Extensive Study on the Influencing Parameters of Sc CO2 Foam Viscosity for Enhanced Oil Recovery and Carbon Sequestration: A Machine Learning Approach

Author:

Bashir Ahmed1,Kasha Ahmed2,Patil Shirish1,Aljawad Murtada Saleh1,Kamal Muhammad Shahzad3

Affiliation:

1. Department of Petroleum Engineering, King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia.

2. Midstream Production Systems Houston, SLB, Houston, Texas, USA.

3. Center for Integrative Petroleum Research, King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia.

Abstract

Abstract Foam flooding has been used to control gas mobility during oil displacement and CO2 sequestration processes in subsurface porous media, mitigating the negative impacts of low gas viscosity, reservoir heterogeneity, and gravity override. In this research, we study the application of machine learning (ML) to develop a data-driven prediction of the effective viscosity of supercritical CO2 foam (Sc-CO2) for enhanced oil recovery (EOR) and CO2 sequestration. The ML approach is used to overcome the challenge of using physical correlations to account for the effect of key experimental parameters on the viscosity of supercritical CO2 foam. The experimental data for evaluating the effective Sc-CO2 foam viscosity were measured using a high-pressure high-temperature foam rheometer (Model 8500) under different temperatures (50-110 °C), pressures (1000-3000 psi), foam qualities (30-90%), and surfactant concentrations (0.1-0.5 wt.%) at shear rates between 100-1450 s−1. A total of 5,552 data points were used as primary data for developing supervised ML regression models. Machine learning algorithms from the Scikit-learn library, such as K-Nearest Neighbors (KNN), Random Forest (RF), Gradient Boosting (GB), and AdaBoosting (AB), were used. The results revealed that machine learning algorithms generated models for the effective viscosity of Sc-CO2 foam with predictive accuracies of 0.989, 0.987, 0.941, and 0.723 for RF, KNN, GB, and AB, respectively. The RF and KNN algorithm demonstrated superior performance among all the other algorithms, with RF being better in terms of accurate viscosity prediction across different viscosity values. This paper provides data-driven approach that can predict the effective foam viscosity under different reservoir conditions which leads to the design of an optimum injection strategy and effectively controls Sc-CO2 mobility for EOR and CO2 sequestration.

Publisher

SPE

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3