Sunteți pe pagina 1din 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301396588

Optimal Waterflood Voidage Management Significantly Increases Oil Recovery


with Minimal Incremental Cost

Conference Paper · January 2014


DOI: 10.2118/171937-MS

CITATIONS READS

3 507

3 authors, including:

Zhouyuan Zhu
China University of Petroleum (Beijing). Beijing, China
15 PUBLICATIONS   37 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Zhouyuan Zhu on 23 December 2017.

The user has requested enhancement of the downloaded file.


SPE-171937-MS

Optimal Waterflood Voidage Management Significantly Increases Oil


Recovery with Minimal Incremental Cost
E. Steven Vittoratos, Consultant to BP; Zhouyuan Zhu and C.C. West, BP

Copyright 2014, Society of Petroleum Engineers

This paper was prepared for presentation at the Abu Dhabi International Petroleum Exhibition and Conference held in Abu Dhabi, UAE, 10 –13 November 2014.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Empirical data, laboratory experiments, and mathematical simulation methods indicate that for many
heavy oil waterfloods the optimal voidage replacement ratio (VRR) is less than one. Many more recovery
mechanisms are activated with VRR⬍1 than with VRR ⫽ 1. Some are readily understood with existing
conventional reservoir engineering concepts, while others invoke emulsion & foamy oil multiphase flows
that are activated under changing reservoir conditions. Not all the consequences of VRR⬍1 are positive
for recovery. In aggregate, however, as displayed in empirical observations, VRR⬍1 can result in a
significant increase in reservoir oil recovery, particularly for the heavier oils.
This paper focuses on the quantification of the relative importance of the mechanisms activated by
VRR⬍1 in saturated or nearly saturated reservoirs. The absolute value of the VRR⬍1 depends primarily
on the oil quality and its associated properties, whereas the optimal time evolution of the VRR depends
on the well spacing and the reservoir heterogeneity of the depositional environment. Numerical simula-
tions have ranged from simple 1D models to a detailed fluvial depositional environment model. The role
of relative permeabilities are demonstrated by the 1D model. The 3D field models of increasing geological
complexity are useful for quantifying the incomplete waterflood sweep, and the role of VRR ⬍ 1 in
activating solution gas drive to deplete the ‘cul-de-sacs’ of bypassed oil. We also propose here a numerical
model to incorporate emulsion flow behavior into heavy oil water flooding, calibrated using data from a
large scale (‘big can’) experimental waterflooding study of a heavy oil prone to emulsion formation. The
methodologies developed in this study show that for heterogeneous heavy oil reservoirs the incremental
recovery expected from an optimized VRR ⬍ 1 process is similar to that of other commercial IOR
processes such as polymer flooding, but with very little incremental cost.
Introduction
For light oil waterflooding, the fundamental paradigm for optimal reservoir management is to fully replace
voidage, i.e., VRR ⫽ 1, both instantaneously and cumulatively. Empirical evidence is accumulating,
however, that this paradigm is sub-optimal for the waterflooding of heavier oils, and that periods of VRR
⬍ 1 are beneficial. For the older basins, such as the North Sea or the North Slope of Alaska, the remaining
resources to be developed are increasingly heavy. Is is urgent, therefore, that these resources be recovered
with a fuller understanding of voidage management. In the last decade, BP has applied empirical analyses,
2 SPE-171937-MS

Figure 1—Conceptual oil recovery vs VRR curve for heavy and light oils [7].

