Sunteți pe pagina 1din 12

The Challenge of Predicting

Field Performance of Air


Injection Projects Based on
Laboratory and Numerical
Modelling
D. gutierrez
Computer Modelling Group Ltd.
R.G. Moore, s.a. mehta, m.g. ursenbach
University of Calgary
F. SKOREYKO
Computer Modelling Group Ltd.

flue gas injection-based IOR and CO2 sequestration, safety as-


Dubert Gutiérrez is a Reservoir Simu- pects of underbalanced drilling operations, gas phase combustion
lation Engineer with the Computer Mod- in porous media, near wellbore stimulation processes and explo-
elling Group (CMG) Ltd. in Calgary. sions and safety aspects associated with the energy industry. He
Previously, he worked for BP Colombia is a member of a number of Boards, including the Canadian En-
in Bogotá as a reservoir simulation engi- ergy Research Institute (CERI), the Petroleum Society and the
neer and for Nexen Inc. in Calgary as a University of Calgary’s International Education Council. He is
reservoir engineer. His current main area also a member of the Distinguished Lecturer Selection Com-
of interest is the numerical simulation of mittee of the SPE and is a registered professional engineer in the
enhanced oil recovery (EOR) processes Province of Alberta.
for the development of conventional and
unconventional oil and gas reservoirs, with an emphasis on EOR Matthew Ursenbach is a Research Engi-
processes that involve chemical reactions. Dubert holds a B.Sc. neer for the In Situ Combustion Research
degree in petroleum engineering from the National University of Group in the Department of Chemical
Colombia at Medellín and an M.Sc. degree with specialization in and Petroleum Engineering at the Uni-
reservoir engineering from the University of Calgary. versity of Calgary, where he studies the
mechanisms of in situ combustion and
R. Gordon Moore is the University Pro- other combustion-related processes. He
fessor in Air Injection-Based Oil Re- holds a Masters of Engineering degree
covery at the University of Calgary since from the University of Calgary.
July 2003. He was Chairman of the De-
partment of Chemical and Petroleum En-
gineering at the University of Calgary Fraser Skoreyko has been working at
from 1992 through 2003. Gordon was a CMG since 1993. From 1990 to 1993, he
founding partner in Hycal Energy, and worked as a consultant in Mallorca, Spain
currently is the President of HOT-TEC on Libyan reservoirs. From 1985 to 1990,
Energy Inc. His prime research area has he worked as a staff reservoir engineer
been the improved recovery of conven- for Alberta Energy Company in Calgary.
tional and heavy oils and bitumens, with particular emphasis on From 1982 to 1985, he worked as a con-
oil recovery by air injection. He is the head of the In Situ Com- sultant in Tripoli, Libya and in California.
bustion Research Group at the University of Calgary. He is a dis- From 1980 to 1982, he worked as a res-
tinguished member of the SPE and served as a SPE Distinguished ervoir engineer for Gulf Canada in Cal-
Lecturer in 2000/2001. Gordon is also an active member of the gary. From 1978 to 1980, he worked for
Petroleum Society, APEGGA, AIChE, CSChE and CHOA, and Schlumberger and Dresser Atlas as a open hole logging engineer.
is a registered professional engineer in the Province of Alberta. He received his B.S. from the University of Calgary in 1978.

S.A. (Raj) Mehta is a Professor of Oil
and Gas Engineering and the Director of
International Programs in the Department Abstract
of Chemical and Petroleum Engineering
at the University vof Calgary. His prime Air injection-based enhanced oil recovery processes are re-
research interests are in the areas of high ceiving increased interest because of their high recovery poten-
pressure combustion including mecha- tials and applicability to a wide range of reservoirs. However,
nisms associated with high pressure air most operators require a certain level of confidence in the poten-
injection-based oil recovery processes, tial recoveries from these (or any) processes prior to committing
resources. This paper addresses the challenges of predicting field
April 2009, Volume 48, No. 4 23
performance of air injection projects using laboratory and numer- each driving mechanism depends on the oil reservoir (i.e. heavy
ical modelling. versus light oils). The process is fuelled by a small fraction of the
Laboratory testing, including combustion tube tests, ramped oil (i.e. 5 to 10% of the oil-in-place), generally considered to be
temperature oxidation and accelerating rate calorimeters can supply the heavier components, which are burned by the advancing com-
data for simple analytical models, as well as providing important bustion front.
insights into potential recovery-related behaviours. These tests are The term In Situ Combustion (ISC) is generally used to describe
less suited to providing detailed kinetic data for direct and reliable air injection in heavy oils and bitumen while the phrase High
use in numerical simulators. Indeed, the oxidation reactions are Pressure Air Injection (HPAI) or Light Oil Air Injection (LOAI)
sufficiently complex that, regardless of how powerful the thermal is normally used to refer to air injection in conventional light oil
reservoir simulator is, its predicting capability will strongly de- reservoirs.
pend on the engineer’s understanding of the process and ability to There are two significant process mechanisms that must be
model the most relevant oxidation behaviours of the particular oil understood by anyone trying to design a successful air injection
reservoir under study. project. These are: 1) the two distinct oxidation reaction regimes
It is proposed that the optimum design cycle for air injection- that exist; and 2) the high microscopic sweep efficiency or ‘bull-
based processes is to perform laboratory testing that would aid in dozing’ effect of the combustion front.
the understanding of the process and in the design and monitoring
of a pilot-scale field operation. Analytical models and simplified,
semi-quantitative reservoir simulation models would be employed Oxidation Reactions
at this stage. If this evaluation stage is successful, a pilot operation There are two principal reaction modes for oil and oxygen that
would be initiated and the data gathered during the pilot, as well are relevant to the air injection process; namely, ‘bond scission’ re-
as laboratory oil property and compositional data, would then be actions and ‘oxygen addition’ reactions. Oxygen addition reactions
used to history match and tune a model for predictions of the full follow the general form of:
field operation.
Hydrocarbon + O2 → Oxygenated compounds + Energy

Introduction
In general, this reaction has oxygen chemically binding with the
This paper has been written in response to questions which hydrocarbon molecules to produce oxygenated species such as al-
many reservoir engineers express when evaluating the feasibility dehydes, alcohols, ketones and hydroperoxides. These compounds
of air injection as an enhanced oil recovery process for their fields. tend to further react and polymerize with each other forming
Questions such as, “What laboratory tests are available? What type heavier, less desirable compounds such as asphaltenes and, even-
of data is provided by each test? How do we use the lab results
tually, coke. On the other hand, bond scission reactions are of the
to predict field performance?” are not uncommon, and, although
form:
there are not straightforward answers, a discussion on the useful-
ness of different lab tests is presented to clarify some of the related
concepts. Hydrocarbon + O2 → Carbon oxides + Water + Energy
This document has also been written in response to the concerns
and comments expressed by many reservoir simulation practitio-
Bond scission reactions represent the traditional combustion re-
ners when matching combustion tube tests and other supporting
actions and involve the destructive oxidation of the hydrocarbon to
oxidation experiments, and trying to predict field performance of
an air injection project based on kinetic parameters obtained from produce carbon oxides (i.e. carbon dioxide and carbon monoxide)
such tests. Questions such as, “How do we use the lab data in the and water.
reservoir simulator? What are the limitations of thermal reservoir Figure 1 shows a schematic of the oxygen uptake rate plotted
simulation when predicting field performance of air injection proj- against temperature, showing the temperature range over which
ects?” are addressed to provide additional feedback and promote each oxidation mode is dominant (i.e. both reaction types occur si-
further discussion. multaneously but one type dominates the other) for ‘typical’ light
Additionally, this manuscript describes some of the combus- and heavy oils. It can be seen from this figure that, for heavy oils,
tion behaviours which have been observed by the In Situ Combus- oxygen addition reactions are dominant at temperatures below
tion Research Group (ISCRG) at the University of Calgary while 300ºC and bond scission reactions become the dominant reac-
performing combustion tube tests and supporting cracking/oxida- tion mechanism at temperatures above 350ºC. In between these
tion experiments, and gives some recommendations to improve two ranges there is the region known as the negative temperature
the modelling of the combustion process using thermal reservoir gradient region where the rate of reaction actually decreases with
simulators. increasing temperature. For light oils, bond scission reactions are
Hopefully, this paper will help the conventional reservoir en- normally the preferred reaction path at most temperature levels,
gineer to understand the differences associated with the less than and oxygen addition reactions only seem to dominate at tempera-
conventional ‘combustion reservoir engineering’ and the inherent tures below 150ºC.
challenges when trying to predict field performance of an air injec-
tion project based on experimental data and numerical reservoir
simulations.

