Article
Energy & Fuels
Paulo Henrique Ranazzi, Xiaodong Luo, Marcio Augusto Sampaio
Summary: This article proposes a novel adaptive localization scheme for ensemble-based history matching in petroleum reservoirs. The scheme combines the strengths of two existing techniques to achieve improved matching performance in a field-scale reservoir model with both local and non-local parameters.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2022)
Article
Computer Science, Interdisciplinary Applications
Fahim Forouzanfar, Xiao-Hui Wu
Summary: The paper introduces a workflow to search for solutions in parameter space by defining constraint domains and developing a constrained history matching algorithm. The constrained algorithm projects the solutions to the feasible domain in each optimization iteration to meet the constraints.
COMPUTATIONAL GEOSCIENCES
(2021)
Article
Computer Science, Interdisciplinary Applications
Michael Gineste, Jo Eidsvik
Summary: This study introduces an ensemble-based method, the iterative ensemble Kalman smoother, to address the challenge of seismic waveform data inversion. By partitioning the data recordings in time windows and processing them sequentially, adaptive estimation with a focus on reliability and efficiency is shown. Different approaches to balancing the contributions from prior and likelihood are formulated and evaluated to achieve robust estimation performance in ensemble-based seismic waveform inversion.
COMPUTATIONAL GEOSCIENCES
(2021)
Article
Engineering, Civil
H. Delottier, R. Therrien, N. L. Young, D. Paradis
Summary: Integrated surface and subsurface hydrologic models are important for simulating baseflow and managing water resources at the regional scale. However, their computational costs and the additional post-processing steps required for estimating baseflow hinder their application. This study presents a hybrid approach that combines a surface water mass balance module with the HydroGeoSphere model to address these challenges and provide an efficient and accurate integrated model for low-flow processes.
JOURNAL OF HYDROLOGY
(2022)
Article
Computer Science, Interdisciplinary Applications
Geir Evensen
Summary: When formulating the history-matching problem using Bayes' theorem, it is important to consider the uncertainty of model parameters and errors in observed data to ensure an improved posterior ensemble of prediction models that better cover actual observations.
COMPUTATIONAL GEOSCIENCES
(2021)
Article
Water Resources
Tianhao He, Nanzhe Wang, Dongxiao Zhang
Summary: The study introduces a surrogate model based on a full convolutional neural network to solve groundwater contamination issues, which demonstrates strong generalization capabilities and reduced time consumption in inverse problem solving.
ADVANCES IN WATER RESOURCES
(2021)
Article
Computer Science, Interdisciplinary Applications
Zidong Pan, Wenxi Lu, Han Wang, Yukun Bai
Summary: A novel ensemble learning search framework using auto extreme gradient boosting tree was proposed to solve groundwater contaminant source identification (GCSI) problem. The framework achieved improved search accuracy and efficiency by employing boosting strategy (BOS) and auto xgboost.
ENVIRONMENTAL MODELLING & SOFTWARE
(2023)
Article
Engineering, Civil
Zidong Pan, Wenxi Lu, Yukun Bai
Summary: In this study, an adaptive-correction iterative ensemble smoother (ACIES) is proposed to adjust the range of unknown variables and improve the estimation accuracy. The ACIES method is applied to groundwater contaminated source estimation (GCSE) and utilizes a surrogate model to reduce calculation cost. A swarm evolutionary algorithm is used to tune the hyper-parameters of the surrogate model, specifically the auto-lightgbm surrogate, to ensure high fidelity and fast tuning time. The results show that ACIES can provide accurate estimation in the presence of vague prior ranges, and the auto-lightgbm surrogate achieves promising generalization accuracy with fast tuning time.
JOURNAL OF HYDROLOGY
(2023)
Article
Environmental Sciences
Hengnian Yan, Chenyu Hao, Jiangjiang Zhang, Walter A. Illman, Guang Lin, Lingzao Zeng
Summary: In this study, a supervised dimension reduction method, active subspace (AS) method, is proposed for groundwater models, addressing the challenge of high-dimensional parameters. By developing a cluster-based gradient-free AS (GFAS) method and combining it with Gaussian process regression, efficiency in data assimilation can be improved. A compensation scheme is also suggested to handle the issue of uncertainty underestimation caused by dimension reduction.
WATER RESOURCES RESEARCH
(2021)
Article
Engineering, Petroleum
Ricardo Vasconcellos Soares, Xiaodong Luo, Geir Evensen, Tuhin Bhakto
Summary: This study demonstrates the practical advantages of a new local analysis scheme in a 4D seismic history-matching problem. Compared to the Kalman gain localization scheme, the proposed local analysis scheme has improved capacity in handling big models and data sets, leading to faster convergence to the same level of data mismatch values.
