利用人工智能和机器学习技术,采用人工神经网络开发并验证了用于油藏模拟历史拟合、敏感性分析和不确定性评估的智能代理模型,将其应用于油藏模拟的两个案例中。第1个案例研究了代理模型在油藏模型历史拟合中的应用,输出结果预测了井的产量;第2个案例研究了基于人工神经网络的代理模型在CO2提高采收率油藏快速建模中的应用,目标为预测油藏压力和相饱和度在注入期间以及注入后的分布,预测效果均良好。相比基础数值模拟模型,智能代理模型运行单次模拟只需几秒钟,总节省98.9%的运算时间。智能代理模型在运算速度、消耗时间以及成本等方面都有巨大的优势。此外,智能代理模型与基础油藏模型模拟结果非常接近。图19参30
SHAHKARAMI Alireza
,
MOHAGHEGH Shahab
. 智能代理在油藏建模中的应用[J]. 石油勘探与开发, 2020
, 47(2)
: 372
-382
.
DOI: 10.11698/PED.2020.02.16
Using artificial intelligence (AI) and machine learning (ML) techniques, we developed and validated the smart proxy models for history matching of reservoir simulation, sensitivity analysis, and uncertainty assessment by artificial neural network (ANN). The smart proxy models were applied on two cases, the first case study investigated the application of a proxy model for calibrating a reservoir simulation model based on historical data and predicting well production while the second case study investigated the application of an ANN-based proxy model for fast-track modeling of CO2 enhanced oil recovery, aiming at the prediction of the reservoir pressure and phase saturation distribution at injection stage and post-injection stage. The prediction effects for both cases are great. While a single run of basic numerical simulation model takes hours to days, the smart proxy model runs in a matter of seconds, saving 98.9% of calculating time. The results of these case studies demonstrate the advantage of the proposed workflow for addressing the high run-time, computational time and computational cost of numerical simulation models. In addition, these proxy models predict the outputs of reservoir simulation models with high accuracy.
[1] MULLER S, MILANO M, KOUMOUTSAKOS P.Application of machine learning algorithms to flow modeling and optimization[J]. Center for Turbulence Research Annual Research Briefs, 1999: 169-178.
[2] ANSARI A, MOHAGHEGH S, SHAHNAM M, et al. Data driven smart proxy for CFD application of big data analytics & machine learning in computational fluid dynamics, part three: Model building at the layer level[EB/OL]. (2018-04-02)[2019-07-02]. https://www. osti.gov/biblio/ 1431303.
[3] RAMPRASAD R, BATRA R, PILANIA G, et al.Machine learning in materials informatics: Recent applications and prospects[J]. npj Computational Materials, 2017, 3(1): 54.
[4] MARTINEZ-MARTINEZ F, RUPEREZ-MORENO M, MARTINEZ- SOBER M, et al.A finite element-based machine learning approach for modeling the mechanical behavior of the breast tissues under compression in real-time[J]. Computers in Biology and Medicine, 2017, 90: 116-124.
[5] ABHISHEK K, SINGH M, GHOSH S, et al.Weather forecasting model using artificial neural network[J]. Procedia Technology, 2012, 4: 311-318.
[6] WANG S, SUN S, XU J.Analysis of deep learning method for protein contact prediction in CASP12[J]. Proteins, 2018, 1: 67-77.
[7] PAGANINI M, de OLIVEIRA L, NACHMAN B. Accelerating science with generative adversarial networks: An application to 3D particle showers in multilayer calorimeters[J]. Physical Review Letters, 2018, 120(4): 042003.
[8] GOH G B, HODAS N O, VISHNU A.Deep learning for computational chemistry[J]. Journal of Computational Chemistry, 2017, 38(15): 1291-1307.
[9] LADICKY L, JEONG S, SOLENTHALER B, et al.Data-driven fluid simulations using regression forests[J]. ACM Transactions on Graphics, 2015, 34(6): 1-9.
[10] YANG C, YANG X, XIAO X.Data-driven projection method in fluid simulation[J]. Computer Animation & Virtual Worlds, 2016, 27(3/4): 415-424.