Overview of the Air Injection Process


Air injection is an enhanced oil recovery process in which com-
pressed air (or oxygen-enriched air) is injected into an oil reser-
voir; the injected oxygen is intended to react with a fraction of
the reservoir oil at an elevated temperature to produce carbon di-
oxide. The oil is ignited ‘in situ’ at or near the injection wellbore
by downhole burners, electric heaters, steam or hot fluid injec-
tion, chemical agents or, while not recommended for heavy oils,
by spontaneous reactions. Oil is displaced by the direct action of
the resulting thermal front, as well as the drive provided by the
FIGURE 1: Crude oil oxidation regions.
combustion gases and hot water and steam; the importance of
24 Journal of Canadian Petroleum Technology
The significance of these competing reactions is their differ- production or steam injection). Then the pore blocking mechanism
ence when attempting to mobilize oil. Oxygen addition reactions would become an advantageous feature in later stages as the local
are very ineffective at mobilizing oil, but rather, these reactions action of pore blocking would reduce the tendency for the reaction
lead to the production of heavier hydrocarbons (i.e. asphaltenes zone to finger. Wherever the process is disproportionately burning,
and eventually coke) and oxidized hydrocarbons that are generally additional oil will be mobilized, increasing the oil saturation ahead
acidic and promote the formation of stable emulsions with water. of the finger and reducing the gas permeability, forcing the air to
On the other hand, bond scission reactions are extremely effective distribute to other regions of the combustion front.
at mobilizing oil and are the desired state of operation for air injec- Conversely, the HPAI process (combustion in light oil reser-
tion and particularly for heavy oil combustion. Consequently, the voirs) has generally been viewed as a gasflood, where the flue
most important key to the success of an air injection project is the gases provide the main driving mechanism for oil production and
need to start and maintain the process in the bond scission mode; the thermal effects (i.e. the bulldozing effect) have very little im-
achieving this involves proper ignition procedures. Once a high pact on the oil recovery mechanism. However, as discussed by
temperature combustion zone is created, a sufficient supply of ox- Montes et al.(3), this bulldozing effect has the ability to mobilize
ygen is required to maintain the oxidation reactions in the tempera- oil that can not be displaced by any of the other driving mecha-
ture range where bond scission reactions are dominant. nisms present (i.e. flue gases, hot water/steam drive). This ‘extra’
Although not really an oxidation reaction (oxygen is not a reac- oil can potentially be produced via the other recovery mechanisms
tant), another reaction mechanism that is worth mentioning is the (i.e. flue gas drive). As with the case of heavy oils, the bulldozing
pyrolisis reaction, which has traditionally been known as the fuel- effect during HPAI causes the rapid mobilization of oil into the
forming reaction during ISC. The pyrolisis reaction of a hydro- downstream pores, which temporarily reduces the gas permea-
carbon (HC) is of the general form: bility, redirects the air flow and promotes better volumetric sweep
of the reservoir. All these features can result in a much longer res-
ervoir life as compared to a conventional immiscible gasflood, pro-
Hydrocarbon(liquid ) Energy
 → HC(liquid and / or solid ) + HC( gas) vided there is a reasonably good economic environment during the
project life.
In summary, although the bulldozing effect represents an advan-
As pointed out by Abu-Khamsin et al.(1), pyrolysis of crude oil tage of the combustion process over many other EOR processes, it
in porous media goes through three overlapping stages: distilla- must be interpreted appropriately. As pointed out by Ursenbach et
tion, visbreaking and coking. During distillation, the oil loses most al.(2), its misinterpretation did lead to failures of several past ISC
of its light ends and part of its medium gravity fractions. During projects, and, as discussed by Montes et al.(3), neglecting it could
visbreaking, the oil mildly cracks into a slightly different product lead to the mismanagement of HPAI projects and the underestima-
and, at high temperatures, the oil remaining in the porous medium tion of reserves.
cracks into a semisolid residue that is rich in carbon. Both cracking
reactions produce gas. Also, most pyrolisis reactions are endo-
thermic, which makes temperature the most influential parameter
that dictates the severity of cracking. On the Laboratory Modelling of the
Pyrolisis plays a significant role during air injection in heavy Combustion Process
oils to the extent that coke is an important source of fuel for the
Laboratory combustion and oxidation tests can provide impor-
bond scission reactions. Given the tendency of light oils to undergo
tant and relatively inexpensive insight into the behaviour of in situ
bond scission reactions at much lower temperatures, pyrolisis is
combustion or HPAI for a given reservoir system.
not as important and coke is rarely seen during the analysis of the
post-test (burned) samples.
Significance of Different Oxidation and
‘Bulldozing’ Effect of the Combustion Front Combustion Tests
It is well-known that the combustion process associated with O xidation and combustion tests are performed mainly for three
air injection yields very high microscopic displacement efficiency. reasons:
Laboratory studies and cores taken from post-burn regions of fields 1. To better understand the oxidation behaviour and exother-
show a complete absence of hydrocarbons in the regions swept by micity of the oil and oil/rock systems;
the combustion zone; approximately 5 to 10% of the oil-in-place is 2. To estimate kinetic parameters of the relevant reactions,
consumed as fuel while the rest is mobilized and available for cap- and;
ture. The main reason for this feature is what we call in this paper 3. To better understand and have a more quantitative idea on
the ‘bulldozing’ effect; the combustion front acts as a bulldozer to the expected recovery performance of the combustion pro-
mobilize most of the oil immediately ahead of it, which could not cess when applied to a particular reservoir.
be previously displaced by the other acting driving mechanisms Unfortunately, the complexity of the process is such that there
(i.e. flue gas sweeping, hot water/steam displacement, etc.). is still not a single test that can provide all the information listed
above. Therefore, different experiments are required to get a com-
Although the presence of this front has been recognized for de-
plete picture of the process. For the purpose of this paper, the
cades, its relevance to the air injection process has not always been
laboratory tests will be classified into three general categories de-
interpreted properly.
pending on their scope: a) screening or ‘fingerprinting’ studies;
For the case of heavy oil and bitumen reservoirs, which are typ- b) quantitative kinetic studies; and, c) combustion performance
ically highly saturated with oil, this feature can be a problem if studies.
not managed appropriately. When air is injected in a reservoir like
this, since the oil closer to the production wells (in the native or
Screening or ‘Fingerprinting’ Studies
‘cold’ region) has low mobility and the thermal front is so effec-
tive at mobilizing oil immediately ahead of it (i.e. the bulldozing The screening or fingerprinting studies allow the estimation of
effect), this unbalance causes an increase in oil saturation in the re- kinetic parameters of the oxidation reactions (e.g. activation en-
gion immediately ahead of the combustion front and creates what ergy, reaction order and frequency factor) but, since they do not
Ursenbach et al.(2) call ‘pore blocking.’ Pore blocking occurs when replicate the flow conditions that occur within the reservoir, they
the liquid saturation of mobilized oil becomes sufficiently high so do not provide any insight on the recovery performance of the
as to reduce the gas saturation below the critical level for flow. process.
The resulting restriction can cause the injected air flux to drop to a The Thermogravimetric Analyzer (TGA), Differential Thermal
level where the high temperature (bond scission) reactions cannot Analyzer (DTA), High Pressure TGA, Differential Scanning Cal-
be sustained. This is why for reservoirs like this, it is preferable to orimeter (DSC), Pressurized Differential Scanning Calorimeter
use air injection as a secondary recovery mechanism (i.e. after cold (PDSC) and Accelerated Rate Calorimeter (ARC) fall into this
April 2009, Volume 48, No. 4 25
category and, due to their relative simplicity, they are the most
widely employed thermal analytical tools to study the oxidation
kinetics associated with the combustion process.
Although this type of test allows the rapid determination of oxi-
dation kinetic parameters, its results only have a qualitative nature.
The kinetic parameters are normally calculated assuming a single
reaction model, which cannot realistically represent the multiple
oxidation reactions experienced by the oil sample that occur si-
multaneously throughout the tests. Nevertheless, these tests have
proven to be effective tools to understand the exothermicity of
crude oils and to evaluate the potential of the air injection process
for a reservoir in a relatively quick and inexpensive manner(4, 5).
Except for the ARC, which can be performed at pressures up to
41 MPa, most of these tests can only be performed at relatively low
pressures (i.e. less than 7 MPa) which limit their practical applica-
tion to certain reservoir settings.
For specific details on these tests, the interested reader is di-
rected to Chapter 3 of Sarathi’s in situ combustion handbook(6) and
its list of references. FIGURE 2: ISC schematic temperature profile for a typical heavy
oil(6).