Article
Mathematics, Applied
Tuan Nguyen Huy, Luu Vu Cam Hoan, Yong Zhou, Tran Ngoc Thach
Summary: This study investigates a Cauchy problem for the stochastic elliptic equation driven by Wiener noise, showing it is not well-posed through a simple illustrative example. To regularize the unstable solution, a regularization method called Fourier truncated expansion method is applied, with further investigation into the convergence rate of the regularized solution.
MATHEMATICAL METHODS IN THE APPLIED SCIENCES
(2021)
Article
Operations Research & Management Science
Alexandre Colaers Andersen, Konstantin Pavlikov, Tulio A. M. Toffolo
Summary: The paper explores the weapon-target assignment problem and applies linearization techniques to solve the nonlinear combinatorial optimization problem. By approximating the problem using linear functions, heuristic and exact solutions are provided. The proposed branch-and-adjust method can handle large-scale problem instances within a reasonable computational runtime.
ANNALS OF OPERATIONS RESEARCH
(2022)
Article
Computer Science, Artificial Intelligence
Y. J. Zhang, X. D. Mu, X. W. Liu, X. Y. Wang, X. Zhang, K. Li, T. Y. Wu, D. Zhao, C. Dong
Summary: This paper presents a quantum circuit solution scheme based on the quantum approximate optimization algorithm for the minimum vertex cover problem. By adjusting the problem Hamiltonian expectation through parametric unitary transformation, the solution probability is improved, leading to exponential acceleration.
APPLIED SOFT COMPUTING
(2022)
Article
Energy & Fuels
Xiaodong Luo, Rolf J. Lorentzen, Tuhin Bhakta
Summary: In order to address the ubiquitous model errors in history matching problems, a new approach utilizing machine learning to characterize model errors was adopted and integrated into an ensemble-based matching framework. Experimental results demonstrated that this new method helped improve the quality of estimated reservoir models and enhanced the accuracy of production data forecasts.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2021)
Article
Environmental Sciences
Simin Jiang, Ruicheng Zhang, Jinbing Liu, Xuemin Xia, Xianwen Li, Maohui Zheng
Summary: Joint estimation of groundwater contaminant source characteristics and hydraulic conductivity is significant for contaminant transport models in heterogeneous subsurface media. This can be achieved through geostatistical modeling and groundwater inverse modeling, both of which accurately characterize hydraulic conductivities.
Article
Geochemistry & Geophysics
Kjersti Solberg Eikrem, Geir Naevdal, Morten Jakobsen
Summary: This work uses the Lippmann-Schwinger equation to model seismic waves in strongly scattering acoustic media, improving efficiency by developing new preconditioners based on randomized matrix approximations and hierarchical matrices. Experimental results demonstrate the excellent performance of the method on two 2-D models.
GEOPHYSICAL JOURNAL INTERNATIONAL
(2021)
Article
Energy & Fuels
Xiaodong Luo, Rolf J. Lorentzen, Tuhin Bhakta
Summary: In order to address the ubiquitous model errors in history matching problems, a new approach utilizing machine learning to characterize model errors was adopted and integrated into an ensemble-based matching framework. Experimental results demonstrated that this new method helped improve the quality of estimated reservoir models and enhanced the accuracy of production data forecasts.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2021)
Article
Computer Science, Interdisciplinary Applications
Xiaodong Luo
Summary: Iterative ensemble smoothers (IES) are advanced methods for solving history matching problems, derived from optimization-theoretic perspective. The article proposes a more generalized cost function to construct a group of novel ensemble data assimilation algorithms, called generalized IES (GIES). Experimental results show that many of the tested GIES algorithms outperform the original IES, highlighting the potential advantages of designing new ensemble data assimilation algorithms using the proposed framework.
COMPUTATIONAL GEOSCIENCES
(2021)
Article
Geochemistry & Geophysics
Kui Xiang, Kjersti Solberg Eikrem, Morten Jakobsen, Geir Naevdal
Summary: In this study, a convergent scattering series solution for the frequency-domain wave equation in acoustic media with variable density and velocity has been derived. Through the homotopy analysis method, an iterative solution for the vectorial Lippmann-Schwinger equation has been obtained, achieving convergence even in strongly scattering media. The computational cost of the developed algorithm scales as N2 and involves a convergence control operator selected using hierarchical matrices.
GEOPHYSICAL PROSPECTING
(2022)
Article
Energy & Fuels
Micheal B. Oguntola, Rolf J. Lorentzen
Summary: The paper introduces a new efficient, robust, and accurate optimal solution strategy for non-linear constrained optimization problems using the exterior penalty function method and the adaptive ensemble-based optimization approach. This strategy aims to address the issues faced with current constraint handling techniques in EnOpt method, showing faster convergence rate, accuracy, and robustness in solving practical optimization problems.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2021)
Article
Energy & Fuels
Dean S. Oliver, Kristian Fossum, Tuhin Bhakta, Ivar Sando, Geir Naevdal, Rolf Johan Lorentzen
Summary: Reservoir simulation models require a large number of parameters to predict future reservoir behavior while being constrained by various factors to reduce uncertainty. The use of seismic surveys to observe changes in reservoir properties between wells offers potential in reducing prediction uncertainty, but faces challenges in practical applications.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2021)
Article
Computer Science, Interdisciplinary Applications
Xiaodong Luo, William C. Cruz
Summary: This work investigates an ensemble-based workflow to handle generic, nonlinear equality and inequality constraints in reservoir data assimilation problems. The proposed workflow is built upon a generalized iterative ensemble smoother (GIES) algorithm and can handle cost functions beyond nonlinear least squares. It treats data assimilation with constraints as a constrained optimization problem and derives a closed form to iteratively update model variables. The study demonstrates the potential of the proposed workflow to improve data assimilation performance in practical reservoir data assimilation problems.
