Summary
Traditionally, the evaluation of CO2-flooding processes is performed with finite-difference compositional-simulation models. However, compositional simulation is impractical for modeling large-scale CO2 floods because of computational run-time restrictions. In cases in which reservoir heterogeneity and fluid mobility dominate the reservoir recovery mechanism, streamline simulation offers a viable alternative to compositional simulation. The"reduced" physics in streamline simulation allows field-scale CO2-flood modeling to be feasible, as long as the streamline pressure/volume/ temperature(PVT) model can be calibrated so that the streamline model will produce accurate results for CO2-injection processes. Using streamline simulation allows for the evaluation of multiple full-field development scenarios that otherwise would not be possible with compositional simulation.
The objective of the study was to provide CO2-flood performance forecasts under various full-field development scenarios for the Midale field. This paper focuses on the methodology and results from the 1,000-well,>400,000-gridblock, 45+-year streamline simulation of the Midale field. In particular, it discusses the construction and history match of the full-field model, the calibration of the streamline model with the compositional model, and the development of the full-field CO2 forecasts.
Introduction
The Midale field, in southeast Saskatchewan, Canada (Fig. 1), was discovered in 1953 and subsequently delineated on 80-acre spacing. The field produced under competitive drainage until unitization in late 1962, after which an inverted nine-spot waterflood scheme was implemented. During the mid-1980s, an extensive vertical infill program was used to modify the waterflood patterns. Horizontal wells in the late 1980s and multi legged perpendicular horizontals in the mid-1990s were used to further improve waterflood conformance. To date, the unit has recovered more than 125 million STB of oil (primarily from waterflood operations), representing approximately 24% original oil in place (OOIP).
Recognizing the large volume of oil that would not be recovered by waterflooding operations, a CO2-flood pilot project was initiated in 1984. This project involved the drilling of 10 closely spaced wells in an area 4.4 acres in size and generated an enormous amount of reservoir and geological information. Results from the CO2 pilot project were used to justify the larger-scale Midale CO2-flood demonstration project, a six-pattern CO2 flood located in the southwestern part of the unit that began operations in 1992.Positive results from the Midale CO2 demonstration project were instrumental in justifying the neighboring Weyburn CO2-flood project, which began operations in2000, and they were also key to the technical justification of a full-field Midale CO2 flood.
Apache's acquisition of the field in 2000 was followed by an aggressive campaign to increase the recovery through infill drilling with horizontal wells at 20- to 40-acre spacing, increased injection and throughput (by a factor of three), and a review of the feasibility of a full-field CO2 flood.
The objective of this study was to assess the commercial CO2-flood potential of the Midale unit within a 5-month study period. Traditionally, a compositional simulator is used to accurately model the pressure-dependent phase behavior of CO2. However, despite advances in computing power and software, compositional simulation is impractical for field-level simulations of large fields such as the Midale unit. An alternative to compositional modeling is streamline simulation. Recent advances in streamline simulation show that in cases in which reservoir heterogeneity and the production/injection coupling dominate, first-order approximations offered by streamline simulation are sufficient for full-field development decisions. Invariably, the development plan is modified as the field is depleted and more information becomes available. Also, full-field streamline simulation allows for optimization of water-alternating-gas (WAG) cycles and pattern-injection timing that would be difficult to evaluate with compositional simulation. The difficulty with streamline simulation is that it lacks the direct PVT model to accurately describe the interaction between the oil and the CO2 at various pressures and temperatures. Detailed compositional models are required when drastic changes in fluid properties occur, such as near the critical point or condensate dropout in retrograde gas reservoirs. When the problem is one of modeling a relatively smooth transition between miscibility and immiscibility at a certain pressure, the modification of a black-oil model by Todd and Longstaff is quite often used because of its significantly faster computational speed.