Search

Subsurface environments host diverse microorganisms in fluid-filled fractures; however, little is known about how geological and hydrological processes shape the subterranean biosphere. Here, we sampled three flowing boreholes weekly for 10 mo in a 1478-m-deep fractured rock aquifer to study the role of fracture activity (defined as seismically or aseismically induced fracture aperture change) and advection on fluid-associated microbial community composition. We found that despite a largely stable deep-subsurface fluid microbiome, drastic community-level shifts occurred after events signifying physical changes in the permeable fracture network. The community-level shifts include the emergence of microbial families from undetected to over 50% relative abundance, as well as the replacement of the community in one borehole by the earlier community from a different borehole. Null-model analysis indicates that the observed spatial and temporal community turnover was primarily driven by stochastic processes (as opposed to deterministic processes). We, therefore, conclude that the observed community-level shifts resulted from the physical transport of distinct microbial communities from other fracture(s) that outpaced environmental selection. Given that geological activity is a major cause of fracture activity and that geological activity is ubiquitous across space and time on Earth, our findings suggest that advection induced by geological activity is a general mechanism shaping the microbial biogeography and diversity in deep-subsurface habitats across the globe.

Summary Smart completions enable physical measurements over space and time, which provides large volumes of information at unprecedented rates. However, optimizing inflow control valve (ICV) settings of smart multilateral wells is a challenging task. Traditionally, ICV field tests, evaluating well performance at different ICV settings, are conducted to observe flow behavior and configure ICVs; however, this is often suboptimal. This study investigated a surrogate-based optimization algorithm that minimizes the number of ICV field tests required, predicts well performance of all unseen combination of ICV settings, and determines the optimal ICV setting and net present value (NPV). A numerical model of a real offshore field in Saudi Arabia was used to generate scenarios involving a two-phase (oil and water) reservoir with trilateral producers. Multiple scenarios were examined with variations in design parameters, mainly well count, placement, and configuration. Eight discrete settings were assumed to match the commonly installed ICV technology, where all possible scenarios were simulated to establish ground truth. The investigation considered three major algorithmic components: sampling, machine learning, and optimization. The sampling strategy compared physics-based initialization, space-filling sampling, and triangulation-based adaptive sampling. A cross-validated neural network was used to fit a surrogate (in this case, machine learning algorithm) dynamically, whereas enumeration was adopted for optimization to avoid errors arising from using common optimizers. This study evaluated two sampling techniques: space-filling and adaptive sampling. The latter was found superior in capturing reservoir behavior with the smallest number of simulation runs (i.e., ICV field tests). Algorithm performance was evaluated based on the number of ICV field tests required to exceed an R2 threshold of 90% on all unseen scenarios and match the optimal ICV settings and NPV. Surface and downhole flow profile prediction and optimization were achieved successfully using this approach. To determine the diminishing value of additional ICV field tests, the triangulation sampling loss was used as a stoppage criterion. When running the algorithm on a single producer for both surface and downhole oil and water flow prediction, the algorithm required only 6 and 11 ICV field tests to achieve 80% and 90% R2 across the different cases of this real reservoir model. Fishbone wellbore configurations were found to pose a more challenging task because changes in any ICV pressure decrease affects multiple laterals simultaneously, which increases the level of interdependence. The resultant surrogate was used to decide on the optimal settings of ICV devices and effectively predict the NPV. Surrogates, in this approach, are statistical proxies of the targeted ground-truth production function. Further improvement was accomplished through adaptively sampling and fitting surrogates to predict NPV explicitly, where NPV predictions were generated with nearly 95% R2 given only 10 ICV field tests. Using adaptive sampling and machine learning proved effective in the prediction of surface and downhole flow profiles and optimization of smart wells. The method further allows for dynamically optimizing field strategy in a reinforcement learning setting where production data are used continuously to further improve the prediction performance.

Recent experience in applying recurrent neural networks (RNNs) to interpreting Permannt Downhole Gauge (PDG) records has highlighted the potential utility of machine learning algorithms to learn reservoir behavior from data. The power of the RNN resides in its ability to retain information in a form of memory of previous patterns and information contained in the previous behavior of phenomena being modeled. This memory plays a role of informing the decision at the present time by using what happened in the past. This property suggests RNNs as a suitable choice to model sequences of reservoir information, even when the reservoir modeler is faced with incomplete knowledge of the underlying physical system. Convolutional Neural Networks (CNNs) are another variant of machine learning algorithms that have shown promise in sequence modeling domains, such as audio synthesis and machine translation. In this study, RNN and CNN were applied to tasks that traditionally would be modeled by a reservoir simulator. This was achieved by formulating the relationship between physical quantities of interest from subsurface reservoirs as a sequence mapping problem. In addition, the performance of a CNN layer as compared to RNN was evaluated systematically to investigate their capabilities in a variety of tasks of interest to the reservoir engineer. Preliminary results suggest that CNNs, with specific design modifications, are as capable as RNNs in modeling sequences of information and as reliable when making inferences to cases that has not been seen by the algorithm during training. Design details and reasons pertaining to the way these two seemingly different architectures process information, and handle memory are also discussed.

Summary Intelligent multilateral well completions provide downhole flow rate, pressure, and temperature measurements at multiple well segments, which allows for a continuous spatiotemporal data stream. Such an extensive data input poses a challenging task to decide on the optimal strategy of manipulating the inflow control valve (ICV) settings over time for best performance. In this study, we investigate using machine learning to analyze and predict well performance given different ICV settings to ultimately maximize the well output. A commercial reservoir simulator was used to generate two synthetic reservoir models: homogeneous (Case A) and heterogeneous (Case B). These synthetic data were used to train, validate, and test machine learning models. The reservoir cases were generated on the basis of a segmented, trilateral producer completed with three ICV devices installed at tie-in segments. The data used were measurements of wellhead and downhole flow rates across ICV segments over a period of 4,000 days. A total of 1,330 experiments were conducted with an 8-day timestep, generating a total of 667,660 sample data points for each of Case A and Case B. Fully connected neural networks were used to fit the data, while model generalizability was enhanced using regularization techniques, namely L2 regularization and early stopping. Both random sampling and Latin hypercube sampling (LHS) methods were evaluated in constructing the training, validation, and testing splits. Trained with different sample sizes drawn from the 1,330 simulated data histories for the two reservoir models, the proposed neural network showed excellent results. Given only 10 simulated choices of ICV settings for training, the network proved capable of predicting oil/water production profiles at surface for both homogeneous and heterogeneous reservoir models with greater than 0.95 coefficient of determination (R2) when evaluated at unseen, test ICV settings. Extending the problem to downhole flow performance prediction, approximately 40 training simulated settings were necessary to achieve 0.95 R2. We observed that LHS was superior to random sampling in both R2 average and confidence interval. We also found that increasing the training and validation sample sizes increased the test R2 when testing against unseen cases. Study results suggest the applicability of machine reinforcement learning to estimate the well output at different ICV settings, where the neural network model depends fully on the real-time well feedback and production measurements. By using a machine learning approach during the operation of a well with multiple ICV settings, it would be feasible to estimate the lateral-by-lateral output at unseen scenarios. Hence, it becomes possible to maximize the well output by using an optimization algorithm to determine the optimal ICV settings.

