Hybrid physics and data-driven modeling for unconventional field development and its application to US onshore basin

被引:18
作者
Park, Jaeyoung [1 ,4 ]
Datta-Gupta, Akhil [1 ]
Singh, Ajay [2 ]
Sankaran, Sathish [3 ]
机构
[1] Texas A&M Univ, Petr Engn Dept, College Stn, TX 77843 USA
[2] Amazon Web Serv Oil & Gas, Houston, TX USA
[3] Xecta Digital Labs, Houston, TX USA
[4] Occidental Petr, Houston, TX 77046 USA
关键词
Unconventional reservoirs; Field development optimization; Machine learning; Well performance prediction; SENSITIVITY-ANALYSIS; RESERVOIRS; OPTIMIZATION;
D O I
10.1016/j.petrol.2021.109008
中图分类号
TE [石油、天然气工业]; TK [能源与动力工程];
学科分类号
0807 ; 0820 ;
摘要
The objective of this study is to develop a hybrid model by combining physics and data-driven approach for optimum unconventional field development. We used physics-based reservoir simulations to generate training datasets. These uncalibrated priors were incorporated into data-driven machine learning (ML) algorithms so that the algorithms can learn the underlying physics between reservoir simulation input and output, cumulative oil production for parent and child wells. The reservoir simulations in a single half-cluster model consists of injection and production phases. This allows the algorithm to relate the impact of completion design, well spacing and timing to the production profiles. After performing a sensitivity analysis to reduce the number of training inputs, more than 20,000 simulations results were generated as the training data. The best accuracy, R2 = 0.94, was achieved with the neural network model after tuning hyper-parameters. Then, we incorporated the hybrid model with the genetic algorithm to perform efficient history matching to calibrate model parameters and predict estimated ultimate recovery (EUR) and net present value with 10% discount rate (NPV10). The great efficiency in the hybrid model takes only several minutes to complete a single well history matching. The calibrated model shows that the prediction from the hybrid model is physically meaningful compared to actual reservoir simulations, capturing the impact of fracture geometry, child well spacing and timing on the well production. With multiple history matching results, we populated value acreage maps (e.g., EUR and NPV10) which can benefit the unconventional field development planning. Lastly, a blind test was conducted in an area of interest to demonstrate that the proposed workflow can effectively predict a well performance in the area without sufficient data, utilizing calibrated model parameters from history matched results of adjacent wells. The proposed model can provide quick and scalable solutions that honor underlying physics to help decision making on unconventional field development based on production profile and economic metrics with various completion designs.
引用
收藏
页数:15
相关论文
共 34 条
  • [1] [Anonymous], 2010, SPE ANN TECHNICAL C
  • [2] Baek S, 2019, UNC RES TECHN C URTE
  • [3] Fracture inference and optimal well placement using a multiscale history matching in a HPHT tight gas reservoir, Tarim Basin, China
    Chen, Hongquan
    Yang, Changdong
    Datta-Gupta, Akhil
    Zhang, Jianye
    Chen, Liqun
    Liu, Lei
    Chen, Baoxin
    Cui, Xiaofei
    Shi, Fashun
    Bahar, Asnul
    [J]. UPSTREAM OIL AND GAS TECHNOLOGY, 2020, 2
  • [4] Artificial Intelligence Applications in Reservoir Engineering: A Status Check
    Ertekin, Turgay
    Sun, Qian
    [J]. ENERGIES, 2019, 12 (15)
  • [5] Quantifying Asymmetric Parameter Interactions in Sensitivity Analysis: Application to Reservoir Modeling
    Fenwick, Darryl
    Scheidt, Celine
    Caers, Jef
    [J]. MATHEMATICAL GEOSCIENCES, 2014, 46 (04) : 493 - 511
  • [6] Numerical Investigation of Effects of Subsequent Parent-Well Injection on Interwell Fracturing Interference Using Reservoir-Geomechanics-Fracturing Modeling
    Guo, Xuyang
    Wu, Kan
    An, Cheng
    Tang, Jizhou
    Killough, John
    [J]. SPE JOURNAL, 2019, 24 (04): : 1884 - 1902
  • [7] Hetz G., 2017, SPE ANN TECHN C EXH
  • [8] Iino A., 2017, SPE/AAPG/SEG Unconventional Resources Technology Conference. Unconventional Resources Technology Conference, P21
  • [9] Iino A, 2019, SPE187116PA
  • [10] Conditioning well data to rule-based lobe model by machine learning with a generative adversarial network
    Jo, Honggeun
    Santos, Javier E.
    Pyrcz, Michael J.
    [J]. ENERGY EXPLORATION & EXPLOITATION, 2020, 38 (06) : 2558 - 2578