laboratory experiments, and mathematical simulations to design voidage management practices to specific
heavy oil reservoirs.
Many more recovery mechanisms are activated with VRR⬍1 than with VRR ⫽ 1. Some are readily
understood with existing conventional reservoir engineering concepts, while others invoke emulsion &
foamy oil multiphase flows that are activated under changing reservoir conditions. Not all the conse-
quences of VRR⬍1 are positive for recovery. In aggregate, however, as displayed in empirical observa-
tions, VRR⬍1 can result in a significant increase in reservoir oil recovery, particularly for the heavier oils.
The VRR⬍1 improved oil recovery process potentially takes advantage of all of the following
mechanisms:
1. Reservoir ‘cul-de-sac’ heterogeneity: heavy oil waterflooding is always associated with non-
uniform sweep that leaves significant amount of oil behind, which can be produced by solution gas
drive.
2. Foamy oil drive: in many heavy oils, exsolved gas remains dispersed in the oil leading to high
critical gas saturations and low gas mobilities resulting in an effective recovery mechanism.
3. Emulsification: water and oil emulsification can occur in heavy oil waterfloods with the help of
fine particles, i.e. clay and silt, and the chemical changes that accompany gas exsolution.
4. Three phase relative permeability interference: when the gas saturation in the oil reservoir
increases, the krw decreases, and the kro increases initially and then decreases. This leads to the
fractional flow fw⫽krw/(kro⫹krw)’s initial decrease and later increase.
A conceptual oil recovery-VRR curve is shown in Fig. 1. A VRR ⫽1 corresponds to a conventional
waterflood; a VRR ⫽ 0 corresponds to pure solution gas drive. Typically – at least for light oils –
waterflood recovery is a few times larger than solution gas drive. For a saturated or a nearly saturaged oil
what would the recovery be for VRR ⬍ 1, a hybrid process where both water & solution gas act to drive
the oil to be produced? For light oils, a VRR ⬍ 1 leads to high gas exsolution and highly mobile gas
saturations which can be ineffective in driving oil out; furthermore, the oil viscosity will increase with gas
exsolution. Thus, conceptually, we can expect that for light oils, oil recoveries decrease with decreasing
VRR, as illustrated by the dashed line in Fig. 1. In contrast, for heavy oils, we expect the optimal VRR
to be less than 1. Any oil viscosity increase caused by gas exsolution as pressure declines below the bubble
point is counterbalanced by the activation of other recovery mechanisms, as indicated in Fig. 2. Foaming
can drive oil out of the un-swept areas (‘dangling ends’) and gas evolution can also promote emulsion
formation in the reservoir. Emulsion flow in the ‘backbone’ of the connected paths between the injector
and the producer can in turn reduce the water cut at the producer. Foamy oil also slows gas production
and retains it within the reservoir keeping reservoir energy in place.
SPE-171937-MS 3

Figure 2—Conceptual recovery mechanisms for VRR<1 improved oil recovery process: reservoir heterogeneity, foaming in the ‘dangling ends’ and
emulsification in the backbone [7].

These mechanisms have been introduced & discussed elsewhere: [1], [2], [3], [4], [5], [6], [7]. These
studies include laboratory testing using a five foot long “big can” in AITF (Edmonton, Canada) [4],
numerical reservoir simulations including a study of optimal VRR for polymer flooding [5], [6], [9], [10]
and empirical studies of production histories in various oil fields [7].
This paper is a simulation study to more rigorously quantify various recovery mechansims activated by
VRR ⬍ 1, which have been described only qualitatively in the previous studies. In this work, we begin
with a simple 1D simulation model that quantifies solution gas drive, 3-phase relative permeability effects,
and the viscosity increase in the VRR⬍1 process. Next, we examine oil recovery from two reservoirs
representative of two different depositional environments. We conduct numerical simulations using type
pattern models (TPM) of shallow marine shoreface and fluvial depositional environments. We chose these
models as representative of relatively low to relatively high reservoir heterogeneity. These two envirom-
enents have shown different results, which will be analyzed in detail for the quantification and decon-
volution of the cul-de-sac mechanism. Finally, this paper proposes and numerically tests the impact of
heavy oil emulsion flow using a model calibrated by the experimental results from the ‘big can’.

Quantification of Recovery Mechanisms in the VRR < 1 Process


Three phase relative permeability interference and solution gas drive in 1D
In this section, we study the fundamental aspects of the VRR⬍1 process in a 1D system, isolating the
mechanisms of solution gas drive and three phase relative permeability, while incorporating the concom-
itant viscosity increases. In the current tested 1D system, we are not limited by practical issues such as
artificial lift BHP (bottomhole pressure) limits.
The first test case is a viscous oil 1D VRR⬍1 simulation test suite, with a critical gas saturation Sgc
⫽ 5%. Motivated by the experimental ‘big can’ data to be described later in the paper, this 1D problem
has small dimensions, 5 feet long, 0.83 feet by 0.83 feet cross section. The model is homogeneous with
a porosity of 0.35 and a permeability of 4000mD. The initial pressure is 1500 psi. We have the injector
on the left side, and the producer on the right side. We do not have the space to enter into a detailed
discussion of the relative permeabities relating to this case. Using the Stone II algorithm, we illustrate
(Fig. 3) the 3-phase relative permeability effects for this case: as the gas saturation increases, the kro
initially increases but then decreases. In order to have positive effect in slowing down the water
breakthrough, i.e. lowering the fw value, we want to keep the gas saturation, Sg, value within a certain
range, not too high, as shown in this plot. This implies that in practice VRR⬍1 should be implemented
for a finite period of time so that the Sg values inside the reservoir are maintained within a lower range.
It should be noted that for this short length 1D model, the pressure drop is very small. In heavy oil
reservoirs pressure gradients are considerably larger, as much as 1, 000 psi from injector to producer, as
4 SPE-171937-MS

Figure 3—fw-Sg curves for the test case’s relative permeability (Sgcⴝ5%) with mobility ratio of 30 and 10: as the gas saturation is increased, the
krw decreases, and the kro increases initially and then decreases. This leads to the fractional flow fwⴝkrw/(kroⴙkrw)’s initial decrease and later
increase.