Nano- and microparticles are expected to have several functionalities, and the ability to control size and shape is an advantage of using nano-/microparticles. This study investigated a possibility that the sizes of nano-/microparticle can be used to extract new information on structures in a fractured medium. Flow experiments were conducted to observe the particle transport in a micromodel on which a single fracture and rock matrix (grain and pore space) was fabricated on a silicon wafer. Water and nano-/microparticles were injected into the micromodel, and the droplets were collected at the outlet. Tunable Resistive Pulse Sensing (TRPS) was used to measure the frequency distributions of particle diameters from each droplet at each time. The result shows that the larger particles were observed only at early time, while the smaller particles were detected at early time and also at late time. This indicates that the larger particles flow in a fracture quickly, while smaller particles migrate through both fracture and matrix over a wider range of time. Particles with different sizes transport through fractured media differently depending on the fracture structures. The tracer response of nano- and microparticles may be useful to evaluate the fracture structures and the flow properties for different flow pathways.

Knowledge of well connectivity in a reservoir is crucial, especially for early-stage field development and water injection management. However, traditional interference tests can often take several weeks or even longer depending on the distance between wells and the hydraulic diffusivity of the reservoir. Therefore, instead of physically shutting in production wells, we can take advantage of deep learning methods to perform virtual interference tests. In this study, we first used the historical field data to train the deep learning model, a modified Long- and Short-term Time-series network (LSTNet). This model combines the Convolution Neural Network (CNN) to extract short-term local dependency patterns, the Recurrent Neural Network (RNN) to discover long-term patterns for time series trends, and a traditional autoregressive model to alleviate the scale insensitive problem. To address the time-lag issue in signal propagation, we employed a skip-recurrent structure that extends the existing RNN structure by connecting a current state with a previous state when the flow rate signal from an adjacent well starts to impact the observation well. In addition, we found that wells connected to the same manifold usually have similar liquid production patterns, which can lead to false causation of subsurface pressure communication. Thus we enhanced the model performance by using external feature differences to remove the surface connection in the data, thereby reducing input similarity. This enhancement can also amplify the weak signal and thus distinguish input signals. To examine the deep learning model, we used the datasets generated from Norne Field with two different geological settings: sealing and nonsealing cases. The production wells are placed at two sides of the fault to test the false-negative prediction. With these improvements and with parameter tuning, the modified LSTNet model could successfully indicate the well connectivity for the nonsealing cases and reveal the sealing structures in the sealing cases based on the historical data. The deep learning method we employed in this work can predict well pressure without using hand-crafted features, which are usually formed based on flow patterns and geological settings. Thus, this method should be applicable to general cases and more intuitive. Furthermore, this virtual interference test with a deep learning framework can avoid production loss.

This study developed an analytical tool for the detection and characterization of scale buildup from well data using deep learning methods. The developed method allows for a sensitive detection of the initiation of scaling as well as an accurate prediction of the magnitude of existing scale buildup. Scale deposition causes tubing ID decreases and therefore results in production declines, so a sensitive approach to detect the scale deposition is valuable to reduce the damage and losses due to this problem. The underlying deep learning methods are both single- and multioutput, deep neural networks that consist of a combination of convolutional, long short-term memory (LSTM) and fully connected layers. We trained the networks on more than 30 sets of well data, with the objective of predicting the presence and the magnitude of scale. We started with cases of full scale deposition over the whole wellbore depth, and then extended our study to partial depth scale deposition. We built up a point-wise neural network model combining two blocks, which each contain several fully-connected layers followed by an LSTM layer specifically focusing on relatively smaller or larger tubing ID changes, corresponding to more or less scale deposition. To characterize the segmented scale deposition, we transformed to a three-dimensional problem, which can be solved by extracting the tubing ID changes and the scale deposition segment length. The multioutput model was able to predict the tubing ID and volume changes at the same time, using a combination of convolutional and LSTM layers with residual network blocks and updated using a loss function that we defined. Tubing ID changes were extracted accurately with metric R-square more than 90%, while the length of the scale deposit could be classified into two classes (high scaling or low scaling) with good accuracy. Though existing physical and chemical methods can be used to analyze scale deposition, the methods are often applied after considerable production decrease has already occurred. By using deep learning algorithms, our study came up with a new way to predict the scaling problem in advance with high sensitivity.

Knowledge of reservoir heterogeneity and connectivity is fundamental for reservoir management. Methods such as interference tests or tracers have been developed to obtain that knowledge from dynamic data. However, detecting well connectivity using interference tests requires long periods of time with a stable reservoir pressure and constant flow-rate conditions. Conversely, the long duration and high frequency of well production data have high value for detecting connectivity if noise, abrupt changes in flow-rate and missing data are dealt with. In this work, a methodology to detect interference from longterm pressure and flow-rate data was developed using multiresolution analysis in combination with machine learning algorithms. The methodology presents high accuracy and robustness to noise while requiring little to no data preprocessing. The methodology builds on previous work using the Maximal Overlap Wavelet Transform (MODWT) to analyze long-term pressure data. The new approach uses the ability of the MODWT to capture, synthesize and discriminate the relevant reservoir response for each individual well at different time scales while still honoring the relevant flow-physics. By first applying the MODWT to the flow rate history, a machine learning algorithm was used to estimate the pressure response of each well as it would be in isolation. Interference can be detected by comparing the output of the machine learning model with the unprocessed pressure data. A set of machine learning, and deep learning algorithms were tested including Kernel Ridge Regression, Lasso Regression and Recurrent Neural Networks. The machine learning models were able to detect interference at different distances even with the presence of high noise and missing data. The results were validated by comparing the machine learning output with the theoretical pressure response of wells in isolation. Additionally, it was proved that applying the MODWT multiresolution analysis to pressure and flow-rate data creates a set of "virtual wells" that still follow the diffusion equation and allow for a simplified analysis. By using production data, the proposed methodology allows for the detection of interference effects without the need of a stabilized pressure field. This allows for a significant cost reduction and no operational overhead because the detection does not require well shut-ins and it can be done regardless of operation opportunities or project objectives. Additionally, the long-term nature of production data can detect connectivity even at long distances even in the presence of noise and incomplete data.

