Solvent-dependent recovery characteristic and asphaltene deposition during solvent extraction of heavy oil

Solvent-dependent recovery characteristic and asphaltene deposition during solvent extraction of heavy oil

Fuel xxx (xxxx) xxxx Contents lists available at ScienceDirect Fuel journal homepage: www.elsevier.com/locate/fuel Full Length Article Solvent-dep...

5MB Sizes 0 Downloads 12 Views

Fuel xxx (xxxx) xxxx

Contents lists available at ScienceDirect

Fuel journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Solvent-dependent recovery characteristic and asphaltene deposition during solvent extraction of heavy oil ⁎

Xuesong Lia, , Steffen Berga, Orlando Castellanos-Diazb, Andreas Wiegmannc, Marco Verlaand a

Shell Global Solution International B.V., The Netherlands Shell Canada Ltd., Canada c MATH2MARKET GmbH, Germany d Shell Kuwait Exploration and Production B.V., Kuwait b

A R T I C LE I N FO

A B S T R A C T

Keywords: Micro-model Solvent (assisted) heavy oil recovery VAPEX Pore scale Propane Pentane Phase behaviour Asphaltenes precipitation and deposition

Most traditional recovery methods for heavy oil, such as thermal recovery processes, are energy consuming and greenhouse gas emission is intense. Non-thermal based solvent extraction methods are attractive because they require much less heat, no water and the produced heavy oil has low viscosity even at surface conditions, making it easier to transport. However, heavy oil recovery by solvent occurs mainly at the moving solvent front which is only a few millimetres width. It is thus challenging for it to be modelled in reservoir scale and leaves a large uncertainty regarding the recovery factor prediction. For any reliable scaling up, the fundamental details of the extraction in millimetre range down to the pore scale need to be understood. In this study, we use a 2D micromodel operated at reservoir pressure and temperature conditions, in order to analyse the extraction processes of bitumen by solvent. Two different solvents, pentane and propane, which are both considered as field-relevant candidates, are selected for comparison and toluene is selected as a fully miscible reference case. By imaging both the displacement pattern on a larger scale of 5 cm and, simultaneously, the details of the flow processes at pore scale, it was observed that both pentane and propane precipitate asphaltenes. But in the case of pentane, the precipitated asphaltenes are solid which significantly reduce the permeability thus resulting in plugging. The result from flow simulation confirmed the influence of formation plugging on the flow and diffusion process. In the case of propane, the precipitated asphaltenic phase remains liquid, hence mobile, which leads to capillary trapping. Consequently, the recovery factor for propane is ~15% higher than for pentane and more stable. In general, this study suggests that heavy oil recovery using solvent is a dynamic process where the pore scale phase characteristic of asphaltene has a strong influence both on the transport of fresh solvent and the production of dissolved crude components. These differences between the solvent are key and should be included in field scale estimations.

1. Introduction Significant heavy oil and extra-heavy oil resources exist in places such as Canada, United States, Venezuela, Russia and the Middle East, estimating around 6–9 trillion barrels in place [1]. These fluids are highly viscous, which makes primary recovery difficult. Currently, thermal EOR is applied to unlock these resources, mainly using steam at 200–350 °C to increase reservoir heating, reduce oil viscosity and displace it towards producer wells. Since reservoir porosities ranges between 20 and 30%, which means that most of the volume of the reservoir is occupied by rock, thermal EOR processes are inherently energy-inefficient. Most of the injected heat will be wasted on the large solid fraction in the reservoir and the surrounding formations instead of



on the targeted oil, leading to intense CO2 emissions. Additionally, the treatment of recirculated water consumes additional energy and increase extra emissions [2]. Heavy oil production can emit up to 100% more greenhouse gases compare with more conventional sources of oil [3]. An alternative for in-situ heavy oil viscosity reduction is dilution with solvents. The viscosity reduction achieved by heating the oil to 200–250 °C using steam can be achieved by diluting the oil with light hydrocarbon solvents at much lower temperatures such as 60–80 °C. Consequently, the energy consumption can be dramatically reduced by completely replacing steam with solvent, while reducing greenhouse gas emissions [4]. Water consumption is also reduced significantly, hence a reduction in capital expenditure on water treatment facilities.

Corresponding author. E-mail addresses: [email protected] (X. Li), steff[email protected] (S. Berg).

https://doi.org/10.1016/j.fuel.2019.116716 Received 16 April 2019; Received in revised form 13 November 2019; Accepted 20 November 2019 0016-2361/ © 2019 Published by Elsevier Ltd.

Please cite this article as: Xuesong Li, et al., Fuel, https://doi.org/10.1016/j.fuel.2019.116716

Fuel xxx (xxxx) xxxx

X. Li, et al.

processes is much more efficient [18,20–22]. Solvent-only oil recovery pilots started in Alberta, with the first ones focusing on dry VAPEX applications, see Fig. 2. These earlier pilots were dismissed due to lower-than-expected oil production, low solvent recovery, and/or improper design and execution of the pilot. Recent trails are more successful, i.e. NSolv’s warm VAPEX trial [23] has demonstrated that these technologies have the potential to replace thermal EOR in an economic manner. The target solvents are light alkanes such as propane. In a 2017 report, the Government of Canada National Energy Board stated, that the use of solvents such as propane is expected to increase from 5 Mb/d in 2016 to over 12 Mb/d in 2040, i.e. more than double within the coming 25 years, even to over 16 Mb/d depending on the scenario [30]. Since the introduction of VAPEX using cold hydrocarbon injection (or “dry VAPEX”), other VAPEX-like technologies have been proposed [31].There are also attempts to use different solvents such as CO2 [32,33], which capture the purpose of reducing the CO2 emission or methane, however as most of the reservoirs are having lower pressures than their vapour pressures, in these cases the gas phases will directly displace the heavy oil phase. The associated larger viscosity ratio between the gas and heavy oil in combination with a relative high permeability reservoir, makes stable displacement a challenge. For commercially implementing warm VAPEX, it is important to have a good estimation on the recovery factor and evaluation on risks such as formation plugging. Therefore, understanding the recovery mechanisms, upscaling to reservoir simulation tools and assessing the impact of in-situ asphaltene deposition are some of the remaining key issues. Warm VAPEX experimental work conducted at laboratory scale since 2000s [15,17,34–43] suggested promising oil recovery factors. However, the recovery processes around the mixing zone, the interplay between solvent, bitumen and the associated asphaltene precipitation in a porous rock structure are not very well understood. As illustrated in Fig. 1, the recovery related moving boundary is in millimetre range, whereas filed scale simulation uses grid blocks in metres and doesn’t directly simulate the physical behaviours. For scaling up the thin mixing zone driven by VAEPX process to a full reservoir level, it is crucial to have a good understanding of the fundamental millimetre scale extraction behaviours. This work addresses the conceptual processes occurring in the mixing zone, including transport processes, solvent-dependent asphaltene precipitation and its impact on the permeability, which eventually affects the recovery rate. The experiments in this study are conducted at warm VAPEX conditions, i.e. field-relevant pressure and temperature. In a vertical fracture-matrix 2D geometry, the macroscopic sweep at a length scale of 5 cm and the microscopic processes at pore level are imaged simultaneously. Recoveries from Canadian bitumen by three solvents are compared, toluene as a fully miscible reference case has no asphaltene precipitation; pentane and propane, two possible deployment candidates, exhibit asphaltene precipitation. Our results suggest that solvent-specific interactions with the asphaltenes in the bitumen impact the permeability, and change both the vertical drainage behaviour and the ultimate recovery. The insights into these fundamental mechanisms provide more physical-based elements for a reliable field scale assessment.

