Assisted History Matching Workflow Research Articles

Machine learning assisted history matching for a deepwater lobe system

High exploration costs resulting in sparse datasets and complicated geological structures in deepwater depositional systems make the reservoir characterization extremely difficult. To meet this challenge, rule-based geostatistical subsurface modeling has been developed to fill this data gap with enhanced integration of geological concepts for deepwater reservoirs based on geological rules constrained by this sparse well data. The rule-based model simulates sediment dynamics through depositional rules to calculate reservoir architecture and the associated rock properties distributions. As a result, rule-based models incorporate conceptual and qualitative information, such as temporal deposition sequence and consequent compensational stacking patterns. Integrating quantitative data, such as fluid production history, is a remaining obstacle to the broad application of rule-based subsurface models. We propose a new machine learning assisted history matching workflow for rule-based models. First, multiple rule-based models are calculated as training data for a Generative Adversarial Network (GAN). The successfully trained GAN enables exploration of the latent reservoir manifold with only a small number of numeric values (i.e., latent random vector); this is a massive reduction in the dimensionality of potential subsurface models. The initial ensemble models, generated by the trained GAN with a range of latent random vectors, are coupled with physics-based reservoir flow simulation to obtain production forecasts. Ensemble Kalman filter (EnKF) updates the latent random vectors of the ensemble by minimizing the misfit (i.e., error norm) between the production forecasts and the production observation over time. Our proposed workflow finds an ensemble of optimized reservoir models that honor realistic geological heterogeneity and production history. Besides, as the GAN's latent random vectors are Gaussian distributed, one of the fundamental mathematical assumptions of EnKF, this workflow effectively alleviates possible artifacts (e.g., filter divergence or overshooting) in EnKF. Moreover, this proposed workflow is more computationally efficient than updating the entire high-dimensional reservoir properties and removes the limitation to only Gaussian simulation models with their limited geological realism. The proposed workflow may be expanded to various reservoir depositional settings. • We propose a machine learning assisted history matching workflow for complex deepwater subsurface models. • We apply generative adversarial networks (GAN) to represent the complex subsurface models by latent random vectors. • We update latent random vectors of the subsurface model to match the production history by the ensemble Kalman filter (EnKF). • By combining GAN and EnKF, we can 1) enhance computational efficiency, 2) remove artifacts, and 3) preserve geologic realism.

A Fast Gridding Method for Capturing Geological Complexity and Uncertainty

SummaryIn this paper, we introduce an efficient method to generate reservoir simulation grids and modify the fault juxtaposition on the generated grids. Both processes are based on a mapping method to displace vertices of a grid to desired locations without changing the grid topology. In the gridding process, a grid that can capture stratigraphical complexity is first generated in an unfaulted space. The vertices of the grid are then displaced back to the original faulted space to become a reservoir simulation grid. The resulting inversely mapped grid has a mapping structure that allows fast and easy fault juxtaposition modification. This feature avoids the process of updating the structural framework, which may be time-consuming. There is also no need to regenerate most of the reservoir properties in the new grid. To facilitate juxtaposition updates within an assisted history matching workflow, several parameterized fault throw adjustment methods are introduced. Grid examples are given for reservoirs with Y-faults, overturned beds, and complex channel-lobe systems.NOTE: This paper is also published as part of the 2021 SPE Reservoir Simulation Conference Special Issue.

Probabilistic history matching of multi-scale fractured reservoirs: Integration of a novel localization scheme based on rate transient analysis

In this paper, a new assisted history-matching workflow is presented, where rate transient analysis (RTA) results are used to constrain not only the initial discrete fracture network (DFN) models, the interpreted flow regimes are also used to formulate a localization scheme for more efficient updating of the pertinent DFN model parameters. The outcome is an ensemble of DFN realizations that are calibrated to both geologic and dynamic production data.RTA interpretations and other pertinent geological data are used to infer the prior probability distributions of the unknown fracture parameters, from which an ensemble of initial DFN models is sampled. The DFN models are subjected to numerical multiphase flow simulation. The fracture parameters are adjusted following an indicator-based probability perturbation method, which is capable of minimizing the objective function and reducing the uncertainties in the unknown fracture parameters simultaneously. A key feature is that the flow regimes identified from RTA are used to formulate a localization strategy, where individual segments of the production data are used to tune only a specified subset of the unknown model parameters.In a case study, the method is applied to characterize the probability distributions of four parameters in a multifractured shale gas well: primary fracture transmissivity, aperture of the secondary fracture, transmissivity of the secondary induced fracture, and global fracture intensity. Their probability distributions are updated following the proposed approach to match the production history. Multiple realizations of the DFN model are sampled.A key novelty is that the proposed probabilistic approach facilitates the representation of uncertainties in fracture parameters via multiple equally-probable DFN models and their corresponding upscaled flow-simulation models. A more comprehensive and robust approach is presented for integrating specific RTA interpretations and estimations into various steps of the history-matching process.