Quantitative Kinetic Studies


On the Physics of the Combustion
These are kinetic studies that replicate the flow conditions Process – Combustion Tube Tests
within the reservoir and allow the quantitative determination of ki-
netic parameters of the oxidation reactions. Nevertheless, they are Having understood some of the basic concepts associated with
not meant to provide quantitative information on the recovery per- the air injection process, this section presents a more detailed dis-
formance of the combustion process (i.e. air requirement, air-fuel cussion of the physics associated with the combustion process,
ratio, etc.) which can be determined from combustion tube tests. based on observations from several types of oxidation tests and
particularly from combustion tube tests. Figure 2 shows an ide-
The most representative experiment in this category is the
alized representation of a forward combustion process developed
Ramped Temperature Oxidation test (RTO)(7, 8). There are different
based on a linear combustion tube test. A discussion of this figure
designs for the RTO apparatus described in the literature(8-11) but,
and each of the regions shown is presented elsewhere(6). This figure
in general, this test involves the controlled heating of oil-satu- will be used for the purpose of this discussion.
rated cores in a one-dimensional plug flow reactor under a flowing
As background, the ISCRG has been performing combustion
stream of air (or oxygen-containing gas) and are normally designed
tube tests, as well as thermal cracking and oxidation kinetics ex-
to operate at the pressure of the reservoir under study. The purpose
periments, since 1974. To the end of 2007, the group has performed
of this test is to study global oxidation behaviour and reaction ki-
in excess of 350 combustion tube tests on over 150 reservoirs from
netics under controlled conditions, with the end goal of providing around the world. Combustion tube tests have been performed on
realistic reaction data which could be used in thermal reservoir crude oils having gravities which vary between 6 and 45ºAPI at
simulators to predict field performance. This test has proven to be pressures which range from essentially atmospheric to 41.4 MPa.
extremely useful at defining the different oxidation regimes (i.e. The majority of the combustion tube tests have involved normal
oxygen addition reactions vs. bond-scission reactions) and under- air (21% oxygen), but in the order of 75 of these test used 95%
standing their impact on oil recovery(7, 8). oxygen-enriched air. Dry combustion is the most common process
tested, especially for light oils, however, a significant number of
Combustion Performance Studies normal wet and super wet tests have been performed on Canadian
reservoirs as this was the preferred mode of operation for many
This type of test provides quantitative information on the com- of the Canadian pilots which were operated in the late 1970s and
bustion performance that can be expected in the field, however, it early 1980s.
does not provide any kinetic data of the occurring reactions. For combustion tube tests where bond scission or combustion re-
The traditional experiment in this category is the combus- actions are the dominant oxygen uptake mode, the rate of propaga-
tion tube test. Combustion tubes are elemental physical simula- tion of the so-called ‘combustion front’ is controlled by the rate of
tors which allow for direct observation of the propagation of the delivery of oxygen to the location where the combustion reactions
combustion front under conditions which approximate those of the are occurring and by the amount of hydrocarbon residual which is
reservoir. Because the nature of the combustion front propagation available for reaction at that location. In an in situ combustion or
is controlled by the chemical reactions and relative permeability high pressure air injection process, the so-called fuel available for
characteristics of the reservoir, combustion tube experiments pro- the combustion reaction is very dependent on the extent of oxida-
vide a means for estimating the net effect of the interrelationships tion of the oil within the steam bank region which exists down-
between the various mechanisms that affect combustion. The main stream of the combustion zone. This, in turn, is dependent on the
drawback of combustion tube tests is that they are unscaled ex- availability of oxygen, the absolute pressure, the surface area of
periments and direct correlation of combustion tube results to the the sand matrix, the composition of the oil, the composition of the
corresponding reservoir is not straightforward. Nevertheless, ex- rock and, of course, the time that the oil in the steam bank region
is exposed to oxygen at temperatures below approximately 300ºC.
perience has shown that, as long as the testing is done under rep-
The exposure time is dependent on the velocity of the combustion
resentative reservoir conditions, the test exhibits good correlation
front, which is controlled by the oxygen flux, fuel availability and
between oxygen and fuel requirements, combustion stability and
oxygen utilization efficiency within the combustion zone. Oxida-
the characteristics of the produced oil, water and gas phases.
tion reactions will occur over essentially the full extent of the core
A properly designed and operated combustion tube test not only that has experienced a temperature rise. Hence, there is a coupling
can provide very useful information about the rock/oil system’s between bond scission (combustion) and oxygen addition reac-
combustion characteristics, but it also provides the necessary data tions over a significant portion of the heated region. The extent of
to engineer and make proper economic projections of an air injec- this coupling is low for tests exhibiting high oxygen utilizations in
tion project(12, 13). the combustion zone and for tests conducted with normal air at low
26 Journal of Canadian Petroleum Technology
pressures. This is why low pressure burns on many heavy oils and field. Unless specifically requested otherwise, combustion tube
bitumen show stable burning characteristics. tests are designed to evaluate oil displacement under conditions
The importance of the coupling between the bond scission re- where the dominant oxygen uptake mode is that of bond scission.
actions (normally referred to as High Temperature Oxidation or As in the field, operating the combustion tube at too low a flux will
HTO) and oxygen addition reactions (normally referred to as Low promote operation at low temperatures which are now well-recog-
Temperature Oxidation or LTO) is dependent on the residual hy- nized as not providing effective displacement of the oxygen con-
drocarbon concentration at the downstream or leading edge loca- tacted oil. Low air flux combustion tube tests can be performed,
tion of the steam bank region. The amount of residual hydrocarbon however the effective heat capacity of the apparatus makes it im-
is governed by the relative permeability characteristics and re- possible to obtain a meaningful estimate of the minimum air flux
sidual oil end point saturation in the unheated portion of the core, (which would be the prime purpose of this type of test) that could
as well as by re-saturation of the so-called ‘oil bank’ due to the dis- be translated to the field. If the intent is to observe operation where
placement of the residual oil (i.e. bulldozing effect). This residual oxygen addition reactions dominate, then there are more economic
hydrocarbon is exposed to hot fluids in the steam bank and to va- tests which can be performed (e.g. RTO test).
pourization/thermal cracking reactions which exists just down- History has shown that the air and fuel requirements as mea-
stream of the location where the dominant bond scission reactions sured in the laboratory are very representative of those measured
are occurring. on a field project, so long as the field project is operating in an
Fluid saturations within the re-saturated oil bank region are crit- effective oil displacement mode. Combustion tube measured air
ical in the sense that they will reduce the gas phase relative per- and fuel requirement parameters for an oil which has shown good
meability in the highly saturated zone which, in turn, will reduce burning performance in the laboratory can be used as input to the
the rate of delivery of oxygen to the combustion zone. Because of analytical models based on the historic work by Nelson and Mc-
the nature of the oxidation kinetics, reduction or interruption of Neil(12) or the correlations of Gates and Ramey(13) to provide what
the rate of delivery of oxygen to the combustion zone will drop is still the best semi-analytical estimate of the performance of a
the reaction temperature. For all oils, the oxidation reactions at field-based process.
low temperatures (from reservoir temperature to a level which is
characteristic of the oil) are those of oxygen addition. These re-
actions will temporarily or permanently immobilize the residual
hydrocarbon depending on the extent to which they are allowed
On the Numerical Modelling of the
to occur. An ISC process will be ineffective if the low flux condi- Combustion Process
tion is allowed to persist. A HPAI process may be less sensitive to
the decreased effectiveness of the elevated temperature region for Historically, the parameters involved in the simulation of a com-
mobilizing oil if the process is operated as a gasflood (i.e. at low bustion tube using a thermal simulator are tuned using combustion
injection rates), however, a significant loss of potential oil produc- tube data for tests operating in the dry (no water injection) mode.
tion could occur. Then, the history matched parameters are up-scaled for the simula-
tion of the process at a larger scale (i.e. field scale)(14-19).
The above discussion is not presented to imply that oxidation re-
actions in the steam bank region totally dominate the in situ com- This section of the paper discusses some of the challenges and
bustion and high pressure air injection processes. For many oils, potential pitfalls associated with history matching combustion
the effects of the coupling of the oxidation and combustion re- tube data with the end purpose of predicting field performance.
actions are overcome by proper ignition and by recognizing the It also addresses some of the intrinsic challenges associated with
importance of the oil mobility downstream of the elevated temper- simulating the combustion process and proposes some solutions to
ature region. It is evident that the best way to minimize feedback overcome them using current thermal simulation technology.
between the oxidation reactions in the steam bank and combustion
reactions is to have high oxygen utilizations by the combustion re- Conceptual Significance of History Matching
actions. This is why air flux at the reaction zone is such an impor- Dry Combustion Tube Tests
tant parameter controlling the nature of air injection processes.
While combustion tube tests are often used as the reference for As implied in previous sections of this paper, the problem with
tuning the parameters in numerical simulators, a one-dimensional matching a single dry combustion tube test is that the laboratory
combustion tube test cannot be expected to yield all of the re- combustion tube is typically operated at conditions where the oxi-
quired information needed for a meaningful simulation of a field- dation/combustion reactions place the process in the correct mode
scale process. It is well understood that one cannot tune a miscible for effective oil displacement. Under these conditions, the rate of
flood simulation based only on a single core flood, so it should not oxygen uptake associated with the propagation of the so-called
be a surprise that auxiliary data are required to develop a mean- combustion zone is controlled by the rate of delivery of oxygen to
ingful match of a combustion tube experiment that could be used the location where the combustion reactions are occurring. Under
for field-scale simulations. Combustion tube tests are primarily in- these ideal conditions, the observed oxygen utilization based on
tended to evaluate the ability of a given crude oil to exhibit stable the product gas is generally close to 100% and post-test core anal-
combustion characteristics under a prescribed set of conditions. ysis shows that the sand matrix contains no residual hydrocarbon.
For a given oil composition, the important conditions are the op- By doing this, we are modelling a best case scenario and no reli-
erating pressure and native core material. The effect of pressure is able predictions can be made as to whether the process goes into
very oil dependent and it is a much more important parameter for the wrong oxidation mode.
light oils as compared to heavy oils or bitumen. With the exception Just like matching a swelling test is not enough to predict the
of the percentage of initial oil recovered, fluid saturations do not performance of a CO2 flood, history matching of a single dry com-
have a significant impact on the measured air and fuel requirement bustion tube test is not sufficient to model the combustion pro-
parameters so long as the oil has sufficient mobility so as not to cess properly. A combustion tube test does not highlight the entire
restrict the gas phase relative permeability within the highly fluid physics of the process; in fact, normally it only highlights the fa-
saturated bank downstream of the steam bank. Heavy oil/bitumen vourable features of it.
tests performed at high initial oil saturations almost always experi- Matching the performance of a normal wet and super wet com-
ence ‘pore plugging’ due to the ‘bulldozing’ effect. To prevent this bustion tube test using the kinetic model used to match the dry
condition from totally dominating the test performance, heavy oil/ combustion test is a good way to test the validity of the kinetic
bitumen cores are generally preheated to a level (generally 50 to model. Normal wet tests often show higher peak temperatures, but
90ºC) which will restrict the excessive pressure drop condition to relatively similar air requirements as the corresponding dry test.
approximately one third of the total burn time. Since super wet tests operate at temperatures which are close to
It is also acknowledged that combustion tube tests are conducted the saturation temperature of the steam at the operating pressure of
at much higher air injection fluxes than will be experienced in the the test, these runs provide an excellent test of the suitability of the
April 2009, Volume 48, No. 4 27
kinetics models for temperatures significantly below those where
combustion of a solid-like fuel occurs.
Variables to Match During Dry Combustion
Tube Simulations
Significance of Kinetic Data Derived From Having understood the limitations of modelling combustion
Combustion Tube History Matches tube tests, the next question is related to the variables to match
in order to have a meaningful combustion tube simulation. These
As explained earlier, combustion tube tests do not yield kinetic would include: combustion front velocity, peak or maximum tem-
data. Hence, it is not possible to estimate reaction kinetics from perature profile, radial heat losses, pressure drop over the length of
combustion tube data alone, and some auxiliary tests are needed the core, fluid (oil, gas and water) production rates, produced-gas
(e.g. RTO experiments). composition and equivalent oil consumed as fuel.
It is acknowledged that very few numerical simulation matches For more detailed studies, additional variables to consider are:
of combustion tube tests are performed assuming only a single post-test hydrocarbon residual composition and distribution, com-
combustion reaction. Multiple reactions are often assumed and, positional changes in produced oil, shape of the temperature pro-
in general, most reaction schemes assume at least one thermal file, compositional changes in produced water and behaviour of
cracking reaction, one or more low temperature oxidation reactions burn during early stages following ignition (i.e. fast or delayed
and often two or more bond scission or combustion reactions. ignition, excessively high temperature in ignition zone, slow ad-
In numerical simulators, these reactions are modelled using an vancement rate of combustion zone).
Arrhenius model of the type: Now, achieving a meaningful combustion tube match to under-
stand the process is not straightforward and there are a few uncer-
 E  m n tainties and difficulties that need to be acknowledged by anyone
Reaction Rate = A exp −  PO C f attempting a meaningful match. Some of them are discussed in the
 RT  2 section below.

