scholarly journals Application of Machine Learning and Artificial Intelligence in Proxy Modeling for Fluid Flow in Porous Media

Fluids ◽  
2019 ◽  
Vol 4 (3) ◽  
pp. 126 ◽  
Author(s):  
Shohreh Amini ◽  
Shahab Mohaghegh
Keyword(s):  
Artificial Intelligence ◽  
Machine Learning ◽  
Fluid Flow ◽  
Simulation Model ◽  
Mole Fraction ◽  
Reservoir Simulation ◽  
Simulation Models ◽  
Short Period ◽  
Proxy Models ◽  
Reservoir Simulation Model

Reservoir simulation models are the major tools for studying fluid flow behavior in hydrocarbon reservoirs. These models are constructed based on geological models, which are developed by integrating data from geology, geophysics, and petro-physics. As the complexity of a reservoir simulation model increases, so does the computation time. Therefore, to perform any comprehensive study which involves thousands of simulation runs, a very long period of time is required. Several efforts have been made to develop proxy models that can be used as a substitute for complex reservoir simulation models. These proxy models aim at generating the outputs of the numerical fluid flow models in a very short period of time. This research is focused on developing a proxy fluid flow model using artificial intelligence and machine learning techniques. In this work, the proxy model is developed for a real case CO2 sequestration project in which the objective is to evaluate the dynamic reservoir parameters (pressure, saturation, and CO2 mole fraction) under various CO2 injection scenarios. The data-driven model that is developed is able to generate pressure, saturation, and CO2 mole fraction throughout the reservoir with significantly less computational effort and considerably shorter period of time compared to the numerical reservoir simulation model.

scholarly journals Assessing the Needs to Incorporate Completion Details in a Petroleum Reservoir Simulation Model

2015 ◽  
Vol 8 (1) ◽  
pp. 16-28 ◽  
Author(s):  
Liang-Biao Ouyang
Keyword(s):  
Fluid Flow ◽  
Simulation Model ◽  
Simulation Study ◽  
Reservoir Simulation ◽  
High Rate ◽  
Petroleum Reservoir ◽  
Well Performance ◽  
Well Completion ◽  
Reservoir Permeability ◽  
Reservoir Simulation Model

Most of the current research and commercial reservoir simulators lack the capability to handle complex completion details like perforation tunnels in a simulation study. In most common applications, the simplified handling of completion complexity in reservoir simulations is not expected to introduce significant error in simulation results. However, it has been found that under certain circumstances, especially in high rate wells that have become more and more common in deepwater oil and profilic gas development, exclusion of the complex completion details in a reservoir simulation model would lead to nontrivial errors. New equations have been proposed to assess the needs to incorporate completion details in a reservoir simulation study based on the understanding of the fluid flow in a formation, the fluid flow along a wellbore and the fluid flow through perforation tunnels if exist. A series of sensitivity studies with different completion options under different flow and reservoir environments has been conducted to provide some guidance to improve well performance prediction through reservoir simulation. Impacts of key parameters like perforation density, perforation diameter, perforation length, wellbore length, borehole diameter, well completion configuration, well placement, reservoir permeability, reservoir heterogeneity, pressure drawdown, etc, have also been investigated.

History to Prediction Transition Calibration of Reservoir Simulation Model Ensemble

2021 ◽  
Author(s):  
Bjørn Egil Ludvigsen ◽  
Mohan Sharma
Keyword(s):  
Simulation Model ◽  
Reservoir Simulation ◽  
History Matching ◽  
Simulation Models ◽  
Productivity Index ◽  
Model Parameters ◽  
Gas Oil ◽  
Well Performance ◽  
Reservoir Simulation Model ◽  
Water Cut

