Steam injectivity during cyclic steam stimulation (CSS) at Cold Lake can be achieved only by injecting at pressures high enough to fail the formation mechanically. The reservoir also exhibits water/oil relative permeability hysteresis. This paper describes enhancements made to a thermal reservoir simulator to incorporate these Cold Lake physics. The geomechanical model allows both localized fracturing and global reservoir deformation (dilation and history-dependent recompaction). This representation allows the simulator to match injection and production pressures that are otherwise difficult to reproduce. The history-dependent relative permeability hysteresis model calculates gridblock relative permeabilities that always lie on or between input imbibition and drainage bounding curves. The model makes it possible to use laboratory-derived relative permeabilities in a simulation and still match field WOR's. A ----------------------~ ,,-- §

Summary Cyclic steam stimulation (CSS) typically is thought of as a single-well process. At Cold Lake, however, where steam injectivity is achieved by fracturing the formation, considerable interwell communication is observed. The result is usually the watering out of a producer by condensed steam from a neighboring injector. These interwell interactions greatly complicate sam-injection scheduling for commercial projects involving hundreds of wells but do not seem to reduce bitumen production in early cycles. Field experience indicates that steaming rows of wells sequentially with 50% overlap in injection time between adjacent rows significantly reduces water transfer between wells, increases the conformance of the injected heat, and reduces the field's tendency to form communicating well couplets. Exploratory numerical simulations show that the impact of steaming strategy on bitumen production is not significant until later cycles. Introduction CSS is a complex recovery process composed of a variety of recovery mechanisms whose relative importance changes with cycle number. For the Cold Lake reservoir, where steam injectivity is achieved by fracturing the formation, modeling the process is particularly challenging. The basis for investment decisions for commercial expansion, therefore, has been largely empirical. Even so, efforts to understand the process, and in particular to delineate the recovery mechanisms, are continuing. CSS generally has been modeled as a single-well process. The injected heat and fluids are envisioned to remain in the vicinity of the wellbore, lowering bitumen viscosity and increasing reservoir pressure. During the production phase, the increased reservoir pressure, along with gravity, drives heated bitumen to the wellbore. No-flow boundary conditions are assumed to encircle the spacing area of the well. Conceptually, such a model is correct for small steam-stimulation volumes. As the steam-stimulation volumes increase, however, the disturbances associated with heat injection become less localized; for very large volumes associated with continuous injection, the process evolves into a multiwell steamflood process. Thus, a continuum exists from single-well CSS to multi-well displacement. At the Leming pilot at Cold Lake, substantial single-well CSS behavior has been observed and successfully modeled for well patterns with large aspect ratios. CSS is not a single-well process. however, for commercial projects with steam-stimulation volumes of 8000 m3 [50,000 bbl] or more and wells on ha [4-acre] spacing with aspect ratios of 1.7 For example, a fraction of the fluids injected in a well may not be produced by the same well but rather by neighboring wells on production, indicating that elements of displacement also exist in the process. Interwell interactions during CSS at the Cold Lake area have been reported. This interwell interference or "communication" affects the production profiles dramatically. Thus, the total performance of the reservoir is not merely the summation of isolated wells whose performance is independent of the steaming sequence. This paper gives examples of the interwell communication observed with CSS at Cold Lake and introduces terminology useful for quantifying the process. Approaches to inferring the properties of the interwell communication path from surface measurements are outlined, and the practical limitations in their use noted. The paper then discusses a steaming strategy to reduce the degree of fluid breakthrough and to increase the reservoir conformance of the injected heat. Finally, results from numerical simulations of the multiwell CSS process are presented and compared with field experience.

The Cold Lake reservoir is an unconsolidated sand containing extremely viscous bitumen. Steam injectivity during cyclic steam stimulation can be achieved only by injecting at pressures high enough to mechanically fail the formation. Simulation of the complex fracturing and reservoir deformation behavior that results is very challenging. In addition, the reservoir exhibits water-oil relative permeability hysteresis, which must also be properly modeled. This paper describes enhancements made to a thermal reservoir simulator to incorporate these Cold Lake physics. Rigorous geomechanical modeling is not economical, so an empirical approach has been developed that is consistent with the behavior of unconsolidated sands. Fracturing is modeled by allowing the permeability in a plane of gridblocks to increase rapidly when the pressure exceeds a specified fracture pressure. Reservoir deformation in all blocks is modeled by first allowing dilation, during which porosity increases when the pressure exceeds a specified failure pressure. Subsequent pressure decline causes the reservoir to recompact, and porosity decreases. However, recompaction is not the reverse of dilation, and a fraction of the total dilation is permanent. While all gridblocks have similar deformation properties, the history of each individual block plays a role in determining its exact behavior. This geomechanical representation allows the simulator to match field observations that are otherwise difficult to reproduce, including injection pressures, flowback times, and production pressures. Also, the model appropriately handles the recompaction process which provides drive energy in the Cold Lake reservoir. The water-oil relative permeability hysteresis model is based upon laboratory measurements. Bounding imbibition and drainage curves are input; gridblock relative permeabilities, which depend upon both saturation and saturation history, are determined such that calculated values always lie on or between the bounding curves. The hysteresis model makes it possible to use laboratory-derived relative permeabilities when simulating cyclic steam stimulation and still match field water-oil ratios.

Canadian Petroleum Ltd. and partners in the Yemen Masila Block have successfully used detailed three-dimensional reservoir modeling and reservoir simulation to optimize the development of the larger oilfields in the Masila area. The models were used to predict reservoir performance and plan additional development drilling which subsequently demonstrated that the models accurately predicted drilling results. The main producing horizon in the Masila area is the Cretaceous Upper Qishn formation, a clastic-dominated transgressive depositional sequence with fluvial sediments at the base, tidal dominated estuarine sediments in the middle, and marine shoals at the top. This variable array of facies presents modeling challenges but the resulting heterogeneous models provide a realistic representation of actual reservoir characteristics. This paper describes the approach used to stochastically distribute both facies bodies and petrophysical parameters, and to upscale the model for reservoir simulation, while preserving the complex reservoir description. The Tawila field was the first Masila field to have wells drilled on the basis of the modeling effort, with very encouraging results. For these new well locations, the model successfully predicted both reservoir development and oil- water contact movements resulting from production from existing wells. This paper presents key conclusions and predictions from the modeling and reservoir simulation, and compares them to the results from subsequent drilling. As a result of the successful development drilling, these models are now an integral part of reservoir management and development planning for all Masila fields. P. 715