Where, A is the frequency factor, E is the activation energy, PO2


is the partial pressure of oxygen, Cf is the fuel concentration, m is On Combustion Tube Matches
the reaction order with respect to the partial pressure of oxygen and
This section discusses some of the inherent challenges associ-
n is the reaction order with respect to the fuel concentration. In the
ated with combustion tube matches.
above equation, the reaction rate may refer to oxygen uptake rate
or hydrocarbon consumption rate as they are coupled by the as-
sumed combustion equation. Combustion Front Velocity: The Role of the
Clearly, the above equation has four unknowns (A, E, m and n), Amount of Fuel
and this assumes that the heat of combustion of the hydrocarbon When matching a combustion tube test, the most direct approach
fraction undergoing combustion is known. This is not a bad as- is to fix the end point residual oil saturation at a level that leaves the
sumption for non-oxidized fuels, but great care is required when measured amount of fuel, and to then convert all of the fuel to com-
the Arrhenius-type reaction scheme is applied to oxygen addition bustion products, which neglects the different reactions that occur
reactions and the reader is directed to the paper by Burger and Sa- in the region ahead of the combustion front and modify the compo-
huquet(9) for their comments on peroxide formation. sition of the oil. If the effects of distillation, thermal cracking and
This means that if a combustion tube model has three reactions oxygen addition reactions, which modify the composition of the
(i.e. one of each kind), there are 12 variables that need to be esti- residual oil, are ignored, the composition of the produced gas (i.e.
mated which, of course, results in an infinite set of solutions and hydrogen/carbon ratio) will correspond to the composition original
limits the usefulness of the match. Nonetheless, the orders of reac- oil. Nevertheless, if the amount of hydrocarbon consumed is forced
tion are normally assumed to be unity, hence, the frequency factor to match, the combustion front velocity should certainly be reason-
and activation energy are the parameters which are normally ad- able, as the rate of fuel consumption is fixed by the rate of delivery
justed. The problem is how to adjust these parameters under condi- of oxygen to the reaction front. However, compositional changes
tions where the global oxygen uptake rate is controlled by the rate in the amount and composition of the oil fractions involved in the
of delivery of oxygen to the reaction location which is the situation process do occur and should be taken into account.
during most dry combustion tube tests. The match is made even
more uncertain due to the fuel concentration term as the fuel con- The Role of the Nature of the Fuel
centration is a predicted value based on the compositional changes
As implied in the previous portions of this paper, ideally the
which occur downstream of the location where the combustion re-
simulation model should incorporate kinetics that describe the
action is occurring. The same could be said of the oxygen partial
oxidation kinetics in both the oxygen addition (low temperature
pressure term, however, it is most uncertain for oxidation reac-
oxidation) and bond scission (combustion) modes. The kinetic
tions which occur downstream of the combustion zone. With the
model must also be capable of recognizing the impact of oil com-
previously mentioned uncertainties in mind, the question is what position on these reactions. For instance, post-test analysis of the
combustion tube measured parameter should be matched while ad- swept (burned) portion of the core following so-called light oil
justing the frequency factor (A) and the activation energy (E) terms tests shows that little or no coke is formed; peak temperatures for
for multiple reactions. these oils are well below those where coke combustion will occur.
The intent of the above discussion is not to imply that the au- Hence, the prime fuel for a burn of this type is primarily the satu-
thors do not believe in numerical simulation as a required tool for rates fraction remaining on the core following distillation at tem-
predicting the performance of a field-based air injection project. peratures approaching 300ºC. Nevertheless, numerical simulations
This is far from the case, as the authors fully recognize that nu- that assume coke as a reactive component when matching the com-
merical simulations, which can reliably predict both success and bustion behaviour of a light oil which show little reactivity above
failure of a field project, are required for management to fully em- 350°C are not uncommon.
brace this technology. Comments on the uncertainties in matching While not definitively confirmed, the ISCRG believes that the
combustion tube tests are primarily directed toward people who saturates fraction undergoes bond scission or combustion reac-
are new to the combustion modelling field and who are tempted tions in the vapour phase, which has been considered in simulation
to tune the reaction parameters in a thermal simulator based on models by a limited number of authors(20, 21). This adds a com-
a single dry combustion tube test. Tuning of complex kinetics plexity to the kinetic models, as it will be necessary to incorporate
models involving a large number of simultaneous reactions simply a flammability limit constraint. A further uncertainty associated
cannot be done based on a single dry test. Therefore, auxiliary data with modelling the vapour phase burn is that the ISCRG has not
is required for such purpose. yet identified the source of the hydrocarbon involved in the vapour
28 Journal of Canadian Petroleum Technology
phase burn. Some tests will suggest that it is a component of the heaters operate in a lead-lag manner; the length of the cycle de-
original oil and, therefore, predictable using phase equilibrium cal- pends on the placement of the thermocouples and the length of
culations. Other tests indicate that it may be a volatile component core is covered by a given heater. Heat is transferred from the core
generated by the oxidation reactions occurring in the liquid/solid during the part of the cycle in which the combustion front is ap-
phase. It should be noted that the ‘heavy oil’ and ‘light oil’ termi- proaching the control thermocouple, and is transferred to the core
nology used by the authors should not be interpreted to imply that as the combustion zone moves downstream of the control point.
all oils fit into these two arbitrary categories. There are many oils For this reason, some adjustment of the radial heat transfer is justi-
that exhibit both, with significant energy generation at tempera- fied. However, the variability which occurs between different nu-
tures in the range of 300ºC, as well as a distinct combustion zone merical simulations would suggest that the heater control strategy
with peak temperatures in excess of 400ºC. It all depends on the used on the laboratory combustion tube is changed depending on
oil composition. The bond scission combustion reactions occur- the reservoir sample being tested. This is not the case.
ring around 300ºC are believed to be associated with vapour phase An important observation with regard to peak temperatures is
burning while the higher temperature zone appears to involve a that they show little variability on repeat tests on the same oils but
solid-like residue. Post-test extraction of cores following tests on they are characteristic to the nature of the oil being tested. What
these intermediate type oils often indicates that the residual is tol- is almost universally observed is that the peak temperatures are
uene soluble which would classify it as an asphaltene fraction. A essentially independent of the operation of the heaters for stable
toluene insoluble or ‘coke’ fraction based on toluene solubility is burning oils. This is not the case for unstable burning oils for
often not present in any appreciable amount. The distinction be- which the common observation is that heaters operated in a rea-
tween asphaltenes and coke is significant as the asphaltene frac- sonable manner cannot stabilize the burn. An additional laboratory
tion can undergo thermal cracking to generate a coke-like residual observation is that when testing the same oil, the agreement in burn
which undergoes heterogeneous combustion reactions (gas-solid), performance between the two main combustion tube systems oper-
a volatile fraction which may undergo gas phase combustion, as ated by the ISCRG is excellent in spite of the fact that each heater
well as a modified maltenes (toluene and pentane soluble) fraction covers 15 cm and 5 cm, respectively.
which is displaced downstream of the high temperature reaction The reference to reasonable operation of the heaters implies that
zone. It is the gas phase combustion of the vapour generated by the heaters are not operated so as to continually drive the core tem-
the cracking of the asphaltene fraction that is believed to result in peratures into the temperature range where bond scission or com-
the slow rising temperatures which quite often occurs after a given bustion reactions occur. Holding the wall temperatures above the
zone exhibits a rapidly rising temperature and a distinct peak tem- core temperatures for extended periods at temperatures in the range
perature. Instead of exhibiting continuous cooling after achieving of 500°C will artificially support the combustion reactions for both
the peak temperature, the zone temperature will drop a few degrees light and heavy oils. Residual hydrocarbon remaining on the core
and then climb slowly. This behaviour is noted in combustion tube will undergo bond scission or combustion reactions at tempera-
reports by reporting both a peak and a maximum temperature. tures above approximately 300°C, but the surface area of this res-
idue is often so low that it controls the oxygen uptake reaction to
Pressure Drop and Oil, Gas and Water sufficiently low levels that an advance of the reaction front will
not sustain without external heat input. This is an aspect that one
Production Rates: The Role of Relative could easily miss when performing combustion tube tests. In fact,
Permeability Curves when reviewing combustion tube literature, a very useful indicator
Assuming the rate of advance of the combustion front is reason- of heaters inadvertently being used to drive the process is sustained
able, the production rates for the different fluid phases can be ad- low levels of CO2 concentrations (i.e. lower than 9% per mole) at
justed by tuning the relative permeabilities. However, one needs temperatures in excess of approximately 350°C.
to remember that multiphase flow occurs at a wide range of tem-
peratures and at different compositions of the flowing stream. Produced Gas Composition: The Role of the
Moreover, the oil re-saturation of the zone swept by flue gases and Nature of the Fuel
steam/hot water due to the ‘bulldozing’ effect of the combustion
Numerical simulations involving multiple carbon oxide and
front also causes a hysteretic effect which introduces more com-
water forming combustion reactions are often encountered. The
plexity to the problem.
multiple reactions often assume the same Arrhenius parameters.
Niz-Velásquez et al.(22) addressed the impact of oil re-saturation
Hence, the purpose of the multiple reactions is primarily to match
and hysteresis during the HPAI process by measuring three phase
the product gas composition and, in some cases, the water pro-
(oil, water and gas) relative permeability curves. This provides fur-
duction. The effect of assuming different fuels on the CO2 con-
ther insight on the physics of the multiphase flow involved during
centration is illustrated in Table 1. Complete combustion of the
oil recovery by air injection in light oils.
fuel is assumed. Hence, the products are CO2 and H2O. Only the
Having matched the combustion front velocity, pressure drop
CO2 concentrations are presented as they illustrate the variability
and fluid rates, what remains to be matched is the temperature pro-
in composition which can be expected as the fuel type changes.
file; more specifically, the peak temperature profile.
Normal air (21% O2) is assumed as the feed gas.
While not shown in the table, mass balance deviations for the
Peak Temperature Profile: The Role of low temperature oxidation of Athabasca bitumen can be up to 30%
Heaters and Radial Heat Losses based on the mass of the initial oil. If this oxidized residual under-
goes complete combustion, the elemental composition of the fuel
In theory, the peak temperature profile should relate to the fuel
consumption rate; hence, energy generation rate, which is con- Table 1: Typical variability of produced-gas
trolled by the oxygen flux. If, as assumed previously, the combus- composition with fuel type.
tion reaction is consuming the whole original oil, in theory, the
upper heat of combustion can be estimated with a fair degree of Produced Gas
certainty(9) or determined in the laboratory using a bomb calorim- H/C Ratio
eter(23) and the peak temperature profile should be in quite good Concentration
agreement with the observed laboratory values. Fuel Type Molar CO2 (%)
Unfortunately, what is not known is the thickness (i.e. volume) Hydrogen Infinite 0
of the reaction zone. Hence, radial heat losses from the combustion Methane 4 11.7
zone are often used as a tuning parameter. This is a procedure which Large carbon number paraffins 2 14.7
causes a high degree of consternation within the laboratory per- Aromatics 1 17.0
sonnel. It is recognized that no combustion tube is truly adiabatic. Typical coke 1 17.5
For near adiabatic units, such as those operated by the ISCRG, the Carbon monoxide 0 34.7