Abstract Well performance calibration after history matching a reservoir simulation model ensures that the wells give realistic rates during the prediction phase. The calibration involves adjusting well model parameters to match observed production rates at specified backpressure(s). This process is usually very time consuming such that the traditional approaches using one reservoir model with hundreds of high productivity wells would take months to calibrate. The application of uncertainty-centric workflows for reservoir modeling and history matching results in many acceptable matches for phase rates and flowing bottom-hole pressure (BHP). This makes well calibration even more challenging for an ensemble of large number of simulation models, as the existing approaches are not scalable. It is known that Productivity Index (PI) integrates reservoir and well performance where most of the pressure drop happens in one to two grid blocks around well depending upon the model resolution. A workflow has been setup to fix transition by calibrating PI for each well in a history matched simulation model. Simulation PI can be modified by changing permeability-thickness (Kh), skin, or by applying PI multiplier as a correction. For a history matched ensemble with a range in water-cut and gas-oil ratio, the proposed workflow involves running flowing gradient calculations for a well corresponding to observed THP and simulated rates for different phases to calculate target BHP. A PI Multiplier is then calculated for that well and model that would shift simulation BHP to target BHP as local update to reduce the extent of jump. An ensemble of history matched models with a range in water-cut and gas-oil ratio have a variation in required BHPs unique to each case. With the well calibration performed correctly, the jump observed in rates while switching from history to prediction can be eliminated or significantly reduced. The prediction thus results in reliable rates if wells are run on pressure control and reliable plateau if the wells are run on group control. This reduces the risk of under/over-predicting ultimate hydrocarbon recovery from field and the project's cashflow. Also, this allows running sensitivities to backpressure, tubing design, and other equipment constraints to optimize reservoir performance and facilities design. The proposed workflow, which dynamically couple reservoir simulation and well performance modeling, takes a few seconds to run for a well, making it fit-for-purpose for a large ensemble of simulation models with a large number of wells.

Bridging the Gap Between Material Balance and Reservoir Simulation for History Matching and Probabilistic Forecasting Using Machine Learning

2021 ◽  
Author(s):  
Nigel H. Goodwin
Keyword(s):  
Machine Learning ◽  
Time Series ◽  
Simulation Model ◽  
Reservoir Simulation ◽  
History Matching ◽  
Material Balance ◽  
Data Driven ◽  
Probabilistic Forecasting ◽  
Reservoir Model ◽  
Reservoir Simulation Model

Abstract Objectives/Scope Methods for efficient probabilistic history matching and forecasting have been available for complex reservoir studies for nearly 20 years. These require a surprisingly small number of reservoir simulation runs (typically less than 200). Nowadays, the bottleneck for reservoir decision support is building and maintaining a reservoir simulation model. This paper describes an approach which does not require a reservoir simulation model, is data driven, and includes a physics model based on material balance. It can be useful where a full simulation model is not economically justified, or where rapid decisions need to be made. Methods, Procedures, Process Previous work has described the use of proxy models and Hamiltonian Markov Chain Monte Carlo to produce valid probabilistic forecasts. To generate a data driven model, we take historical measurements of rates and pressures at each well, and apply multi-variate time series to generate a set of differential-algebraic equations (DAE) which can be integrated over time using a fully implicit solver. We combine the time series models with material balance equations, including a simple PVT and Z factor model. The parameters are adjusted in a fully Bayesian manner to generate an ensemble of models and a probabilistic forecast. The use of a DAE distinguishes the approach from normal time-series analysis, where an ARIMA model or state space model is used, and is normally only reliable for short term forecasting. Results, Observations, Conclusions We apply these techniques to the Volve reservoir model, and obtain a good history match. Moreover, the effort to build a reservoir model has been removed. We demonstrate the feasibility of simple physics models, and open up the possibility of combinations of physics models and machine learning models, so that the most appropriate approach can be used depending on resources and reservoir complexity. We have bridged the gap between pure machine learning models and full reservoir simulation. Novel/Additive Information The approach to use multi-variate time series analysis to generate a set of ordinary differential equations is novel. The extension of previously described probabilistic forecasting to a generalised model has many possible applications within and outside the oil and gas industry, and is not restricted to reservoir simulation.