This study investigated the ability to produce accurate multiphase flow profiles simulating the response of producing reservoirs, using Generative Deep Learning (GDL) methods. Historical production data from numerical simulators were used to train a GDL model that was then used to predict the output of new wells in unseen locations.This work describes a procedure in which data analysis techniques are used to gain insight into reservoir flow behavior at a field level based on existing historical data. The procedure includes clustering, dimensionality reduction, correlation, in addition to novel interpretation methodologies that synthesize the results from reservoir simulation output, characterizing flow conditions. The insight was then used to build and train a GDL algorithm that reproduces the multiphase reservoir behavior for unseen operational conditions with high accuracy. The trained algorithm can be used to further generate new predictions of the reservoir response under operational conditions for which we do not have previous examples in the training data set.We found that the GDL algorithm can be used as a robust multiphase flow simulator. In addition, we showed that the physics of flow can be captured and manipulated in the GDL latent space after training to reproduce different physical effects that did not exist in the original training data set. Applying the methodology to the problem of determining multiphase production rate from new producing wells in undrilled locations showed positive results. The methodology was tested successfully in predicting multiphase production under different scenarios including multiwell channelized and heterogeneous reservoirs. Comparison with other shallow supervised algorithms demonstrated improvements realized by the proposed methodology, compared to existing methods.The study developed a novel methodology to interpret both data and GDL algorithms, geared towards improving reservoir management. The method was able to predict the performance of new wells in previously undrilled locations without using a reservoir simulator.

Author(s): Fu, P; Schoenball, M; Ajo-Franklin, JB; Chai, C; Maceira, M; Morris, JP; Wu, H; Knox, H; Schwering, PC; White, MD; Burghardt, JA; Strickland, CE; Johnson, TC; Vermeul, VR; Sprinkle, P; Roberts, B; Ulrich, C; Guglielmi, Y; Cook, PJ; Dobson, PF; Wood, T; Frash, LP; Huang, L; Ingraham, MD; Pope, JS; Smith, MM; Neupane, G; Doe, TW; Roggenthen, WM; Horne, R; Singh, A; Zoback, MD; Wang, H; Condon, K; Ghassemi, A; Chen, H; McClure, MW; Vandine, G; Blankenship, D; Kneafsey, TJ | Abstract: The final version of the above article was posted prematurely on 16 July 2021, owing to a technical error. The final, corrected version of record will be made fully available at a later date.

Efficient simulation of multiphysics problems is a challenging task. This is often due to the multiscale nature of the physics and nonlinear coupling between the different processes. One approach to this problem is to solve the entire multiphysics problem simultaneously in a fully coupled manner. However, due to the strong coupling and multiphysics interactions, it is difficult to design and analyze fully coupled solvers, which often entail the construction and solution of global fully coupled Jacobian systems. Another approach is the sequential-implicit method, whereby the full multiphysics problem is split into different subproblems. Each subproblem is constructed and solved separately; then the solutions of the subproblems are stitched together in a specific sequence. Isolation of each subproblem allows for the design of specialized solvers that tackle the complexity of the particular subproblem efficiently. The sequential-implicit approach offers wide flexibility and extensibility. However, these advantages are often offset by slow convergence rates when there is strong coupling between the subproblems. This slow convergence rate of the overall coupled system is directly linked to the use of a fixed-point outer-loop iteration in sequential-implicit methods. We present a Sequential Implicit Newton (SIN) approach, whereby the sequence of implicit subproblems is wrapped with an outer full Newton scheme. We demonstrate that the SIN formulation improves the overall convergence rate from linear to quadratic. The SIN method uses the same sequential-implicit scheme as the fixed-point method, but after each sequential iteration a Newton update is computed. Wrapping a Newton loop around the traditional sequential-implicit scheme leads to significant improvements in the convergence rate. The SIN method allows for the ability to split a multiphysics problem into individual subproblems while taking advantage of the quadratic convergence rate of the Newton method. We demonstrate the effectiveness of SIN using two different multiphysics porous-media problems: flow-thermal in geothermal reservoir simulation and flow-mechanics in geomechanics reservoir simulation. Just as with the fixed-point iteration version of the sequential-implicit method, SIN benefits from careful design of the constraints used to stitch the sequence of the subproblems. Our numerical experiments show that the SIN approach improves the overall convergence rate for all the nonlinear multiphysics problems considered. For specific cases where there is strong nonlinear coupling between the subproblems, we see up to two orders of magnitude decrease in the number of sequential iterations when using SIN compared with the fixed-iteration scheme.