April 2009, Volume 48, No. 4 29


can be approximated as C1H1O0.7 which would yield a mole per- versus time history of the field, the laboratory observed ignition
cent CO2 in the produced gas of approximately 22%. characteristics will not translate to the field.
Combustion tube tests on a good burning oil often show CO2 Needless to say, this limitation also applies to numerical models;
concentrations in the range of 12 to 14% and CO concentrations in spontaneous ignition is not something that can be accurately pre-
the range of 0 to 4%. The observed apparent H/C ratio will often dicted by the use of current numerical reservoir simulators.
be in the range 1 to 1.7 for a heavy oil/bitumen. Matching of the
product gas composition is, therefore, not difficult if it is assumed The Negative Temperature Gradient Region
that more than one hydrocarbon type is undergoing combustion
reactions. The observed differences in the burning characteristics of so-
An interesting aspect of the product gas composition observed called heavy and light oils is the reason that the authors use the
during a combustion tube test is that CO is generally present in the term bond scission reactions to describe the combustion reactions
produced gas. Low pressure dry tests will normally show CO2/CO for which the products are carbon oxide and water. These were
historically referred to as high temperature oxidation reactions.
ratios in the range of 3 but the ratio drops with increasing pressure
As implied previously, combustion tube tests on light oils showed
and with water injection. While not common, some combustion
that this terminology is a misnomer as the bond scission reactions
tube tests exhibit no CO. These are generally wet or high-pressure
for light oils can occur below 300ºC, a temperature which is his-
tests. For reference, field projects often show little or no CO until
torically associated with low temperature oxidation reactions. For
the production well is in communication with an elevated tempera-
this reason, the authors prefer to use the definition ‘High Tempera-
ture front.
ture Region (HTR)’ and ‘Low Temperature Region (LTR)’ to de-
A common misperception is that the CO2/CO ratio is a good in-
scribe the temperature ranges above and below what is known as
dicator of the reaction temperature associated with an air injection
the ‘Negative Temperature Gradient Region or NTGR’ (see Figure
process. This belief has been explored by the ISCRG with partic-
1). The NTGR is the temperature range which generally extends
ular reference to comparing the behaviour of dry, normal wet and
from about 280 to 350ºC (for a typical Canadian heavy oil/bi-
super wet tests, and no correlation has been found to date. tumen) where the oxygen uptake and energy generation rates de-
crease with increasing temperature. Recognizing the existence of
the NTGR region is key to understanding combustion behaviour.
Challenges During Combustion Belgrave et al.(25) were the first to introduce the NTGR into the ki-
Simulations netic model used for matching combustion tube tests.
For heavy oils/bitumen, the low energy generation rates which
This section addresses some of the inherent challenges associ- occur within the NTGR act as a barrier to the transition of the re-
ated with combustion numerical simulations in general. action temperatures from the Low Temperature Region to the High
Temperature Region. On the other hand, the NTGR promotes the
The Role of the Minimum Air Flux transition from the HTR to the LTR whenever the energy gener-
ation rate in the HTR drops below a level required to offset the
The concept of a minimum air flux for combustion operations is conductive and convective heat losses. For a typical light oil, the
not new as it is based on the minimum combustion front velocity NTGR will have an impact on the reaction temperatures after the
and extinction radius concepts first introduced by Nelson and Mc- vapour phase combustion reaction is essentially complete. Since
Neil(12), but unfortunately, it is not yet possible to define the min- vapour phase combustion is believed to be the dominant oxida-
imum flux for a given oil reservoir from the laboratory. Hence, it tion mode for the light oils, the NTGR is not as significant as for
is suggested to assume a value [i.e. 0.5 m3(ST)/m2h] as a starting heavy oils.
point(12, 24) or, if detailed/quantitative kinetic data are available, de-
termine it using a thermal reservoir simulator.
Field-scale simulation models often employ very low activation
Estimation of Reaction Kinetic Parameters
energies (or even zero, which makes the reactions independent As discussed before, deriving kinetic parameters from combus-
of temperature) for the different oxidation/combustion reactions. tion tube tests is not possible and auxiliary data is required for such
The danger of this approach is that these models are likely to pro- a purpose (i.e. kinetic studies). The difficulty with this is that the
duce better performance forecasts at lower air injection rates than estimation of these parameters is not necessarily straightforward
those indicated by the necessary minimum flux to sustain combus- and the kinetic studies must be interpreted with care, particularly
tion. Trusting such models (i.e. forgetting the significance of the when using the so-called screening or fingerprinting tools.
minimum flux) not only could lead to under sizing the compres- For example, the authors have noticed that the ARC test is be-
sion facilities, but also to operating the process in the wrong oxi- coming popular to estimate Arrhenius parameters to be used for
dation mode which could potentially translate into the failure of combustion simulations, particularly in light oils. When doing this,
the project. the resulting exothermic response is normally matched using a
single reaction model which could result in reaction orders with
Spontaneous Ignition respect to the fuel concentration, which are much greater than one
(up to 20)(5, 26) and do not have much physical meaning. These re-
A couple of important questions which are normally associated ally large reaction orders result from the fact of trying to match
with HPAI processes are, “Will the reservoir spontaneously ig- multiple chemical occurrences with a single chemical reaction,
nite?” and, if so, “How long will it take?” The latter is an important and not only do they have little physical meaning, but they could
question, but it is one for which the ISCRG has, to date, not been cause numerical difficulties when used for combustion numer-
able to develop a suitable test to provide a definitive answer. The ical simulations due to the sharper changes in reaction rate with
major problem in the laboratory is the effective heat capacity of the small changes in fuel concentration in the region of higher fuel
apparatus, which essentially ‘soaks’ up the energy generated by the concentrations.
low temperature reactions that are responsible for spontaneous ig- Another typical example is the use of reaction orders of zero (or
nition. This heat load prolongs the time period over which the low very close to zero) for the fuel concentration term, which causes
temperature oxidation reactions occur and this, in turn, results in a even more numerical problems due to the sharp changes in reac-
greater modification to the oil composition than would occur under tion rate with small changes in fuel concentration when the fuel
the more adiabatic environment of a large reaction zone in a field concentration is approaching zero. This practice is believed to
setting. The degree of oil composition modification that occurs in come from the fact that experimentalists like to use reaction or-
the laboratory often appears to result in the formation of a skin on ders of zero when estimating reaction energies from tests such as
the oil, which causes the oxygen uptake rates and, therefore, en- the ARC(27, 28), but this does not imply that the reaction rate is in-
ergy generation rates to decrease with time. This means that, unless dependent of the fuel concentration. It simply assumes that there is
the laboratory experiment can replicate the self-energy generation always an excess of fuel for the reaction to occur. Using a reaction
30 Journal of Canadian Petroleum Technology
air is very small and the heavy oil viscosity may be very large. The
remedy for this ailment is to use a nine-point connection formula-
tion in the areal direction. Comparative runs of combustion in an
inverted five spot pattern using 5-point and 9-point are shown in
Figure 3, which illustrates that grid orientation effects are indeed
large in this situation (temperature profiles correspond to the same
grid layer at the same time).

