Layer | Fill | Outline |
---|
Map layers
Theme | Visible | Selectable | Appearance | Zoom Range (now: 0) |
---|
Fill | Stroke |
---|---|
Collaborating Authors
Results
Accelerating Tight Reservoir Workflows With GPUs
Mukundakrishnan, K.. (Stone Ridge Technology) | Esler, K.. (Stone Ridge Technology) | Dembeck, D.. (Stone Ridge Technology) | Natoli, V.. (Stone Ridge Technology) | Shumway, J.. (Stone Ridge Technology) | Zhang, Y (Stone Ridge Technology) | Gilman, J. R. (iReservoir.com, Inc.) | Meng, H.. (iReservoir.com, Inc.)
Abstract There are numerous complex characteristics that impact the long-term decline behavior of wells in tight oil and gas reservoirs. Numerical simulation with fine spatial discretization is required to capture important characteristics such as: the presence of natural fractures (stimulated and unstimulated); changes in total fluid mobility and compressibility; heterogeneous matrix and fracture properties (including permeability loss due to compaction); early flow transients; and nonuniform stimulation during hydraulic fracturing. This discretization requirement, when combined with the need to include multiple wells to capture interference effects, can result in model sizes in the tens of millions of cells over a few 640-acre sections. While these model sizes can sometimes be addressed with current-generation simulators, the excessive run time limits the ability to simulate multiple realizations for history matching and sensitivity analysis. In practice, lower fidelity simulations are often substituted. We present our efforts to help eliminate this tradeoff by building a fully-implicit black-oil simulator that combines recent advances in simulation algorithms with the high performance of GPUs. All major computational tasks are executed on GPU, including property evaluation, Jacobian construction and assembly, and linear solution with CPR-AMG. This approach allows models with many millions of cells to be simulated within minutes on a single workstation with multiple GPUs. For example, on a tiled SPE 10 model with 55 million cells and 250 wells, we simulate 2000 days of production in 20 minutes using eight GPUs. We summarize our approach to address the challenges in building a fine-grained, scalable simulator. We discuss two challenges in particular: the need to expose massive parallelism while retaining the robustness of linear solvers; and managing the complexity of the many features required by engineers for practical application. We discuss how we apply this fully-accelerated technology to increase fidelity and throughput on tight reservoir workflows, improving our understanding of the complex nature of production decline and possible long-term well interference. Furthermore, we illustrate workflows for simulations based on detailed reservoir/fracture descriptions.
- Geophysics > Seismic Surveying (1.00)
- Geophysics > Borehole Geophysics (0.68)
- North America > United States > Texas > Permian Basin > Yeso Formation (0.99)
- North America > United States > Texas > Permian Basin > Yates Formation (0.99)
- North America > United States > Texas > Permian Basin > Wolfcamp Formation (0.99)
- (22 more...)
Abstract Two-phase flow occurs during the production of oil and gas in the wellbores. Modeling this phenomenon is important for monitoring well productivity and designing surface facilities. Since transient time period in the producing well is usually shorter than reservoir time steps, stabilized flow is assumed in the wellbore. As such, semi-steady state models are used for modeling wellbore dynamics. However, in the case that flow variations occur in a short period of time (i.e., gas kick during drilling) the use of a transient two-phase model is crucial. A great deal of research has been conducted to study transient two-phase flow in wellbores. However, there is lack of a comprehensive two-fluid model in the petroleum literature. In this paper, we present an implementation of a pseudo-compositional, thermal, fully implicit, transient two-fluid model for two-phase flow in wellbores. In this model, we solve gas/liquid mass balance, gas/liquid momentum balance, and two-phase energy balance equations to obtain five primary variables: liquid velocity, gas velocity, pressure, holdup, and temperature. This simulator can be used as a stand-alone code or can be used in conjunction with a reservoir simulator to mimic wellbore/reservoir dynamic interactions. In our model, we consider stratified, bubbly, intermittent and annular flow regimes using appropriate closure relations for inter-phase and wall shear stress terms in the momentum equations. In our simulation, we found that the inter-phase and wall shear stress terms for different flow regimes can significantly affect the model's results. In addition, the inter-phase momentum transfer terms mainly influence the holdup value. The outcome of this research leads to a more accurate simulation of multiphase flow in the wellbore and pipes, which can be applied to the surface facility design, well performance optimization, and wellbore damage estimation.
