Geological Uncertainty Research Articles

Capturing Geological Uncertainty in Salt Cavern Developments for Hydrogen Storage

Capturing Geological Uncertainty in Salt Cavern Developments for Hydrogen Storage

Intelligent optimization of horizontal wellbore trajectory based on reinforcement learning

Ultra-long horizontal wells are crucial for improving the production and developmental benefits of shale oil and gas wells. Currently, advanced downhole semi-closed-loop steering drilling technology depends on a dual communication mechanism between the surface and downhole, as well as the empirical decisions of human experts. However, it is difficult to acutely control the trajectory of ultra-long horizontal open hole drilling in a reservoir owing to geological uncertainties related to shale oil and gas, the uncertainty of the drilling tool's building capacity, and the lag of measurement information while drilling. This report proposes an adaptive target detection and design method for ultra-long horizontal well trajectories based on reinforcement learning. The developed approach enables dynamic identification of reservoir sequences according to logging-while-drilling data and prediction of horizontal targets in real-time, with a recognition accuracy of 90.1%. Additionally, measurement-while-drilling data are used to accurately characterize the depth, inclination, and azimuth of the target-entering process. The dynamic interaction mechanism between the bit and the target environment is simulated by defining the bit action, reward function, and an update mechanism. The drilling engineering conditions restrict the interactions, such that the bit automatically decides the direction in which to drill the targets, demonstrating 100% accuracy for the wellbore trajectory toward the target. The experimental application results indicate that the developed method can be applied to determine the dynamic target area of a horizontal well in real time and make intelligent downhole decisions regarding ultra-long horizontal section borehole trajectory adjustments while drilling and measuring. The drilling rate of high-quality reservoirs is therefore significantly improved. The results discussed herein provide insights to support further developments of downhole full-closed-loop intelligent autonomous decision-making borehole trajectory control.

Modelling hazards impacting the flow regime in the Hranice Karst due to the proposed Skalička Dam

Abstract. This study examines the hydrogeological hazard associated with the construction of the proposed Skalička Dam in the vicinity of the Hranice Karst. Prompted by the catastrophic regional floods in 1997 and 2010, the design of the dam aims to mitigate floods along the Bečva River downstream of the reservoir. However, concerns have been raised regarding the potential disturbance of the natural groundwater regime in the Hranice Karst and the source of mineral waters for the Teplice spa. This is due in particular to the dam's location in an area with limestone outcrops potentially susceptible to surface-water infiltration. Previous studies have also highlighted the strong correlation between the water level in the Bečva River and the water level in karst formations such as the Hranice Abyss, Zbrašov Aragonite Caves, and other caves in the locality. To address these concerns, a nonlinear reservoir-pipe groundwater flow model was employed to simulate the behaviour of the Hranice Karst aquifer and specifically the effects of the dam reservoir's impoundment. The study concluded that the lateral variant of the dam would have a practically negligible impact on the karst water system, with the rise in water level being only a few centimetres. The through-flow variant was found to have a more significant potential impact on water levels and the outflow of mineral water to the spa, with a piezometric rise of about 1 m and an increase in the karst water discharge to the Bečva River of more than 50 %. Based on these results, recommendations for further investigations concerning the design of the dam and its eventual construction have been formulated to reduce geological uncertainties and to minimize the potential impact of the hydraulic scheme on the hydrogeology of the karstic system.

Optimization of underground mining production layouts considering geological uncertainty using deep reinforcement learning

Mineral extraction plays a key role in the global raw materials supply chain, however the exhaustion of shallow deposits and typical scarcity of sampled data during exploration activities creates challenges in mine planning and design, where decision-making is highly sensitive to uncertainty in geology and mineral grade prediction. Geostatistical techniques are commonly used to generate a set of equiprobable simulated numerical models to capture these uncertainties, however incorporating these simulated models within a mine planning and design framework remains a major challenge. The purpose of this paper is to propose a novel approach to decision-making in underground mine design that can use information from an ensemble of numerical realizations of a mineral resource to improve the financial performance of the asset. A deep reinforcement learning (DRL) framework, using the proximal policy optimization (PPO) algorithm, is developed for the design of underground mining production level layouts. A case study is presented using a gold mineral resource characterized by an ensemble of 100 numerical realizations to verify the advantages of the proposed method, considering a baseline consisting of an industry standard automated design method. The DRL approach achieved an 8.3% higher expected profit, a 3.4% more gold mined than the baseline, and has the added functionality of considering uncertainty in mineral grades.

