A number of solvent-based processes for the recovery of heavy oil have been proposed in recent years. One of the phenomena that characterizes all such processes, to varying degrees, is viscous fingering. This paper describes the results of a combined experimental/simulation study aimed at characterizing viscous fingering under conditions typical of heavy oil recovery (very high ratios of oil to solvent viscosity). The study also sheds light on other phenomena that are part of such processes. We describe a set of four experiments carried out in heavy oil saturated sand packs contained within a 30 cm x?60 cm x?1.4 cm visual cell. Three of the experiments involved injection of a miscible, liquid solvent at the bottom of the sand pack, with subsequent upward displacement of the heavy oil; the fourth involved top-down injection of a gaseous solvent. The miscible liquid displacements were dominated by a single solvent finger, which broke through quickly to a producing well at the other end of the sand pack. Observed breakthrough times were consistent with a correlation that describes reported results at lower viscosity contrast. The gaseous solvent experiment exhibited fingering but also had features of a gravity-driven VAPEX process in its later stages. Numerical simulations using a commercial reservoir simulator have been successful in reproducing key features of the experiments. Realistic fingering patterns are produced in the simulations by assuming small, random spatial variations of permeability. The correct modelling of dispersion is crucial in matching the observed phenomena. For gaseous fingering and VAPEX processes, capillary effects are significant and should be included in simulations. Introduction Solvent-based processes for the recovery of heavy oil have attracted increasing attention in recent years. Much of this attention has focused on the Vapour Extraction or "VAPEX" process(1), a solvent analogue of steam assisted gravity drainage (SAGD). However, it has been suggested that for thin reservoirs, and particularly primary-depleted reservoirs, a cyclic solvent process might be preferred. Whereas VAPEX is analogous to SAGD, a cyclic solvent process would be analogous to the cyclic steam stimulation process. A concept for a cyclic process is shown in Figure 1. In this concept, solvent would be injected for a period of time, then oil produced from the same well; and this process would be repeated. A number of questions may be asked about the basic mechanisms of a cyclic solvent process and the resulting efficiency of oil recovery. The work reported here was aimed particularly at understanding the phenomenon of viscous fingering, which characterizes any such process in which a low viscosity solvent is injected into a high viscosity oil. Viscous fingering is an instability phenomenon which occurs when one fluid is displaced by another fluid of lower viscosity. The displacing fluid is said to "finger" into the resident fluid. The two fluids may be either miscible or immiscible, and the displacement may take place in a porous medium or even a Hele- Shaw cell(2, 3).

A solvent-assisted gravity drainage process (Vapex) for recovery of heavy oil or bitumen offers high recoveries and promising rates of oil production. In order to predict process performance, data on solvent-oil solubility and on the viscosity of solvent-oil mixtures must be obtained. The solubility of selected gasses in Lloydminster Aberfeldy oil and in a Cold Lake oil was measured. The gasses used were: CH4, C2H6, C3H8 and CO2. Measurements were done at reservoir temperature.The data were regressed using the Peng-Robinson equation of state, which was used to generate k-values expressing the solubility of the gas-oil systems. Regressing the Peng-Robinson equation to the measured data generated interaction coefficients for the systems measured. These coefficients were used with the equation to generate k-value or solubility tables at other conditions. Measured viscosity data were used to confirm the usefulness of the Puttagunta viscosity correlation for propane-based heavy oil systems. The work confirmed the formation of two liquid phases in the oil-propane system at high solvent loading. The measurements also confirmed the large viscosity reductions available (100:1–200:1) by saturating oil with light hydrocarbons. A viscosity increase in one oil-propane system was observed at high solvent loading, suggesting possible asphaltene precipitation and/or deposition on the walls of the capillary viscometer tube. These observations confirmed the need to study phase behaviour and asphaltene deposition in the oils at high solvent loading, as well as obtaining solubility and viscosity measurements. The data have been used to perform numerical simulations of Vapex and other solvent-based processes, and to perform predictions of field process performance. Introduction Thermal recovery processes have been used successfully on many Alberta bitumen and heavy oil reservoirs. Some reservoirs, however, are not suited to thermal processes. This may be due to depth, unfavorable mineralogy, bottom water, thin pay sections, or a combination of these factors. For these reservoirs, a non-thermal process may be more suitable.The most likely candidate is a Vapex-type process, where oil is contacted by solvent vapour. The vapour dissolves in the oil, and diluted oil drains to a production well. The application of this technology to heavy oil recovery requires confident prediction of the process performance for a field-scale operation. This in turn requires knowledge of the mechanisms active in the process, and of the magnitude of each of these mechanisms. Mechanisms identified to date include solubilization of the solvent in oil, mass transfer from vapour to liquid phases by diffusion, mixing of diluted and undiluted oil by diffusion and dispersion, reduction of the oil viscosity by solvent dilution, and upgrading of the oil by asphaltene precipitation and deposition. This work measured solubility and viscosity of several oil-solvent systems.