Figure 4 —1D VRR<1 simulation test suite, and the test case’s 0D nature in pressure, and the comparison to realistic problem in the actual reservoir
(much larger pressure drop, ~1000 psi from injector to producer).

illustrated in Figure 4. Thus, our 1D model has almost uniform pressure, which is different than actual
field conditions.
To implement the VRR⬍1 process, one may cut the injection rate or increase the production rate. It
is not uncommon that commercial waterflood projects are injectivity limited; in that situation, simulations
that achieve VRR ⬍ 1 by increasing production would be representative of commercial realities. This does
raise, however, the possibility that the response may be at least in part an acceleration of production rather
than a true incremental recovery; the agreement of both ways of achieving VRR ⬍ 1, indicates that
acceleration effects are insignificant in this 1D model.
We observe that in the 1D system most of the production occurs prior to water breakthrough (Fig. 5).
It is important to keep in mind, however, that in commercial heavy oil waterfloods most of the production
SPE-171937-MS 5

Figure 5—WOR vs Cum Oil Produced, Cum Oil Produced vs cumulative water injected for the viscous 1D VRR<1 simulation case with a gas relative
permeability with Sgcⴝ5%.

occurs after water breakthrough. The 1D model will thus tend to underestimate the role of relative
permeability interference between the phases compared to the commercial recovery process. We also
make the following observations:
1. The total oil production prior to water breakthrough is significantly larger for VRR ⬍ 1 than with
VRR ⫽ 1. The increase equals approximately 5% of the OOIP (~1400 cc), suggesting that it is the
5% Sgc that primarily drives the extra oil.
2. For a given quantity of water injected, VRR ⬍ 1 recovers significantly more oil than VRR ⫽ 1.
On a time basis, however, the recovery with VRR ⬍ 1 will be less than with VRR ⫽ 1 because
of the slower injection of the displacing water. In an injectivity limited commercial system, this
cannot be considered a limitation of the process.
6 SPE-171937-MS

Figure 6 —WOR vs Cum Oil Produced and Cum Oil Produced vs cumulative water injected for the viscous oil 1D VRR<1 simulation case with a
gas relative permeability with a Sgcⴝ2%.

We have also considered a case with the smaller Sgc value of 2%, while keeping everything else the
same. The results are shown in Fig. 6. We see a smaller improved oil recovery; the delay in water
production is much smaller, and the incremental recovery with VRR ⬍ 1 for a given quantity of water
injection is marginal early on, but after breakthrough the VRR ⫽ 1 is a more efficient process. This again
points out that in voidage management, VRR ⬍ 1 should be of finite duration and early on in the process.
The negative impact of longer duration VRR ⬍ 1 is likely related to the increased oil viscosity as the
pressure is reduced.
In summary, the 1D system is controlled by three mechanisms: Sgc, the three phase relative
permeability interference, and viscosity increase with the decline of pressure with VRR ⬍ 1. We do not
include in this case emulsion flow behavior. Our results indicate that Sgc is the most critical parameter
for oil recovery and controls most of the observations. The details of the relative permeability curves and
their three phase interference are secondary if not insignificant. This may not, however, be a general result
SPE-171937-MS 7

for other relative permeability curves, particularly


for foamy behavior where a high Sgc is accompa-
nied with suppressed krg values for a large range of
Sg beyond Sgc. We will discuss this subject further
elsewhere. Losing solution gas makes the oil more
viscous. This may result in a negative effect on the
fractional flow of the system, by raising the equiv-
alent mobility ratio. These three effects in aggregate
Figure 7—Influence diagram for the 1D VRR simulation study. The
control the final effect of VRR⬍1 and improved green mechanisms increase recovery, whereas the red effect reduces oil
recovery in our 1D simulations, as shown in Fig 7. recovery.