As a secondary intended effect, light paraffinic solvents may precipitate asphaltenic-like material, retaining heavy metals, carbon, and sulphur content within the reservoir. Producing the bitumen without the asphaltenes can be regarded as a type of in-situ upgrading, i.e. with more favourable chemistry and consequently increasing the value of the produced oil for global markets and being more environmental friendly [5]. In addition, the viscosity of the oil produced by solvent injection has been reduced significantly, thus saving the cost of surface oil transportation to the refinery [6]. The idea of using solvent-only process dates back to the mid 1970s. Farouq Ali [7,8] proposed hot miscible displacement recovery methods utilizing solvents. Allen and Redford [9] and Allen [10] patented “huff and puff” recovery methods using a variety of solvents and non-condensable gases. Nenniger [11] patented a process in which heated soluble gases were injected at pressures above the dew point to reduce the viscosity of the oil in place. In principle, the proposed concepts looked promising at the time; however, preliminary assessment of the data proved the concept to be feasible but not economical [12]. In the early 1980s, with advancements in horizontal drilling technology, thermal processes with gravity drainage as the primary recovery mechanism started to be investigated [13–15]. In the late 1980s, solvent injection revived based on gravity drainage processes. In the early 1990s [16,17] Butler, Dunn, Nenniger and Rajan introduced the concept of vapour extraction (VAPEX) where a light hydrocarbon is injected in the gas phase at reservoir temperature and pressure conditions. In this case, the solvent dilutes the bitumen by diffusion and production occurs by the reduction of oil viscosity. One of the most promising techniques is warm VAPEX. Typically, a light alkane is injected at temperatures slightly higher than its boiling point and at reservoir pressure. The selection of solvents, such as propane or butane, is based on finding candidates that have saturation conditions at reservoir pressure and temperatures around 60–80 °C. After sufficient volume of heated solvent is injected into the reservoir, a hot solvent vapour chamber forms inside the reservoir, as shown in Fig. 1 [18]. The extraction of bitumen occurs in the thin mixing zone between the solvent chamber and the bitumen, where the deposition of asphaltenes can also happen. Mullins reviewed in a filed scale this asphaltene solution chemistry and thermodynamics along with the principles of reservoir fluid geodynamics [19]. At the boundary between the solvent envelop and the heavy oil, solvent vapour is condensed onto the colder bitumen. After condensation, the now liquid solvent diffuses into the heavy oil and forms a thin layer of solvent-bitumen mixture. This mixture has lower viscosity than the pure bitumen and higher density than the solvent vapour, hence it migrates due to gravity to a lower point where the producer is located. In terms of cross-phase diffusion process in dry VAPEX (gas phase solvent interacts with liquid phase bitumen), this solvent-vapour extraction occurs between two liquid phases, solvent and bitumen, where the diffusion and drainage

2. Materials and methods 2.1. Materials Dewatered and degassed heavy crude oil from the Peace River region in Canada is used in this study. The API gravity is around 10° and it has a viscosity of around 20,000 cP at room temperature. Two types of VAPEX relevant solvents, propane and pentane, were obtained from Linde, with purity > 99.5%. Toluene (SIGMA-ALDRICH, purity ≥ 99.5%) is used as the reference solvent for comparison since it does not precipitate asphaltenes. Between different experiments,

Fig. 1. Production process schematics of warm VAPEX. A solvent chamber is formed and solvent vapour condenses on the bitumen. Dissolution of desired components and intentional asphaltene precipitation occurs in a thin layer. 2

Fuel xxx (xxxx) xxxx

X. Li, et al.

Fig. 2. Timeline of solvent recovery field pilots in Alberta, Canada (grey: failed or inconclusive; yellow: weak results; green: successful; blue: planned) [23-29]. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

cleaning of the micromodel and flowlines is done with toluene.

of the fracture is 21 µL. The rectangular matrix area is comprised of a repeat pattern with an area of 2.625 mm2, composed by randomly populated bimodal distribution of circles: 41 grains of diameter 300 μm and 59 grains of diameter 200 μm, see Fig. 3 (right). The void space has a depth of 50 μm. The minimum width of the fluid passage between grains is greater than 15 µm. The repeat patterns were designed with overlap to prevent highly permeable streaks. The above micromodel pattern is etched into a glass surface. During the experiment, the glass block with the pattern is compressed (by the sleeve pressure) onto another block with a smooth surface. In this way, the flow channel is generated. The outer boundaries of the glass blocks are then sealed with silicone rubber to prevent the invasion of sleeve liquid into the micromodel.

2.2. Micromodel design Micromodel, which is widely used for studying pore space oil recovery mechanisms [44–57], is chosen to investigate the mechanisms of VAPEX in detail. Elementary requirements are high pressure and temperature range in order to reproduce the field-relevant conditions. The micromodel domain mimics the thin mixing zone (rectangular area in Fig. 1), which requires taking account for gravity-driven convention, mixing, dispersion and the moving boundary between the condensed solvent and the bitumen. Since gravity is the key driving force for the migration [2,58–60], the micromodel is thus vertically mounted and designed as a 2D matrix (42.565 mm × 38.797 mm, 50 µm depth), as shown in Fig. 3 (left). To initiate the boundary between fresh solvent and the bitumen, a fracture with 3 mm width is created along one of the vertical edges of the micromodel structure. Solvent flows through the fracture from top to bottom, in parallel with the migration direction of the (solvent + heavy oil) mixture phase in the reservoir. In micromodel geometry, the fracture functions as both the injector and the producer in the field. There are four ports in total on the micromodel, two are connected to the micromodel matrix (used to initially saturate the matrix with bitumen) and the other two are the inlet and the outlet of the fracture. For a typical heavy oil formation, the Canada Peace River field is chosen as the reference field for the micromodel design. In this area, the reference Bluesky reservoir porosity is between 25 and 30%, and the permeability is ranging from 0.1 millidarcy to 10 darcy [61]. For most of the feasible solvent VAPEX applicable reservoirs, the permeability is suggested to be above 3 darcy on average. To reduce the experimental time and avoid potential asphaltene plugging, the permeability of the micromodel matrix is designed as 12.4 darcy and with a porosity of 35%. The total pore volume of the matrix area is 28 µL and the volume

2.3. Flow setup The micromodel assembly is contained in a pressure cell equipped with sapphire windows for optical access. By heating the jacket surrounding the pressure cell, the working temperature is regulated ( ± 0.05 °C) up to maximum of 200 °C. The solvent and crude oil are being delivered to the micromodel from piston vessels driven by the ISCO pumps. Additionally, two other pumps, one for the confining pressure control and another for cleaning, are added to the system. Compared with other types of micromodels, this design operates at pressures up to the maximum confining pressure of 300 bar. A sketch of the experimental setup is shown in Fig. 4. A more detailed explanation of the system has been presented elsewhere [62]. 2.4. Imaging and image processing Imaging is performed in transmission mode, using two cameras (selected via a beam splitter) where Camera 1 (Basler scA1400-30gm GigE camera) images the whole field of view, while Camera 2 (Basler,

Fig. 4. Sketch of the experimental setup consisting of the micromodel, the flow infrastructure with pumps and two cameras visualizing the whole micromodel pattern.

Fig. 3. Micromodel design (left) and repeat pattern in the micromodel matrix (right). Grains are in white and pore space is in grey. Solvent injected to the fractures on the right side next to the matrix flowing from top to bottom. 3

Fuel xxx (xxxx) xxxx

X. Li, et al.

the flowlines and the micromodel matrix are cleaned with a volatile solvent (toluene) and vacuum is applied for more than 20 h. After cleaning, with the flowlines internal pressure maintained at vacuum condition, the crude oil is injected from the bottom port of the matrix with a constant flow rate of 1 µL/min until the pressure reaches the experimental pressure of 10/20 bar. The next step is to inject solvent in liquid phase from the top of the fracture and collecting the effluent from the bottom port of the fracture. During this process, both ports are connected to an ISCO pump. Flow rates of 0.1 µL/min or 0.2 µL/min were used to relate to a typical field relevant value. Subsequently, the ISCO syringe pump model 100DM with a flow range of 0.01 µL/min to 30 mL/min and a displacement resolution of 4.8nL is chosen. It has a flow accuracy of 0.3% at the setpoint, which ensures sufficient accuracy of the flow rate [64]. Also, the back pressure regulation was realised via an ISCO pump. The inlet pump is working with constant pressure mode, which is equal to the experimental pressure and the outlet pump is working with constant volume flow rate mode, extracting fluid from the system. This control scheme was substantially more accurate and stable than when operating at constant injection rate and constant back pressure. The solvent injection rate is designed with a 10:1 solvent to bitumen ratio of the expected oil production rate, which relates to typical field conditions. This avoids production limitations and allows the production rate to be constrained only by mass transfer. In turn, the expected oil rate is estimated based on a correlation developed by Nenniger and Dunn [65], who discovered that the oil production can be simplified as a function of drainage area, dead bitumen viscosity, permeability and porosity only as expressed by

Fig. 5. Grey level of the grain area before (left) and after (right) solvent flooding. Due to the small contrast between the solvent in the pore space and glass grains, red dotted lines are added in the solvent saturated image to indicate where the glass is.

