東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Deep and Machine Learning Surrogate ...

Tang, Haoyu,

Linked to FindBook

Google Book

Amazon

博客來

Deep and Machine Learning Surrogate Models for Optimization of Well Placement and Control in Subsurface Operations /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Deep and Machine Learning Surrogate Models for Optimization of Well Placement and Control in Subsurface Operations // Haoyu Tang.
Author:	Tang, Haoyu,
Description:	1 electronic resource (181 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-05, Section: A.
Contained By:	Dissertations Abstracts International86-05A.
Subject:	Problem solving. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31710848
ISBN:	9798346570202

Deep and Machine Learning Surrogate Models for Optimization of Well Placement and Control in Subsurface Operations /
Tang, Haoyu,

Deep and Machine Learning Surrogate Models for Optimization of Well Placement and Control in Subsurface Operations /Haoyu Tang. - 1 electronic resource (181 pages)

Source: Dissertations Abstracts International, Volume: 86-05, Section: A.

The efficient design and operation of subsurface flow processes, such as oil and gas production, geological CO2storage, geothermal energy extraction, and freshwater aquifer management, require solving complex nonlinear optimization problems. These optimizations can involve determining optimal well locations, drilling schedules, and time-varying settings for each well. Well placement optimization is challenging due to the presence of multiple local optima, making gradient-based methods less effective. Consequently, metaheuristic global stochastic search algorithms are often applied, but these are computationally expensive due to the large number of multiphase flow simulations required to evaluate potential candidate solutions. These high computational costs pose challenges, necessitating the development of efficient and accurate surrogate models to accelerate the optimization process.In this dissertation, we first apply machine-learning techniques to improve the efficiency of well placement optimization in oil production problems. This entails the development of an optimization framework that leverages low-fidelity (LF) models constructed from high-fidelity (HF) geomodels using a global transmissibility upscaling procedure. Tree-based machine-learning methods, specifically random forest and light gradient boosting machine, are used to estimate the error in the objective function (net present value, NPV) computed using LF models. During the preprocessing stage, differential evolution (DE) optimizations are conducted on LF models, and a representative set of well configurations is selected for training. In the runtime stage, DE optimization is performed with machine-learning-corrected LF models. This approach is tested on two cases involving vertical wells in 3D geomodels. Our method demonstrates significant computational speedup compared to HF models while maintaining accuracy, making it a viable option for efficient well placement optimization.Next, we propose a graph network surrogate model (GNSM) for optimizing well placement and controls. The GNSM converts the flow model into a computational graph using an encoding-processing-decoding architecture. Separate networks provide global predictions for pressure and saturation state variables. GNSM performance is enhanced by incorporating the single-phase steady-state pressure solution as a feature. A multistage multistep training strategy is employed. The trained GNSM is applied to predict oil-water flow responses in a 2D unstructured model of a channelized reservoir. Results are presented for numerous test cases, with injection and production wells placed randomly. The GNSM achieves low median relative errors in pressure and saturation and demonstrates reasonably accurate predictions for new permeability realizations. The GNSM is then used to optimize well locations and controls. GNSM-based optimization results for NPV are shown to be comparable to those from simulation-based optimization. Runtime speedup of about a factor of 36 is achieved with the GNSM procedure.Finally, we extend the GNSM methodology for application to CO2 storage operations. The GNSM framework is enhanced to handle 3D structured models with horizontal CO2injection wells that are perforated in multiple grid blocks. Additional features and modified training strategies are incorporated to treat these more challenging models. The GNSM is trained to predict global pressure and saturation fields, as well as bottom-hole pressures (BHPs) for all wells, at a set of time steps.

