Scholarlink

<p>This study develops a cognitive computing framework for reservoir characterisation and exploration optimisation in the Gabo Field, Niger Delta, with a focus on enhancing predictive accuracy for porosity, permeability, and fluid saturation. This research work aims to overcome the limitations related to geological complexity, shortage of available data, and complexity-related heterogeneity in the deltaic depositional environments. This work proposes a workflow that combines physics-based petrophysical calculations, Random Forest regression (RF), and Deepseek-R1 cognitive perception to improve predictive abilities, estimate uncertainties, and provide actionable intelligence related to reservoir management. Exploratory data analysis involved Akima spline interpolation and Isolation Forest algorithms. The RF model achieved superior predictive performance, with R² values above 0.98 for all predicted properties, and RMSE values below accepted thresholds. Predicted porosity values ranged between 0.18 and 0.25, clustering at 0.22–0.24, while permeability extended up to ~ 5230 mD, with several zones exceeding 500 mD, highlighting strong flow potential. Water saturation ranged between 0.25 and 0.45, suggesting favourable hydrocarbon saturation. Uncertainty quantification revealed low prediction errors (0.0062 v/v for porosity, 0.0040 log(mD) for permeability, and 0.0106 for saturation), confirming robustness and reliability. Deepseek-R1 cognitive evaluation identified potential bypassed pay zones and provided recommendations for enhanced recovery, including infill drilling and targeted waterflooding in high-permeability intervals. The integration of physics-based calculations, advanced machine learning, and cognitive computing demonstrates significant improvements in reservoir characterisation in geologically complex settings. This study delivers not only high predictive accuracy but also expert-level recommendations for exploration and field development. The proposed workflow contributes to digital intelligence in oil and gas exploration.</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

Cognitive artificial intelligence for automated reservoir analysis and prediction of porosity, permeability, and fluid saturation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月4日

查看全文
  • Fossong Guilianno,
  • Kingsley Onyekwere Okengwu,
  • Ugochi Adaku Okengwu

摘要

This study develops a cognitive computing framework for reservoir characterisation and exploration optimisation in the Gabo Field, Niger Delta, with a focus on enhancing predictive accuracy for porosity, permeability, and fluid saturation. This research work aims to overcome the limitations related to geological complexity, shortage of available data, and complexity-related heterogeneity in the deltaic depositional environments. This work proposes a workflow that combines physics-based petrophysical calculations, Random Forest regression (RF), and Deepseek-R1 cognitive perception to improve predictive abilities, estimate uncertainties, and provide actionable intelligence related to reservoir management. Exploratory data analysis involved Akima spline interpolation and Isolation Forest algorithms. The RF model achieved superior predictive performance, with R² values above 0.98 for all predicted properties, and RMSE values below accepted thresholds. Predicted porosity values ranged between 0.18 and 0.25, clustering at 0.22–0.24, while permeability extended up to ~ 5230 mD, with several zones exceeding 500 mD, highlighting strong flow potential. Water saturation ranged between 0.25 and 0.45, suggesting favourable hydrocarbon saturation. Uncertainty quantification revealed low prediction errors (0.0062 v/v for porosity, 0.0040 log(mD) for permeability, and 0.0106 for saturation), confirming robustness and reliability. Deepseek-R1 cognitive evaluation identified potential bypassed pay zones and provided recommendations for enhanced recovery, including infill drilling and targeted waterflooding in high-permeability intervals. The integration of physics-based calculations, advanced machine learning, and cognitive computing demonstrates significant improvements in reservoir characterisation in geologically complex settings. This study delivers not only high predictive accuracy but also expert-level recommendations for exploration and field development. The proposed workflow contributes to digital intelligence in oil and gas exploration.