The Use of Reservoir Simulation in Deterministic Proved-Reserves Estimation

2016 ◽  
Vol 19 (03) ◽  
pp. 358-366 ◽  
Author(s):  
A. D. Jones ◽  
F. R. Denelle ◽  
W. J. Lee ◽  
D. G. MacDonald ◽  
B. J. Seiller
Keyword(s):  
Simulation Model ◽  
Reservoir Simulation ◽  
Simulation Models ◽  
Sufficient Evidence ◽  
Data Quality Control ◽  
Business Decisions ◽  
Independent Evidence ◽  
Physical Description ◽  
Static Data ◽  
Reservoir Simulation Model

Summary This paper proposes an approach for assessing a reservoir-simulation model for use in estimating reserves. A simulation model can integrate complex static data, the physical description of displacement processes, production constraints, and schedules. Hence, it can provide important information for business decisions and reserves estimation. Confidence in simulation predictions depends on the strength of evidence for the input data, quality control of the model, robustness of the history match, and whether there is independent evidence supporting predictions. We explain the principles for evaluating a simulation model and propose requirements for simulation predictions to be considered as proved reserves. This involves evaluation against different strands of evidence, such as static and dynamic characterization, wells and facilities description, reservoir performance, and analogs. Simulation models are often built to support business decisions by use of the best technical estimates for inputs. There can be instances where a simulation model may be reasonable and reliable but it only represents a “best technical” outcome. There may not be sufficient evidence to count the whole predicted recovery as proved reserves. We propose how such a model may be modified to also provide proved-reserves estimates. The approach can be used with different available data and at different stages of field life. It is illustrated through a case study that shows how the principles may be applied.

Hybrid Fluid Flow Simulation Combining Full Physics Simulation and Artificial Intelligence

2021 ◽  
Author(s):  
Mokhles Mezghani ◽  
Mustafa AlIbrahim ◽  
Majdi Baddourah
Keyword(s):  
Artificial Intelligence ◽  
Reservoir Simulation ◽  
History Matching ◽  
Hybrid Approach ◽  
Flow Simulation ◽  
Latin Hypercube Sampling ◽  
Simulation Models ◽  
Fine Scale ◽  
Simulation Results ◽  
Optimization Loop

Abstract Reservoir simulation is a key tool for predicting the dynamic behavior of the reservoir and optimizing its development. Fine scale CPU demanding simulation grids are necessary to improve the accuracy of the simulation results. We propose a hybrid modeling approach to minimize the weight of the full physics model by dynamically building and updating an artificial intelligence (AI) based model. The AI model can be used to quickly mimic the full physics (FP) model. The methodology that we propose consists of starting with running the FP model, an associated AI model is systematically updated using the newly performed FP runs. Once the mismatch between the two models is below a predefined cutoff the FP model is switch off and only the AI model is used. The FP model is switched on at the end of the exercise either to confirm the AI model decision and stop the study or to reject this decision (high mismatch between FP and AI model) and upgrade the AI model. The proposed workflow was applied to a synthetic reservoir model, where the objective is to match the average reservoir pressure. For this study, to better account for reservoir heterogeneity, fine scale simulation grid (approximately 50 million cells) is necessary to improve the accuracy of the reservoir simulation results. Reservoir simulation using FP model and 1024 CPUs requires approximately 14 hours. During this history matching exercise, six parameters have been selected to be part of the optimization loop. Therefore, a Latin Hypercube Sampling (LHS) using seven FP runs is used to initiate the hybrid approach and build the first AI model. During history matching, only the AI model is used. At the convergence of the optimization loop, a final FP model run is performed either to confirm the convergence for the FP model or to re iterate the same approach starting from the LHS around the converged solution. The following AI model will be updated using all the FP simulations done in the study. This approach allows the achievement of the history matching with very acceptable quality match, however with much less computational resources and CPU time. CPU intensive, multimillion-cell simulation models are commonly utilized in reservoir development. Completing a reservoir study in acceptable timeframe is a real challenge for such a situation. The development of new concepts/techniques is a real need to successfully complete a reservoir study. The hybrid approach that we are proposing is showing very promising results to handle such a challenge.