Summary Permanent downhole gauges (PDGs) provide a continuous record of pressure, temperature, and, sometimes, flow rate during well production. The continuous record provides rich information about the reservoir and makes PDG data a valuable source for reservoir analysis (e.g., pressure-rate deconvolution for reservoir-model identification). It has been shown in previous work that the convolution-kernel (CK) -based data-mining approach is a promising tool to interpret flow-rate and pressure data from PDGs. The CK method denoises and deconvolves the pressure signal successfully without an explicit-breakpoint detection. However, the bottlenecks of computational efficiency and the incomplete recovery of reservoir behaviors limit the application of the method to interpret real-PDG data. In this paper, three different machine-learning techniques were applied to flow-rate/pressure interpretation. We formulated the machine-learning techniques into a linear regression (LR) on parameters that connect the nonlinear flow-rate features with pressure targets. Such a formulation leads to a closed-form solution, which speeds up the computation dramatically. The machine-learning algorithms that were formulated using LR were shown to have the same learning quality as the CK method, and they outperformed it with much less computational effort. Next, the kernel method was applied to address the issue of the incomplete recovery of reservoir behaviors, because it efficiently expanded the dimension of the feature space without an explicit representation of the features, but it led to overfitting. Finally, kernel ridge regression (KRR) used the expanded features given by the kernel function to capture the more detailed reservoir behaviors, while controlling the prediction error using ridge regression (RR). It was shown that KRR recovers the full reservoir behaviors successfully (e.g., wellbore-storage effect, skin effect, infinite-acting radial flow, and boundary effect). Some potential uses of temperature data from PDGs are also discussed in this paper. Machine learning was shown to be able to model the temperature and pressure data recorded by PDGs, even if the actual physical model is complex. This originates from the fact that, by using features as an approximation of model characteristics, machine learning does not require a perfect knowledge of the physical model. The modeling of pressure using temperature data was extended to two promising applications: pressure-history reconstruction using temperature data, and the cointerpretation of temperature and pressure data when flow-rate data are not available.

Smart completions enable physical measurements over space and time, which provides large volumes of information at unprecedented rates. However, optimizing inflow control valve (ICV) settings of smart multilateral wells is a challenging task. Traditionally, ICV field tests, evaluating well performance at different ICV settings, are conducted to observe flow behavior and configure ICV's, however this is often suboptimal. This study investigated a surrogate-based optimization algorithm that minimizes the number of ICV field tests required, predicts well performance of all unseen combination of ICV settings, and determines the optimal ICV setting and net present value (NPV). A numerical model of a real offshore field in Saudi Arabia was used to generate scenarios involving a two-phase (oil and water) reservoir with trilateral producers. Multiple scenarios were examined with variations in design parameters, mainly well count, placement and configuration. Eight discrete settings were assumed to match the commonly installed ICV technology, where all possible scenarios were simulated to establish ground truth. The investigation considered three major algorithmic components: sampling, machine learning, and optimization. The sampling strategy compared physics-based initialization, space-filling sampling, and triangulation-based adaptive sampling. A cross-validated neural network was used to fit a surrogate dynamically, while enumeration was adopted for optimization to avoid errors arising from using common optimizers. This study evaluated two sampling techniques: space-filling and adaptive sampling. The latter was found superior in capturing reservoir behavior with the smallest number of simulation runs, i.e. ICV field tests. Algorithm performance was evaluated based on the number of ICV field tests required to: 1) surpass an R2 threshold of 0.9 on all unseen scenarios, and 2) match the optimal ICV settings and NPV. Surface and downhole flow profile prediction and optimization were achieved successfully using this approach. To determine the diminishing value of additional ICV field tests, the triangulation sampling loss was used as a stoppage criterion. When running the algorithm on a single producer for both surface and downhole oil and water flow prediction, the algorithm required six and 11 ICV field tests only to achieve 80% and 90% R2 across the different cases of this real reservoir model. Fishbone wellbore configurations were found to pose a more challenging task as changes in any ICV pressure drop affects multiple laterals simultaneously, which increases the level of interdependence. The resultant surrogate was used to decide on the optimal settings of ICV devices and also predict the NPV effectively. Further improvement was accomplished through adaptively sampling and fitting surrogate to rather predict NPV explicitly where NPV predictions were generated with nearly 95% R2 given only ten ICV field tests. Using adaptive sampling and machine learning proved effective in the prediction of surface and downhole flow profiles, and optimization of smart wells. The method further allows for dynamically optimizing field strategy in a reinforcement learning setting where production data are used continuously to further improve the prediction performance.

This work investigated deep learning (DL) algorithms to forecast wells' economic potential using wellhead surface measurements. The results of DL algorithms were compared to existing solutions in the literature that aim at solving the same problem. The performance of algorithms trained on multiple input representations, resulting from applying unsupervised DL feature extraction methods, was compared to algorithms trained on raw inputs. In this work, available wellhead measurements that can be collected at surface were used in modeling their relationship to well production rate. A search method was used to select the best combination of measurements for modeling. In addition, features were extracted automatically using autoencoders to generate additional inputs. The algorithms used for forecasting were designed using standard and customized DL layers. The results of both DL and existing solutions from the literature were compared systematically to determine the usefulness of the DL methodology developed. Previous work has shown the limitations of correlation-based methods in offering a reliable solution when applied to data outside of the historically observed range. In addition, multiple models are sometimes required depending on operational settings, e.g. choke opening, making the modeling process inefficient. Moreover, mass balance-based methods require information, such as phase density, that might not be readily available when modeling is performed, affecting the reliability of the solutions. We found that using DL algorithms offers a flexible solution that can help overcome these limitations and opens a new path to solve similar problems. We found that algorithms trained properly on relevant features are useful tools to use in assessing the economic potential of wells, using already available surface measurements. The study developed a novel methodology for forecasting economic potential using wellhead surface measurements and demonstrated the application to real field data. The method was found to be an efficient utilization of data that are readily available and do not require well intervention to measure.

The high frequency and long duration of well production data have high value for inferring reservoir properties, monitoring reservoir conditions and detecting changes in the well flow conditions. Unlike traditional well test data where the well flow-rate and the pressure response are carefully monitored, long-term pressure data is subject to abrupt changes in flow-rate, noise caused by hardware failures, missing data and physical changes in the reservoir. In this work, methodology to interpret long-term flow-rate and pressure data was developed using wavelet multiresolution analysis in combination with deep learning algorithms. The methodology requires little to no data preprocessing, no explicit knowledge of a physical model and has high robustness to noise.The proposed methodology was found to exploit the ability of wavelets to capture and synthetize the relevant behavior of the reservoir performance history at multiple scales. This was achieved by applying the Maximal Overlap Wavelet Transform (MODWT) to long-term flow-rate data and using the output of the MODWT as the input for a Recurrent Neural Network (RNN) which creates a function map between flow-rate and pressure as a function of time. For the training of the RNN, synthetic well data were used.The application of wavelet transforms allowed the neural network to perform an automatic noise filtering and simplified the training process while requiring no knowledge of noise levels, little to no manual preprocessing of data and no knowledge of the reservoir's physical model. Moreover, the mathematical formulation of the wavelet transforms allows the neural network to take advantage of the multiresolution properties, opening the door to executing the whole analysis procedure in a simplified way.The proposed methodology aids in the inference of effects such as changes in skin factor, pressure depletion or changes in the reservoir model from the available production data, even when those data are noisy or incomplete. By using production data, the inference can be done without loss of production and used to support oil production enhancement operations. Moreover, the coupling of a neural network with multiresolution analysis eliminates the need for feature engineering and performs automatic denoising of the data, which were found to be extremely desirable properties for a methodology to be scalable and applicable to real well datasets.