Heavy oil is commonly produced in the form of water-in-oil emulsions. It has long been debated whether the emulsions are formed in the reservoir, and if so, what effect they have on the recovery process. This work examines the conditions under which water-in-oil emulsions can form in situ, and their flow properties. A number of experiments have been carried out in which water and oil were injected as separate phases into a sandpack, and the produced oil analyzed for emulsified water content We find that emulsified water content is small at low injection rates, but that above some threshold rate, the water content rises rapidly. Dependence of the threshold rate on oil type, oil viscosity, water:oil ratio, pack permeability and pack wettability have all been examined. Our results suggest that oil phase capillary number, Nc, defined as (oil velocity)x(oil viscosity)/(oil-water IFT) is a natural dimensionless parameter to describe the emulsification threshold. For oil viscosity much greater than that of water, and permeability of the order of a few darcies, threshold Nc is between 10–4 and l0–3. Such capillary numbers are commonly encountered in near-wellbore flow during heavy oil recovery operations. Experiments comparing a reservoir oil with mineral oils show similar threshold capillary numbers. The threshold increases with permeability of the medium and is significantly lower in oil-wet compared with water-wet sand. Preliminary results indicate that a significant reduction in effective mobility of the oil phase occurs when emulsification takes place. Introduction A majority of the oil produced from heavy oil reservoirs is in the form of emulsions, most often of the water-in-oil (w/o) type. It has also been proposed that oil-in-water emulsions be injected to improve mobility control or achieve better conformance by plugging more permeable zones of the reservoir. In order to predict the effects of either naturally formed emulsions or injected emulsions on a recovery process, it is necessary to characterize the formation and flow of emulsions in reservoirs. The emulsions produced in heavy oil recovery processes have given rise to several debates. One issue has been whether the emulsions are produced by fluid flow through the reservoir, or simply by the pumps and valves used for production. A body of evidence now suggests that emulsification takes place, at least partly, in situ, although the emulsification process is still poorly understood. P. 675

In 1998, Butler and Mokrys proposed a "Closed-Loop Extraction Method for the Recovery of Heavy Oils and Bitumens Underlain by Aquifers." The process has potential application to many Alberta and Saskatchewan heavy oil reservoirs. The objective of our work was to produce an experimental evaluation of solvent-assisted process options for bottom water reservoirs. The current work is entirely experimental, and provides data that may be used to back up a numerical simulation effort. The experimental series modelled a bottom water process in order to determine its feasibility for a field-scale oil recovery scheme. A series of five experiments were run in an acrylic visual model. Pujol and Boberg's scaling criteria(1) were used to produce a lab model scaling a field process by a geometric ratio of 100:1, and compressing field time by a ratio of 10,000:1. The model simulated a slice of a 30 m thick reservoir with a 10 m thick bottom water zone, containing a pair of horizontal wells at the oil-water interface, offset by 25 m. For field prediction, experimental results were scaled up to represent a 30 m thick reservoir (20 m thick oil zone) with 500 m horizontal wells. The experimental rates were negatively impacted by continuous low permeability layers and by oil with an initial gas content. The lower effective diffusion rates required that the surface area exposed to solvents be increased in order to achieve commercial oil recovery rates. The Bottom Water Process described in this report offers the opportunity to do just that, as the large surface area of the oil water interface between the wells will provide contact for solvent by injecting gas at the interface. Given an appropriate well spacing, high production rates should be possible. Introduction The Alberta Research Council (ARC) has done several years of investigative work into solvent-assisted heavy oil recovery processes(2, 3). The present report describes a particular contribution