English

ISBN: 9798346570202Subjects--Topical Terms:

516855
Problem solving.

Deep and Machine Learning Surrogate Models for Optimization of Well Placement and Control in Subsurface Operations /
LDR:04966nmm a22004213i 4500 001 2432474
005 20250707081510.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798346570202
035 $a (MiAaPQD)AAI31710848
035 $a (MiAaPQD)Stanfordgb544kk0409
035 $a AAI31710848
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Tang, Haoyu, $e author. $3 3821708
245 1 0 $a Deep and Machine Learning Surrogate Models for Optimization of Well Placement and Control in Subsurface Operations / $c Haoyu Tang.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (181 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-05, Section: A.
500 $a Advisors: Durlofsky, Louis Committee members: Horne, Roland; Volkov, Oleg; Bent, Stacey F.
502 $b Ph.D. $c Stanford University $d 2024.
520 $a The efficient design and operation of subsurface flow processes, such as oil and gas production, geological CO2storage, geothermal energy extraction, and freshwater aquifer management, require solving complex nonlinear optimization problems. These optimizations can involve determining optimal well locations, drilling schedules, and time-varying settings for each well. Well placement optimization is challenging due to the presence of multiple local optima, making gradient-based methods less effective. Consequently, metaheuristic global stochastic search algorithms are often applied, but these are computationally expensive due to the large number of multiphase flow simulations required to evaluate potential candidate solutions. These high computational costs pose challenges, necessitating the development of efficient and accurate surrogate models to accelerate the optimization process.In this dissertation, we first apply machine-learning techniques to improve the efficiency of well placement optimization in oil production problems. This entails the development of an optimization framework that leverages low-fidelity (LF) models constructed from high-fidelity (HF) geomodels using a global transmissibility upscaling procedure. Tree-based machine-learning methods, specifically random forest and light gradient boosting machine, are used to estimate the error in the objective function (net present value, NPV) computed using LF models. During the preprocessing stage, differential evolution (DE) optimizations are conducted on LF models, and a representative set of well configurations is selected for training. In the runtime stage, DE optimization is performed with machine-learning-corrected LF models. This approach is tested on two cases involving vertical wells in 3D geomodels. Our method demonstrates significant computational speedup compared to HF models while maintaining accuracy, making it a viable option for efficient well placement optimization.Next, we propose a graph network surrogate model (GNSM) for optimizing well placement and controls. The GNSM converts the flow model into a computational graph using an encoding-processing-decoding architecture. Separate networks provide global predictions for pressure and saturation state variables. GNSM performance is enhanced by incorporating the single-phase steady-state pressure solution as a feature. A multistage multistep training strategy is employed. The trained GNSM is applied to predict oil-water flow responses in a 2D unstructured model of a channelized reservoir. Results are presented for numerous test cases, with injection and production wells placed randomly. The GNSM achieves low median relative errors in pressure and saturation and demonstrates reasonably accurate predictions for new permeability realizations. The GNSM is then used to optimize well locations and controls. GNSM-based optimization results for NPV are shown to be comparable to those from simulation-based optimization. Runtime speedup of about a factor of 36 is achieved with the GNSM procedure.Finally, we extend the GNSM methodology for application to CO2 storage operations. The GNSM framework is enhanced to handle 3D structured models with horizontal CO2injection wells that are perforated in multiple grid blocks. Additional features and modified training strategies are incorporated to treat these more challenging models. The GNSM is trained to predict global pressure and saturation fields, as well as bottom-hole pressures (BHPs) for all wells, at a set of time steps.
546 $a English
590 $a School code: 0212
650 4 $a Problem solving. $3 516855
650 4 $a Error correction & detection. $3 3480646
650 4 $a Permeability. $3 915594
650 4 $a Neural networks. $3 677449
650 4 $a Water. $3 516619
650 4 $a Optimization algorithms. $3 3683616
650 4 $a Aquifers. $3 551140
650 4 $a Geology. $3 516570
650 4 $a Finance. $3 542899
650 4 $a Hydrologic sciences. $3 3168407
650 4 $a Petroleum engineering. $3 566616
650 4 $a Water resources management. $3 794747
690 $a 0372
690 $a 0800
690 $a 0508
690 $a 0388
690 $a 0765
690 $a 0595
710 2 $a Stanford University. $e degree granting institution. $3 3765820
720 1 $a Durlofsky, Louis $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-05A.
790 $a 0212
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31710848