Tidal analysis provides a cost-effective way of estimating aquifer properties. Tidal response models that link aquifer properties with tidal signal characteristics, such as phase and amplitude, hav...

This study proposed and examined a practical technique for analysing CO2 storage responses using offshore pressure transients affected by the ocean tide. The gravitational attractions of the solar-system bodies cause ocean tides, and the pore pressure exhibits diurnal and semidiurnal fluctuations in response to such tidal phenomena. The pressure-fluctuation amplitude is related to the loading efficiency, which is a function of reservoir elastic properties and fluid saturations. Therefore, the loading efficiency can be used to estimate the in situ pore compressibility and the CO2 saturation. Applying the tidal-signal analysis in a time-lapse manner, one may see temporal changes in CO2 saturation and consequently describe the dynamic behavior of sequestered CO2. In the analysis at the offshore CO2 storage site in Tomakomai, Japan, a temporal decrease in CO2 saturation was detected during the shut-in period, which is caused primarily by CO2 migration away from the well. The proposed methodology essentially requires only continuous pressure data, which are routinely available during CO2 storage operations, and thus, can be a cost-effective and labour-saving monitoring technique.

Conventional experiments using natural rock samples have trouble in observing rock structures and controlling fracture properties. Taking advantage of 3D printing technologies, a complex fracture network was made by using a 3D printer. This approach allowed us to control the properties of the fracture networks and to prepare identical geometries for both simulation and experiment. A tracer response curve from the flow experiment was obtained and compared with numerical simulations. The result of the computational fluid dynamics (CFD) simulation based on the Navier–Stokes equations was in good agreement with experimental result, which suggested that the results of experiment and the CFD simulation are reliable. On the other hand, comparison with an equivalent permeability model based on the cubic law showed a discrepancy from the experimental result. This validation approach enabled discussion of the limitation of the flow model. Because 3D printed fracture networks could reduce uncertainty between numerical simulation and laboratory experiment, they will be useful for understanding more detailed and more complicated phenomena in fracture networks.

The maximum expected earthquake magnitude is an important parameter in seismic hazard and risk analysis because of its strong influence on ground motion. In the context of injection-induced seismicity, the processes that control how large an earthquake will grow may be influenced by operational factors under engineering control as well as natural tectonic factors. Determining the relative influence of these effects on maximum magnitude will impact the design and implementation of induced seismicity management strategies. In this work, we apply a numerical model that considers the coupled interactions of fluid flow in faulted porous media and quasidynamic elasticity to investigate the earthquake nucleation, rupture, and arrest processes for cases of induced seismicity. We find that under certain conditions, earthquake ruptures are confined to a pressurized region along the fault with a length-scale that is set by injection operations. However, earthquakes are sometimes able to propagate as sustained ruptures outside of the zone that experienced a pressure perturbation. We propose a faulting criterion that depends primarily on the state of stress and the earthquake stress drop to characterize the transition between pressure-constrained and runaway rupture behavior.

In the past, the problem of predicting the multiphase flow rate in producing wells has been tackled using the Gilbert correlation. The Gilbert correlation offers a quick solution and requires quantities that are readily available at the surface, making it a staple among solutions in many publications. However, the correlation is limited in applicability to local data, and in some cases, multiple parameter settings are required when flow conditions change. To address these points, we introduced a systematic comparison of applying Gilbert correlation and Deep Learning (DL) algorithms to the problem of liquid flow prediction. We determined that some of the issues, specifically the need for multiple models, can be alleviated using DL methods. In this work, available wellhead measurements that can be collected at surface were used in modeling their relationship to well production rate. In addition, features were extracted automatically using autoencoders to generate additional inputs. The algorithms used for forecasting were designed using standard and customized DL architectures. The results of both DL and existing solutions from the literature were compared systematically to determine the usefulness of the DL methodology developed. The study developed a novel methodology for forecasting well liquid and multiphase restricted flow rate using wellhead surface measurements and demonstrated the application to real field data. The method was found to be an efficient utilization of data that are readily available and do not require well intervention to measure.

Financial risk due to geological uncertainty is a major barrier for geothermal development. Production from a geothermal well depends on the unknown location of subsurface geological structures, such as faults that contain hydrothermal fluids. Traditionally, geoscientists collect many different datasets, interpret the datasets manually, and create a single model estimating faults' locations. This method, however, does not provide information about the uncertainty regarding the location of faults and often does not fully respect all observed datasets. Previous researchers investigated the use of stochastic inversion schemes for addressing geological uncertainty, but often at the expense of geologic realism. In this paper, we present algorithms and open-source code to stochastically invert five typical datasets for creating geologically realistic structural models. Using a case study with real data from the Patua Geothermal Field, we show that these inversion algorithms are successful in finding an ensemble of structural models that are geologically realistic and match the observed data sufficiently. Geoscientists can use this ensemble of models to optimize reservoir management decisions given structural uncertainty.

During hydraulic stimulation treatment in an enhanced geothermal (EGS) reservoir, it has been suggested that new fractures are initiated from stimulated preexisting fractures and propagate through the reservoir. In this stimulation mechanism, a fracture propagation from a preexisting natural fracture and the interaction of newly formed fractures and preexisting natural fractures play an important role in the creation of a complex fracture network . In this study, we developed a physics-based numerical model that couples fluid flow between fracture surfaces , fracture deformation , and fracture propagation driven by fluid injection to better understand how a fracture network is created by a hydraulic stimulation treatment. The results showed that the created fracture network complexity is affected by the fracture intersection angle, stress state, and injection rates . Having flow path complexity is advantageous in EGS because it increases the heat exchange areas , the fracture surface to rock volume ratio, and the general reservoir permeability . Therefore, the implication from the results of this study to make a better EGS reservoir is to stimulate an EGS reservoir with 1) well-oriented fractures, 2) a high stress ratio, and 3) a low injection rate. These reservoir conditions and stimulation design will likely make more flow paths and create a more complex fracture network.