COMPUTATIONAL GEOSCIENCES
(2022)
Article
Biophysics
Geir Naevdal, Einar K. Rofstad, Kjetil Soreide, Steinar Evje
Summary: Pancreatic cancer has a propensity for early metastasis, even for small early stage tumors. A computer model simulating tumor progression suggests high interstitial fluid pressure (IFP) as a possible driver for metastasis, shedding light on the clinical aggressiveness of pancreatic cancer.
JOURNAL OF BIOMECHANICS
(2022)
Article
Mechanics
Xin-Lei Zhang, Heng Xiao, Xiaodong Luo, Guowei He
Summary: In this work, an ensemble Kalman method is proposed to learn a nonlinear eddy viscosity model using a tensor basis neural network. By training the neural-network-based turbulence model with indirect observation data, the method proves to be effective in correctly learning the underlying turbulence models and predicting flows in similar configurations.
JOURNAL OF FLUID MECHANICS
(2022)
Article
Geochemistry & Geophysics
Kui Xiang, Morten Jakobsen, Kjersti Solberg Eikrem, Geir Naevdal
Summary: The distorted Born iterative method reduces a nonlinear inverse scattering problem to (ill-posed) linear inverse scattering problems that can be solved using a regularized least-squares formulation. It has been applied to electromagnetic and acoustic problems in three dimensions and to seismic problems for moderately large two-dimensional models.
GEOPHYSICAL PROSPECTING
(2023)
Article
Engineering, Mechanical
Steinar Evje, Hans Joakim Skadsem, Geir Naevdal
Summary: Conservation laws of the generic form c(t)+f(c)(x)=0 are important in engineering, but identifying unknown flux functions from observation data is challenging. This study explores a Bayesian method combined with iterative ensemble Kalman filtering to learn unknown nonlinear flux functions. Experimental results demonstrate the method's strong ability to identify unknown flux functions.
NONLINEAR DYNAMICS
(2023)
Article
Green & Sustainable Science & Technology
Kjersti Solberg Eikrem, Rolf Johan Lorentzen, Ricardo Faria, Andreas Storksen Stordal, Alexandre Godard
Summary: When planning wind farms, optimizing the layout is crucial for increasing production and reducing costs. This paper focuses on minimizing the levelized cost of energy (LCOE) for a floating wind farm using wind data in Porto Santo, Portugal. The ensemble based optimization (EnOpt) method is employed to handle the layout problem with multiple constraints, and the extended version called EPF-EnOpt outperforms other methods in reducing LCOE and computational cost. It is also found that EPF-EnOpt performs the best in maximizing annual energy production without considering costs.
Article
Energy & Fuels
William Chalub Cruz, Xiaodong Luo, Kurt Rachares Petvipusit
Summary: In order to improve reliability and reduce uncertainties, reservoir models need to be conditioned on field data through history matching. However, the joint history matching of production and inter-well tracer data remains challenging due to the lack of a coherent quantitative workflow. This study proposes a non-intrusive and derivative-free ensemble history matching workflow that successfully integrates reservoir models based on both production and inter-well tracer data.
GEOENERGY SCIENCE AND ENGINEERING
(2023)
Article
Engineering, Petroleum
Ricardo Vasconcellos Soares, Xiaodong Luo, Geir Evensen, Tuhin Bhakto
Summary: This study demonstrates the practical advantages of a new local analysis scheme in a 4D seismic history-matching problem. Compared to the Kalman gain localization scheme, the proposed local analysis scheme has improved capacity in handling big models and data sets, leading to faster convergence to the same level of data mismatch values.
Article
Geosciences, Multidisciplinary
Chuan-An Xia, Xiaodong Luo, Bill X. Hu, Monica Riva, Alberto Guadagnini
Summary: In this study, we used the MEs-EnKF approach to investigate the relationship between conductivity estimates and the type of available hydraulic head information in a heterogeneous groundwater field. Our results show that monitoring wells of Type A provide the best quality estimates, while Type B and C wells yield similar quality estimates. Additionally, inflating the measurement-error covariance matrix can improve conductivity estimates in simplified flow models.
HYDROLOGY AND EARTH SYSTEM SCIENCES
(2021)