GigE) equipped with long working distance microscope lens (INFINITY Model K2 DistaMax) images a small region of interest at high resolution, where the fluid interfaces are visible in the pore space. During the experiment, monochrome images of the micromodel are recorded at 30second intervals. Image processing is conducted with the open source software package ImageJ [63]. For the solvent front tracking, the original 8-bit images from the experiment are first subtracted by the 100% oil saturation image to enhance the contrast. The post-process involved filtering, thresholding and finally edge detecting. The oil recovery is calculated from the average transmission images (grey level). Toluene flooding is used as the reference case for the concentration calculation. A high permeability of 12.4 darcy results in 100 h of continued flooding through the finite volume saturated with oil, as a volume fraction of 1:40 for crude oil and toluene is reached. It is estimated, with an experimental uncertainty, that at the end of the experiment, oil is mostly recovered. Even nanoscale structures of the oil may exist in the pore (see Fig. 5, images from the toluene experiment before and after the solvent flooding in a pore scale). Some uncertainty is brought in by the dark spots on the images, caused by the impurity in the heavy oil and the microscale dust in the imaging path. This uncertainty is estimated by comparing the grey level with the surrounding clear area. Assuming that 10 worst-case spots exist in the image zone, the resulting calculation error is below 0.25%, which is negligible in this case. The average oil contents in the micromodel is estimated from the area coverage of the oil versus the transparent solvent from the light transmission, using

Gt − GxB ΔGx RF = = E x B ΔGToluene GToluene − GToluene

k ϕ mO" = 43550 ⎛⎜ o ⎞⎟ ⎝ μo ⎠

0.51

(2)

where “mo” stands for bitumen mass flux per unit of drainage area in an hour [g/m2·h], and the permeability and viscosity are in darcy and centipoise respectively. In this correlation, the physical basis is that the oil production rate is related to the surface area of the shock front, so solvent flow direction is transverse to the production area. In our micromodel, the production area is the perimeter area between the ‘’matrix’ and the ‘fracture’ (i.e. height times width: 42.565 mm × 50 μm). Note that Equation (2) is developed for vapour extraction and, in this particular case, it is used as a reference for liquid injection. Based on this correlation, the expected bitumen drainage rate is calculated to be around 0.02 µL/min, using the parameters listed in Table 1. Hence, the injection rate of the solvent used in the experiments is 0.2 µL/min. Furthermore, 0.1 µL/min is used for comparison. 2.6. List of experiments The experiments in this study are carried out at conditions where solvents are in a liquid state i.e. below the bubble point. In the solvent recovery process, the injected pure solvent will always be in contact with the material within the formation, which brings impurity to the solvent phase. Due to this, the solvent saturation temperature will be reduced in-situ. Taking this into account, the designed experimental

(1)

where “RF” represents the recovery factor, “G” represents the sum of the grey level of the 8-bit images (capturing the entire micromodel), “B” and “E” represent the beginning and the end of the flooding period. At “B” the micromodel matrix is saturated with heavy oil, while at “E”, the solvent is saturated in the matrix. “x” represents the type of solvent to be analyzed, and the images from the “Toluene” experiment are used as the reference for the “RF” calculation of pentane or propane. In this grey level, to calculate the oil production curve process, the original images from the experiment with 50-minute intervals are used for the whole flooding period analysis. To calculate the recovery rate, a 5-point rolling average has been applied.

Table 1 Nenniger correlation input and calculation.

2.5. Injection scheme The temperature is regulated to the target experimental value (44 °C/55 °C) before crude oil/solvent is injected into the system. First, 4

Property

Unit

Value

oil viscosity temperature permeability porosity height depth Nenniger mass flux volume rate

cP °C darcy – mm mm g/m2·h µL/min

20,000 50 12.4 0.35 42.5 0.05 589.6 0.02

Fuel xxx (xxxx) xxxx

X. Li, et al.

on the right, and a region with the original bitumen at bottom-left which has not yet been reached by the solvent. The second and third images in the figure show the solvent front tracking and front contours, respectively. It is noticed that this experimental front contour evolves almost the same way as what is suggested in some mathematical models, as presented in Qiong’s work [68]. Which is also generally recognised as the conceptual shape of solvent envelope for VAPEX in a reservoir scale.

Table 2 List of the conditions of the experiments. Every pentane or propane experiment is repeated to evaluate the repeatability of the experimental observations. Experiment no.

Solvent type

Temperature (°C)

Flow rate (µL/min)

Pressure (bar)

Toluene_0.2 Pentane_0.11 Pentane_0.11r Pentane_0.2 Pentane_0.2r Propane_0.2 Propane_0.2r

Toluene Pentane Pentane Pentane Pentane Propane Propane

55 55 55 55 55 40 40

0.2 0.1 0.1 0.2 0.2 0.2 0.2

10 10 10 10 10 20 20

3.1. Frontal advancement and asphaltene precipitation for pentane and propane In Fig. 7, the results of the four pentane experiments are displayed. In all cases, asphaltene precipitation occurs in a form of dark bands behind the solvent front and accumulates at the lower part of the matrix next to the production point. It is important to note that the asphaltene precipitation does not occur directly next to the solvent front but is separated from the bulk bitumen by a gap. As the injection of solvent continues, part of the asphaltenes deposit locally. Also, part of the asphaltenes migrate with the flow and accumulate at the bottom-right corner of the formation. The disturbance of the flow fields by the asphaltene accumulation at the outlet is reflected by overlapped solvent front contours at the bottom-right corner. That becomes more severe for the high solvent flow rate cases, where asphaltenes eventually plug up the outlet and oil production is ended. There is literature [69] suggested that the plugging issue is mainly caused by the more polar and less soluble fraction in the asphaltenes, which tend to have stronger aggregation and early precipitations comparing with the less polar structures. It is also noticeable that the front profiles of the low flow rate experiments are closer to straight lines, but the front profiles of the high flow rate experiment are “S” shape. Without considering the asphaltene precipitation, the shape of the solvent front is a combined result of solvent diffusion rate and gravity drainage rate. During the experiment, a higher solvent injection rate (refreshing rate) maintains a higher solvent concentration at the injection boundary, which results in a higher solvent diffusion rate to the micromodel formation both vertically and horizontally. For the experiments with different solvent injection rates (solvent diffusion rate), it is assumed that the gravity drainage rate in the vertical direction, which is driven by the density difference between the (solvent + bitumen), blend and solvent are similar. As a result, high and low solvent injection rates give different front shape. However, at the meantime, the asphaltene deposition rate is also observed positively correlated to the solvent injection rate, which for the high solvent flow rate case promotes more plugging at the bottom part of the formation. For propane, there are distinct differences compared with pentane, see Fig. 8. First, the frontal advancement displayed in the grey level