Understanding flow mechanisms in fractured media is essential for geoscientific research and geological development industries. This study used 3-D printed fracture networks in order to control the properties of fracture distributions inside the sample. The accuracy and appropriateness of creating samples by the 3-D printer was investigated by using a X-ray CT scanner. The CT scan images suggest that the 3-D printer is able to reproduce complex three-dimensional spatial distributions of fracture networks. Use of hexane after printing was found to be an effective way to remove wax for the posttreatment. Local permeability was obtained by the cubic law and used to calculate the global mean. The experimental value of the permeability was between the arithmetic and geometric means of the numerical results, which is consistent with conventional studies. This methodology based on 3-D printed fracture networks can help validate existing flow modeling and numerical methods.

Mechanisms for the delayed triggering between the Mw 4.8 foreshock and Mw 5.6 main shock of the 2011 earthquake sequence near Prague, Oklahoma, USA were investigated using a coupled fluid flow and fault mechanics numerical model. Because the stress orientations, stress magnitudes, fault geometry, and earthquake source mechanisms at the Prague site have been well-characterized by previous studies, this particular earthquake sequence offered an opportunity to explore the range of physical processes and in-situ fault properties that might be consistent with the 20 hour delayed triggering effect observed at the site. Our numerical experiments suggest that an initial undrained response resulting from elastic stress transfer from the foreshock followed by transient fluid flow along the fault may have contributed to the earthquake nucleation process. The results of the numerical experiments were used to constrain fault compliance and fault transmissivity for the fault that hosted the Mw 5.6 event. Relatively compliant behavior in response to changes in normal stress, corresponding to Skempton pore pressure coefficients near 1, was consistent with the field observations. Fault transmissivity was estimated to range from 10−18 to 10−15 m3. This study has implications for understanding hydraulic properties, frictional properties, and faulting behavior of basement faults in Oklahoma that are large enough to host damaging earthquakes.

The rapid development of machine learning algorithms and the massive accumulation of well data from continuous monitoring has enabled new applications in the oil and gas industries. Data gathered from well sensors are a foundation of the oilfield digitization and data-driven analysis. Here, we describe a deep learning approach to predict the long-term well performance based on a moderate duration of well monitoring data. In this study, we first developed the data processing procedures for oilfield time series data and determined the proper selection of data sampling frequency, parameter combinations and data structures for deep learning models. Then we explored how Deep Learning (DL) models can be employed for well data analysis and how can we combine physics and DL models. Recurrent Neural Network (RNN) is a type of sequential DL model, which can be utilized for time series data analysis. This approach preserves preceding information and yields current response with memory of prior well behavior. Two candidate RNN models were tried to determine how well they were able to improve the accuracy and stability of well performance estimates. These two methods are Gated Recurrent Unit (GRU) and Long Short Term Memory (LSTM). In addition, a novel combination of RNN with Convolutional Neural Networks (CNNs), Long- and Short-term Time-series network (LSTNet), was also investigated. These various models were tested and compared based on the public production datasets from Volve Field. Both GRU and LSTM achieved higher accuracy in performance prediction compared to the simple RNN. In the case of frequent well shut-in and opening, the failure in capturing fast pressure responses and the extreme fluctuations with the simple RNN ultimately leads to high error. In contrast, LSTNet is more stable to frequent or significant well variations. With advanced deep learning structures, engineers can interpret long-term reservoir performance information from responses estimated by deep learning models, instead of performing costly well tests or shut-ins.

Characterizing the fractures is an important task to improve the understanding and utilization of hydraulic fracturing. As an approach to augment and improve on the existing methods, time-lapse electric potential measurements could be used to characterize subsurface features. In this study we investigated the characterization of fracture length and fracture density by using time-lapse electric potential data. A new borehole ERT (electric resistivity tomography) method designed specifically for hydraulic fracture characterization is proposed to better capture reservoir dynamics during hydraulic fracturing. This method uses high resolution electric potential data by implementing electrodes in or near boreholes and monitor electric potential distribution near the horizontal fracture zone. The time-lapse electric potential data generated by this tool were simulated and subsequently used to analyze fracture characteristics. Inverse analysis was then performed on the electric potential data to estimate fracture length and fracture density. Last, we performed sensitivity analysis to examine the robustness of the estimates in nonideal environments. The results of this work show that time-lapse electric potential data are capable of capturing flow dynamics during the fracturing process. Using the proposed borehole ERT method we successfully estimated the true fracture length and true fracture density of a constructed fracture model. We were able to determine the best locations in the constructed reservoir to place the electrodes, and through sensitivity analysis we found the maximum noise level of the electric potential data that can still allow the proposed method to make robust fracture length and fracture density estimates. Our proposed method offers a new approach to make robust estimates of fracture length and fracture density. Electric potential data have been used mostly for well logging in the past. This study demonstrates a novel way of using electric potential data in unconventional development and opens possibilities for more applications such as production monitoring.

There have been many reports on laboratory experiments measuring TEG (thermoelectric generator) power output at different flow rates of water, different temperatures, and other different conditions, but there have been few field tests utilizing geothermal wells. To this end, we designed a geothermal-TEG apparatus that has a six-layer modularized design to allow expanded power production by increasing the number of layers. After demonstrating the TEG's performance in the laboratory, we tested the apparatus in the field at a well located in the Bottle Rock geothermal field in The Geysers, CA, USA. The whole six-layer TEG device could generate about 500 W electricity with a temperature difference of about 152 °C between the hot and cold fluid manifolds, while each individual TEG chip could generate about 3.9 W. The steam pressure at the inlet of the TEG apparatus was about 122 psi, close to the wellhead pressure of 125 psi. After optimizing the field infrastructure, the TEG device could generate electricity without any leak at the wellhead pressure of 125 psi and the temperature over 176 °C (349 °F). The field test of the six-layer TEG device at Bottle Rock geothermal power plant was considered successful, and plans have begun to design and build a TEG device that could produce power up to 20 kW.