Surrogate model for reservoir performance prediction with time-varying well control based on depth generative network

This paper proposes a novel intelligent method for defining and solving the reservoir performance prediction problem within a manifold space, fully considering geological uncertainty and the characteristics of reservoirs performance under time-varying well control conditions, creating a surrogate model for reservoir performance prediction based on Conditional Evolutionary Generative Adversarial Networks (CE-GAN). The CE-GAN leverages conditional evolution in the feature space to direct the evolution of the generative network in previously uncontrollable directions, and transforms the problem of reservoir performance prediction into an image evolution problem based on permeability distribution, initial reservoir performance and time-varying well control, thereby enabling fast and accurate reservoir performance prediction under time-varying well control conditions. The experimental results in basic (egg model) and actual water-flooding reservoirs show that the model predictions align well with numerical simulations. In the basic reservoir model validation, the median relative residuals for pressure and oil saturation are 0.5% and 9.0%, respectively. In the actual reservoir model validation, the median relative residuals for both pressure and oil saturation are 4.0%. Regarding time efficiency, the surrogate model after training achieves approximately 160-fold and 280-fold increases in computational speed for the basic and actual reservoir models, respectively, compared with traditional numerical simulations. The reservoir performance prediction surrogate model based on the CE-GAN can effectively enhance the efficiency of production optimization.

Optimal monitoring design for uncertainty quantification during geologic CO[formula omitted] sequestration: A machine learning approach

An effective monitoring design is crucial to ensure the safe and permanent geologic storage of CO2. Optimal monitoring design involve an optimal placement of monitoring wells, and optimal monitoring measurement data (pressure, CO2 saturation, temperature, etc.). We developed a filtering-based data assimilation approach to design an optimal monitoring strategy for well placement and monitoring data design. To efficiently solve the optimization problem and reduce computational costs, Artificial Neural Networks are used to develop computationally efficient reduced-order models based on full-physics numerical simulations of CO2 injection in saline aquifers. We demonstrate our approach in two scenarios of CO2 leakage through legacy or abandoned wellbores where an optimal monitoring strategy are devised to reduce the uncertainty in cumulative CO2 leakage in the geologic CO2 sequestration (GCS) site. The optimal monitoring design resulted in an uncertainty reduction in the cumulative leakage of CO2 of approximately 73% and 62% in each case, respectively. The proposed approach is efficient in developing monitoring designs under geologic uncertainty and enables safe geologic carbon sequestration operations.

Adaptive Simultaneous Stochastic Optimization of the Escondida Mining Complex, Chile

AbstractThis paper presents the application of adaptive simultaneous stochastic optimization with a representative branching framework to generate a strategic mining plan for the Escondida mining complex, the world’s largest copper-production operation. This adaptive stochastic optimization considers geological uncertainty while integrating investment and operational alternatives in the production schedule of a mining complex. Mining complexes are comprised of interconnected components affected by multiple sources of uncertainty. Thus, they must be optimized simultaneously in order to maximize their value, manage environmental impacts, and minimize risk. Additionally, due to the extensive lives of assets and the dynamic and uncertain environment in which mining complexes operate, it is not reasonable to assume that the current strategic plan will remain optimal. Thus, an operationally feasible method to embed alternatives in the mine plan is used. The method utilized provides a strategic plan with representative branches for future possible investment decisions. Adaptive decisions are made sequentially over time, activating costs and effects over the model. The optimization process chooses the optimal strategic production plan accordingly, as well as the investments made and their timing. The Escondida mining complex is a multi-element, multi-pit operation with nine different processing destinations. Investment options considered herein are truck and shovel fleet sizing, adding a secondary crusher in one of the plants, and investing in a main crusher assigned to one of the pits. Additionally, operational alternatives at the mine and plant levels are included. The adaptive solution shows a substantial probability that the mine plan might change its design substantially due to geological uncertainty, presenting an increased expected net present value when compared to the previously developed stochastic mathematical programming formulation that does not consider adaptive decisions, thus generating a single static strategic production plan for the related mining complex. Further studies at the Escondida mining complex can consider adoptive solutions integrating capital investments pertinent to climate change issues.