- Europe (0.67)
- North America > United States > Texas (0.47)
Abstract A new analytical expression has been developed for the equivalent well block radius for three dimensional (3D) flow. The new equation has the same structure as Peaceman's equation with two new parameters added. Therefore, implementation in a reservoir simulator is straight forward. Numerical experiments have shown that Peaceman's equation for two dimensional (2D) flow can cause high rate errors for the partially penetrating wells. The new formula reduces these errors to a few percent. In this paper, derivation was presented for a vertical well, steady-state single phase flow with uniform and nonuniform grids. Multiphase, transient flow, horizontal wells and partial completion within a grid block is under development and will be presented in a new paper.
Abstract The well known PUNQ-S3 reservoir model represents a synthetic problem which was formulated to test the ability of various methods and research groups to quantify the uncertainty in the prediction of cumulative oil production. Previous results reported on this project suggest that the randomized maximum likelihood (RML) method gives a biased characterization of the uncertainty. A major objective of this paper is to show that this is incorrect. With a correct implementation of the RML method within a Bayesian framework, we show that RML does an adequate job of sampling the a posteriori distribution for the PUNQ problem. In particular, the true predicted oil production lies within the band of predictions generated with the RML method and is not biased. Very recently, the Ensemble Kalman Filter has gained notoriety, because it is very easy to couple with any reservoir simulator, allows one to continuously assimilate dynamic data as the forward simulation run is done and allows one to characterize uncertainty in performance predictions. When applied to the PUNQ data set, we show that this method also gives a reasonable quantification of the uncertainty in performance predictions with an uncertainty range similar to the one obtained with RML.
- Reservoir Description and Dynamics > Reservoir Simulation > History matching (1.00)
- Reservoir Description and Dynamics > Reservoir Simulation > Evaluation of uncertainties (1.00)
- Production and Well Operations > Well & Reservoir Surveillance and Monitoring (1.00)
- (3 more...)
Abstract In numerical reservoir simulation, the well index is used to relate the flowing well pressure to the numerical wellblock pressure. The well index is usually calculated from the equivalent wellblock radius. A previous paper presented an exact relationship for the equivalent wellblock radius for multiple wells with arbitrary rates. In this paper, a new, simpler, and more general, equation is presented for calculating the well index for all nw wellblocks in a reservoir. Arbitrary wellblock flow rates and interactions between wellblocks are fully accounted for. The data required for the new equation may be obtained in a preprocessor by calculating nw single-phase pressure distributions. Then an accurate WI can be calculated for each wellblock at each timestep, even under conditions where well rates vary with time. Introduction In reservoir simulation, well models are used to relate the rate of flow of fluid to or from a well to the difference between the wellblock pressure, pwb, and the flowing wellbore pressure, pwf. For general multiphase multicomponent flow, the mass flow rate of each component in each phase into the well may be writtenEquation (1) where WI is the well index. As pointed out by Wolfsteiner et al, the well index accounts for the geometry of the gridblock, location and orientation of the well segment in the gridblock, and rock properties. For single-phase flow, Eq. 1 simplifies toEquation 2 where q is volumetric flow rate into the well. Other authors have used another term for the well index, i.e., "connection transmissibility factor" or CF.
