Scholarlink

<p>Accurate characterization and forecasting of Bottom-Hole Closed-In Pressure (BHCP) are critical for effective reservoir management, especially in highly heterogeneous carbonate reservoirs like the Mishrif Formation in the West Qurna-1 oilfield, southern Iraq, where complex pressure depletion behavior has been observed during prolonged production. However, existing studies in Iraqi carbonate reservoir have largely focused on localized or short–term pressure analysis and lack an integrated framework that simultaneously reconstruct long–term spatio–temporal behavior while explicitly quantifying predictive uncertainty. This study tackles the challenge of reconstructing the historical spatio-temporal evolution of BHCP and forecasting its future behavior amidst ongoing production and injection. Utilizing annual BHCP data from 2010 to 2023, the historical pressure distributions were reconstructed through an analysis of several geostatistical interpolation techniques, namely Simple Kriging (SK), Ordinary Kriging (OK), and Empirical Bayesian Kriging (EBK). A Monte Carlo Simulation (MCS) framework coupled with a Random Forest machine-learning algorithm was then developed to forecast BHCP distributions for the period 2024–2030 under varying production and injection scenarios. The findings indicate that EBK consistently surpassed conventional kriging techniques in historical BHCP mapping, achieving the lowest prediction errors (with RMSE values as low as 60.17 psi). Meanwhile, the Random Forest–based MCS model delivered outstanding predictive accuracy (R² = 0.99 for training and 0.99 for testing, with RMSE below 11.00 psi), demonstrating its strong capability to capture nonlinear pressure dynamics. Forecast results indicate a gradual but persistent decline in average BHCP from approximately 2943 psi in 2024 to 2874 psi in 2030, despite ongoing water injection, highlighting spatially variable pressure depletion across the reservoir. These results highlight that existing injection strategies inadequately compensate for production-driven pressure decline, with reservoir heterogeneity exerting significant control over pressure behavior. The integrated geostatistical–machine-learning framework presented in this study provides a robust tool for pressure monitoring, forecasting, and optimization of spatially targeted reservoir management strategies aimed at mitigating depletion and improving long-term recovery.</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

Modeling the bottom-hole closed-in pressure of the Mishrif formation at west Qurna-1 oilfield, Iraq

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Mustafa A. Abdulhasan,
  • Amna M. Handhal,
  • Alaa M. Al-Abadi

摘要

Accurate characterization and forecasting of Bottom-Hole Closed-In Pressure (BHCP) are critical for effective reservoir management, especially in highly heterogeneous carbonate reservoirs like the Mishrif Formation in the West Qurna-1 oilfield, southern Iraq, where complex pressure depletion behavior has been observed during prolonged production. However, existing studies in Iraqi carbonate reservoir have largely focused on localized or short–term pressure analysis and lack an integrated framework that simultaneously reconstruct long–term spatio–temporal behavior while explicitly quantifying predictive uncertainty. This study tackles the challenge of reconstructing the historical spatio-temporal evolution of BHCP and forecasting its future behavior amidst ongoing production and injection. Utilizing annual BHCP data from 2010 to 2023, the historical pressure distributions were reconstructed through an analysis of several geostatistical interpolation techniques, namely Simple Kriging (SK), Ordinary Kriging (OK), and Empirical Bayesian Kriging (EBK). A Monte Carlo Simulation (MCS) framework coupled with a Random Forest machine-learning algorithm was then developed to forecast BHCP distributions for the period 2024–2030 under varying production and injection scenarios. The findings indicate that EBK consistently surpassed conventional kriging techniques in historical BHCP mapping, achieving the lowest prediction errors (with RMSE values as low as 60.17 psi). Meanwhile, the Random Forest–based MCS model delivered outstanding predictive accuracy (R² = 0.99 for training and 0.99 for testing, with RMSE below 11.00 psi), demonstrating its strong capability to capture nonlinear pressure dynamics. Forecast results indicate a gradual but persistent decline in average BHCP from approximately 2943 psi in 2024 to 2874 psi in 2030, despite ongoing water injection, highlighting spatially variable pressure depletion across the reservoir. These results highlight that existing injection strategies inadequately compensate for production-driven pressure decline, with reservoir heterogeneity exerting significant control over pressure behavior. The integrated geostatistical–machine-learning framework presented in this study provides a robust tool for pressure monitoring, forecasting, and optimization of spatially targeted reservoir management strategies aimed at mitigating depletion and improving long-term recovery.