Robust Life-Cycle Production Optimization With a Support-Vector-Regression Proxy

SPE Journal ◽  
2018 ◽  
Vol 23 (06) ◽  
pp. 2409-2427 ◽  
Author(s):  
Zhenyu Guo ◽  
Albert C. Reynolds
Keyword(s):  
Life Cycle ◽  
Simulation Model ◽  
Support Vector Regression ◽  
Reservoir Simulation ◽  
Production Optimization ◽  
Forward Model ◽  
Support Vector ◽  
Well Control ◽  
Steepest Ascent ◽  
Reservoir Simulation Model

Summary We design a new and general work flow for efficient estimation of the optimal well controls for the robust production-optimization problem using support-vector regression (SVR), where the cost function is the net present value (NPV). Given a set of simulation results, an SVR model is built as a proxy to approximate a reservoir-simulation model, and then the estimated optimal controls are found by maximizing NPV using the SVR proxy as the forward model. The gradient of the SVR model can be computed analytically so the steepest-ascent algorithm can easily and efficiently be applied to maximize NPV. Then, the well-control optimization is performed using an SVR model as the forward model with a steepest-ascent algorithm. To the best of our knowledge, this is the first SVR application to the optimal well-control problem. We provide insight and information on proper training of the SVR proxy for life-cycle production optimization. In particular, we develop and implement a new iterative-sampling-refinement algorithm that is designed specifically to promote the accuracy of the SVR model for robust production optimization. One key observation that is important for reservoir optimization is that SVR produces a high-fidelity model near an optimal point, but at points far away, we only need SVR to produce reasonable approximations of the predicting output from the reservoir-simulation model. Because running an SVR model is computationally more efficient than running a full-scale reservoir-simulation model, the large computational cost spent on multiple forward-reservoir-simulation runs for robust optimization is significantly reduced by applying the proposed method. We compare the performance of the proposed method using the SVR runs with the popular stochastic simplex approximate gradient (StoSAG) and reservoir-simulations runs for three synthetic examples, including one field-scale example. We also compare the optimization performance of our proposed method with that obtained from a linear-response-surface model and multiple SVR proxies that are built for each of the geological models.

A Proxy Model for Predicting SAGD Production From Reservoirs Containing Shale Barriers

2018 ◽  
Vol 140 (12) ◽  
Author(s):  
Jingwen Zheng ◽  
Juliana Y. Leung ◽  
Ronald P. Sawatzky ◽  
Jose M. Alvarez
Keyword(s):  
Simulation Model ◽  
Network Model ◽  
Reservoir Simulation ◽  
Oil Sands ◽  
Training Set ◽  
Steam Assisted Gravity Drainage ◽  
Proxy Model ◽  
Reservoir Simulation Model ◽  
The Individual ◽  
Operational Constraints

Artificial intelligence (AI) tools are used to explore the influence of shale barriers on steam-assisted gravity drainage (SAGD) production. The data are derived from synthetic SAGD reservoir simulations based on petrophysical properties and operational constraints gathered from the Suncor's Firebag project, which is representative of Athabasca oil sands reservoirs. The underlying reservoir simulation model is homogeneous and two-dimensional. Reservoir heterogeneities are modeled by superimposing sets of idealized shale barrier configurations on this homogeneous reservoir model. The individual shale barriers are categorized by their location relative to the SAGD well pair and by their geometry. SAGD production for a training set of shale barrier configurations was simulated. A network model based on AI tools was constructed to match the output of the reservoir simulation for this training set of shale barrier configurations, with a focus on the production rate and the steam-oil ratio (SOR). Then the trained AI proxy model was used to predict SAGD production profiles for arbitrary configurations of shale barriers. The predicted results were consistent with the results of the SAGD simulation model with the same shale barrier configurations. The results of this work demonstrate the capability and flexibility of the AI-based network model, and of the parametrization technique for representing the characteristics of the shale barriers, in capturing the effects of complex heterogeneities on SAGD production. It offers the significant potential of providing an indirect method for inferring the presence and distribution of heterogeneous reservoir features from SAGD field production data.

Sign in / Sign up

Export Citation Format

Share Document