- Reservoir Description and Dynamics > Reservoir Simulation (1.00)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Drillstem/well testing (1.00)
- Production and Well Operations > Well & Reservoir Surveillance and Monitoring > Production logging (1.00)
- Reservoir Description and Dynamics > Reservoir Fluid Dynamics > Flow in porous media (0.71)
Abstract Well calibration after a history match in reservoir simulation ensures that the production wells will give realistic rates during the prediction phase. This process is usually very time consuming. The amount of trial-and-error work increases dramatically as the number of wells and/or the well productivity indices increase. In addition, reservoir simulation models are getting bigger and may contain a greater number of wells as computer hardware expands its limits. A reservoir model with hundreds of high productivity wells would take months to calibrate using conventional approaches. This paper describes a systematic well calibration procedure that yields very high degree of accuracy and requires only a fraction of the usual time. It was demonstrated that, with simple planning in the application of the PFV calibration process, a reservoir model with 500 high productivity wells could be calibrated within a couple days after obtaining a successful history match. The PFV procedure is general and it can be applied to any reservoir simulators. Introduction Calibration can be defined as the adjustment of model well parameters to match observed production rates at specified backpressures. These backpressures can either be flowing bottomhole pressure (FBHP) or flowing wellhead pressure (WHP). When specifying FBHP, the model is not concerned whether or not the fluids can be lifted to the surface. Therefore, tubing performance and surface conditions are not considered to be a constraint and are not modeled. When specifying WHP, the wellbore is modeled using a tubing performance curve, or flow table. This assumes that the surface conditions and wellbore hydraulics are critical to the prediction of flow rates. The calibration against FBHP is straightforward. Therefore, it is not discussed here. This paper focuses on the calibration against WHP which is very complicated. The process is done by trial-and-error. In addition, there might not be a solution between the flow table of the wellbore and the inflow performance curve of the reservoir. However, the well rate calculated in the prediction phase is more reliable. Past experience in Saudi Aramco has shown that the well calibration procedure is tediously long. On average, it takes about 3.3 weeks to calibrate a reservoir model with 150 wells. Moreover, reservoir simulation models are getting bigger and contain hundreds and even thousands of wells due to the rapid development in computer hardware. The time for calibrating the large number of wells could take months. Furthermore, the procedure for calibration varies for each of the different reservoir simulators. Therefore, it would be most beneficial to develop an efficient procedure that can be applied to all reservoir simulators available in-house. Current Situation There are three reservoir simulators being used in production mode in Saudi Aramco. They are MARS, CHEARS and ECLIPSE. Because each simulator has its own unique features, the calibration procedure is done differently. The most vigorous procedure was performed manually requiring many trial-and-error reservoir simulation runs. A typical well calibration procedure is depicted in Figure 1. A simulation engineer receives a set of best estimate flow tables from the Facilities Department. He inserts these flow tables in his reservoir simulation data, and he specifies the wells using measured wellhead pressures. After making a simulation run, he checks to see if the calculated rates match with the well test rates. If they do not matched, he then adjusts the well PI multipliers. If there is no solution from any flow tables, he requests new sets of flow table. New simulation runs are made, new flow tables requested if necessary, and PI multipliers adjusted until all the calculated rates match with the test rates.
- North America > United States > Texas (1.00)
- Asia > Middle East > Saudi Arabia (1.00)
- Government > Regional Government > Asia Government > Middle East Government > Saudi Arabia Government (1.00)
- Energy > Oil & Gas > Upstream (1.00)
- North America > United States > Texas > Permian Basin > Yeso Formation (0.99)
- North America > United States > Texas > Permian Basin > Yates Formation (0.99)
- North America > United States > Texas > Permian Basin > Wolfcamp Formation (0.99)
- (21 more...)
