Scholarlink

<p>The efficient development of tight sandstone gas reservoirs is severely challenged by profound heterogeneity, which leads to complex pore structures and highly variable permeability, making sweet-spot prediction extremely difficult. Traditional porosity-permeability models often fail in such reservoirs. This study proposes a novel workflow that integrates the hydraulic flow unit (HFU) concept with the XGBoost machine learning algorithm to accurately characterize and predict sweet-spots in the Shaximiao Formation (J<sub>2</sub>s) of the ZT Gas Field, Sichuan Basin. First, four distinct HFUs were classified based on the Flow Zone Indicator (FZI) derived from core data, effectively capturing the pore-throat characteristics controlling fluid flow. Subsequently, the XGBoost model was trained to predict HFU categories using conventional well logs (DEN, CNL, AC, RT) and derived parameters. The model achieved a high prediction accuracy (&gt;92%), enabling the application of HFU-based permeability models across non-cored intervals and wells. The spatial distribution of HFUs revealed that sweet-spots (HFUs I &amp; II) are predominantly developed in specific sedimentary microfacies. Critically, production data confirmed that high-rate wells are exclusively located within these predicted sweet-spot areas, validating the reliability of our method. This integrated approach provides a robust and practical tool for sweet-spot evaluation, which is crucial for optimizing well placement and development strategies in heterogeneous tight gas reservoirs.</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

Sweet-Spot Evaluation in a Highly Heterogeneous Tight Sandstone Gas Reservoir

  • 期刊论文

  • 发表日期：

    2026年1月1日

查看全文
  • Jian Cao,
  • Run Shu,
  • Jing Luo,
  • Xin Hu,
  • Shiyi Xie,
  • Xiao Tang,
  • Xuli Wang,
  • Qin Sang

摘要

The efficient development of tight sandstone gas reservoirs is severely challenged by profound heterogeneity, which leads to complex pore structures and highly variable permeability, making sweet-spot prediction extremely difficult. Traditional porosity-permeability models often fail in such reservoirs. This study proposes a novel workflow that integrates the hydraulic flow unit (HFU) concept with the XGBoost machine learning algorithm to accurately characterize and predict sweet-spots in the Shaximiao Formation (J2s) of the ZT Gas Field, Sichuan Basin. First, four distinct HFUs were classified based on the Flow Zone Indicator (FZI) derived from core data, effectively capturing the pore-throat characteristics controlling fluid flow. Subsequently, the XGBoost model was trained to predict HFU categories using conventional well logs (DEN, CNL, AC, RT) and derived parameters. The model achieved a high prediction accuracy (>92%), enabling the application of HFU-based permeability models across non-cored intervals and wells. The spatial distribution of HFUs revealed that sweet-spots (HFUs I & II) are predominantly developed in specific sedimentary microfacies. Critically, production data confirmed that high-rate wells are exclusively located within these predicted sweet-spot areas, validating the reliability of our method. This integrated approach provides a robust and practical tool for sweet-spot evaluation, which is crucial for optimizing well placement and development strategies in heterogeneous tight gas reservoirs.