temperature of pentane in this study is 55 °C (the saturation temperature of pure pentane which is around 120 °C at 10 bar) and for propane it is 40 °C (saturation temperature for pure propane which is around 58 °C at 20 bar). The experimental conditions are listed in Table 2. In total, seven experiments are carried out. The toluene experiment is designed as the reference case for comparison. Two experiments are conducted for pentane, one with a flow rate of 0.1 µL/min and another with a flow rate of 0.2 µL/min. For propane, one experiment is conducted at a flow rate of 0.2 µL/min. The repeat experiments indicated with an “r” are conducted to assess the repeatability. 3. Results and discussion The experimental results in this section are grouped into three parts. First, we compare all the solvents based on the macroscopic displacement behaviour, observing the solvent front evolution characteristics and the asphaltene precipitation traces. Then we zoom in and look at the pore scale details regarding the asphaltene deposition and permeability change. Lastly the respective recovery factors and rates for the three solvents are compared and discussed in relation to the image observations. Here, it is necessary to restate the definition of asphaltenes, which is the portion of crude oil insoluble in n-alkanes, such as propane, n-pentane, yet soluble in benzene or toluene [66]. Varying compositions by one crude oil to another, asphaltene contents in certain crude oil is a broad distribution of molecular structures. At molecular level, asphaltenes are characterized by fused ring aromaticity, small aliphatic side chains, and heteroatom containing functional groups. According to a study on the chemistry of reference Alberta bitumen, if the solvent solubility parameter is above about 17.1 MPa1/2, the asphaltene in it become completely soluble [67]. The ranking of solvent solubility parameter at our experimental conditions is toluene > 17.1 MPa1/2 > pentane > propane. In the reference toluene experiment shown in Fig. 6, the first grey level image is a typical miscible solvent front that consists of a clean region in the upper-right half, which is next to the inlet of the fracture

Fig. 6. Solvent front tracking steps illustrated with the result from toluene. Step 1: subtracting the image during the solvent flooding by the initial bitumen saturated image peer to solvent injecting, then apply thresholding; step 2: solvent front edge detection; step 3: all different front-edge saturation times are projected into one image with 250-minute intervals.

5

Fuel xxx (xxxx) xxxx

X. Li, et al.

Fig. 7. Grey level images and front tracking (250-minute interval) for pentane at two different solvent flow rates, and respective repeat experiments. From the grey level image, it is observed that the brighter solvent flooded area is not clear due to the asphaltene precipitation, especially the bottom-right part closer to the outlet of the solvent outlet in the fracture. As a result, the propagation of the solvent fronts overlapped in this area. For the high solvent injection rate case, poorer repeatability of the experiment is an indication of unstable recovery processes.

separated with clear gaps. With the same solvent injection rate of 0.2 μL/min, the propane front is more a straight line instead of the characteristic “S” shape. Due to the higher density difference between the propane and bitumen, a bigger gravity draining force is expected. Also, in a vertical direction, there is no obvious asphaltene precipitation at the bottom-right corner blocking the flow path. The reduced plugging in the propane case provides a more stable/ smooth flow of the solvent in the matrix. Using the gap and time interval between the contours in the middle of the flooding, a u = 0.03 mm/minute solvent front moving speed is calculated. Also, the calculated propane diffusion coefficient is around D = 8 × 10−9m2/s, which is in the same range as in the literature [70]. With a frontal width of L ~ 3 mm, this translates into a Péclet number Lu Pe = D 0.15 < 1 meaning that the flow is diffusion dominated and concentration gradients can be expected. Overall, the mechanism of asphaltenes precipitation was controlled by dispersion interactions between the asphaltenes and solvent, which is solvent dependent. As a result, the front advancement of either pentane or propane are observed very different from the fully miscible case as toluene, which indicated that using a fully miscible concept to describe the general VAPEX process is not sufficient. As in a filed scale research suggested, the type of the solvent in such processes is an important factors that influences how and where asphaltene is deposited in the reservoir [71]. In the propane repeat experiment in Fig. 8, a darker asphaltene area appeared next to the fracture in the grey level image and, as a result, the front contours at the same level inward appeared to overlap occasionally. This is due to an unexpected “asphaltenic phase slug” entering the matrix, see Fig. 9. At an earlier stage of the experiment, after the breakthrough of propane in the fracture, propane started to diffuse into the matrix. But around 1,100 min later, a slug of black oil phase coming from the inlet of the fracture, entered the matrix. It imbibed quickly into the matrix and spread faster than the solvent front moving speed. In a pore scale, the mobility of the fluid is controlled by the pore size, the fluid viscosity, the diffusion coefficient and the capillary forces. A more viscous phase moving faster than the less viscous phase in the same pore structure is most likely an indication that the thicker phase is the wetting phase. There is evidence that this phase has been in contact with propane,

Fig. 8. Grey level image and front tracking (250-minute interval) for the propane experiment. Form the grey level image, the brighter area which has been flooded by solvent is cleaner compared with pentane, and the solvent front is more obviously identifiable. There are some darker asphaltene precipitations in the flooded zone, but the contrast between the precipitation and the sound area have bigger contrast with sharper interfaces. The front counters are separated except in the repeat experiment, an “asphaltenic phase slug” accidently entered the matrix from the fracture, which prohibited the solvent diffusion in the matrix, as a result, the “asphaltenic phase slug” projected area has a few front conjunctions on the front contours.

images is more distinguishable than in the pentane case. Some dark bands in the solvent-rich zone are observed, which indicate asphaltene precipitation. But here, the asphaltene precipitation did not show a heavy plugging effect towards the outlet. The front contours are well 6

Fuel xxx (xxxx) xxxx

X. Li, et al.

Fig. 9. An “asphaltenic phase slug” from the fracture entered the matrix and developed during the flooding in the propane repeat experiment (at t = 5800, green rectangle: area with “slug” trace; red square: area without clear trace from the “slug”). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the fracture and the downstream of the flow path. However, for propane flooding, the asphaltene precipitation occurs intermittently with the dissolution front. The precipitation occurred in the form of smaller patches with a surrounding sharp interface. Additionally, these precipitated patches migrated along the slope of the solvent front to a lower formation forming a stream-like narrower band. This process is suggesting a much less risk of formation plugging. From a phase behaviour point of view, a light paraffinic solvents and bitumen mixture under different (solvent:bitumen) ratios generates a phase split of a heavy asphaltenic material. The study from Yarranton et al. [76,77] and Akbarzadeh et al. [78] indicated that the liquid–liquid onset for pentane occurs at around 50 wt% solvent; the propane onset is at a lower solvent concentration, around 25 wt%. Inside of the micromodel matrix, solvent concentration decreases from the fracture side (boundary on the right) to the solvent front side. As a result, the asphaltene precipitation for propane is appearing closer to the solvent front compared with pentane.

but when trapped in the flowline it is becoming mobile, because on entering the fracture, a sharp interface between the phase and the surrounding propane preserved. After entering the pore matrix, it spread and further dissolved/recovered by the solvent and eventually left a mark at its entrance in the flow pattern as seen in the green rectangle in the last frame. This “asphaltenic phase slug” observed from propane experiment has liquid-like mobility. Most likely it is a asphaltene rich phase but also include part of the most polar fraction of resins [72,73], the presence of resins in this asphaltene rich phase prevented further aggregation of asphaltene constituents as a separate phase [74,75]. This also explains why propane recovery is more stable than pentane. A black phase slug also gets produced mostly from the formation. Also, in all the flooded zone, most of the asphaltene-rich phase remains localised inside the matrix instead of clogging and plugging, which is supported by the relatively stable and well-distributed dark bands displayed in the figure. This raises the more general question of the dynamics of asphaltenes precipitation, migration and deposition which will eventually impact the mass transfer of solvent, the production of oil and formation plugging. As a step forward in answering this question, in Fig. 10, differences between every two frames with a 50-minute interval are taken to illustrate the areas where bitumen is dissolved and the areas where asphaltenes are precipitated. All three experiments have the same solvent injection rate of 0.2 μL/min. For toluene, since the recovery mechanism is nearly 100% dissolution, there is only dissolution and the blue dissolution band follows the front tracking from Fig. 6. There is no sign of asphaltene precipitation. For pentane flooding, both experiments show that asphaltene deposition is spatially separated from the dissolution zone, i.e. asphaltenes are deposited in the flooded zone at a distance from the dissolution front. Precipitation bands are wide and spread and they appear predominantly at the lower side of the formation. While the flooding continues, part of the precipitations migrates further downwards. The observed pressure difference between the inlet and the outlet increased for one magnitude during the recovery process, which is another evidence of asphaltene precipitation in