Cul-de-sac Mechanism Deconvolution and


Quantification
We next analyze the distribution and magnitude of the cul-de-sac mechanism, deconvolving the cul-de-sac
mechanism from other effects in VRR⬍1. Traffic flow metaphors of the cul-de-sac mechanism are shown
in Fig. 8. Consider flow through a connected neighborhood: there are through roads (backbone) and
cul-de-sac regions (dangling ends) attached to the through roads. If there are no dead-end roads, the entire
connected region gets swept at the end. The cul-de-sacs neighbourhoods, however, remain unswept
without some internal displacement power to move the contents to the main thoroughfares. The VRR⬍1
process activates the solution gas drive and foamy oil mechanisms within the cul-de-sacs so that additional
recovery may be achieved.
The two simulation models we use in this section are the TPM for a shallow marine shoreface viscous
oil reservoir and the TPM for a fluvial heavy oil reservoir with foamy oil behavior. The permeability and
well configurations of the two TPM simulation models are shown in Fig. 9. In the viscous oil model we
have two horizontal tri-lateral producers and two vertical injectors perforated to multiple zones. To
simplify and make the problem more transparent, we flood only one flow unit: one the middle sands in
this TPM. Thus, it is equivalent to having two single lateral horizontal producers and two single zone
injectors. In the heavy oil reservoir model, we have one single lateral horizontal injector and two single
lateral horizontal producers. In both cases, the well spacing between producers is 2000 feet.
The heavy oil model is undergoing polymer flood with ~60 cp viscosity injectant. The viscous oil
model is undergoing normal water flood. The two models have similar order of magnitude mobility ratios
(heavy oil polymer flood versus viscous oil normal water flood). All VRR⬍1 effects (heterogeneity,
solution gas drive, three phase relative permeability interference) are represented in these models (though
not emulsion flow). The heavy oil model has a gas relative permeability that mimics foamy oil drive, while
the viscous model has a gas relative permeability that mimics light oil solution gas drive.
We conduct the simulation of the VRR ⫽ 1 and the VRR ⬍ 1 process on these two models, with their
VRR history shown in Fig. 10. We observe that the viscous oil model was able to sustain VRR ⫽ 0.7
production by cutting the injection rate to 70% for the initial 20 years. In the next 30 years, the producers
hit the minimal BHP control and the production wells change to a lower rate, with VRR ⫽ 0.7 no longer
sustained. In the heavy oil model, the VRR⫽0.6 was sustained throughout the life of the field.
Differences in VRR⬍1 performance are shown in the figures: Fig. 11 shows the Cum Oil vs Time
curves for VRR⫽1 and VRR⬍1 processes; Fig. 12 shows the actual Cum Oil vs Cum Water Injected
(equivalent to Pore Volume Injected, PVI) curves. The heavy oil model displays improved oil recovery
on both time and PVI basis. The incremental oil recovery on PVI basis is significant. The viscous oil
model shows no improved recovery on a time basis and only slight improvement in the first 10 years of
production on the PVI basis. Overall, the two models show different response to the implementation of
VRR⬍1 process.
8 SPE-171937-MS

Figure 8 —Metaphors of Cul-de-sac Mechanism in Traffic Flow.

Figure 9 —TPM simulation models used in this study and the permeability distribution, and the well configurations. All VRR<1 effects (heteroge-
neity, solution gas drive, three phase relative permeability interference) are represented in these models.

Potential explanations of the difference in the VRR⬍1 process incremental recovery between the two
models include these factors:
1. Heterogeneity (the heavy oil model has more cul-de-sac type permeability features concomitant
with its fluvial depositional environment);
2. Foamy oil effect (heavy oil models has stronger solution gas drive, with low krg and high Sgc);
3. Three phase relative permeability effects (heavy oil model has potentially stronger three phase
relative permeability interference effects with higher Sgc value).
We will next examine these differences in more detail, beginning with heterogeneity:
● Viscous model’s depositional environment is shallow marine, with a permeability variation of 3-4
orders of magnitude.
● Heavy oil model’s depositional environment is fluvial, with a greater permeability variation of 6
orders of magnitude.
SPE-171937-MS 9

Figure 10 —The actual VRR history for VRR ⴝ 1 and VRR < 1 processes for the viscous & heavy oil models

Figure 11—The Cum Oil vs Time curves for VRR ⴝ 1 and VRR < 1 processes for the viscous and heavy oil models

Fig. 13 shows the heterogeneity comparison between viscous oil TPM and heavy oil TPM in the
horizontal plane. Fig. 14 shows the heterogeneity comparison between viscous oil TPM and heavy oil
TPM in the vertical plane. Fig. 15 shows the permeability cumulative distribution of all the gridblocks in
the two TPM models. Fig. 16 shows the permeability cumulative distribution of all the gridblocks in the
viscous oil and heavy oil TPM models (larger heterogeneity variation in the heavy oil model, with 4
different facies).
For the gas-liquid relative permeability, the heavy oil model has very low krg values at small Sg values,
simulating the foamy oil drive. The Sgc critical gas saturation in the viscous oil model is very small, 1.5%.
With the higher Sgc value, we expect the heavy oil model to also show stronger three phase relative
permeability interference effects. As the gas saturation is increased, the kro increases initially, and then
10 SPE-171937-MS

Figure 12—The Cum Oil vs Cum Water Injected (equivalent to PVI) curves for VRR ⴝ 1 and VRR < 1 processes for viscous & heavy oil models