Stochastic Modeling of Two-Phase Transport in Fractured Porous Media Under Geological Uncertainty Using an Improved Probabilistic Collocation Method

Summary Simulation of multiphase transport through fractured porous media is highly affected by the uncertainty in fracture distribution and matrix block size that arises from inherent heterogeneity. To quantify the effect of such uncertainties on displacement performance in porous media, the probabilistic collocation method (PCM) has been applied as a feasible and accurate approach. However, propagation of uncertainty during the simulation of unsteady-state transport through porous media could not be computed by this method or even by the direct-sampling Monte Carlo (MC) approach. Therefore, with this research, we implement a novel numerical modeling workflow that improves PCM on sparse grids and combines it with the Smolyak algorithm for selection of collocation points sets, Karhunen-Loeve (KL) decomposition, and polynomial chaos expansion (PCE) to compute the uncertainty propagation in oil-gas flow through fractured porous media in which gravity drainage force is enabled. The effect of uncertainty in the vertical dimension of matrix blocks, which are frequently an uncertain and history-matching parameter, on simulation results of randomly synthetic 3D fractured media is explored. The developed numerical model is innovatively coupled with solving governing deterministic partial differential equations (PDEs) to compute uncertainty propagation from the first timestep to the last timestep of the simulation. The uncertainty interval and aggregation of uncertainty in ultimate recovery are quantified, and statistical moments for simulation outputs are presented at each timestep. The results reveal that the model properly quantifies uncertainty and extremely reduces central processing unit (or CPU) time in comparison with MC simulation.

Uncertainty propagation from crustal geologies to rock-site ground motion with a Fourier Neural Operator

Seismic hazard analyses in the area of a nuclear installation must account for a large number of uncertainties, including limited geological knowledge. Combining the accuracy of physics-based simulations with the expressivity of deep neural networks can help to quantify the influence of crustal geological uncertainties on surface ground motion. This work uses a Factorised Fourier Neural Operator (F-FNO) to learn the relationship between 3D crustal heterogeneous geologies and time-dependent wavefields at the Earth’s surface up to a 5 Hz maximal frequency. The F-FNO is pretrained on a generic database and then, fine-tuned with only 250 samples targeted to the Le Teil region (South-Eastern France). The F-FNO correctly predicts the wave arrival times and the main wavefronts. As quantified by Goodness-Of-Fit (GOF) criteria, 83% of predictions have excellent phase GOF and 75% have very good envelope GOF. The Peak Ground Velocity and Pseudo-Spectral Acceleration are also accurate, especially for geologies with low-to-moderate heterogeneities. Thanks to the F-FNO speed, ground motion distributions can be easily computed and provide safety margins compared to 1D simulations. These results show that the F-FNO is an efficient surrogate model to quantify the range of ground motion a nuclear installation could face in the presence of geological uncertainties.

The Value of Drilling—A Chance-Constrained Optimization Approach

Managing uncertainty is a core challenge in mine planning. Mine planners often represent various planning variables, such as equipment performance and geological parameters, as random variables due to inherent uncertainties. This paper looks at geological uncertainty and its impact on mine planning. Some traditional approaches to manage this uncertainty include using conditional simulations or mathematical programming in the planning process. Drilling additional holes, despite its cost, is a common method to reduce uncertainty using additional samples to reduce deposit variance. In this paper, we first outline an ore blending optimization model which uses chance-constrained programming to manage property limit risk when selecting the order of ore feed into a processing facility. In coal mining, in tactical planning horizons, the order of coal seam removal is usually predetermined, allowing a blending model to ensure optimal feed properties. Using chance-constrained programming allows us to blend the uncertainties from geological models to maximize plant output while adhering to property constraints. We use the chance-constrained blending model to determine the value of additional information from infill drilling. The model prioritizes drilling locations that reduce uncertainty and improve blending outcomes. A case study on a coking coal mine in Queensland, Australia, demonstrates the model’s application, highlighting significant improvements in blending by reducing the variance of high-quality blocks. The study concludes that targeting high-quality blocks for variance reduction can better accommodate lower-quality material, offering a more valuable approach than the traditional focus of reducing uncertainty in low-quality blocks. This approach provides insights for improving mine planning strategies and showcases the potential of chance constraints in optimizing ore blending under uncertainty.

Incorporating Operational Modes into long-Term Open-Pit Mine Planning Under Geological Uncertainty: An Optimization Combining Variable Neighborhood Descent with Linear Programming

Incorporating Operational Modes into long-Term Open-Pit Mine Planning Under Geological Uncertainty: An Optimization Combining Variable Neighborhood Descent with Linear Programming

Simulation of geological uncertainty based on improved three-dimensional coupled Markov chain model

The natural stratigraphic distribution is uncertain due to the geological tectonic movements and sedimentation. It is a challenge for traditional deterministic models to capture geological uncertainty. The objective of this study is to develop an improved three-dimensional coupled Markov chain (ICMC3D) model based on limited boreholes. The proposed method enhances the traditional three-dimensional coupled Markov chain model by modifying the transition probability equation for predicting unknown elements. The dip data is added to estimate the transition probability matrix with the borehole information. On this basis, a back analysis method is proposed for estimating the horizontal transition probability matrics. The method makes the predicted results more suitable for the possible stratigraphic distribution. The accuracy of the proposed method is verified through some applications. In this paper, an improved three-dimensional coupled Markov chain model and transition probability matrix estimation method are illustrated by the borehole data collected from a tunnel project in Qingdao. The results show that the presented method can reflect the asymmetry, continuity and anisotropy of the three-dimensional stratum. The exceedance probability is proposed to predict the fault zone development range in order to reduce the geological uncertainty.