3.2. Asphaltene deposition at the pore scale and permeability change Pore scale images displayed in Fig. 11 indicate, that there are more fundamental differences in the way how asphaltenes are precipitated in different solvents, which impacts the dissolution behaviour, and the advancement of the front. For the reference toluene case, at the end of the experiment, the matrix is saturated with clear transparent solvent and the heavy oil is being diluted and produced, which is consistent with the literature [79]. For both pentane and propane, asphaltene precipitation is observed in the pore space but with a very different appearance. For pentane, the deposited asphaltene is dispersed in solid, which can block the flow path. The precipitation pattern on a larger scale, as illustrated in Fig. 7 and Fig. 10, suggested that this pore scale precipitation partly transports and accumulates at the bottom-right corner of the formation. In that way, the earlier flow path is getting plugged and needs to re-configure to the un-plugged regions, which at the end of the flooding, asphaltene deposits higher up in the formation, 7

Fuel xxx (xxxx) xxxx

X. Li, et al.

Fig. 10. Difference between adjacent frames with 50-minute intervals visualising dissolved heavy oil (blue) and precipitated (red) asphaltenes at the same solvent flooding flow rate. Toluene case has no precipitated asphaltene, pentane case has precipitated asphaltene widely spread at the bottom-right side which is closer to the drainage port. Propane case has asphaltene precipitation at the solvent front and formed a steam-like migration channel. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the glass surface. They form pendular rings around the glass with visible meniscus in between the narrow pores. The small contact angles suggest a strong affinity of the dark phase towards the glass surface. Note that in field application, water is also present as an initial wetting phase which changes the characteristics of the formation surface. However, literature suggested that asphaltenes can diffuse through thin water layers and/or break water films which would lead to the same configuration as observed in this study [81]. In practical situations,

significantly above the producing point at the bottom-right corner. But an interesting observation here is that in individual asphaltene deposited area in pore space, there is no obvious orientational depositing preference. This deposition characteristic is also observed from scanning electron micrographic visualization on a field core after light hydrocarbon charge [80]. As opposed to pentane, the deposited asphaltene in the propane experiment appeared as a liquid phase, also the wetting phase toward

Fig. 11. Pore scale images at the end of solvents flooding, zoomed into the centre of the matrix. Red dotted circles are added in the toluene images to indicate where the glass is. In case of Toluene, there is no asphaltene deposition in the pore space. For pentane, the pore is blocked by solid-like deposited asphaltene. In case of propane, the precipitated asphaltene is liquid and has a clear interface with the surrounding solvent phase. Meanwhile, the asphaltene-rich phase is the wetting phase toward the glass surface. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

8

Fuel xxx (xxxx) xxxx

X. Li, et al.

wettability effects may have an influence on how much asphaltenic phase is retained in a particular configuration which may also have an impact on the ultimate recovery. In general, when using n-alkanes as the solvent, with a fixed (solvent: bitumen) ratio, the fraction of asphaltene precipitation is expected to increase with decreasing carbon number [82]. Moreover, the asphaltenes precipitated by heavy normal alkanes are more polar than asphaltenes precipitated by light normal alkanes. So the pentane asphaltenes is more polar than the propane asphaltenes [83]. The state of asphaltene, either solid in pentane or liquid in propane, confirmed the hypothesis from some phase behaviour studies in the literature [19]. Also, at a slightly different pressures, 9 bar (10 bar in our experiment), on the heavy oil samples from Lloydminster area, Canada, Luo [73] observed liquid-like asphaltene after aging for 2 days an originally semi-solid like C3-asphaltenes from flashed-off process, but the C5-asphaltenes remained as brown powder at ambient pressure. His explanation is that resins in C3-asphaltenes peptized the asphaltenes, so they were redissolved into the maltenes after some time. Luo also observed that the microscopic structures of C3-asphaltenses (at 9 bar) under scanning electron microscope is smoother and more amorphous than those of C5-asphaltenes. In this study, C5-asphaltene in the original heavy oil was around 20 wt%. However, in a phase partitioning test, the yield of heavy phase is close to 30 wt% for pentane and up to 55 wt% for propane. It indicated that the precipitations from propane should have a lower viscosity due to its higher lighter component fraction. In the micromodel experiment, the propane-bitumen mixture is visible as liquid whereas the pentane heavy phase comprises of glass-like particles [84,85]. Also similar filed scale well observations of this different asphaltene state was suggested in the literature [86]. These fundamental differences in the pore scale are eventually driving the solvent recovery characteristic in a field scale and determine the ultimate oil recovery. In the pentane case, precipitation of solid asphaltenes and subsequent fines migration can cause plugging and, consequently, permeability and solvent diffusivity will be reduced. To assess the influence of asphaltene precipitation more quantitatively, flow simulations are conducted to estimate the impact on permeability and diffusivity. The 3D geometry is regenerated from the 2D images as shown in Fig. 12, with a depth of 50 μm in the z-direction. Note that, here, the black colour represents pore space instead of oil phase, the grey pixels represent micro porous asphaltene deposits, and the white colour represents solids. The image on the left is binary, including white glass and open pores. The image on the right is post-processed from the experimental images, where grey scale represents micro porosity. Using a Kozeny-Carman correlation (permeability ~ porosity3) the micro porosity is translated into permeability ranging from 0 to 1e-7 m2. Here 1e7 m2 is used to represent open pores because the computed permeability of glass beads with pores of such high permeability agrees with the Stokes result for completely empty pores all the way to 1e-7 m2 and

Fig. 13. Comparison of flow in the pore before and after asphaltene deposition in the pentane flooding case. Significant flow prohibition and permeability reduction due to the asphaltene deposition. The high velocity channel in the clean pore image (left) corresponds to a more severe asphaltene deposition (right) in these channels.

starts deviating at 1e-8 m2. The velocity of solvent in the fracture for this selected experiment is 0.01 mm/s. At this low rate, it is expected that solvent flow in the micro porous asphaltene deposits is laminar and the combined flow in open pore space and micro porous asphaltene deposits can be described by a Stokes-Brinkman equation, ⇀ → − μΔ→ u + μK−1u + ∇p = f Δ→ u =0

(3) ⇀

where μ is the fluid viscosity, u is the fluid flow velocity, k is the permeability, p is the pressure, and f is a force density. The simulation is set up to cause flow from top to bottom with a mean flow rate of 6.7·10−5 mm/s, which is representative for experiments with asphaltene plugging. Symmetric boundary conditions are assigned to the inlet, outlet, left and right surfaces. No-slip boundary conditions are assigned to the front and back walls. The simulation grid uses the same dimensions in every direction with a side length of 1.4 μm. The simulation results displayed in Fig. 13 show, that the asphaltene plugging causes a significant change in the flow field. Plugging seems to occur mainly in originally high velocity flow paths, consistent with a fines migration and filtration mechanism that reduces permeability. The respective permeability in the vertical direction decreased from 8.5 darcy to 1.2 darcy, with the same pressure drop of 3.8 Pascal in the simulation setting. Micro porous asphaltene filtration not only reduces the permeability but also impacts the mixing process of bitumen with solvent which involves diffusion. For the diffusion simulation in the clean pore structure, the Laplace equation is applied since the minimum pore size is large enough and the diffusion is mostly controlled by molecular collisions. In the case of asphaltene precipitated pores, the diffusivity inside these pores are modelled by the equation below:

D=

D 0ϕ τ2

(4)

where D is the diffusivity, ϕ is the porosity and τ is the so-called tortuosity factor. Due to lack of information regarding the precise microstructure of the asphaltene deposits, the relationship was concluded by choosing

τ 2 = ϕ−1 or τ 2 = ϕ−2

(5)

So, without asphaltene, we solve div (D grad u) = D Δu = 0 in the pore space with concentration u specified on the left and right faces of the domain, while in the presence of asphaltene, we solve 0

div (D grad u) = 0

Fig. 12. Geometries used in the simulation. Left: modelled numerical structure with empty pore space; right: experimental results with plugged pores. The white colour represents glass and the black colour represents open pore area. The grey area is partially porous depending on the grey level.