to solvent-assisted oil recovery technology; a comparative scaled physical model study of bottom water process options. The results of this work showed scaled field rates of 25.3 m3/d and a live oil scaled rate of 16.5 m3/d, both at 25 m offset well spacing. Mechanisms of the Bottom Water VAPEX Process The Bottom Water VAPEX Process, illustrated in Figure 1, is a recovery process depending on the interplay of several mechanisms for its success. The solubility of the gas in the oil is controlled by the k-values of the oil/solvent system. Diffusion, hydrodynamic dispersion, and swelling also play a role in the movement of gas into the reservoir oil. The oil flow is enabled by viscosity reduction due to the dissolution of solvent in the oil. Oil-solvent contact is further augmented by capillary pressure moving some oil into the vapour chamber zone, as was observed in Experiment #2. Heterogeneity of the reservoir sand further increased the surface produced by capillary action, but excessive layering can hinder the movement of oil, as was shown by Experiment #3.

A series of steam corefloods has been performed utilizing X-ray CT imaging to monitor phase saturations during the floods. A numerical simulator was used to analyze the experimental results. The steamfloods of initially water-filled sand packs were carried out at low enough flow rates that capillary pressure was comparable in magnitude with viscous pressure drop. Numerically marching the observed pressure drops provided a significant test of the assumed steam-water capillary relationship. A capillary pressure curve of Leverett j-function form was found suitable for the simulations. The steam relative permeability curve was found by numerically matching both pressure drop and CT saturation data. The expected steam frontal behaviour was seen in the CT images, but the images also revealed unexpectedly large variations in water saturation transverse to the steam propagation direction. Although core porosity derived from the CT images showed no significant variation across the core, it was hypothesized that small variations in steam-water capillary pressure could have resulted from more subtle core packing variations. The expected type of core heterogeneity was subsequently confirmed by subjecting the core to petrographic image analysis. Numerical simulations embodying a realistic degree of core heterogeneity were able to reproduce the transverse saturation variations as a consequence of capillary crossflow. Introduction The steam displacement or ‘steam drive’ method is one of the major techniques for thermal recovery and has been intensively studied both theoretically and experimentally. Particularly important from a theoretical standpoint was the work of Miller(1) showing that the frontal propagation of steam was far more stable than that of a non-condensible gas. Although many analytic methods have been developed and used for simplified predictions(2, 3, 4), in most practical situations a numerical simulator is employed to model steam processes. A fundamental physical property required for the simulation of steam-based processes is the relative permeability of steam. Customarily, numerical models treat steam flow in the same manner as the flow of any ocher reservoir gas. As noted by Sanchez and Schechter(5) there is little direct evidence to support this assumption. Indeed, the issue of steam relative permeability has generated debate in the literature right up to the present. For example, Sanchez and Schechter(5) and Closmann and Vinegar(6) present experimental results confirming the similarity between steam and non-condensible gas relative permeabilities. Verma et al.(7) however, find that steam relative permeability is significantly higher than that of a non-condensible gas. The latter authors further argue that it is reasonable to expect such differences because of the possibility of phase transformation in the flow channels. Because of the diversity of opinion surrounding the basic issue of relative permeability, it is important to investigate steam flow using a range of tools. In particular, the technique of X-ray computed tomography (CT) provides the possibility of monitoring fluid saturations while processes are taking place in a laboratory coreflood experiment. This relatively new technique in petroleum research has already proven itself to be very useful in a wide range of experimental studies(6, 8, 9, 10).