Figure 13—Heterogeneity comparison between viscous oil TPM and heavy oil TPM in the horizontal plane

decreases. This leads to the fractional flow fw⫽krw/(kro⫹krw)’s initial decrease and later increase. If
implemented properly, it will lead to a decrease in water cut and improved oil recovery.
Finally, we visualize and quantify the VRR⬍1/cul-de-sac effects in these two simulation models. First,
we try to identify & quantify all the VRR ⬍ 1 grid blocks in the simulation model. We use the
methodology shown in Fig. 17. For cells with only pure water displacement and without gas presence, we
shall have (Sw-Swi) ⫽ -1⫻(So-Soi). For cells with VRR⬍1 (all of the VRR⬍1 mechanisms, except
emulsion), the presence of gas leads to -1⫻(So-Soi) ⬎ (Sw-Swi), i.e. more oil is displaced than water
saturation change in that cell. The summation of (-1⫻(So-Soi) - (Sw-Swi)) for all the entire VRR⬍1 cells
shall be the total VRR⬍1 improved recovery oil amount. We can also define the pure Cul-de-sac effect
region to be the subset of the entire VRR⬍1 cells region that has circa zero water saturation change,
identifying the cells not swept by water. Account needs to taken of numerical diffusion causing the
presence of water in grid blocks unswept by water. Thus the unswept cul-de-sac regions are larger & more
SPE-171937-MS 11

Figure 14 —Heterogeneity comparison between viscous oil TPM and heavy oil TPM in the vertical plane

Figure 15—Permeability cumulative distribution of all the gridblocks in the viscous oil and heavy oil TPM models

pervasive than those identified merely by a zero water saturation increase (above connate); we, therefore,
identify the unswept regions by a band rather than a line in Fig. 17.
Fig. 18 is the VRR⬍1/Cul-de-sac effects plot for the VRR⫽1 process in the viscous & heavy TPMs.
For VRR ⫽ 1, we rarely see any cells in the VRR⬍1/Cul-de-sac zone. Fig. 19 is the VRR⬍1/cul-de-sac
effects plot for the VRR⬍1 process in the viscous & heavy oil TPMs. We see minimal effects of
VRR⬍1/cul-de-sac for the viscous oil model. For heavy oil model, we clearly observe the large number
of grid blocks in the VRR⬍1/cul-de-sac zone. This illustrates why we have much higher incremental
recovery for VRR⫽0.6 in the heavy oil model.
We also calculate the pure cul-de-sac effect region and oil recovery amount in Fig. 20. It is estimated
that of the total oil recovery from this model, about 5% comes from the pure Cul-de-sac effect. The rest
9% comes from solution gas drive and three phase relative permeability effects in the water swept region.
Fig. 21 and Fig. 22 finally show the 3D view of the pure Cul-de-sac cells in the heavy oil model (pressure
and oil saturation profile). The Cul-de-sac cells are mostly cells above the horizontal layer where the
injector and producer reside, and the cells around the producer.
The viscous oil model’s cul-de-sac regions have also been visualized. Fig. 23 shows the gas saturation
Sg in the small cul-de-sac zone in the viscous oil model. The low pressure areas around the producer
12 SPE-171937-MS

Figure 16 —Permeability cumulative distribution of all the gridblocks in the viscous oil and heavy oil TPM models (larger heterogeneity variation
in the heavy oil model, with 4 different facies)

indicates that solution gas drives small amount of


oil towards the producer (Sg~Sgc⫽1.5%). From the
comparison between heavy oil model and viscous
oil model, we see that solution gas drive is one of
the key factors in the impact of the cul-de-sac mech-
anism.
Heavy oil waterflood emulsion and its
modelling
The conventional conceptualization of oil & water
Figure 17—Plot of (Sw-Swi)-(So-Soi) for every grid block in the simu-
flow is that the phases slip past each other as de- lation models. If a pure water flood, we shall have (Sw-Swi) ⴝ -1ⴛ(So-
scribed mathematically by the Buckley-Leverett Soi). For cells with VRR<1 effect, we shall have -1ⴛ(So-Soi) > (Sw-
(B-L) theory. In some heavy oil reservoirs, how- Swi), i.e. more oil displaced than water saturation change in that cell. We
define the entire VRR<1 region (grey color) and pure Cul-de-sac region
ever, empirical evidence has accumulated that the (green color) in this plot.
phases flow by embedding themselves within each
other forming various forms of emulsions. Under
some conditions emulsion flow may contribute to
the improved oil recovery in the VRR⬍1 process.
Vittoratos [8] proposed that the sequential flow regimes observed during the steam stimulation the
Cold Lake bitumen field could be explained by emulsion formation. Vittoratos et al. suggested that
emulsion flow is the possible cause of the constant WOR ~ 1 regime in heavy oil waterflooding [1]. To
verify and better understand the emulsion flow physics in heavy oil water flooding, a series of experiments
were performed at AITF [4]. Fig. 24 shows one of the sample bottles from the heavy oil water flooding
experiment, showing the oil-in-water and water-in-oil emulsions formed using flooding conditions similar
to the field conditions. In practice, a detailed chemical analysis of the exact water and oil content is
required to accurately measure the water cut in such a sampling bottle. Fig. 25 further shows the conducted
big can VRR⬍1 experiment’s cumulative oil production vs time curve for a 12 API Alaska heavy oil, with
an in-situ viscosity of circa 2, 000 cp. We clearly observe the improved oil recovery for the two cases with
VRR ⬍ 1. Fig. 26 finally shows the microscopic images of the produced fluid during the VRR⫽1 test for
the heavy oil water flood experiment shown in Fig. 25. The images here were taken at atmospheric
SPE-171937-MS 13

