Assisted History Matching of a 1.4-Million-Cell Simulation Model for Judy Creek--A Pool Waterflood/HCMF Using a Streamline-Based Workflow
- Roderick Panko Batycky (StreamSim Technologies, Inc.) | Andrew C. Seto (Pengrowth Corp.) | Darryl Hyde Fenwick (Streamsim Technologies, Inc.)
- Document ID
- Society of Petroleum Engineers
- SPE Annual Technical Conference and Exhibition, 11-14 November, Anaheim, California, U.S.A.
- Publication Date
- Document Type
- Conference Paper
- 2007. Society of Petroleum Engineers
- 4.1.2 Separation and Treating, 2.2.2 Perforating, 5.5.8 History Matching, 5.5 Reservoir Simulation, 5.4 Enhanced Recovery, 4.3.4 Scale, 5.4.2 Gas Injection Methods, 5.5.3 Scaling Methods, 5.4.1 Waterflooding, 5.4.9 Miscible Methods, 5.1.5 Geologic Modeling, 1.6.9 Coring, Fishing, 4.1.5 Processing Equipment, 5.4.7 Chemical Flooding Methods (e.g., Polymer, Solvent, Nitrogen, Immiscible CO2, Surfactant, Vapex)
- 1 in the last 30 days
- 490 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 8.50|
|SPE Non-Member Price:||USD 25.00|
We illustrate what is practical for history matching of a large and complex field like Judy Creek. Due to the size of the geomodel, the limited time available, the large amount of historical data, and the 300+ wells, manual history matching methods were ruled out. Instead, we applied a streamline-based method to update the model based on well rate mismatches and identification of streamline bundles in 3-D space that contributed to local mismatches. Knowing the bundles and the magnitude of the mismatches, we computed cell-level multipliers for any gridblock and timestep and automatically update the geomodel. This eliminated tedious manual updates as in the conventional "box multiplier?? approach. However, due to the nonlinear nature of history matching, our method is not automatic but assisted, requiring engineering judgment to identify wells to focus on for matching and which timesteps to use.
Approximately 60+ flow simulations and geomodel updates were required to improved the well-level match within the time-frame allowed. Results showed that while our focus was improving the history match for the waterflood period, the field-level hydrocarbon miscible flood match was also improved. Furthermore, starting with a good geologic model led to changes that were within geological realism, even though we did not include geological modeling within each model update. Lastly, conventional history matching changes such as adjustment to relative permeabilities, solvent PVT mixing parameters, and well perforations, were also required.
Today, million cell flow simulation models can be easily created, even after upscaling. For large models like these, streamline-based flow simulators can efficiently simulate long historical production periods on standard hardware. See Thiele1 for a recent review of the details of streamline-based flow simulation. For overviews of streamline field applications, see Baker et al.2 or Samier et al.3
For large models, a primary bottleneck in the history matching process remains how to efficiently examine and then update the geological properties. Automated or semi-automated methods to update the geology are required. As an added benefit of streamline-based flow simulation, the streamlines can be used to guide the well-level history matching [HM] process as they can tell where and by how much to change the geology. Because of this convergence with large models, history matching, and streamlines, there is substantial activity in the area of streamline-based flow simulation and history matching.
Emmanuel and Milliken4 presented an early assisted history matching method using streamlines, whereby they used the streamlines to identify well regions and then computed Dykstra-Parson coefficients for each region. They then decided manually by how much to increase or decrease this coefficient, and the flow simulation was then repeated. So while they used the streamlines to suggest where to make changes, they were not used to determine by how much to make changes.
In this paper we use the approach outlined by Fenwick et al.5, in that we use streamlines to tell us where to make grid-level changes and by how much the changes should be. We have one variation from their method in that like Emmanuel and Milliken, we are not trying to maintain geological consistency when updating the geology to improve the well-level matches. Instead we map gridblock multipliers directly to the grid along their associated streamline-bundles in a "smart-painting?? type of approach. The reasons for ignoring this fundamental component of Fenwick et al.'s method are discussed in the HM workflow overview.
|File Size||1 MB||Number of Pages||10|