Evaluation of Key Reservoir Drive Mechanisms in the Early Cycles of Steam Stimulation at Cold Lake

Abstract

Summary. A history-matched, 2D, single-well numerical model was used to evaluate the contributions of four key drive mechanisms to early cyclic-steam-stimulation (CSS) oil recovery at Cold Lake, Alta. Formation compaction was found to be by far the dominant producing mechanism. Solution-gas drive was the most important of the remaining mechanisms. Fluid expansion had a relatively minor role. Gravity drainage accounted for little of the oil produced in the first two cycles, but increased in importance in subsequent cycles. Introduction Esso Resources Canada Ltd. (ERCL) has conducted experimental bitumen-recovery pilot operations in the Cold Lake oil sands since 1964 and commercial operations since 1985. The current bitumen production rate is about 90,000 B/D [14,000 M3 /d]. ERCL has established that CSS is a viable method to produce this resource economically. The evolution of the CSS operating strategy at Cold Lake has been largely empirical. Many different well configurations, completion strategies, and operating parameters were pilot-tested, and the overall operating strategy was modified as dictated by pilot results. Despite the wealth of pilot experience, however, questions remained about the role played by various producing mechanisms in the CSS process. This study was undertaken to improve understanding of the contributions of known producing mechanisms at Cold Lake to provide a basis for optimizing current operating strategies and for planning future development. This paper documents our evaluation of the relative contributions of four key reservoir drive mechanisms: formation compaction, solution gas, fluid expansion, and gravity. Cold Lake Reservoir Description The main target of steam stimulation at Cold Lake is the Clearwater formation, a member of the Lower Cretaceous Mannville group. The Clearwater formation was deposited as near-shore deltaic sands, with interbeds of silt and shale. The sands are thick, with gross pay up to 160 ft [49 m] and net pay up to 150 ft [46 m]. The average depth of the Clearwater is 1,450 ft [442 m]. The Grand Rapids shales overlie the Clearwater, and water sands of varying thickness partly underlie the oil-sands deposit. The Clearwater @ are unconsolidated, clean, well-sorted, fine-to-medium-grained, and have a porosity between 30 and 35 %. The in-situ absolute permeability ranges between 0.5 and 2.0 darcies. The oil saturation averages 10% of the bulk weight, or about 70% PV. The oil-sands deposit at Cold Lake contains roughly 150 billion bbl [25 × 109 M3] of bitumen. The in-situ viscosity of the bitumen is about 100,000 cp [100 Pa.s], but by heating to 500 degrees F [2600C], it can be lowered to less than 5 cp [5 mPa.s]. Oil viscosity and FVF as functions of temperature are given in Table 1. CSS Process Oil and water mobilities in the unheated bitumen-saturated sands are essentially negligible. To achieve the steam injectivity required in the commercial development at Cold Lake (about 1,320 B/D [2 10 M3/d] of cold water equivalent per well), the formation must be fractured. Early cycle injection pressures are typically 1,450 psia [10 MPa]. Both vertical and horizontal fractures have been created during steaming at Cold Lake. The observed fracture orientation generally has been horizontal in areas of the field that have undergone intensive steam injection. Injection periods typically last from 30 to 50 days. The production phase of the stimulation cycle ranges from around 120 days in the first cycle to more dm 400 days by the eighth cycle. In rich-oil-sand areas, the average production day oil rate per well declines from 155 B/D [25 M3 /d] in the first cycle to about 40 B/D [6 M3 /d] in the eighth cycle. A relatively small fraction of the injected water is produced back during the first two cycles, and the reservoir is in a state of overinjection; however, by the fourth cycle, total fluid production per cycle begins to exceed the volume of steam injected per cycle and the reservoir eventually becomes underinjected. Methodology Before the mechanism study began, it was necessary to develop a numerical simulator model of a typical CSS well and history match it over the early cycles of steam stimulation. A commercially available fully implicit thermal simulator was used with a standard black-oil, solution-gas treatment. To improve the representation of the physics of the CSS process, the simulator was enhanced to model both water/oil-relative-permeability-hysteresis behavior and the expansion/compaction of porosity that occurs during CSS operations at Cold Lake. The approach taken in developing the water/oil-relative-permeability-hysteresis model is discussed in Ref. 8. Relative permeabilities used in the investigation are based on laboratory data taken on preserved Cold Lake cores at elevated temperature levels representative of mean operating conditions. The two-phase water/oil and gas/oil bounding curves used are given in Figs. 1 and 2, respectively. The three-phase relative permeabilities are calculated with Stone's first method. The porosity expansion/compaction subroutine is consistent with the classical critical state concept of the geomechanical behavior of unconsolidated sands in response to shear failure, dilatancy, and consolidation. Ref. 9 describes the deformation model in detail. Actual parameter values used in the model are provided in Table 2. The values were established by history matching observed production behavior of a group of Cold Lake wells. Both the relative-permeability-hysteresis and reservoir-deformation enhancements were necessary to achieve a history match that was sufficiently accurate to enable us to proceed with the mechanism study. After a history-matched model was developed, the relative contributions of the four mechanisms to CSS oil recovery were evaluated. A base case was obtained by simulating five cycles of CSS operation with all four drive mechanisms operative. Two independent estimates of the individual contribution of each mechanism to oil production were then obtained. For the first estimate, one of the four mechanisms was systematically disabled and the modified simulator was used to recalculate the base case (the disabling of the mechanisms is discussed below). The second estimate was intended to be made from a series of base-case simulations in which all but one of the four mechanisms had been disabled. However, the second series of simulations could be done only in an approximate way, as discussed later. SPERE P. 207⁁