Figure 18 —VRR<1/Cul-de-sac effects plot for the VRRⴝ1 process in the viscous & heavy oil TPMs.

Figure 19 —VRR<1/Cul-de-sac effects plot for the VRR<1 process in the viscous oil TPM and heavy oil TPM.

pressure. We also observe the micron sized water droplets dispersed in the aqueous phase, which
constitutes the water-in-oil emulsion fluid. The water droplets are small enough to move through a typical
pore throat in unconsolidated sand heavy oil reservoirs. This makes the emulsion flow almost like a single
phase flow that is similar as generating quasi-miscibility of water and oil in the reservoir under dynamic
flow conditions.
14 SPE-171937-MS

Figure 20 —Calculation of the oil displacement volume from the cul-de-sac zone and the percentage in total oil recovery from the cul-de-sac Zone
in the heavy oil model.

Here we propose a simplified model for modeling heavy oil in-situ emulsion flow. We assume for
certain levels of shear and chemical conditions, the water component can be dispersed as small droplets
into the oleic phase to form an oil emulsion and the same for oil dispersed into an aqueous phase.
Furthermore, the dispersed water droplets move at the same speed as the oleic phase, and the same for oil
dispersed in the aqueous phase. Considering a specific block, we start from 100% pure oil and gradually
add water into it. Up to a certain fraction limit, all the water can be dispersed into the oleic phase,
maintaining a single phase. Then above a certain fraction, we will start forming another free aqueous
SPE-171937-MS 15

Figure 21—View of Pure Cul-de-sac Cells in heavy oil Model (Pressure profile).

Figure 22—3D View of Pure Cul-de-sac Cells in heavy oil Model (So oil saturation).

phase, which also has some oil in it. By continuing to add water, eventually the oleic phase will disappear,
with single aqueous phase left in the block. We can further keep adding the water and reducing oil until
at the end we have 100% pure water. Fig. 27 shows one possible sequence of phase behavior of emulsion
16 SPE-171937-MS

Figure 23—Visualization of the gas saturation Sg in the small Cul-de-sac Zone in the viscous oil Model.

formation and flow in a block of the simulation


model: a maximum of 30% water that can be dis-
persed in the oil, the formation of two emulsions at
a water content between 30 and 75% water, and the
existence of only the aquaeous emulsion for higher
water fracitons. These two fraction limits in reality
shall be a function of shear and oil chemistry, which
could be calibrated from experiments. For the Buck-
ley-Leverett flow, on the other hand, we assume
water and oil are completely immiscible, i.e., the
water component stays only in the aqueous liquid Figure 24 —Sample bottle from the heavy oil waterflooding experiment,
phase and the oil component only in the oloeic showing the oil-in-water & water-in-oil emulsions formed using water-
phase. flooding conditions mimicking the field conditions.

To simplify the discussion, we here present a 1D


analytical formulation of emulsion flow. There are two phases, aqueous and oleic phase (water emulsion
phase and oil emulsion phase). Water and oil components can exist in both liquid phases within a certain
ratio. The two phase saturations are S1 and S2. The fractional flow functions for aqueous and oleic phases
are f1 and f2. Assuming incompressibility, plus aqueous and oleic phase viscosities constant, the transport
equations of water and oil components are:

Here, the water and oil component concentrations are:


SPE-171937-MS 17

Figure 25—‘big can’ VRR<1 experiment’s cumulative oil production vs time curves for a heavy oil

Figure 26 —Microscopic images of the produced fluid for the VRRⴝ1 run for the heavy oil waterflood experiments shown in Fig. 25. Images are taken
at atmospheric pressure.

And the fluxes for water and oil components are:

A modified black oil model is proposed here to model the emulsion flow for future field simulation.
In actual field simulation, the time scale of the flow transport will be much larger than the emulsion
18 SPE-171937-MS

Figure 27—Phase behavior example of emulsion flow in heavy oil water flooding.