0

(6)

again, with a concentration difference in u specified on the left and right faces of the domain. The result for the effective diffusivity that can be computed from the 9

Fuel xxx (xxxx) xxxx

X. Li, et al.

Table 3 Concentration distribution setting for the asphaltene deposited pores.

Equation Porosity Effective diffusivity

Empty pores

τ2 = ϕ−1

τ2 = ϕ−2

div (D0 grad u) = 0 0.363 1.77 e−1

div (D0ϕ grad u) = 0 0.174 1.98 e−2

div (D0ϕ2 grad u) = 0 0.174 7.95 e−3

Fig. 14. Comparison of diffusion in the pore before and after asphaltene deposition in the pentane flooding case. The asphaltene deposited case has the effective diffusivity which is one to two orders lower than the clean pore case.

to the plugging. The high initial recovery rate with respect to toluene for the pentane experiments is likely due to the higher diffusion and dispersion rate. The recovery of propane is much more stable thus smoother recovery curves are observed. The 5% recovery factor difference between the two experiments is caused by the “asphaltene slug”. Outside that region, the same 70% recovery factor is obtained (see Fig. 9). The recovery rate of propane is comparable to toluene and much more stable than pentane. In the repeat experiment, the oil slug acted as a buffer between the fracture and the matrix, preventing the solvent to enter the matrix, and slowed down the recovery rate. After a certain volume of propane injection, around 30% liquid asphaltene remained trapped in the matrix and was not recoverable with additional solvent flooding. From Fig. 9, it’s also observed that, due to severe asphaltene plugging (i.e. experiment: Pentane_0.2r) or asphaltene entering a flooded zone (i.e. experiment: Propane_0.2r), the maximum recovery factors are reached after almost twice much the pore volumes of solvent being injected comparing with their equivalent cases. In field scale, this type of asphaltene precipitation could potentially reduce the recovery efficiency of per volume of solvent injected to the reservoir, adding additional capital lost.

detailed concentration distribution [87] is shown in Table 3. From the diffusion flux pattern comparison in Fig. 14, asphaltene deposition can significantly reduce the effective diffusivity in the mild level plugging simulation case. As a result, the different concentration gradient around the blocked area will further drive an unstable recovery process on a larger scale. 3.3. Recovery factor and recovery rate The micromodel observations discussed above indicate that the recovery process is strongly influenced by solvent extraction induced polar elements precipitation from the heavy oil. When pentane is applied, strong polar components including asphaltene and likely part of the hard resin are precipitate together, which formed solid state depositions in the pore space. But when propane is applied, due to its weaker solvency power i.e. smaller solubility parameter compared to pentane, more asphaltene including those with weaker polarity, most likely together with more resins are precipitated together as a separate phase. Due to the larger range of polarities of the asphaltene, and also the interactions between resins and asphaltenes in this phase, it appeared as a liquid with mobility [88]. In principle, this phase has smaller molecular weight and lower viscosity comparing with the pentane precipitation. As a result, the recovered oil phase from propane is also having higher grade (higher economical value) comparing with pentane. In Fig. 15, the recovery factor and recovery rate for pentane and propane experiments are compared with the toluene reference case. For better comparison and a more accurate calculation of the rate (derivative of the production curve), smoothing by local polynomial regression was applied (solid lines). The ultimate recovery for toluene is close to 100%. One can clearly see that, on average, propane (70%, 65%) has a higher recovery than the pentane (50% and 67%). This not only means that asphaltene precipitation reduces the recovery factors of propane and pentane compared with the fully miscible toluene case, but also that the different asphaltene precipitation mechanisms in propane and pentane have an impact. For pentane experiments, in terms of flow rate dependency, the lower flow rate gives higher final heavy oil recovery and a more stable displacement with less fluctuations indicating a lighter influence from asphaltene precipitation. This is reflected by the recovery rate as well, which first increases to a pronounced maximum and then decreases due

4. Summary and conclusions The VAPEX process of field relevant bitumen solvent systems was studied under reservoir conditions. In comparison with a fully miscible reference case, the solvents of practical choices, propane and pentane, showed different asphaltene precipitation mechanisms which resulted in different levels of plugging risks, having a strong impact on the recovery factor and recovery rate. The elementary recovery processes are illustrated in Fig. 16 by comparing the difference between propane and pentane. Fresh solvent enters the matrix and mixes with bitumen by diffusion. The mixture has a higher density than solvent and flows downwards by gravity-driven flow. Depending on the type of solvent at a critical (solvent:bitumen) ratio, asphaltenes start to precipitate. Part of the precipitation migrated to the lower formation and part of it deposited locally. For the pentane case, the asphaltene deposition are solid and blocked specially the low part of the formation, hence reducing the solvent transportation and gravity draining process, which leads to an unstable displacement process. As discussed in a field scale observation [89], this asphaltene 10

Fuel xxx (xxxx) xxxx

X. Li, et al.

Fig. 15. Recovery factors (two top plots) and recovery rates (two bottom plots) for pentane (two left plots) and propane (two right plots) in comparison with the toluene (fully miscible) reference (black line). For the pentane case, higher solvent flooding rates result in a lower heavy oil recovery factor. In general, from the pentane experiment, the recovery rate is unstable. For the propane case, the recovery factor is higher than pentane and the recovery rate change is smoother and stable.

Fig. 16. Conceptual differences between pentane and propane during solvent recovery. Three key differences: where asphaltene precipitation occurs, where asphaltene tends to deposit, how this will change the solvent front shape and solvent imbibition into heavy oil. 11

Fuel xxx (xxxx) xxxx

X. Li, et al.

migration to low points in the reservoir coupled with gravity drainage can further accelerate the concentration of asphaltene at the base of the reservoir where solvent:bitumen ratio is even lower, i.e., a much lover solvency capacity of the solvent. For propane, a liquid asphaltene-rich phase is formed and no plugging appears, but since this asphaltenic phase is the wetting phase towards the glass in presence of the solvent phase, so the process carries multiphase flow features. Overall, propane has a more stable displacement also a higher recovery factor of 70% compared with pentane ~55%, which is different from the phase partitioning ratios (only 45% for propane but 70% for pentane) as the phase behaviour test would suggest. In summary, the warm-VAPEX recovery is a coupled process where the asphaltene-solvent phase characteristics affect the pore scale flow behaviour which, in turn, control the transport of fresh solvent and production of dissolved crude components. In a field scale, this would change the shape of the solvent front, influence the finial recovery factor associated with the plugging risks and determine the chemical compunctions of the recovered oil. This study identified the key steps in a warm-VAPEX process and suggests some of the critical elements to be included in a reservoir scale assessment.

