Estimation of Real-Time Bottomhole Parameters in CO2 Injection Wells During Operations by Means of an Ensemble of Neural Networks

Author:

Figueroa J.1,Baraldi P.1,Chouybat I.1,Ursini F.2,Vignati E.2,Zio E.3

Affiliation:

1. Department of Energy, Politecnico di Milano, Milan, Italy

2. ENI, Milan, Italy

3. Department of Energy, Politecnico di Milano, Milan, Italy / MINES Paris-PSL, CRC, Sophia Antipolis, France

Abstract

Abstract Effective management of CO2 storage operations in subsurface reservoirs demands real-time estimation of well performance to identify deviations from expected operating conditions and ensure compliance with the requirements of Carbon Capture and Storage monitoring plans. To overcome the limitations of analytical tools in accurately estimating bottomhole flowing parameters, particularly during two-phase flow, this work aims to develop an empirical mapping between monitored parameters at various well sections and the bottomhole flowing parameters. Specifically, an ensemble of neural networks (NNs) is developed for the estimation of the bottom hole pressure (FBHP) and temperature (FBHT) in a CO2 injection well during operations. A feature selection approach based on trial-and-error is employed for the selection of the model inputs among the monitored parameters. The data used to train the NNs are generated using a thermo-fluid dynamic numerical simulator. Diversity between the NNs is achieved during the training phase by using different subsets of data and different weight initializations. The final FBHP and FBHT estimations are the median of the individual NN outcomes. The methodology has been applied to several cases in all flow regimes and different regions of the pure CO2 phase diagram, including single-phase, supercritical and two-phase transition. The proposed ensemble of NNs achieves accurate estimations of FBHP and FBHT, with root mean square errors (RMSE) below 1 bar and 0.1 °C, respectively, significantly outperforming an individual NN. This is due to the diversity among the NNs of the ensemble, which allows obtaining individual models that compensate for their errors, and the use of the median for the aggregation of the NN estimations, which make the ensemble robust towards the presence of possible outlier estimations provided by individual NNs. To our knowledge, this is the first application of machine learning solutions to real-time estimate bottomhole properties in CO2 injection wells. The satisfactory performance achieved on simulated data in a wide range of operating conditions, including two-phase transition, paves the way for the deployment of an ensemble of NNs for field application during operations.

Publisher

SPE

Reference19 articles.

1. Directive 2009/31/EC of the European Parliament and of the Council of 23 April 2009 on the geological storage of carbon dioxide and amending Council Directive 85/337/EEC, European Parliament and Council Directives 2000/60/EC, 2001/80/EC, 2004/35/EC, 2006/12/EC, 2008/1/EC and Regulation (EC) No 1013/2006;Parliament,2009

2. Action, Implementation of directive 2009/31/EC on the geological storage of carbon dioxide -- Guidance document 2, characterisation of the storage complex, CO2 stream composition, monitoring and corrective measures;Commission,2011

3. A. Bilogan , V.Elichev, H.Binn-Tahir, R.Khabibullin, and K.Litvinenko, ‘A real-time well integrity monitoring workflow based on artificial intelligence and dynamic flow modeling’, in BHR International Conference on Multiphase Production Technology, 2019, p. BHR-2019.

4. A. Shchipanov , G.Namazova, and K.Muradov, ‘Well Performance Metrics Suitable for Automated Monitoring’, in SPE EuropEC-Europe Energy Conference featured at the 84th EAGE Annual Conference & Exhibition, 2023.

5. S.K. Sahu , C.Lauwerys, Z.Abdullah, M.J.Muhammad, and D.Agrawal, ‘Journey of effectively using real-time production surveillance tool in digital transformation and well, reservoir and facility management improvements’, in SPE Asia Pacific Oil and Gas Conference and Exhibition, 2021, p. D021S014R001.

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