Figure 28 —Mechanism of the proposed emulsion flow modified black oil model for heavy oil water flooding.

formation and decomposition process. Therefore, it would be reasonable to neglect the kinetic transient
process and assume equilibrium is reached instantaneously. Here, we use the emulsion phase behavior as
described previously. Fig. 28 shows the mechanism of the proposed modified black oil model for heavy
oil water flooding. Now, we can write the three conservation equations for the water, oil and gas
components:

In this new formulation, the oil component does not only stay in oleic phase, and the same for water
component. The solution gas behavior is still described by the Rs function (solution gas/oil ratio function).
When calculating phase velocity, we need to consider the viscosity changes due to phase emulsification.
The fraction Xw, Xo, fw and fo functions are crucial for the success of these simulations. All the
emulsification effects have been packaged into these functions. They will depend on local conditions in
the grid block, for example: the local shear rate, the solution gas effect, the oil chemistry, the concentration
of particulates and concentration of surfactant. Extensive standardized experiments are required to
calibrate these two functions to understand at a certain shear rate and chemical conditions how much water
can be dispersed into the oleic phase and how much oil can be dispersed into the aqueous phase.
Furthermore, the phase viscosities for water emulsion and oil emulsion phase need to be measured under
different conditions. And it could be shear rate and water content dependent. Once we obtain these
SPE-171937-MS 19

Figure 29 —Improved water cut match using the proposed emulsion flow modified black oil model for viscous oil water flooding big can experiment.
The top is the match using the traditional black oil formulation. The bottom is the match using the modified formulation that considers emulsion
formation. We assume stronger oil emulsion formed with the onset of VRRⴝ0.7 in this experiment at 7 hours [4].

correlations from the laboratory, we can apply them to the field simulation and compare the history match
with Buckley-Leverett model.
Finally, we show one example of improved history match through the use of modified black oil
formulation which considers emulsion formation. This is the example of the history match of one of the
big can experiment conducted for a viscous oil [4]. Fig. 29 shows the improved water cut match using the
proposed emulsion flow modified black oil model for viscous oil water flooding big can experiment. We
assume stronger oil emulsion formed with larger water content, from the start of VRR⫽0.7 in this
experiment at 7 hours [4]. The water cut match is clearly improved in this exercise. We have used the
chemical reaction model in CMG STARS to model the water and oil emulsion phase behavior in this
history match. Fig. 30 finally shows the improved cumulative oil recovery match using the proposed
emulsion flow modified black oil model for viscous oil water flooding big can experiment.
20 SPE-171937-MS

Figure 30 —Improved cumulative oil recovery match using the proposed emulsion flow modified black oil model for viscous oil water flooding big
can experiment. The top is the match using the conventional black oil formulation. The bottom is the match using the modified formulation that
considers emulsion formation. We assume stronger oil emulsion formed with the onset of VRRⴝ0.7 in this experiment at 7 hours [4].

Process Optimization
The VRR⬍1 process is well suited for oil fields with large well spacings, e.g., water flooding for offshore
and remote developments. This is because for fields developed with large well spacing, we usually have
large amount of oil bypassed between the injector and producer wells. The activation of solution gas drive,
cul-de-sac and other mechanisms in the VRR⬍1 process, can help to increase the oil recovery in a more
profound way in these applications. We are currently working with such oil fields demonstrating the
commercial value of VRR⬍1 process.
Time evolution is an important parameter in the optimization and implementation of the VRR⬍1
process. In practice, we would like to limit the amount of solution gas that is lost for the VRR⬍1 process
and keep the oil viscosity within a certain range. Here we present an example that improves the VRR⬍1
process performance by conducting time evolution optimizations. We use the same example test case of
SPE-171937-MS 21

Figure 31—Oil recovery vs time curve for viscous oil VRR<1 process. The VRRⴝ0.7 process is run for the entire field life. We observe initial
improved oil recovery in the first 10 years, followed later, however, by a lower oil recovery.

Figure 32—Actual VRR vs time curve for the viscous oil VRR<1 process. The VRRⴝ0.7 cannot be maintained after circa 2011 because of producer
BHP limits.

the viscous oil water flood in the previous section of “Cul-de-sac Mechanism Deconvolution and
Quantification”. The VRR⬍1 process is implemented by increasing the total production rate in the
producers, which is different than the previous examples. Fig. 31 shows the oil recovery vs time curve for
the viscous oil VRR⬍1 process, in which the VRR⫽0.7 process is run for the entire field life. We observe
initial improved oil recovery in the first 10 years; afterwards, however, the oil recovery become less
effective. Fig. 32 shows the actual VRR vs time curve for the viscous oil VRR⬍1 process which is run
for the entire field life, though it is not fully supported in the time after 2015, due to the producers reaching
BHP pressure limits. As shown here, even though we achieve initial benefit, the final recovery is lower
with VRR⫽0.7. This is because we lose much solution gas for the entire duration of the process, thus
making the oil in place more viscous. The increase in viscosity changes the oil/water fractional flow curve
and makes the final oil recovery lower. One solution to this scenario is to shorten the duration of VRR⬍1
process, i.e. catch up with VRR⫽1 after certain number of years of VRR⬍1. When we implement
VRR⫽0.7 only for the first 7 years of the process followed by VRR ⫽ 1, we obtain the oil recovery vs
time curve shown in Fig. 33. We observe improved oil recovery in the first 10 years, and later not much
difference in oil recovery till the end. VRR⬍1 process has successfully accelerated the oil recovery in the
first 10 years, which may increase the commercial project’s net present value NPV. This relatively simple
example illustrates the impact of time evolution optimization of the VRR⬍1 process.
22 SPE-171937-MS