Characterizing hydraulic and natural fractures properties in shale oil well in Permian basin using assisted history matching

We performed assisted history matching (AHM) to an actual shale oil well in Permian basin using Embedded Discrete Fracture Model and neural network Markov chain Monte Carlo algorithm. Three main scenarios were investigated including hydraulic fractures only, hydraulic fractures with one realization of natural fractures, and hydraulic fractures with natural fractures. We integrated data from diagnostic fracture injection test analysis, PVT (Pressure Volume Temperature) report, nearby-well core data and petrophysical interpretation as initial input for AHM workflow. The uncertain parameters consist of fracture geometry, fracture conductivity, matrix permeability, matrix and fracture water saturation, and relative permeability curves. For the case with natural fractures, we included number, length and conductivity of natural fractures as the additional uncertain parameters. From history matching, we obtained the posterior distribution of each fracture parameter and reservoir property for all three scenarios. We found that the presence of natural fractures affects the posterior distribution of matrix permeability and fracture geometry of history match solutions, when comparing to case without natural fractures, by shifting to lower value distributions because natural fractures enhance the flow besides hydraulic fractures. Multiple realizations were then used for a probabilistic production forecast. We might underpredict oil production and overpredict gas production if the actual system was assumed with only hydraulic fractures. Lastly, the benefits from the study are that hydraulic fractures, natural fractures and reservoir properties are characterized in a probabilistic manner from production data. The results from the study can be used to improve future fracturing design and optimize well spacing.

Harmony search optimization applied to reservoir engineering assisted history matching

Based on the analysis of characteristics and advantages of HSO (harmony search optimization) algorithm, HSO was used in reservoir engineering assisted history matching of Kareem reservoir in Amal field in the Gulf of Suez, Egypt. HSO algorithm has the following advantages: (1) The good balance between exploration and exploitation techniques during searching for optimal solutions makes the HSO algorithm robust and efficient. (2) The diversity of generated solutions is more effectively controlled by two components, making it suitable for highly non-linear problems in reservoir engineering history matching. (3) The integration between the three components (harmony memory values, pitch adjusting and randomization) of the HSO helps in finding unbiased solutions. (4) The implementation process of the HSO algorithm is much easier. The HSO algorithm and two other commonly used algorithms (genetic and particle swarm optimization algorithms) were used in three reservoir engineering history match questions of different complex degrees, which are two material balance history matches of different scales and one reservoir history matching. The results were compared, which proves the superiority and validity of HSO. The results of Kareem reservoir history matching show that using the HSO algorithm as the optimization method in the assisted history matching workflow improves the simulation quality and saves solution time significantly.

Assisted history matching in shale gas well using multiple-proxy-based Markov chain Monte Carlo algorithm: The comparison of K-nearest neighbors and neural networks as proxy model

We performed assisted history matching (AHM) in a real shale gas well using uncertain parameters such as fracture geometry, fracture conductivity, matrix permeability, matrix and fracture water saturation, and relative permeability curves. We also investigated the performance of two proxy models including K-nearest neighbors (KNN) and neural networks (NN) to be used in multiple-proxy-based Markov chain Monte Carlo (MCMC) algorithm. We emphasized the performance of both proxy models by comparing the number of history matching solution found and elapsed time. While, KNN required less elapsed time by half than NN, we found that NN performed better in terms of accuracy and predictability than KNN. In other words, NN required a smaller number of simulations by half than KNN in order to obtain the same number of history matching solutions. Therefore, it depends on what is more important to each problem either number of simulations or elapsed time. For history matching result, both proxy models in the multiple-proxy-based MCMC algorithm have similar results of posterior distribution of uncertain parameters. This confirms the robustness of the proposed history matching algorithm. The benefits of this study are that we can characterize fracture geometry and reservoir properties in a probabilistic manner. These multiple realizations can be further used for a probabilistic production forecast, future fracturing design, and well spacing optimization and planning. This AHM workflow can be applied to any hydraulic-fractured wells with historical production data.