Strategic Geosteering Workflow with Uncertainty Quantification and Deep Learning: Initial Test on the Goliat Field Data

Continuous integration of real-time logging-while-drilling data into a subsurface model with relevant geological uncertainties enables strategic geosteering: a field-level optimization of the well-placement strategy. Model errors arising from oversimplified conceptual geological models and imperfect simulation of measurements result in unreliable subsurface-model updates. The model errors are particularly pronounced when synthetic measurements are approximated with a fast but imperfect model, such as a deep neural network (DNN).#xD;We present a practical data-assimilation workflow consisting of offline and online phases. The offline phase involves DNN training and building an uncertain prior near-well geo-model. The online phase utilizes the flexible iterative ensemble smoother (FlexIES) to perform real-time assimilation of extra-deep electromagnetic data while accounting for the model errors in the approximate DNN model. We demonstrate the proposed workflow on historic well-log data from the Goliat Field (Barents Sea). #xD;The median of our probabilistic estimation is on par with proprietary inversion, regardless of the number of layers in the chosen prior or the approximate DNN model. By estimating model errors, FlexIES automatically quantifies the uncertainty in the boundaries and resistivities of layers, which is not standard in proprietary inversion. #xD;This capability allows us to capture uncertainties more efficiently, thus providing input for future quantitative decision support methods. We demonstrate the potential of quantitive decision support by visually estimating the ahead-of-bit risk of reservoir exit that has occurred during the considered operation.

Influence of geological uncertainty and soil spatial variability on tunnel deformation and their importance evaluation

In geotechnical engineering, geological uncertainty and the inherent spatial variability of soil properties are two primary sources of uncertainty, significantly impacting embedded geo-structures. While the influence of the soil spatial variability and geological uncertainty has been separately investigated, the coupling effect of these two types of uncertainties on tunnel is still limited. Meanwhile, little effort has been devoted to quantifying the degree of importance of these two types of uncertainties. This study quantitatively evaluates the effects of geological uncertainty and soil spatial variability on tunnels, and first assess the importance of these two uncertainties based on a novel approach. Improved coupled Markov chain model was utilized to simulate the geological uncertainty based on sparse and limited boreholes. The coefficient of variation (COV) was adopted to characterize the different levels of soil spatial variability in this study. The results illuminated the role of geological uncertainty and spatial variability in tunnel probabilistic analysis by comparing different boreholes schemes and COV. In addition, the geological uncertainty index (GUI) was presented to quantitatively estimate the degree of uncertainty in the simulated strata. A novel approach for evaluating the significance of geological uncertainty and soil spatial variability on tunnel was proposed. Based on the Kendall correlation coefficient, the classification was also given, divided into high, strong, moderate, weak and no correlation. The geological uncertainty should be considered in high and strong correlation groups. Both two types of uncertainties should be simultaneously taken into account in moderate and weak correlation divisions.

Geothermal field development optimization under geomechanical constraints and geological uncertainty: Application to a reservoir with stacked formations

In this work, numerical optimization based on stochastic gradient methods is used to assist geothermal operators in finding improved field development strategies that are robust to accounted geological uncertainties. Well types, production rate targets and well locations are optimized to maximize the economics of low-enthalpy heat recovery in a real-life case with stacked reservoir formations. Significant improvements are obtained with respect to the strategy designed by engineers. Imposing fault stability constraints impacts significantly the optimal configurations, with coordinated well rates and placement playing a key role to boost efficiency of geothermal production while keeping stress change effects to acceptable limits.