Figure 33—Oil recovery vs time curve for the viscous oil VRR<1 process. The VRRⴝ0.7 process is run for the first 7 years, and then switched to
VRRⴝ1. We observe initial improved oil recovery in the first 10 years, and which is substantially maintained in the later field life.

Conclusion
This paper has developed a methodology for the quantification of the relative importance of the
mechanisms activated by VRR ⬍ 1. The absolute value of the VRR ⬍ 1 depends primarily on the oil
property, as well as on the solution gas drive, three phase relative permeability and the reservoir
heterogeneity of the depositional environment. Numerical simulations have been performed for these
studies. The Sgc is a dominant parameter governing the performance of solution gas drive and three phase
relative permeability effects in the 1D VRR⬍1 process. In the 3D field problem, we have examined the
static geological models of depositional environments of increasing complexity for quantifying the
incomplete waterflood sweep, and the depletion from the ‘cul-de-sacs’ of bypassed oil by the activation
of solution gas drive by VRR ⬍ 1. We have demonstrated that the benefits of VRR ⬍ 1 are much more
pronounced in the more heterogeneous fluvial model. We have also proposed a numerical model to
incorporate the emulsion flow behavior into heavy oil water flooding, backed by an experiment study of
the in-situ natural emulsion formations. The study in this paper presents the quantitative ground for
evaluation of the impact of the VRR⬍1 process for field implementations.

Acknowledgement
Conducting the study of VRR⬍1 process has been a rewarding challenge. We would like to acknowledge
and thank our colleagues who contributed to the success of this project through their comments and
suggestions: Brian Vanderheyden, Pedro San Blas, Gary Jerauld, Xiuxu Ning, Bradley Brice, Tom Gould,
Amitabh Garg, Peter Clifford, Frank Paskvan, and many others in the Heavy Oil Flagship, Upstream
Technology function and Reservoir Development function.

References
1. E.S. Vittoratos, et alet al: “Optimizing Heavy Oil Waterflooding: Are the Light Oil Paradigms
Applicable?”, 1st WHOC paper 2006-688, Beijing, November 2006
2. E.S. Vittoratos, C.C. West and C.J. Black: “Flow Regimes of Heavy Oils under Water Displace-
ment”, 14th European Symposium on Improved Oil Recovery, Cairo, Egypt, 2007
3. E. Vittoratos and C.C. West, BP: “Optimal Heavy Oil Waterflood Management May Differ From
that of Light Oils”, SPE EOR Conference at Oil & Gas West Asia, SPE 129545, Muscat, April
2010
4. E. Vittoratos, R. Coates and C. C. West: “Optimal Voidage Replacement Ratio for Communi-
cating Heavy Oil Waterflood Wells”, SPE Heavy Oil Conference and Exhibition, SPE 150576,
Kuwait, Dec 2011
SPE-171937-MS 23

5. D. E. Delgado: “Optimal Voidage Replacement Ratio for Viscous and Heavy Oil Reservoirs”,
Master Thesis, Stanford University, 2012
6. D. E. Delgado, E. Vittoratos, and A. R. Kovscek: “Optimal Voidage Replacement Ratio for
Viscous and Heavy Oil Waterfloods”, SPE-165349, SPE Western Regional & AAPG Pacific
Section Meeting, 2013 Joint Technical Conference, Monterey, Apr 2013
7. E. Vittoratos, and C.C. West: “VRR ⬍ 1 Is Optimal for Heavy Oil Waterfloods”, SPE-166609,
SPE Offshore Europe Conference & Exhibition 2013, Aberdeen, Sep 2013
8. E. Vittoratos: “Flow Regimes During Cyclic Steam Stimulation At Cold Lake”, Journal of
Canadian Petroleum Technology, 1991
9. P.A. San Blas and E.S. Vittoratos: “The Polymer in Polymer Flooding: Is its Value Overesti-
mated?”, SPE 170104, Calgary, June 2014.
10. B. Brice, S. Ning, A. Wood, G. Renouf: “Optimum Voidage Replacement Ratio and Operational
Practices for Heavy Oil Waterfloods”, SPE 170099, Calgary, June 2014.

View publication stats

S-ar putea să vă placă și