Lack of Kinetic Models and Complete


Understanding of the Combustion Process
It is acknowledged that the main challenge when predicting the
FIGURE 3: Areal temperature profiles illustrating grid orientation field performance of an air injection project is due to the still lim-
effects during ISC simulations—9-point scheme (left) versus 5-point ited understanding of the oxidation/combustion process and lack
scheme (right).
of comprehensive kinetic models that allow the prediction of the
expected oxidation behaviours at different conditions. While the
order of zero for numerical simulations has very little meaning as
phase behaviour of a crude oil can normally be modelled and pre-
the reaction rate becomes independent of the fuel concentration.
dicted accurately using cubic equations of state by knowing its
To the authors, making the reaction orders equal to unity is a good
composition and phase behaviour of its individual components (or
practice(10, 25) unless there are strong reasons to believe otherwise,
as may be the case for the reaction order with respect to partial pseudo components), the oxidation behaviour of crude oils do not
pressure of oxygen during LTO reactions(10, 25, 29). enjoy such an advantage(29). Although significant advancements
In order to better estimate kinetic parameters to be used in nu- have been made in the understanding of the oxidation behaviour of
merical simulation, quantitative studies such as the RTO are re- pure hydrocarbons, and the different fractions of the oil (i.e. SARA
quired, but even these have their own limitations. The historic fractions: saturates, aromatics, resins and asphaltenes) and their
RTO experiments clearly show run times where heat generation, impact on the overall oxidation behaviour of the oil mixture(29, 31-
33), predicting the oxidation behaviour of the oil is still not possible
oxygen uptake rates and carbon consumption rates are significant.
However, when the reactor is equipped with only a single thermo- and different tests are required to evaluate the oxidation/combus-
couple, there is significant uncertainty as to the reaction tempera- tion behaviour at different conditions.
ture corresponding to the oxidation reaction which is generating For example, Figure 4 illustrates how the oxidation behaviour
the observed product gas, and there is even greater uncertainty as of a medium oil changes with variations in pressure. At low pres-
to the volume of the reactor in which the reactions generating a sures (i.e. 110.3 kPa), the PDSC records higher exothermic activity
product gas at a given time are occurring. Although it is straight- in the high temperature region than the low temperature region.
forward to determine Arrhenius-type parameters describing the However, as pressure is increased, exothermicity in the LTR in-
overall reaction rate or to fit assumed reaction models based on creases and, at higher pressures (i.e. 4,136 kPa), the sample ex-
multiple reactions, the uncertainty in the nature of the reactions hibits higher exothermic activity in the LTR as compared to the
still makes it difficult to assign significance to the activation ener- HTR. This seems to indicate that the saturates fraction is playing
gies and the uncertainty in both temperature and reaction volume an important role in the oxidation behaviour at high pressure while
makes it almost impossible to apply any credibility to the pre-ex- the rest of the fractions are more important at lower pressure. This
ponential factors. type of behaviour is currently not easy to predict with analytical
Another historic difficulty associated with the reaction kinetics models.
is how to keep the Arrhenius coefficients from becoming coupled This limitation in the understanding of the oxidation/combus-
with the residual end point saturation. Most combustion simulation tion process obviously limits the predicting capabilities of numer-
models use the standard Arrhenius equation to model the chemical ical reservoir simulators. Numerical reservoir simulators are not
reactions. These are non-equilibrium (irreversible) reactions which
specifically designed to model the combustion process as such;
will always occur as long as there are enough reactants (fuel and
these are ‘empirical’ simulators that provide tools to help the engi-
oxygen). In other words, there will be a point where some of the
grid blocks will not have any hydrocarbon left (i.e. of any kind) neer model the specific chemical process he or she desires. A res-
which is not realistic. However, as will be explained later, this ap- ervoir simulator can only predict behaviours that are captured in
parent limitation can be overcome by using the appropriate fea- the kinetic model used. In other words, kinetics developed from
tures of the numerical reservoir simulator. a match of a dry combustion tube test cannot be expected to give
meaningful predictions of the performance of enriched air or super
wet combustion processes.
Grid Size Sensitivity and Grid Orientation
Effects
It is well-known that the kinetic data obtained from combustion
tube test simulations cannot be used directly for field-scale simu-
lations(16). The Arrhenius equation shows that the reaction rate is
dependent on temperature and time through the fuel dependency.
The dependency on temperature makes the reaction sensitive to
grid block size as large grid blocks need more heat to achieve a cer-
tain temperature level as compared to smaller grid blocks. Hence,
the kinetic parameters obtained using a fine grid need to be up-
scaled for use in coarser field-scale grids unless the field model is
run using a fine grid similar to that of the combustion tube model,
which would be ideal to capture the physics of the process (i.e.
width of the combustion front is in the order of a centimeter), but
could require extremely long run times.
The other difficulty is related to grid orientation effects(30). It
has long been known that grid orientation effects are greatest when
the mobility difference between injected fluid and reservoir fluid is
large (i.e. unfavourable mobility ratios)(30). Air injection in a heavy
FIGURE 4: PDSC results for medium oil at different pressures(33).
oil reservoir represents the worst case scenario as the viscosity of
April 2009, Volume 48, No. 4 31
A good example of potential pitfalls that could be made by mis- For example, for the case shown in Figure 4, at low pressures,
using the reservoir simulator to perform sensitivities is the case of the frequency factors of the reactions happening in the LTR would
enriched air combustion. If one was to predict the performance of be lower than those used in the reactions of the HTR. At high pres-
an enriched air combustion process in Athabasca bitumen based sures, however, the trend would be the opposite. This feature pro-
on kinetics derived from a normal air combustion test, the simu- vides reaction rate dependence beyond that naturally occurring
lator could easily predict a much better performance as compared through concentration factors.
to the normal air test (i.e. process is more efficient due to the ab-
sence of N2, which is an inert gas). However, laboratory tests at the
ISCRG have shown that this is not necessarily the case, and the use Dynamic Gridding
of oxygen-enriched air can result in increased oxygen addition re- A problem that has plagued the field-scale simulation of com-
actions, resulting in a decreased burn stability(34). Therefore, what bustion projects is the requirement of small grid blocks to model
may seem intuitive to us (and hence to the numerical simulator) is the process correctly, thereby requiring the use of large model sizes
not really representing the physical reality. that result in very long run times. This problem can be offset by
Regarding the kinetic models, considerable efforts have been the use of a dynamic gridding option(37) that automatically uses
made over the years to better represent the oxidation/combustion
small grid blocks in the areas of large temperature and mole frac-
behaviour of crude oils and allow proper modelling in numerical
simulators(25, 35, 36). Nevertheless, there is still work ahead and ef- tion changes, and in the areas of small changes large grid blocks
forts continue to be made in those areas. For now, the engineer can be used. The use of this dynamic grid option has been shown to
should realize that the predicting capability of the reservoir simu- speed up the simulation run times from 2 to 12 times faster and is
lator does strongly depend on the engineer’s understanding of the becoming more popular for combustion simulations(38, 39).
process and ability to model the most relevant oxidation behaviour
of the particular oil reservoir under study.
Proposed Workflow for Engineering Air
Injection Projects
Possible Solutions Using Current
As implied by this whole discussion, predicting the performance
Simulation Technology of an air injection process is a difficult task as there are many un-
Although reservoir simulators are not explicitly designed to certainties associated with the combustion process itself. However,
model the combustion process, some of the challenges faced when this should not prevent the reservoir engineer from evaluating the
performing numerical simulations of the combustion process may feasibility of the process.
be overcome by using existing capabilities or features of the res- It is proposed that the optimum design cycle for air injection-
ervoir simulators, some of which will be discussed below. For the based processes is to perform laboratory testing that would aid in
purpose of this discussion, reference will always be made to the
the understanding of the process and in the design and monitoring
Computer Modelling Group’s reservoir simulator, STARS.
of a pilot-scale field operation. Analytical models and simplified,
semi-quantitative reservoir simulation models would be employed
Partial Equilibrium Reactions at this stage. If this evaluation stage is successful, a pilot operation
For example, Sequera et al.(36) used the partial equilibrium re- would be initiated and the data gathered during the pilot, as well
actions feature in STARS to model the delay in coke formation as laboratory oil property and compositional data, would then be
observed during oxidation experiments run at low temperature used to history match and tune a model for predictions of the full
conditions. In the partial equilibrium reaction approach, a min- field operation.
imum mole fraction of a reacting component is required before the The proposed design approach is based on the design and pre-
reaction can occur (or stop). This minimum mole fraction can be
diction methods presented by Nelson and McNeil(12) and Gates and
dependent on pressure, temperature, composition of another com-
Ramey(13). A comprehensive step-by-step example of this process
ponent, or all these properties together. This feature can be very
useful to decouple the kinetic parameters from the residual oil satu- has been presented elsewhere(24), and this simple type of procedure
ration, which is one of the challenges mentioned previously. has been followed by other authors(19). In fact, these ‘simplified’
methods have been used for the design and development of two of
Velocity Dependent Reactions the world’s most successful ISC projects: the Suplacu de Barcau
field in Romania(40) and the Santhal and Balol fields in India(41).
Another example is the phenomenon of combustion front fin- On the other hand, it is acknowledged that the semi-analytical
gering in the direction of high air flux rates and extinction in the models mentioned above were developed for heavy oil reservoirs
direction of low air flux rates. Extinction is caused by the air flux
and may not be appropriate for some of the processes considered
rate falling below a minimum value required to maintain the com-
for some reservoir settings (e.g. oil sands). Hybrid processes such
bustion front advance (i.e. a minimum flux). This phenomenon
could be modelled using a velocity dependence of the reaction. as steam or solvent injection plus air injection, along with different
Extinction can be modelled with a minimum critical velocity of well configurations, are being considered by many operators. For
the air that must be achieved before the reaction can occur. Com- these cases, semi-quantitative simulation runs are proposed to as-
bustion front fingering can be modelled with velocity dependence sess the potential benefit of air injection.
by which the reaction rate increases as the air flux (i.e. velocity) For the semi-quantitative simulation runs, it is proposed to start
increases. with the use of simple reaction models (i.e. few components and
few reactions) and evaluate best case scenarios (i.e. using only
Pressure Dependence of Reactions HTO reactions and using residual oil as fuel) for different well or
pattern configurations. If a particular best case scenario does not
Although the pressure dependence of the reactions is apparently
captured through the partial pressure of the oxygen term in the Ar- show any promise, then that case can be discarded. Conversely, if
rhenius equation, what this variable captures is the dependency of it shows a significant incremental benefit, additional complexities
reaction rate on oxygen concentration and this may not be enough would be added to make the case more realistic. Additional labora-
to model the dependency of pressure of the oxidation of crude oils, tory tests that would aid in the design of the particular process and
such as the case shown in Figure 4. To model this case, one could well configuration under study could also be performed. This type
use another simulator feature to enter frequency factors as a func- of semi-quantitative approach coupled with Gates and Ramey’s
tion of pressure. correlation was used successfully by other authors(18).
32 Journal of Canadian Petroleum Technology
Conclusion
The prediction of field performance of air injection projects is a
difficult task for which there are still more questions and uncertain-
In May’s Issue…
ties than definitive answers.
Laboratory testing can provide data for simple analytical models Production and Facilities Optimization
as well as provide important insights into potential recovery-re-
lated behaviours. However, these tests are less suited to providing
detailed kinetic data for direct and reliable use in numerical sim- Optimizing production and its surface treatment
ulators. Indeed, the oxidation reactions are sufficiently complex
that, regardless of how powerful the thermal reservoir simulator is,
reduces costs and improves the bottom line. This
the still limited understanding of the physics of the process makes issue addresses the problems of production and
its predicting capability to be strongly dependent on the engineer’s
understanding of the process and ability to model the most relevant facilities optimization, including topics such as
oxidation behaviours of the particular oil reservoir under study. stimulation and workovers, bottomhole pumping,
It is proposed that the optimum design cycle for air injection-
based processes is to perform laboratory testing that would aid in fluid separation and measurement and innovations
the understanding of the process and in the design and monitoring in surface facilities and processing.
of a pilot-scale field operation. Analytical models and simplified,
semi-quantitative reservoir simulation models would be employed
at this stage. If this evaluation stage is successful, a pilot operation
would be initiated and the data gathered during the pilot, as well
as laboratory oil property and compositional data, would then be
used to history match and tune a model for predictions of the full The Petroleum Society has the following
field operation. publications for sale. Please go to our website
at www.petsoc.org where you will find the order
instructions under the PUBLICATIONS tab.
Acknowledgements
Determination of Oil and Gas Reserves: 

