Reservoir simulation

Proper Simulation of Chemical-EOR Pilots—A Real Case Study

A critical step in proper design and optimization of any chemical-enhanced-oil-recovery (CEOR) process is appropriate and precise numerical simulations.

Abstract image of molecules over chemical formulas
Getty Images

A critical step in proper design and optimization of any chemical-enhanced-oil-recovery (CEOR) process is appropriate and precise numerical simulations. Addition of chemical species to the material-balance equations, along with finer resolution requirements for CEOR simulations compared with waterfloods, often makes it impractical to run full-field CEOR simulations to the required accuracy. Sector models, by their definition, are naturally suited for modeling of pilots. This paper presents a case study for appropriate simulation of a CEOR pilot.

Introduction

Chemical flooding (polymer flooding, surfactant/polymer flooding, and alkali/surfactant/polymer flooding) has improved significantly over the past decade because of vast research efforts to find practical solutions to specific field applications, improvements in the manufacturing and synthesis of new chemicals, field trials, and implementations by the industry.

Because of the complexity of CEOR and the inherent uncertainties regarding success of a particular CEOR process, extensive evaluations are required before decisions about full-field implementation can be made.

Numerical simulations are critical in CEOR evaluation and provide a realistic performance estimate, assuming that the underlying Earth model is reliable and appropriately calibrated. Usually, chemical floods are performed on candidate fields with a successful waterflood history, but CEOR simulations are more complex and challenging compared with water­flood modeling and history matching.

Sector-Model Generation

For the subject asset, which is a brownfield, the primary requirement in CEOR modeling is a representative, quality-controlled waterflood-history-matched full-field (or sector) model. Such a model should replicate the field performance (pressures and flow rates), at least on an overall basis and, preferentially, on a well-by-well basis. The model used for this study had high well density, was sufficiently reliable, and was used routinely for field development and performance prediction. The paper demonstrates the preparation and use of a sector model from this readily available waterflood full-field model for a ­polymer-flood pilot area.

The area of interest (AOI) for the pilot should be selected during an ­alternatives-analysis phase by a multifunctional team of field engineers and scientists and CEOR subject-matter experts after a project-framing exercise. In this case, this area is a seven-spot pattern containing a central injector and six surrounding producers. This is a small area compared with the full-field model. A local-grid-refinement (LGR) option was considered to create a sufficiently fine grid for CEOR simulation, but the resulting model was still deemed practically prohibitive.

It was then decided to use a sector model containing the AOI in the middle to evaluate the CEOR project. In order to minimize the effect of boundary on the AOI, one seven-spot pattern was added in each direction surrounding the sector model to be studied. Generally, when choosing the extent of the sector model, one can leverage the presence of sealing faults or other no-flow boundaries to minimize the required sector-model area. Unfortunately, in this case, there were no sealing faults or no-flow boundaries in the immediate vicinity of the AOI. Flow-diagnostics tools that use streamline simulations were used to ensure that all the streamlines connecting to the AOI are contained in the selected sector model. This is an important step in ensuring that all areas of the field in communication with the AOI are included in the sector model, which is critical for obtaining a representative CEOR forecast with this model.

Boundary-Condition Optimization

Although an attempt was made to minimize the effect of the sector boundary on the AOI by choosing a substantially larger sector model compared with the AOI, it is still critical to recreate the full-field model fluxes at the sector-­model boundary and evaluate the effect of boundary conditions on the wells in the AOI. There are many ways to reproduce the full-field model fluxes at the boundary of the sector model:

  • Adjust the contribution of the wells on the boundary to the sector model on the basis of a corrected AOI of the boundary wells.
  • If the preceding technique is not sufficient to reproduce the fluxes, one can use material-balance regions in the full-field model to extract the fluxes at the cells on the boundary of the sector model. This information can be used to tune injection/production rate or bottomhole pressure of pseudowells placed at the boundary of the sector model to reproduce the full-field model fluxes.
  • Pore-volume multipliers at the boundary of the sector model are another way of adding pressure support and mitigating the boundary effects.