[11] SHAHKARAMI A, MOHAGHEGH S D, HAJIZADEH Y.Assisted history matching using pattern recognition technology[R]. SPE 173405, 2015.
[12] SHAHKARAMI A, AYERS K, WANG G, et al.Application of machine learning algorithms for optimizing future production in Marcellus shale: Case study of southwestern Pennsylvania[R]. SPE 191827, 2018.
[13] SHAHKARAMI A. Database for PUNQ-S3 case study[EB/OL]. (2019-06)[2019-07-02]. https://www.researchgate.net/publication/ 334139645_Smart_Proxy_Modeling_-_PUNQ-S3_HM_Database?channel=doi.
[14] SHAHKARAMI A. Database example for SACROC case study[EB/OL]. (2019-07)[2019-07-02]. https://www.researchgate.net/ publication/334139922_Sample_of_database_for_the_case_study_of_SCAROC_CO2_EOR_and_Storage?channel=doi&linkId=5d1a2e5aa6fdcc2462b69ae2&showFulltext=true.
[15] SHAHKARAMI A.Artificial intelligence assisted history matching: Proof of concept[D]. Morgantown, USA: West Virginia University, 2012.
[16] FLORIS F, BUSH M, CUYPERS M, et al.Methods for quantifying the uncertainty of production forecasts[J]. Petroleum Geoscience, 2001, 7: 87-96.
[17] Petroleum Engineering & Rock Mechanics Group. Standard models: PUNQ-S3 model[EB/OL]. [2019-07-02]. http://www.imperial.ac.uk/ earth-science/research/research-groups/perm/standard-models/.
[18] SHAHKARAMI A. PUNQ-S3 reservoir model (CMG format) for history matching study[EB/OL]. (2019-07)[2019-07-02]. https://www. researchgate.net/publication/334140032_PUNQ-S3_Reservoir_Model_CMG_format_for_History_Matching_Study.
[19] ERWIG M.The graph Voronoi diagram with applications[J]. Networks, 2000, 36(3): 156-163.
[20] GOMEZ Y, KHAZAENI Y, MOHAGHEGH S D, et al.Top down intelligent reservoir modeling[R]. SPE 124204, 2009.
[21] SHAHKARAMI A, MOHAGHEGH S, GHOLAMI V, et al.Modeling pressure and saturation distribution in a CO2 storage project using a surrogate reservoir model (SRM)[J]. Greenhouse Gases: Science and Technology, 2014, 4(3): 289-315.
[22] Computer Modelling Group LTD. Advanced compositional reservoir simulator[EB/OL]. [2019-07-02]. http://cmgl.ca/gem.
[23] RAINES M.Kelly-Snyder fields/SACROC unit[J]. West Texas Geological Society: Oil and Gas Fields in West Texas, 2005, 8: 69-78.
[24] GHOLAMI V, MOHAGHEGH S D, MAYSAMI M.Smart proxy modeling of SACROC CO2-EOR[J]. Fluids, 2019, 4(2): 85.
[25] HAN W.Evaluation of CO2 trapping mechanisms at the SACROC northern platform: Site of 35 years of CO2 injection[D]. Socorro, New Mexico: The New Mexico Institute of Mining and Technology, 2008.
[26] Intelligent Solution Inc. IDEATM[EB/OL]. [2019-07-02]. http://www. intelligentsolutionsinc.com/Products/IDEA.shtml.
[27] MOHAGHEGH S D, ABDULLA F A S. Production management decision analysis using AI-based proxy modeling of reservoir simulations: A look-back case study[R]. SPE 170664, 2014.
[28] CARUANA R, NICULESCU-MIZIL A.An empirical comparison of supervised learning algorithms[R]. Pittsburgh, USA: The 23rd International Conference on Machine Learning, 2006.
[29] SUGANUMA M, SHIRAKAWA S, NAGAO T.A genetic programming approach to designing convolutional neural network architectures[R]. Berlin, Germany: The 2017 Genetic and Evolutionary Computation Conference, 2017.
[30] SNOEK J, LAROCHELLE H, ADAMS R P.Practical Bayesian optimization of machine learning algorithms[R]. Lake Tahoe, Nevada: Advances in Neural Information Processing Systems 25, 2012.