An Improved Numerical Modeling and Simulation Study of the Aquistore CO2 Storage Project

The Plains CO2 Reduction Partnership (PCOR) Program, led by the Energy & Environmental Research Center (EERC), regularly updates a numerical model of the Aquistore storage site and conducts simulation work as more injection and monitoring data become available for integration. The significance of the latest work presented in this paper is related to the integration of periodic monitoring data (CO2 flow profile at the injection well and CO2 breakthrough profile at the observation well) into the history-matching workflow using an assisted history-matching workflow to calibrate the simulation model further and provide a better predictive model to help optimize CO2 storage performance.

A practical and efficient iterative history matching workflow for shale gas well coupling multiple objective functions, multiple proxy-based MCMC and EDFM

A manual history matching is a time-consuming process and too risky for an investment decision because it cannot capture the uncertainty of subsurface information and production forecast. Therefore, Assisted History Matching (AHM) was adopted to obtain multiple reservoir realizations for the probabilistic production forecast. Markov Chain Monte Carlo (MCMC) is one of the AHM in the Bayesian statistics that have benefits of quantifying uncertainty without bias, not being trapped in any local minima. However, the AHM in shale reservoirs has not been widely studied due to the unconventional shale's complexities such as fracture modeling. The continuum approach cannot accurately model the connection between hydraulic fractures and matrix blocks in shale reservoirs. While, Discrete Fracture Modeling (DFM) by using local grid refinement can overcome the issue of the continuum approach, there is still a limitation of computational efficiency. Therefore, Embedded Discrete Fracture Modeling (EDFM) has been developed to combine the benefits of dual-continuum model and DFM. While DFM is an intrusive technique to model fractures, EDFM can account the mass transfer between matrix cells and fractures without being intrusive to the model. Besides, EDFM provides more efficient computational time than DFM. In this study, we proposed a practical and efficient iterative AHM workflow by integrating the benefits of using multiple objective functions, multiple proxy-based MCMC algorithm, EDFM and commercial reservoir simulator. Then, we demonstrated an application of the proposed workflow to a real shale gas well by matching both gas and water rates. The field application shows that the proposed AHM workflow efficiently finds the history matching solutions of 64 out of 300 simulation runs (21%) with a total computational time of one day. Furthermore, the workflow captures the uncertainty of reservoir and fracture parameters in terms of posterior distribution. Also, we performed the probabilistic production forecast and obtained P10-50-90 estimated ultimate recovery of gas production.

An efficient MCMC history matching workflow using fit-for-purpose proxies applied in unconventional oil reservoirs

History matching is a very challenging task because tremendous amount of simulations runs are required to achieve an acceptable accuracy target. In this paper, we develop and evaluate an assisted history matching (AHM) workflow using proxy-based Markov chain Monte Carlo (MCMC) algorithm, which is fully-automated and efficient for application in unconventional reservoirs. We use a synthetic simple case to compare the accuracy between four types of proxies: quadratic polynomial, cubic polynomial, k-nearest neighboring (KNN), and kriging. Also, we investigate the accuracy of proxy in different contexts such as different measurement errors and different quantity of initial simulation points. The results show that kriging proxy is more accurate than KNN proxy and cubic proxy. The quadratic proxy was the least accurate in our evaluations. However, if larger measurement error is introduced, the distinction between accuracy of proxy of the four proxies becomes less clear in spite of their different computational costs. The appropriate number of initial simulation points is another key to improve the AHM workflow efficiency. Incorporating these findings, we implement the AHM workflow on a synthetic field case and successfully perform history matching and uncertainty quantification.