Fault zones affect the flow paths of fluids in groundwater aquifers and geological reservoirs. Fault-related fracture damage decreases to background levels with increasing distance from the fault core according to a power law. This study investigated mass transport in such a fault-related structure using nonlocal models. A column flow experiment is conducted to create a permeability distribution that varies with distance from a main conduit. The experimental tracer response curve is preasymptotic and implies subdiffusive transport, which is slower than the normal Fickian diffusion. If the surrounding area is a finite domain, an upper truncated behavior in tracer response (i.e., exponential decline at late times) is observed. The tempered anomalous diffusion (TAD) model captures the transition from subdiffusive to Fickian transport, which is characterized by a smooth transition from power-law to an exponential decline in the late-time breakthrough curves. This article is protected by copyright. All rights reserved.

Summary Hydraulic fracturing is a widely applied well-stimulation technique. Recently, the application of multiple hydraulic fractures along horizontal wells has made possible the exploitation of unconventional reservoirs, such as shale gas and shale oil. However, the process of horizontal-well fracturing is uncertain and it is not fully understood yet. There exist many questions about the number and position of created fractures and whether the fracture stays contained within the main reservoir thickness. This work presents the modeling and analysis of temperature data during and after the creation of multiple fractures along a horizontal well. Our model accounts for fracture growth and closure, the well effects, and interaction between multiple fractures. The interference between multiple fractures growing simultaneously and the presence of reservoir-permeability heterogeneity were investigated. Capabilities and limitations of information carried by temperature data are presented through different geometry analyses. A special consideration of fracture growth out of zone was addressed. For this case, it was considered that one of the hydraulic fractures activates a pre-existing fault and interconnects the reservoir with other zones hundreds of feet away. Our work shows the advantage of the use of continuous-distributed-temperature data in comparison with the traditional single-point-pressure analysis. The local characteristic of temperature allows us to identify not only which of the created hydraulic fractures has interconnected the main reservoir with a different zone, but also the location of this zone. We also account for the effect of the interconnection on the growth of the other hydraulic fractures along the horizontal wellbore. The simultaneous multiple fracture growth shows that falloff-distributed-temperature data can predict the number of created fractures. More than that, in the presence of heterogeneity, the interaction between the contained fractures can be captured by the temperature analysis. Different from the pressure analysis, distributed-temperature data can differentiate between heterogeneity locations along the wellbore. Throughout this study, we present a series of examples in which the temperature can provide information that would not be obtained from the traditional pressure analysis.

One of the key issues to a successful EGS (Enhanced or engineered Geothermal System) is the creation of a great density of fractures. The detection and characterization of the created fractures is crucial in evaluating the geothermal energy resources in such EGS projects as well as in reservoirs for CO 2 sequestration and storage. Methods to evaluate the fractures after stimulations are few, and limited in application. To this end, an approach to detecting and evaluating the fractures using resistivity data measured at different frequencies was developed in this study. The effects of fractures on resistivity measurements at different frequencies have been investigated as a function of water saturation in rocks with different porosity, permeability and lithology. Different rocks (Berea sandstone, and greywacke from The Geysers geothermal reservoir) were used in this study. The permeability of the samples ranged from 0.5 to over 1000 md for the matrix. The frequency ranged from 100 to 100,000 Hz. It was found that the effect of frequency on resistivity is different in rocks with and without fractures, especially in the range of low water saturation. The validity of the Archie equation depends on the existence of fractures, the frequency, and the range of water saturation. The relationship between resistivity and water saturation did not follow the Archie equation at low water saturation in some rocks with fractures. Models for characterizing different types of rocks with specific fracture patterns have been established using the resistivity data measured at different frequencies and different water saturations.

The enthalpy of two-phase geothermal fluids is typically monitored at the surface to understand the performance of a geothermal field. Surface enthalpy, however, cannot reflect real reservoir conditions downhole because of the heat loss along the wellbore. In addition, with only the surface enthalpy, the energy contributions from separate feed zones cannot be evaluated individually. A technique for determining downhole enthalpy in two-phase geothermal wells using surface inputs and chloride concentrations was developed in this study, including analytical models, experimental work and numerical studies, so that the downhole flowing enthalpy, the energy contribution from each feed zone and the heat loss along the wellbore can be evaluated. Chloride always stays in the liquid phase and becomes more concentrated as the geothermal fluid ascends to the surface and boils, and the change in chloride concentration along the wellbore can be utilized to calculate the change in the flowing steam fraction and the enthalpy. Analytical models to implement this idea were established for geothermal wells with a single feed zone or multiple feed zones, and the calculation procedures are elaborated with detailed flow diagrams and examples. The analytical models involve mass balance, energy balance and an assumption that the chloride concentration in the liquid from the feed zone is determinable. Inputs to the analytical model include two-phase mass flow rates and enthalpy at the surface and chloride concentrations at the surface and in the downhole. Temperature or pressure measurements are utilized to apply the energy balance in the model for multiple feed zones. Experiments were conducted to test the feasibility of an accurate determination of chloride concentration in two-phase fluids and validate the assumption in the analytical model. Chloride concentration distribution in the mixing area at the inlet of the feed zone was determined and visualized in the experiments. At the same time, similar mixing processes were visualized by a numerical simulation using ANSYS Fluent, and results from the numerical simulation are consistent with those from the experiments. Therefore, both the experiments and the simulation support the assumption in the analytical model that chloride concentration in the liquid from the feed zone can be determined accurately. It is demonstrated that the proposed method provides a fast and cost-effective way to determine the downhole two-phase flowing enthalpy in geothermal wells with multiple feed zones and estimate flow rates and energy contribution from each individual feed zone.

Technology using thermoelectric generators (TEG) has many advantages such as compactness, quietness, and reliability because there are no moving parts. One of the great challenges for TEG to be used for power generation is large-scale utilization. It is difficult to manufacture a TEGS system even at the scale of a few kilowatts (kW). To this end, we have designed and built a five-layer TEG apparatus with 90 individual power-producing TEG modules that can be installed with modularized units. Such a system with a layered structure could be expanded in power, something similar to solar Photovoltaics (PV). In this study, laboratory experiments were conducted using the built TEG apparatus to measure the power output and efficiency at different flow rates of water, different temperature, and different temperature differences between hot and cold sides. The effects of these parameters on voltage, power output, and efficiency were investigated and analyzed. The five-layer TEG device could generate about 45.7 W electricity with a temperature difference of 72.2°C between the cold and hot sides. The power of each module was about 0.51 W at this temperature difference. The experimental data can be applied to the design of commercial TEG systems. The expandable TEG system with layered structure provides a possible solution to scaling up TEG power generation to a commercial size.