Well Trajectory Optimization under Geological Uncertainties Assisted by a New Deep Learning Technique

Summary The large number of geological realizations and well trajectory parameters make field development optimization under geological uncertainty a time-consuming task. A novel deep learning-based surrogate model with a novel well trajectory parametrization technique is proposed in this study to optimize the trajectory of wells under geological uncertainty. The proposed model is a deep neural network with ConvLSTM layers to extract the most salient features from highly channelized and layered reservoirs efficiently. ConvLSTM layers are used because they can extract spatiotemporal features simultaneously since layered reservoirs can be regarded as a time series of spatially distributed reservoir properties. The proposed surrogate model could predict the individual objective function with a coefficient of determination of 0.96. After verifying the validity of the surrogate model, four approaches were used to optimize well trajectories. Two of the approaches consumed all available realizations (surrogate model-based and simulation-based approaches), while the remaining two used a subset of realizations. The selection of the subset was based on the cumulative oil production (COP) and the diffusive time of flight (DTOF). Results showed that although the surrogate model used all realizations, it could provide similar results to the simulation-based optimization with only a 5% computational cost of the simulation-based approach. The novelty of this work lies in its proposal of an innovative surrogate model to improve the analysis of channelized and layered reservoirs and its introduction of a novel well trajectory optimization framework that effectively addresses the challenge of optimizing well trajectories in complex three-dimensional spaces, a problem not adequately tackled in previous works.

Large-deformation study of T-bar penetration in spatially variable sediments

The T-bar penetrometer is extensively used in offshore field investigation and laboratory testing to evaluate the undrained shear strength of soft clay because of its reliability and repeatability in measurements. However, most previous studies on T-bar penetration have adopted a deterministic approach and focused solely on homogeneous soils, neglecting the geological uncertainty caused by the spatial variability of soil. To remedy this situation, this study aimed to investigate the effect of soil spatial variation on the T-bar penetration process in soft clay by employing large deformation random finite element modeling. The findings indicate that the variability in soil strength substantially modifies the flow pattern and failure mechanism of the soil, consequently influencing the penetration resistance factor of the T-bar. It was observed that the resistance factors follow the log-normal distribution. A relationship between the probability of failure as well as the factor of safety was determined, allowing for a clear assessment of the reliability level associated with varying factors of safety. To allow for this effect, a practical framework is developed to correct T-bar penetration data to estimate soil strength.

An echo state network approach to data-driven modeling and optimal control of carbonate reservoirs with uncertainty fields

Previous studies of reservoir simulation have shown that it is required to have every inherent carbonate reservoir physics available at the model development phase. However, this process while ideal is computationally intensive and time consuming especially for complex geological fields. In this study, Echo State Network (ESN) is proposed to predict two cases of reservoir waterflooding with integrated uncertainty properties in unstructured physical dimension and log-normal permeability fields. The generalization efficiency of ESN with respect to uncertainties in injection and production data is also investigated by tuning the ESN reservoir size (N). Consequently, adjoint based method is used to optimize the Net Present Value (NPV) via obtained injection well controls. Observation revealed that the developed ESN exhibits superior predictive accuracy in scenarios of low geological uncertainties, achieving an average test accuracy of 90.79%, and then declining to 86.19% as the uncertainty level is increased. Also, N is found to be significant in terms of ESN generalization efficiency with N = 40 achieving higher performance compared to N = 20, thereby validating its influence on ESN reservoir state to adapt to current data history during training. Lastly, the optimized scenario, derived from the initial well control predictions, demonstrated a higher NPV when compared with the base scenario of the developed oil reservoir model.

Simulation of Geological Uncertainty Using Coupled Markov Chain: A Case Study at a Manually Filled Loess Site

Simulation of Geological Uncertainty Using Coupled Markov Chain: A Case Study at a Manually Filled Loess Site

End-to-end dimensionality reduction and regression from 3D geological uncertainties to estimate oil reservoir simulations

Managing geological uncertainties in reservoir engineering involves significant challenges mainly due to the prohibitive computational costs of traditional simulation methods. These simulations, essential for generating geological models, often require extensive computational resources and can take days or weeks to complete. The computational load limits the number of viable models while the high dimensionality of properties compounds the challenge, and the abundance of producing wells, each associated with different objective functions, further complicates the problem. Current methodologies have focused on training an Artificial Neural Network (ANN) for each producer well to create a proxy model. Instead, we propose the development of a single ANN designed to simultaneously predict the behavior of multiple objective functions. This work proposes a novel end-to-end deep neural network that can handle 3D geological uncertainties and replicate the simulator’s response for several cumulative production curves. This approach leverages 3D convolutions to process spatial and depth dimensions within heterogeneous reservoirs, including diverse geostatistical realizations. The end-to-end solution was successfully implemented in a Brazilian offshore field and accurately replicated the simulator’s behavior and yielded results that outperformed state-of-the-art methods. The results indicate correlations that exceeded 0.95 between the proxy response and the numerical simulator. Additionally, our approach demonstrated robustness even when trained with a limited number of simulation models, and it notably reduced computational costs. The findings highlight our architecture’s capacity to integrate dimensional reduction and regression analysis within a unified framework, effectively predicting different fluid behaviors in the reservoir and showcasing robustness against high dimensionality and sparse data.