Application of proxy-based MCMC and EDFM to history match a shale gas condensate well

History matching is an important process for uncertainty quantification of reservoir performance. Hundreds of reservoir simulation cases, or more, are typically required to sample the posterior probability density function of the uncertain parameters. The process is extremely challenging for shale reservoirs due to large uncertainties including reservoir heterogeneity, wide range of permeability, gas desorption, and fracture properties. Direct application of stimulated rock volume or dual continuum model fails to capture fracture connectivity and fracture conductivity that dominate the fluid transport in most shale developments. Embedding discrete fractures in reservoir simulation model is thus a preferable method to attain more realistic reservoir behavior. However, the option of using local grid refinement to embed the fractures is computationally expensive as well as intrusive to a reservoir model. Even more challenging is generating a huge set of fracture-embedded reservoir cases during the history matching process. Nevertheless, recent developments in a methodology called Embedded Discrete Fracture Model (EDFM) have successfully overcome the computational complexity to include discrete fractures in reservoir simulations. We demonstrate the implementation of the assisted history matching workflow coupled with EDFM to history match a real field case of shale gas condensate reservoir in Duvernay Shale, Canada. In the workflow, the EDFM preprocessor is fully coupled with a commercial reservoir simulator and proxy-based Markov chain Monte Carlo (MCMC) algorithm. Moreover, to further reduce the computational cost, the single-fractured submodel is used to represent the full reservoir model for this case study. Using the proposed assisted history matching workflow and the appropriate submodel enables efficient and successful uncertainty quantification of this shale play.

Application of assisted-history-matching workflow using proxy-based MCMC on a shale oil field case

History matching is a crucial step for reservoir simulation and for decision-making process in field development under uncertainty. The history-matching task is well known to be technically and computationally challenging. Several algorithms have been studied for more than a decade to assist history matching. One of the algorithms used is Markov chain Monte Carlo (MCMC), which is capable of providing accurate posterior probability density (PPD) of the history-matched realizations. While several researchers have applied and studied MCMC for assisted history matching (AHM) in many conventional reservoirs, only few studies have been performed on unconventional reservoirs. Since the difference in physics of the two reservoir types are important, it is worthwhile investigating the performance of AHM in unconventional reservoirs. For this purpose, we apply an AHM workflow using proxy-based MCMC on a shale oil well in Vaca Muerta formation to demonstrate application of the workflow and highlight the lessons learnt. The direct MCMC is also performed on the same field case to compare accuracy and efficiency of the first method. In this study, design of experiment (DOE) is used for selecting the most influential uncertain parameters before performing either of the two MCMC methods. It is found that the direct MCMC cannot find enough solutions to construct the statistically meaningful PPD in an efficient manner. By contrast, the proxy-based MCMC is less computationally demanding than the direct MCMC and efficient enough to construct the PPD. The tested workflow was then used to probabilistically forecast the cumulative oil and water production as well as the oil recovery factor for the Vaca Muerta well.

A Direct Overparameterize and Optimize Method for Stratigraphically Consistent Assisted History Matching of Object-Based Geomodels: Algorithm and Field Application

Summary Object-based static models are typically constructed for stratigraphically complex reservoirs. In this approach, the stratigraphic architecture is represented by use of distinct objects with specific geometric attributes and petrophysical characteristics. It is a major challenge to condition such models to production data while simultaneously maintaining the geologic realism and static conditioning. A novel work flow is developed for the assisted history matching (AHM) of object-based geomodels where the uncertain object locations and attributes are directly modified without resorting to (post-geomodeling) reparameterization techniques. It contains an object-modeling algorithm for channels and levees that provides direct access to the preraster geomodeling parameters for individual object locations and attributes. These parameters are gradually modified subject to physical constraints to achieve a history match. A fully integrated protocol is used in this process that automatically couples the static modeling algorithm with the reservoir simulator. The resulting work flow is moderated by a massively parallel and highly efficient iterative data-integration algorithm. In the AHM work flow, static and dynamic conditioning operations are respectively driven by separate objective functions and are performed at the iteration level in a sequential fashion. The static-conditioning step may add and remove objects in the geomodel, which changes the number of active AHM parameters over the course of the iterative search. The work flow handles such operations with minimal impact on the robustness of the search. A potential application of the direct AHM work flow for object-based geomodels is the identification of locations and attributes of channels in deepwater turbidite reservoirs, where the channels are typically below the resolution of the seismic data; the well spacing is typically larger than the characteristic object dimension, yet the production data exhibit strong sensitivity to channel connectivity. The concept of gradually adjusting the channel locations with the information in the production data (while maintaining static conditioning) is demonstrated on a real data set for a deepwater channelized-turbidite reservoir. The models proposed by the new AHM work flow not only improve the difficult-to-history-match injected-gas-breakthrough profiles but also provide geologically based explanations for them, taking into account the channel connectivity. The proposed AHM work flow ensures consistency across static and dynamic models by integrating multidisciplinary data with an easily auditable and replicable capability.