Nanoparticle tracers are being investigated as a potential tool to measure temperature distributions in subsurface reservoirs. If the temperature distributions could be measured more precisely, this would greatly enhance the ability to estimate other reservoir properties, which would in turn inform reservoir engineering and field management decisions. Work toward two distinct but parallel objectives are described here. The first is to rank the informativity of various tracer candidates using modeling. The second is to develop temperature sensitive tracers experimentally, such that they would be capable of characterizing fracture properties and temperature distribution. The design of temperature-sensitive tracers is built around the design of the temperature sensing mechanism. In other words, the mechanism by which temperature is measured and the form and resolution of the resulting data, or response, have a profound impact on how informative that tracer can be about thermal breakthrough. Therefore, it is important to model the responses of candidate tracers in the context of an inverse problem to determine their relative informativity. In order to quantify tracer informativity for the simplified case in which the reservoir consists of a single parallel-plate fracture, one-dimensional analytical models were used to calculate the responses of four tracer types: conservative solute tracer, dye-releasing nanotracer, threshold nanoreactor, and temperature–time tracer (TTT). Inverse modeling was performed for various scenarios, and thermal breakthrough curves were calculated for each solution found. Tracers investigated in this study were found to perform very well, with small ranges of uncertainty for both thermal breakthrough forecasting and fracture property estimation. This was because the problem was highly constrained by defining the reservoir as a single fracture. For more realistic reservoir models consisting of complex fracture networks, it is expected that the uncertainty ranges would be larger and that all three temperature-sensitive tracers investigated here will outperform conservative solute tracers because their responses contain information about the temperature distribution. Experiments were performed to evaluate several potential nanosensor candidates for their temperature-sensitivity and to investigate particle mobility through porous and fractured media. Temperature-sensitive particles investigated in this study include the irreversible thermochromic, dye-attached silica and silica-protected DNA particles. A combined heat and flow test confirmed the temperature-sensitivity of the irreversible thermochromic particles by observing the color change. A detectable change in the fluorescent emission spectrum of the dye-attached silica particles upon heating was observed. The processing and detection of silica-encapsulated DNA particles with hydrofluoric acid chemistry was tested. A protocol to release the DNA by dissolving the silica layer without completely destroying the DNA was established. The silica-encapsulated DNA particles were flowed through a porous medium at high temperature. Some dissolution of silica particles was observed, leading to a reduction in their size. This research showed that synthesizing particles to respond to a specific reservoir property such as temperature is feasible. Using particles to measure reservoir properties is advantageous because particles can be transported to areas in the reservoir that would not be accessible by other means and therefore provide measurements deep within the formation.

Functionalized particles are being investigated as a potential tool to measure temperature distribution in fractured reservoirs. Acquiring reservoir temperature data within the formation could be used to correlate such information to fracture connectivity and geometry. Existing reservoir characterization tools allow temperature to be measured only at the wellbore. Temperature-sensitive nanosensors would enable in situ measurements within the reservoir. Such detailed temperature information enhances the ability to infer reservoir and fracture properties and inform reservoir engineering decisions. This study provides the details of the experimental work performed in the process of developing temperature nanosensors. Specifically, silica particles mobility through fractured media was investigated. Experimental results showed that the recovery of the particles was dependent on the particle size and suspension concentration. The particle size has a direct effect on its recovery. The controlling mechanisms for transport of silica particles were identified. Among all, transport by gravitational sedimentation (gravity settling) was prominent. Results also showed that the existence of the fracture facilitated the particle transport. Particles were found to flow with the fast-moving streamlines within the fracture.

The rich information contained in permanent downhole gauge (PDG) data has drawn attention from researchers. Previous work has demonstrated feature-based machine learning to be promising for PDG data analysis. The recent rise of deep learning was powered by techniques including recurrent neural networks (RNNs), which have been shown useful for processing sequential information. In this work, we explored how RNN can be utilized to analyze PDG data for better reservoir characterization and modeling, by examining two specific RNN structures: nonlinear autoregressive exogenous model (NARX) and standard RNN.RNN is a special type of artificial neural network designed for sequential data processing. Unlike a nonrecurrent neural network, the hidden layers in RNN take memories of previous computations (e.g. hidden layers or outputs computed in previous time) into account. In that way, the information contained in the measurements prior to a certain time can be used to model the response at that time, i.e. the convolutional effects are modeled by the recurrent structure of the RNN. Another favorable property of RNN is that it requires no assumptions on physics in advance. Both the inputs and outputs come directly from the raw measurements, and no handcrafted features need to be extracted from the forward physics model. Compared with feature-based machine learning, RNN is more powerful for the modeling where the forward model has not been determined.In this work, RNN was first tested on a series of synthetic and real flow rate-pressure datasets. The patterns learned by RNN from those data helped correctly identify the reservoir model and forecast the reservoir performance. RNN was also applied on temperature transient data, to demonstrate its advantage over feature-based machine learning with no assumptions on the physics model. The study also showed that RNN has the noise tolerance and computational efficiency that make it a promising candidate to analyze PDG data in practice.

Wettability alteration and interfacial tension (IFT) reduction are two important mechanisms for enhanced oil recovery (EOR). The introduction of nanotechnology from the fields of Biology and Material Science to the application in EOR is emerging because nanoparticles have the potential to alter formation factors like wettability and fluid properties like IFT and viscosity. However, a systematic literature review shows that ambiguity exists regarding whether nanoparticles can change wettability and IFT or not and which component in nanofluid plays a role. In this work, we investigated the effects of bare silica nanoparticles on wettability and IFT using a contact angle goniometer. The results showed that the contact angle measurement on quartz plates had relatively large uncertainty while those on calcite plates showed a clear trend that the smaller the nanoparticle size and the larger the nanofluid concentration, the smaller the contact angle. In addition, silica nanoparticles did not have an effect on IFT. Core flooding experiments showed an increase of 8.7% in oil recovery factor by the use of silica nanoparticles, which support the oil recovery mechanism of wettability alteration.