[13] Rezaei N, Mohammadzadeh O, Chatzis I. Improving the Performance of Vapor Extraction of Heavy Oil and Bitumen Using the Warm VAPEX Process. Canadian Unconventional Resources and International Petroleum Conference. Society of Petroleum Engineers; 2010. [14] Upreti SR, Lohi A, Kapadia RA, El-Haj R. Vapor extraction of heavy oil and bitumen: a review. Energy Fuels 2007;21(3):1562–74. [15] James LA, Rezaei N, Chatzis I. VAPEX, Warm VAPEX, and Hybrid VAPEX-The State of Enhanced Oil Recovery for In Situ Heavy Oils in Canada. Canadian International Petroleum Conference. Petroleum Society of Canada; 2007. [16] Butler RM, Mokrys IJ. Recovery of Heavy Oils Using Vapourized Hydrocarbon Solvents: Further Development of the Vapex Process. J Can Pet Technol 1993;32(06). [17] Butler RM, Mokrys IJ. A new process (VAPEX) for recovering heavy oils using hot water and hydrocarbon vapour. J Can Pet Technol 1991;30:01. [18] Pourabdollah K, Mokhtari B. The VAPEX process, from beginning up to date. Fuel 2013;107:1–33. [19] Mullins OC, Primio RD, Uchytil S, Zuo JY, Dumont H, Mishra V, et al. Bitumen and Tar Deposition and Tar Mat Formation Accounted for by Multiple Charging, Trap Filling and Fluid Geodynamics. SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers; 2016. [20] James LA, Rezaei N, Chatzis I. VAPEX, Warm VAPEX and Hybrid VAPEX – The State of Enhanced Oil Recovery for In Situ Heavy Oils in Canada. 8th Canadian international petroleum conference (58th annual technical meeting) Calgary, Alberta, Canada 2008. [21] Jiang T, Zeng F, Jia X, Gu Y. A new solvent-based enhanced heavy oil recovery method: cyclic production with continuous solvent injection. Fuel 2014;115:426–33. [22] Lin L, Ma H, Zeng F, Gu Y. A Critical Review of the Solvent-Based Heavy Oil Recovery Methods. SPE Heavy Oil Conference-Canada. Calgary, Alberta, Canada: Society of Petroleum Engineers; 2014. [23] NSolv Corporation. NSolv Corporation Announces 25 BoOPfiBP, in: https://www. newswire.ca/news-releases/n-solv-corporation-announces-25000-barrels-of-oilproduction-from-its-best-pilot-516218141.html, 2017. [24] Gupta S, Gittins S, Picherack P. Field implementation of solvent aided process. Canadian International Petroleum Conference. Petroleum Society of Canada; 2002. [25] Bayestehparvin B, Ali SMF, Abedi J. Case Histories of Solvent Use in Thermal Recovery. SPE Western Regional Meeting. Society of Petroleum Engineers; 2017. [26] Devon Canada Corporation. 2014 Performance Presentation CSAN, October, 2014. [27] Imperial Oil. 2017 Annual Performance Review CLAa, September, 2017. [28] Nexen. Long Lake Kinosis Oil Sands Project APP, April, 2017. [29] Suncor Energy Ltd. Suncor Energy begins Radio Frequency Pilot at DOVER site iwscnn-r. [30] https://www.neb-one.gc.ca/nrg/ntgrtd/ftr/2017lsnds/index-eng.html. [31] Zhao DW, Wang J, Gates ID. Optimized solvent-aided steam-flooding strategy for recovery of thin heavy oil reservoirs. Fuel 2013;112:50–9. [32] Du Z, Peng X, Zeng F. Experimental investigation of cyclic solvent injection based on mixture solvent of carbon dioxide and propane. Fuel 2018;225:646–54. [33] Zhou X, Yuan Q, Peng X, Zeng F, Zhang L. A critical review of the CO2 huff ‘n’ puff process for enhanced heavy oil recovery. Fuel 2018;215:813–24. [34] Nenniger JE, Dunn SG. How fast is solvent based gravity drainage? Canadian International Petroleum Conference. Petroleum Society of Canada; 2008. [35] Denney D. Technology Development for Solvent-Based Recovery of Heavy Oil. J Petrol Technol 2012;64(03):91–2. [36] Rezaei N, Chatzis, I. Warm VAPEX; a Thermally Enhanced Vapor Extraction Process. Canadian International Petroleum Conference. Calgary, Canada; 2008. [37] Das SK, Butler RM. Mechanism of the vapor extraction process for heavy oil and bitumen. J Petrol Sci Eng 1998;21(1-2):43–59. [38] James LA. Mass Transfer Mechanisms during the Solvent Recovery of Heavy Oil. University of Waterloo; 2009. [39] Sie C-Y, Nguyen B, Verlaan M, Castellanos-Diaz O, Adiaheno K, Nguyen QP. Viscous oil recovery and in situ deasphalting in fractured reservoirs: Part 1. The effect of solvent injection rate. Energy Fuels 2018;32(1):360–72. [40] Yazdani A, Maini BB. Modeling of the VAPEX process in a very large physical model. Energy Fuels 2008;22(1):535–44. [41] Yazdani A, Maini BB. Further Investigation of Drainage Height Effect on Oil Production Rate in Vapex. SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers; 2006. [42] Yazdani A, Maini BB. Effect of height and grain size on the production rates in the vapex process: experimental study. SPE Reservoir Eval Eng 2005;8(03):205–13. [43] Karmaker K, Maini BB. Experimental Investigation of Oil Drainage Rates in the Vapex Process for Heavy Oil and Bitumen Reservoirs. SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers; 2003. [44] Pei H, Zhang G, Ge J, Jin L, Liu X. Analysis of microscopic displacement mechanisms of alkaline flooding for enhanced heavy-oil recovery. Energy Fuels 2011;25(10):4423–9. [45] Maaref S, Ayatollahi S, Rezaei N, Masihi M. The effect of dispersed phase salinity on water-in-oil emulsion flow performance: a micromodel study. Ind Eng Chem Res 2017;56(15):4549–61. [46] Yao C, Lei G, Hou J, Xu X, Wang D, Steenhuis TS. Enhanced oil recovery using micron-size polyacrylamide elastic microspheres: underlying mechanisms and displacement experiments. Ind Eng Chem Res 2015;54(43):10925–34. [47] Zhang H, Chen G, Dong M, Zhao S, Liang Z. Evaluation of different factors on enhanced oil recovery of heavy oil using different alkali solutions. Energy Fuels 2016;30(5):3860–9. [48] Bai Y, Shang X, Wang Z, Zhao X, Dong C. Experimental investigation of nanolaponite stabilized nitrogen foam for enhanced oil recovery. Energy Fuels

Author contributions Steffen Berg, Orlando Castellanos-Diaz and Marco Verlaan conceived the original experimental idea and designed the micromodel structure used in this study. Xuesong Li carried out the experiment, processed the images, interpreted the results and wrote the manuscript with the input from all authors. Andreas Wiegmann developed the numerical simulation theory and supported Xuesong Li performed the computations. Marco Verlaan helped supervise the project. All authors discussed the results and contributed to the final manuscript. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments This research was supported by Shell Global Solution International B.V., The Netherlands. Sjaam Oedai is gratefully acknowledged for conducting test experiments and identifying improvement steps to the setup. References [1] Alboudwarej H, Felix J, Taylor S, Badry R, Bremner C, Brough B, et al. Oilfield Rev 2006;18:34–53. [2] Guo K, Li H, Yu Z. In-situ heavy and extra-heavy oil recovery: a review. Fuel 2016;185:886–902. [3] Lattanzio RK. Canadian oil sands: life-cycle assessments of Green House emissions. CRS Report R42537 2014. [4] NEB. Her Majesty the Queen in Right of Canada as represented by the National Energy Board. Energy Futures Supplement 2018. [5] Mokrys IJ, Butler RM. In-situ upgrading of heavy oils and bitumen by propane deasphalting: the vapex process. SPE production operations symposium. Oklahoma City, Oklahoma: Society of Petroleum Engineers; 1993. [6] Banerjee DK. 5.2.5 Vapor-assisted petroleum extraction. Oil sands, heavy oil and bitumen – from recovery to refinery. PennWell. [7] Ali SMF, Abad B. Bitumen recovery from oil sands, using solvents in conjunction with steam. J Can Pet Technol 1976;15(03). [8] Ali SMF, Snyder SG. Miscible thermal methods applied to a two-dimensional, vertical tar sand pack, with restricted fluid entry. J Can Pet Technol 1973;12(04). [9] Allen JC, Redford DA. Combination solvent-noncondensible gas injection method for recovering petroleum from viscous petroleum-containing formations including tar sand deposits. 4109720 US. United States; 1976. [10] Allen JC. Gaseous Solvent Heavy Oil Recovery. Canada; 1988. [11] Nenniger EH. Hydrocarbon Recovery, Hatch & Associates. Canada; 1979. [12] Smith DH. Promise and problems of miscible-flood enhanced oil recovery. ACS Symposium Series. American Chemical Society; 1988:2-37.