The authors wish to thank Computer Modelling Group (CMG)
Ltd. for permission to publish this paper. Our gratitude is also ex- Monograph Number 1 - Second Edition
pressed to Jason Close (CMG) and Armando Montes (Univer- Horizontal Wells for the Recovery of Oil,

sity of Calgary) for reviewing the manuscript and providing their Gas and Bitumen: Monograph Number 2
comments.
A special acknowledgement is made to the members and ex-
members of the In Situ Combustion Research Group at the Uni-
versity of Calgary.

References Sproule is a diversified, worldwide petroleum consulting firm.


1. ABU-KHAMSIN, S.A., BRIGHAM, W.E. and RAMEY, Jr., H.J., We have been in business for over 50 years and have
Reaction Kinetics of Fuel Formation for In-Situ Combustion; SPE experience in all aspects of the energy sector throughout
Reservoir Engineering, Vol. 3, No. 4, pp. 1308-1316, November
1988. North America and the world.
2. URSENBACH, M.G., MOORE, R.G. and MEHTA, S.A., Air Injec-
tion in Heavy Oil Reservoirs—A Process Whose Time Has Come
(Again); paper 2007-217 presented at the Petroleum Society’s Cana-
dian International Petroleum Conference, Calgary, AB, 12-14 June
2007.
3. MONTES, A., MOORE, R.G., MEHTA, S.A., URSENBACH, M.G.
and GUTIERREZ, D., Is High Pressure Air Injection (HPAI) Simply
a Flue-Gas Flood?; paper 2008-180 presented at the Canadian In-
ternational Petroleum Conference/SPE Gas Technology Symposium
2008 (59th Annual Technical Meeting), Calgary, AB, 17-19 June Engineering | Geology | Geophysics
2008.
4. YANNIMARAS, D.V. and TIFFIN, D.L., Screening of Oils for In- Reserve Evaluations Resource Assessments
Situ Combustion at Reservoir Conditions by Accelerating-Rate Securities Commission Filings Geological/Petrophysical Evaluations
Calorimetry; SPE Reservoir Engineering, Vol. 10, No. 1, pp. 36-39, Mergers/Acquisitions/Dispositions Geophysical Interpretations
February 1995.
Fair Market Valuations Geo-Model Development
5. SARMA, H.K., YAZAWA, N., MOORE, R.G., MEHTA, S.A., OKA-
ZAWA, N.E., FERGUSON, H. and URSENBACH, M.G., Screening Heavy Oil/Mining & In-Situ Oilsands Unproved Land Valuations
of Three Light-Oil Reservoirs For Application of Air Injection Pro- Coalbed Methane/Shale Gas Reservoir Engineering
cess by Accelerating Rate Calorimetric and TG/PDSC Tests; Journal Field Development/Optimization Plans Reservoir Simulation
of Canadian Petroleum Technology, Vol. 41, No. 3, pp. 50-61, March Project Management Services Education Courses
2002.
Expert Witness Testimony Corporate Advice
6. SARATHI, P.S., In-Situ Combustion Handbook – Principles and
Practices; National Petroleum Technology Office, U.S. Department of
Energy, Tulsa, OK, January 1999. For more information please
7. MOORE, R.G., LAURESHEN, C.J., URSENBACH, M.G., MEHTA, call toll free 1-877-777-6135
S.A. and BELGRAVE, J.D.M., Combustion/Oxidation Behavior of or visit www.sproule.com
Athabasca Oil Sands Bitumen; SPE Reservoir Evaluation & Engi-
neering, Vol. 2, No. 6, pp. 565-571, December 1999.
8. MOORE, R.G., URSENBACH, M.G., LAURESHEN, C.J., BEL-
GRAVE, J.D.M. and MEHTA, S.A., Ramped Temperature Oxidation
Analysis of Athabasca Oil Sands Bitumen; Journal of Canadian Pe-
troleum Technology, Special Edition, Vol. 38, No. 13, pp. 1-10, 1999.
April 2009, Volume 48, No. 4 33
9. BURGER, J.G., Chemical Aspects of In-Situ Combustion – Heat of 28. GREAVES, M. and BENTAHER, A.H., Exothermicity Characteris-
Combustion and Kinetics; SPE Journal , Vol. 12, No. 5, pp. 410-422, tics of Wolf Lake Heavy Oil, Athabasca Tar Sand and Clair Medium
October 1972. Heavy Oil; Journal of Canadian Petroleum Technology, Vol. 46, No.
10. FASSIHI, M.R., BRIGHAM, W.E. and RAMEY, Jr., H.J., Reaction 9, pp. 16-19, September 2007.
Kinetics of In-Situ Combustion: Part 1 – Observations; SPE Journal, 29. FREITAG, N.P. and VERKOCZY, B., Low-Temperature Oxidation
Vol. 24, No. 4, pp. 399-407, August 1984. of Oils in Terms of SARA Fractions: Why Simple Reaction Models
11. KISLER, J.P. and SHALLCROSS, D.C., Oxidation Kinetics of Very Don’t Work; Journal of Canadian Petroleum Technology, Vol. 44, No.
Light Crude Oil; paper presented at the IEA Collaborative Project on 2, pp. 54-61, February 2005.
EOR, 16th International Workshop and Symposium, Chiba, Japan, 30. MATTAX, C.C. and DALTON, R.L., Reservoir Simulation; SPE
October 1995. Monograph Series, Vol. 13, Richardson, TX, 1999.
12. NELSON, T.W. and MCNEIL, J.S., How to Engineer an In Situ Com- 31. KOK, M.V. and KARACAN, C.O., Behavior and Effect of SARA
bustion Project; Oil & Gas Journal, pp. 58-65, 6 June 1961. Fractions of Oil During Combustion; SPE Reservoir Evaluation &
13. GATES, C.F. and RAMEY, Jr., H.J., A Method for Engineering Engineering, Vol. 3, No. 5, pp. 380-385, October 2000.
In-Situ Combustion Oil Recovery Projects; Journal of Petroleum 32. LI, J., MEHTA, S.A., MOORE, R.G., URSENBACH, M.G., ZAL-
Technology, Vol. 32, No. 2, pp. 285-294, February 1980. EWSKI, E., FERGUSON, H. and OKAZAWA, N.E., Oxidation and
14. KUMAR, M., Simulation of Laboratory In-Situ Combustion Data Ignition Behaviour of Saturated Hydrocarbon Samples With Crude
and Effect of Process Variations; paper SPE 16027 presented at the Oils Using TG/DTG and DTA Thermal Analysis Techniques; Journal
SPE Symposium on Reservoir Simulation, San Antonio, TX, 1-4 Feb- of Canadian Petroleum Technology, Vol. 43, No. 7, pp. 45-51, July
ruary 1987. 2004.
15. KUMAR, M., A Cross-Sectional Simulation of West Heidelberg In- 33. LI, J., MEHTA, S.A., MOORE, R.G., ZALEWSKI, E., URSEN-
Situ Combustion Project; SPE Reservoir Engineering, Vol. 6, No. 1, BACH, M.G. and VAN FRAASSEN, K., Investigation of the
pp. 46-54, February 1991. Oxidation Behaviour of Pure Hydrocarbon Components and Crude
16. MARJERRISON, D.M. and FASSIHI, M.R., A Procedure for Scaling Oils Utilizing PDSC Thermal Technique; Journal of Canadian Petro-