Using these techniques can help establish the appropriate boundary con­dition that results in backward boundary-­condition optimization, meaning that, if the boundary conditions of the sector model are close enough to those of the full-field model, initializing the sector model to initial conditions (of the full-field model) and simulating the history of the field should result in saturation and pressure distributions and injection/production histories that are very similar to those of the same area/wells in the full-field model.

Grid-Size Optimization

To mitigate the numerical-dispersion problem, it is recommended to have approximately 15 cells between wells for numerical simulation of CEOR using single-point upstream fully implicit numerical reservoir simulators. To keep the artificial mixing in check, it is recommended to use cell sizes on the order of 30 ft or smaller. These are simply rules of thumb and should be tested for every reservoir and process. Choosing a practical cell size for a given problem requires balancing the level of accuracy required and the simulation runtime that can be afforded, given the number of simulations needed to optimize the slug design and perform required sensitivity simulations.

A detailed study was performed to evaluate various process and reservoir parameters affecting the optimal grid size for polymer and surfactant/polymer processes. A tool was then developed that could provide an educated initial guess for the appropriate grid size for the process of choice. After the grid-­optimization exercise for the problem at hand was performed, it was concluded that a 5×5 grid refinement (in the x- and y‑­direction) is required for appropriate simulation of this polymer flood. The performed level of refinement puts the cell size count of the sector model close to 400,000 cells.

Polymer-Flood Study

The two important design parameters for a polymer flood are processing rate and (endpoint) mobility ratio. The higher the polymer viscosity, the lower the endpoint mobility ratio and, thus, the better the sweep efficiency. On the other hand, processing rate (defined as the fraction of the reservoir pore volume that can be flooded on an annual basis) may also be proportional to the endpoint mobility ratio. Processing rate is a very important parameter for project economics, and the higher the processing rate, the higher the net present value of the CEOR project.

To find the polymer concentration that optimizes both processing rate and endpoint mobility ratio (best achievable sweep efficiency), polymer-concentration sensitivities were conducted using ­appropriate injection-pressure and ­production-rate constraints.

Conclusions

CEOR simulations are more complex and tend to be more computationally expensive compared with waterflood simulations. The complexity in CEOR simulations is because of additional ­material-balance equations (for the chemicals) that need to be solved, while the computational burden is the result of the finer grid resolution that is required to mitigate numerical dispersion and artificial chemical dilution, plus smaller timestep requirements of CEOR models because of highly nonlinear phase behavior and rheology of the chemicals. Thus, CEOR simulations are a trade-off between numerical accuracy and computational expediency. Massively parallel computing, dynamic LGR, and sector modeling are potential options to achieve the trade-off successfully. This paper focused on sector modeling and explained its challenges and its analysis and resolution options. For successful CEOR modeling,

  • Ensure that the sector-model boundary conditions (fluxes and pressure) adequately mimic full-field performance.
  • Get the grid resolution right, a critical step that ensures that

o The reservoir sweep is honored.o The chemicals perform per laboratory expectations in the reservoir.
o There is a reasonable trade-off between simulation accuracy and expediency.

  • Set aside repopulating the fine-grid properties by running Earth-modeling work flows on the fine-grid models.

Even if the fine-grid properties are inherited from the coarse grid, the flow-field heterogeneity of the fine-grid models could be very different from that of the coarse-grid models. The effect of this difference on the recovery efficiency predicted by the models is a function of how displacement and sweep efficiencies are affected by the numerical dispersion/artificial mixing and flow-field heterogeneity, respectively.
This article, written by Special Publications Editor Adam Wilson, contains highlights of paper SPE 179659, “Proper Simulation of Chemical-EOR Pilots—A Real Case Study,” by Nariman Fathi Najafabadi, SPE, and Adwait Chawathe, SPE, Chevron, prepared for the 2016 SPE Improved Oil Recovery Conference, Tulsa, 11–13 April. The paper has not been peer reviewed.