12

Fuel xxx (xxxx) xxxx

X. Li, et al.

[69] Spiecker PM, Gawrys KL, Kilpatrick PK. Aggregation and solubility behavior of asphaltenes and their subfractions. J Colloid Interface Sci 2003;267(1):178–93. [70] Talebi S, Abedini A, Lele P, Guerrero A, Sinton D. Microfluidics-based measurement of solubility and diffusion coefficient of propane in bitumen. Fuel 2017;210:23–31. [71] Pfeiffer T, DiPrimio R, Achourov V, Mullins OC. TAR MAT Formation on Baffles in the Middle of the Oil Column. SPWLA 57th Annual Logging Symposium. Society of Petrophysicists and Well-Log Analysts; 2016. [72] Luo P, Gu Y. Characterization of a heavy oil–propane system in the presence or absence of asphaltene precipitation. Fluid Phase Equilib 2009;277(1):1–8. [73] Luo P, Wang X, Gu Y. Characterization of asphaltenes precipitated with three light alkanes under different experimental conditions. Fluid Phase Equilib 2010;291(2):103–10. [74] Andersen SI, Speight JG. Petroleum resins: separation, character, and role in petroleum. Pet Sci Technol 2001;19(1–2):1–34. [75] Speight J. Petroleum asphaltenes-Part 1: Asphaltenes, resins and the structure of petroleum. Oil Gas Sci Technol 2004;59(5):467–77. [76] Johnston KA, Schoeggl FF, Satyro MA, Taylor SD, Yarranton HW. Phase behavior of bitumen and n-pentane. Fluid Phase Equilib 2017;442:1–19. [77] Mancilla-Polanco A, Schoeggl FF, Johnston K, Richardson WDL, Yarranton HW, Taylor SD. The Phase Behavior of Heavy Oil and Propane Mixtures. SPE Canada Heavy Oil Technical Conference. Calgary, Alberta, Canada: Society of Petroleum Engineers; 2017. [78] Akbarzadeh K, Sabbagh O, Beck J, Svrcek WY, Yarranton HW. Asphaltene Precipitation From Bitumen Diluted With n-Alkanes. Canadian International Petroleum Conference. Calgary, Alberta: Petroleum Society of Canada; 2004:4. [79] Mitchell DL, Speight JG. The solubility of asphaltenes in hydrocarbon solvents. Fuel 1973;52(2):149–52. [80] Pfeiffer T, Di Primio R, Achourov V, Mullins OC. Scanning electron micrographs of tar-mat intervals formed by asphaltene phase transition. Petrophysics 2017;58(02):141–52. [81] Gonzalez V, Taylor SE. Asphaltene adsorption on quartz sand in the presence of preadsorbed water. J Colloid Interface Sci 2016;480:137–45. [82] Wu J, Prausnitz JM, Firoozabadi A. Molecular-thermodynamic framework for asphaltene-oil equilibria. AIChE J 1998;44(5):1188–99. [83] Goual L, Firoozabadi A. Measuring asphaltenes and resins, and dipole moment in petroleum fluids. AIChE J 2002;48(11):2646–63. [84] Rastegari K, Svrcek WY, Yarranton HW. Kinetics of asphaltene flocculation. Ind Eng Chem Res 2004;43(21):6861–70. [85] Speight JG. The Chemistry and Technology of Petroleum. CRC Press; 1999. [86] Dumont H, Mishra V, Zuo J, Mullins O. Permeable tar mat formation within the context of novel asphaltene science. Paper SPE-163292 presented at the SPE Kuwait International Petroleum Conference and Exhibition, Kuwait City, Kuwait, 10í12 December. 2012. [87] Wiegmann A, Zemitis A. EJ-HEAT: A Fast Explicit Jump Harmonic Averaging Solver for the Effective Heat Conductivity of Composite Materials. 2006. [88] Rogel E. Simulation of interactions in asphaltene aggregates. Energy Fuels 2000;14(3):566–74. [89] Qureshi A, Zuo JY, Seifert DJ, Zeybek M, Mullins OC. Mobile Heavy Oil and Tar Mat Characterization Within a Single Oil Column Utilizing Novel Asphaltene Science. SPE Kuwait International Petroleum Conference and Exhibition. Society of Petroleum Engineers; 2012.

2018;32(3):3163–75. [49] Zhang C, Oostrom M, Wietsma TW, Grate JW, Warner MG. Influence of viscous and capillary forces on immiscible fluid displacement: pore-scale experimental study in a water-wet micromodel demonstrating viscous and capillary fingering. Energy Fuels 2011;25(8):3493–505. [50] Lu T, Li Z, Li S, Li B, Liu S. Performances of different recovery methods for orinoco belt heavy oil after solution gas drive. Energy Fuels 2013;27(6):3499–507. [51] Amirian T, Haghighi M, Mostaghimi P. Pore scale visualization of low salinity water flooding as an enhanced oil recovery method. Energy Fuels 2017;31(12):13133–43. [52] Nguyen P, Fadaei H, Sinton D. Pore-scale assessment of nanoparticle-stabilized CO2 foam for enhanced oil recovery. Energy Fuels 2014;28(10):6221–7. [53] Seyyedi M, Sohrabi M. Pore-scale investigation of crude oil/CO2 compositional effects on oil recovery by carbonated water injection. Ind Eng Chem Res 2017;56(6):1671–81. [54] Wei B, Li Q, Jin F, Li H, Wang C. The potential of a novel nanofluid in enhancing oil recovery. Energy Fuels 2016;30(4):2882–91. [55] Cheraghian G, Kiani S, Nassar NN, Alexander S, Barron AR. Silica nanoparticle enhancement in the efficiency of surfactant flooding of heavy oil in a glass micromodel. Ind Eng Chem Res 2017;56(30):8528–34. [56] Gong H, Zhang H, Xu L, Li K, Yu L, San Q, et al. The synergistic effect of branchedpreformed particle gel and hydrolyzed polyacrylamide on further-enhanced oil recovery after polymer flooding. Energy Fuels 2017;31(8):7904–10. [57] Sun Q, Li Z, Li S, Jiang L, Wang J, Wang P. Utilization of surfactant-stabilized foam for enhanced oil recovery by adding nanoparticles. Energy Fuels 2014;28(4):2384–94. [58] Boerrigter PM, Verlaan M, Yang D. EOR methods to enhance gas oil gravity drainage. Society of Petroleum Engineers. [59] Bychkov A, Verlaan M, Boerrigter PM, van Heel AP, van Dorp JJ. Steam injection into fractured carbonates – the physical recovery mechanisms analyzed and upscaled. Society of Petroleum Engineers. [60] Niz-Velasquez E, Bagheri SR, van Dorp JJ, Verlaan ML, Jennings JW. Modelling development of a thermal gas/oil gravity-drainage process in an extraheavy-oil fractured reservoir. [61] Bachu S, Underschultz JR. Regional-scale porosity and permeability variations, Peace River arch area, Alberta, Canada. AAPG Bulletin (American Association of Petroleum Geologists); (United States) 1992: Medium: X; Size: Pages: 547-62. [62] Javadpour F, Fisher D. Nanotechnology-based micromodels and new image analysis to study transport in porous media. [63] Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, et al. Fiji: an open-source platform for biological-image analysis. Nat Methods 2012;9(7):676–82. [64] http://www.teledyneisco.com/en-us/Documents/High-precision%20High-pressure %20Syringe%20Pumps/Technical%20Bulletin/Metering%20Syringe%20Pumps %20-%20Operation%20Theory%20and%20Maintenance%20Tech%20Bulletin.pdf. [65] Nenniger JE, Dunn SG. How Fast is Solvent Based Gravity Drainage? : Petroleum Society of Canada. [66] Long RB. The concept of asphaltenes. Chemistry of asphaltenes. Am Chem Soc 1982:17–27. [67] Strausz OP, Lown EM. The chemistry of Alberta oil sands, bitumens and heavy oils. Alberta Energy Research Institute Calgary, Alberta, Canada; 2003. [68] Wang Q, Jia X, Chen Z. Mathematical modeling of the solvent chamber evolution in a vapor extraction heavy oil recovery process. Fuel 2016;186:339–49.

13