAT + PNN / ensembles) to upscale pore-scale classes to seismic-scale facies/flow-unit models. The findings suggest that porosity in the reservoir is best characterised by a multiphase model comprising karstification-based, matrix-based, and fracture-based components. The analysis indicates that matrix-based porosity contributes approximately 40–45% of total porosity, followed by karstification features (35–40%) and fracture networks (15–20%). Methodologically, our approach differs from seismic-only or logs-only ML by starting at the pore scale to capture carbonate microfacies heterogeneity and then linking quantitatively to the field scale via the validation and upscaling pipeline; recent case studies in attribute fusion, resolution/bandwidth enhancement, automatic interpretation, diffraction-aware imaging, and AI-based permeability prediction reinforce this route-to-practice. This framework has practical implications for sweet-spot ranking, pattern-aware completions, and risk-aware screening, and it aligns with Sustainable Development Goal 7 (Affordable and Clean Energy) by supporting the more efficient development of underexplored carbonate systems.