- Reservoir Description and Dynamics > Reservoir Simulation (1.00)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Drillstem/well testing (1.00)
- Production and Well Operations > Well & Reservoir Surveillance and Monitoring > Production logging (0.67)
Abstract One of the objectives of reservoir performance prediction is to determine the number of wells needed for future development of a field. It is known that most of the wells are deviated, especially in offshore environment. It is also known that conventional simulators cannot simulate the performance of slanted wells with acceptable accuracy in many eases. Many simulators cannot calculate the flowing bottom hole pressure of the wells accurately, even for vertical and horizontal wells in which well block aspect ratios are far from unity or where the well completion interval is short or the well is close to the reservoir boundaries. This paper presents formulations with which the flowing bottom hole pressure of wells (hence their productivity) can be accurately calculated for all wells. These will include vertical, horizontal and slanted wells and generally any well whose trajectory is not parallel with the simulation grid. A methodology is presented by which a conventional finite difference simulator can be used together with an analytical model that we have developed to calculate accurate well pressure and productivity for many types of wells. At this stage the solutions we offer cover a single well in an infinite slab reservoir or in a fixed boundary rectangular box shaped reservoir with a single phase fluid and only for homogeneous or anisotropic uniform permeability reservoirs. Research is in progress for other situations. Sensitivity studies on variables such as reservoir rock and fluid properties, simulation gridding, reservoir geometry and well geometry have been conducted. The validity of the new methodology is demonstrated by comparing the results of our analytical method with numerical solutions for cases in which there is no doubt on validity of the numerical results. Comparison is made between our analytical method and those available in the literature. The results demonstrate the superiority of the method. Furthermore, we have developed a new computer program which has coupled our analytical program with a numerical simulator in one package by which the necessary skin factor can be calculated in one run. It is demonstrated that the use of this program eliminates the need for a refined grid and it is less time consuming and much more economical than local grid refinement, where local grid refinement is used only to improve the numerical accuracy and not for inclusion of more detailed geological and petrophysical data. Introduction The flow equation for a bounded radial reservoir which contains compressible liquid may be written as: P. 273
- North America > United States (0.28)
- Europe > United Kingdom (0.28)
- Europe > Netherlands (0.28)
Holmes, J.A., Exploration Consultants Ltd. Member SPE-AIME Abstract A strongly coupled well model is described for use in a fully implicit 3-phase reservoir simulator. The model uses three variables for each well, instead of the usual single variable representing the well's bottom hole pressure. The two extra variables describe the contents of the wellbore, and provide a means of modeling crossflow of fluid between reservoir layers through the wellbore. For each of the standard modes of well control, it is possible to express the well flow rates in terms of the three well variables alone. This provides a convenient means of handling fully implicitly the collective interactions between wells that occur with group and field production rate control and re-injection or voidage production rate control and re-injection or voidage replacement operations. An example simulation is presented, to illustrate the capabilities of the fully implicit collective well control algorithm. Introduction The strongly coupled well model is an established technique in many implicit reservoir simulators. It couples the well rates to the reservoir conditions by creating an extra variable for each well, corresponding to the well's bottom hole pressure, which is solved simultaneously with the reservoir grid block pressures and saturations. The strong coupling between the source/sink terms and the grid block conditions enables the well to meet its production targets precisely, while maintaining numerical stability over long time steps. This type of well model was initially developed for use in single-well coning simulators, where an implicit numerical scheme is highly desirable for stability and robustness. Fully implicit general purpose black oil simulators were subsequently developed, and the concept of the strongly coupled well model was extended to handle any number of multiply-completed wells with a comprehensive range of well controls. This type of well model provides a robust means of handling traditionally provides a robust means of handling traditionally difficult situations, for example when a well is completed in several poorly-communicating layers of a highly stratified reservoir. It is therefore not surprising to find it now employed in many implicit reservoir simulators. We have extended the concept of the strongly coupled well model, by adding two extra solution variables for each well, making three in all. This extension brings two advantages. Firstly, it enables the simulator to handle, fully implicitly, situations where crossflow is occurring between reservoir layers through the wellbore. Secondly, it provides a convenient strategy for the fully implicit calculation of well flow rates when the operating targets of certain wells are directly influenced by the behaviour of other wells in the field. This extension has been implemented in our fully implicit black-oil/ gas-condensate reservoir simulator ECLIPSE100. THE STRONGLY COUPLED 3-VARIABLE WELL MODEL The state of each well is described by three variables instead of just one. The purpose of the two extra variables is to keep track of the contents of the wellbore at formation level. The variables represent the flowing fractions of water and gas in the wellbore mixture. The flows can be weighted to keep the gas fraction comfortably away from unity under normal producing conditions in oil wells. For example, an arbitrary set of weighting factors, g, such that gg = .01 go, may be used to give ......... (1a) ...............(1b) An alternative choice for the two extra variables would be the water and gas saturations in the wellbore at bottom hole conditions. The conditions in the wellbore would then be represented in an equivalent manner to the grid block conditions. p. 255
- Well Drilling > Pressure Management > Well control (1.00)
- Reservoir Description and Dynamics > Reservoir Simulation (1.00)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Drillstem/well testing (1.00)
- (2 more...)