Heavy-Oil Combustion Tube Results to a Field Model; paper SPE leum Technology, Vol. 45, No. 1, pp. 48-53, January 2006.
24175 presented at the SPE/DOE Enhanced Oil Recovery Sympo- 34. MOORE, R.G., BENNION, D.W., BELGRAVE, J.D.M., GIE, D.N.
sium, Tulsa, OK, 22-24 April 1992. and URSENBACH, M.G., New Insights Into Enriched-Air In-Situ
17. MARJERRISON, D.M. and FASSIHI, M.R., Performance of Morgan Combustion; Journal of Petroleum Technology, Vol. 42, No. 7, pp.
Pressure Cycling In-Situ Combustion Project; paper SPE 27793 pre- 916-923, July 1990.
sented at the SPE/DOE Improved Oil Recovery Symposium, Tulsa, 35. JIA, N., MOORE, R.G., MEHTA, S.A. and URSENBACH, M.G.,
OK, 17-20 April 1994.
Kinetic Modelling of Thermal Cracking and Low Temperature Oxi-
18. AMBASTHA, A.K. and KUMAR, M., New Insights into In-Situ
dation Reactions; Journal of Canadian Petroleum Technology, Vol.
Combustion Simulation for Heavy Oil Reservoirs; paper SPE 56543
45, No. 9, pp. 21-28, September 2006.
presented at the SPE Annual Technical Conference and Exhibition,
36. SEQUERA, B., MOORE, R.G., MEHTA, S.A. and URSENBACH,
Houston, TX, 3-6 October 1999.
M.G., Numerical Simulation of In Situ Combustion Experiments
19. PASCUAL, M., CROSTA, D., LACENTRE, P. and COOMBE, D.,
Operated Under Low Temperature Conditions; paper 2007-189
Air Injection into a Mature Waterflooded Light Oil Reservoir. Labo-
presented at the Petroleum Society’s 8th Canadian International
ratory and Simulation Results for Barrancas Field, Argentina; paper
Petroleum Conference (58th Annual Technical Meeting), Calgary, Al-
SPE 94092 presented at the SPE Europec/EAGE Annual Conference,
berta, Canada, 12-14 June 2007.
Madrid, Spain, 13-16 June 2005.
37. CHRISTENSEN, J.R., DARCHE, G., DECHELETTE, B., MA, H.,
20. TINGAS, J., GREAVES, M. and YOUNG, T.J., Field Scale Sim-
and SAMMON, P.H., Applications of Dynamic Gridding to Thermal
ulation Study of In-Situ Combustion in High Pressure Light Oil
Simulations; paper SPE 86969 presented at the SPE International
Reservoirs; paper SPE 35395 presented at the SPE/DOE Improved
Oil Recovery Symposium, Tulsa, OK, 21-24 April 1996. Thermal Operations and Heavy Oil Symposium and Western Re-
21. KUHLMAN, M.I., Simulation of Light-Oil Air Injection Into Vis- gional Meeting, Bakersfield, CA, 16-18 March 2004.
cous-Dominated and Gravity-Stable Reservoirs; paper SPE 59331 38. DE ZWART, A.H., BAKKER, P., GLANDT, C.A., BROOKS, A.D.
presented at the SPE/DOE Improved Oil Recovery Symposium, Tulsa, and VAN DORP, J., A Thermal Recovery Method for Medium-Heavy
OK, 3-5 April 2000. Oil Reservoirs; paper SPE 112876 presented at the SPE North Africa
22. NIZ-VELASQUEZ, E., VAN-FRAASSEN, K., MOORE, R.G. and Technical Conference and Exhibition, Marrakech, Morocco, 12-14
MEHTA, S.A., An Experimental Study on Three-Phase Flow in High March 2008.
Pressure Air Injection (HPAI); paper 2008-091 presented at the Ca- 39. DE ZWART, A.H., VAN BATENBURG, D.W., BLOM, C.P.A.,
nadian International Petroleum Conference/SPE Gas Technology TSOLAKIDIS, A., GLANDT, C.A. and BOERRIGTER, P., The
Symposium 2008 (59th Annual Technical Meeting), Calgary, AB, Modeling Challenge of High Pressure Air Injection; paper SPE
17-19 June 2008. 113917 presented at the SPE/DOE Symposium on Improved Oil Re-
23. MENDEZ KUPPE, G.J., MEHTA, S.A., MOORE, R.G., URSEN- covery, Tulsa, OK, 20-23 April 2008.
BACH, M.G. and ZALEWSKI, E., Heats of Combustion of Selected 40. PANAIT-PATICA, A., SERBAN, D. and ILIE, N., Suplacu de Barcau
Crude Oils and Their SARA Fractions; Journal of Canadian Petro- Field—A Case History of a Successful In-Situ Combustion Exploita-
leum Technology, Vol. 47, No. 1, pp. 38-42, January 2008. tion; paper SPE 100346 presented at the SPE Europec/EAGE Annual
24. MOORE, R.G., LAURESHEN, C.J., MEHTA, S.A. and URSEN- Conference and Exhibition, Vienna, Austria, 12-15 June 2006.
BACH, M.G., Observations and Design Considerations for In Situ ��������������������������������������������������������������
41. DORAIAH, A., SIBAPRASAD, R. and GUPTA, P., In-Situ Combus-
Combustion Projects; Journal of Canadian Petroleum Technology, tion Technique to Enhance Heavy-Oil Recovery at Mehsana, ONGC
Special Edition, Vol. 38, No. 13, pp. 1-9, 1999. – A Success Story; paper SPE 105248 presented at the SPE Middle
25. BELGRAVE, J.D.M., MOORE, R.G., URSENBACH, M.G. and East Oil and Gas Show and Conference, Kingdom of Bahrain, 11-14
BENNION, D.W., A Comprehensive Approach to In-Situ Combus- March 2007.
tion Modelling; SPE Advanced Technology Series, Vol. 1, No. 1, pp.
98-107, 1993.
26. FLORES CABRERA, C., GONZALES GUEVARA, J.A., MAR-
TINEZ GARCIA, B., FLORES AVILA, F. and BERLANGA Provenance—Original Petroleum Society manuscript, The Challenge of
GUTIERRES, J.M., Inyección de Aire para la Incorporacion de Predicting Field Performance of Air Injection Projects Based on Labo-
Reservas en Yacimientos Naturalmente Fracturados; paper presented ratory and Numerical Modelling (DAS-2008-158), first presented at the
at the Second International Conference and Exhibition of Petroleum, 9th Canadian International Petroleum Conference (the 59th Annual Tech-
Veracruz, Mexico, 28-30 June 2007. nical Meeting of the Petroleum Society), June 17-19, 2008, in Calgary,
27. SUTHERLAND, R.B., HAMMAWA, H., MOORE, R.G., MEHTA, Alberta. Abstract submitted for review January 7, 2008; editorial com-
S.A. and URSENBACH, M.G., Mitigating Explosion Risks in High ments sent to the author(s) February 4, 2009; revised manuscript received
Pressure Air Injection Compressors; Journal of Canadian Petroleum February 9, 2009; paper approved for pre-press February 9, 2009; final ap-
Technology, Vol. 46, No. 8, pp. 55-61, August 2007. proval March 2, 2009.
34 Journal of Canadian Petroleum Technology

S-ar putea să vă placă și