Abstract A comprehensive well management program developed for use with current state-of-the-art black oil simulators during the prediction phase of reservoir performance studies, is presented. The reservoir is modelled as a hierarchical system possessing field, group, and well levels of possessing field, group, and well levels of control at which production and injection specifications and constraints can be imposed. The program enables wellhead rate changes, shut-in and reopening of wells, workovers, stimulations, artificial lift, and the drilling of new wells to be executed automatically to maintain production and injection levels. It also has the capability to automatically maintain the average reservoir pressure within specified limits through injection and production rate changes. This program incorporates production rate changes. This program incorporates a high degree of functional logic and many new concepts, strategies, and options not reported previously in the literature. previously in the literature. In this paper, the logical structure of the program, its optional features, and the strategies program, its optional features, and the strategies employed at different levels of control to meet the reservoir management objectives are described and are illustrated through specific numerical examples. Introduction The well management computer program reported here is designed to be used in conjunction with current state-of-the-art black oil simulators during the prediction phase of reservoir performance studies. For ease of presentation this performance studies. For ease of presentation this program will subsequently be referred to as WELMAN. program will subsequently be referred to as WELMAN. Its function is to maintain the production and injection targets set by the user for individual wells, groups of wells, and the field as a whole while meeting different operating constraints specified by the user at each of these levels. These constraints may be based on equipment capacity limitations, and economic, regulatory or technical considerations in the production of a reservoir. In the absence of a well management routine, the execution of the simulator has to be halted by the user whenever a target or constraint is not met, and the program has to be restarted after the implementation of some appropriate remedial action. In a multi-objective simulation of a reservoir having even a moderate number of wells, the total number of conditions to be satisfied becomes very large, and their simultaneous satisfaction through user intervention becomes an insuperable task. The primary purpose of a well management program, then, is to allow lengthy reservoir performance prediction studies to be made with only a minimum number of manual interruptions by the user. A secondary, but important, reason for using a computer program for this purpose is to systematize the decisions and calculations so that the same set of corrective actions is consistently implemented for the same set of constraint violations under similar conditions. If these decisions are left to a user burdened with a large number of decisions, it is likely that different remedies would be implemented at different times for the same problem. Finally, if the capability to conveniently activate different remedial measures during different simulations is provided, then the program would enable alternative provided, then the program would enable alternative strategies for producing a reservoir to be conveniently and objectively studied. The remedial actions referred to here are those operations usually implemented in the field to maintain production and injection levels. These include rate changes through opening or closing of wellhead valves, complete shut-in of flowing wells, and the retesting of shut-in wells for return to flowing status.
- Reservoir Description and Dynamics > Reservoir Simulation (1.00)
- Reservoir Description and Dynamics > Formation Evaluation & Management > Drillstem/well testing (0.88)
- Production and Well Operations > Well & Reservoir Surveillance and Monitoring > Production logging (0.72)
- Well Completion > Completion Installation and Operations > Perforating (0.67)