SPE 165429

The Role of the Phase Behavior Model within the Numerical Simulation of
In Situ Combustion Processes
J. M. Palma, SPE; E. Rodrguez, SPE; Ecopetrol.

Copyright 2013, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Heavy Oil Conference Canada held in Calgary, Alberta, Canada, 1113 June 2013.

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

In situ combustion is a method with a large potential for recovery of heavy oil reserves, worse yet has many complexities
that make it difficult its field scale development. One of these difficulties is precisely understand the different kinds of
reactions taking place once the process starts. One tool that helps the engineer to understand what is happening during the
process is the numerical simulation and specifically, for the case of combustion, the kinetic model. This model has two
essential parts: the reactions and kinetic parameters. The reactions are developed based on the pseudocomponents used to
characterize the fluid and the kinetic parameters were obtained from measurements made in the laboratory tests: Ramped
temperature oxidation (RTO) and Accelerating Rate Calorimeter (ARC). The initial characterization of a fluid throws several
components, which are subsequently lumped to the least amount possible without losing crude representation. These
pseudocomponents must correctly represent the PVT behavior of oil. This matching to the actual behavior is achieved using
suitable software. Based on the set of tests, it stars to match the behavior of the main fluid properties, for this, have
previously chosen the most relevant parameters that may require matching to obtain a more representative model. Within the
main tuning parameters are the critical properties of the fluid, accentric factors, lumping and splitting schemes and viscosity
coefficients.

This paper proposes a scheme that helps to build a model that represents the phase behavior when the fluid is subjected to
various conditions during the development of the combustion process. Matched the fluid model, this can be used in different
scale models: CTO(Combustion Tube Tests), RTO or field. In this case, the successful development of the reactions involved
in the kinetic model for in situ combustion process is linked the phase behavior model and the molecular weight of the
pseudocomponents.

Introduction

Numerical simulation is a useful tool in the previous study the performance of an enhanced oil recovery technique. In the
case of an air injection process, where the process becomes difficult to represent from physical scale models, since it is
tedious and complicated represent the different mechanisms involved in the process, further gain importance using this
technique. For proper development of a simulation model, it is necessary to properly characterize the fluid of the field for
representing, considering the tests made for it. The phase behavior model is developed from the PVT analysis of it and then
be exported to the simulator STARS from the CMG company.

It is necessary define the number of pseudocomponents to use for representing the crude oil, considering using the fewest
possible without loosing representativeness. In this stage is critical stablish the lumping of that pseudocomponentes having
account the equilibrium constants for each one. Once developed the phase behavior model, proceed to generate the model of
reactions, using an appropriate number to represent most possible processes involved in the combustion process. To load the
kinetic parameters are initially used data reported in the literature and then make the model fit. A representative phase
behavior model and a reaction scheme consistent with the process will assess the application scenarios of the situ combustion
which leads to increase the recovery factor and performance optimization of the technique.
2 SPE 165429

In situ combustion

In situ combustion is an enhanced oil recovery technique that identifies not only as a thermal process, but contains various
processes such as displacement with steam, carbon dioxide stream, oil flow miscible, immiscible displacement gas (N2) and
water displacement (hot and cold), which allows to recover much of the heavy oil present in a reservoir after completion of
primary production and / or secondary. It is based on the injection of air for heat generation in the reservoir through a
combustion front, thus avoiding energy losses by making the pipe and reducing the oil viscosity, allowing this to flow easily
toward the producing well (Burger, 1985).

The in situ involves combustion process consist in injecting air or oxygen-containing gas into the formation to create a
temperature rise in the reservoir (Figure 1). Upon contact with the hydrocarbon gas injected, triggers a series of oxidation and
pyrolysis reactions that release energy and provide the conditions for generating a combustion front either natural or artificial.
In oxidation reactions, the injected air reacts with oil in situ to the first trigger low temperature reactions (LTO, Light
Temperature Oxidation) and pyrolysis reactions, which will produce coke or fuel for combustion, later as the temperature
increases occur breaking reactions or high temperatures (HTO, Hight temperature Oxidation) in which fuel is being
consumed generating this combustion front (Sarathi, 1999).

Once ignition has been achieved, the combustion front begins its slow movement through the undisturbed reservoir portion
toward producing wells. As the front progresses, it causes a decrease in the oil viscosity and a partial improvement of their
properties, as known upgrading process, a process that undergoes the oil is a consequence of the reactions of pyrolysis and
oxidation at high temperatures above (Ursenbach, 1993).

Figure 1. In situ combustion process Scheme

A complex process such as in situ combustion requires a careful and detailed study of all effects and phenomena that occur
during development. So in the laboratory are several tools and parameters that have to set some parameters and the
relationship between them, so as to allow having a clearer idea of the process behavior. The techniques generally can be
classified as: quantitative and qualitative. The information gained from these techniques is of great importance as the amount
of available fuel and air requirements, vital to the technical and economic evaluation of the process. For this case, will take
into account two quantitative tests: the combustion tube tests (CTO) and the Ramped Temperature Oxidation (RTO).

The combustion tube tests allow a representation of the behavior of a combustion process in the reservoir, representing the
flow or combustion front propagation. Such tests are not scalable, ie, can not be directly field case, but give an indication of
the process behavior. The stoichiometry of the reactions in a combustion tube test is the same of the reservoir (figure 2). In
the case of the RTO, sometimes called effluent gas analysis (EGA), studied the oxidation behavior of the reaction kinetics
and a rock-fluid system under controlled conditions (figure 3). To practice it is necessary to use a reference cell, dry core,
comparing the temperatures obtained in the test cells and to establish reference conditions endothermic and exothermic
reactions (Joya et Al, 2012).
SPE 165429 3

The above tests can represent them numerically, but to be valid it is necessary to adjust the variables involved in the
simulation of these, because the kinetic parameters obtained from fitting the models help to better represent the combustion
process in a larger model (field model). To achieve this adjustment is necessary initially from a base model that allows
setting a standard of comparison and then evaluating the differences between this model and the model results to variations
raised. Here we review the aspects concerning the simulation of combustion processes in situ.

Figure 2. Combustion tube tests scheme.

Figure 3. Ramped temperature oxidation tests scheme.

Numerical simulation for combustion tube tests

The in situ combustion is a complex process for representing by physical model at laboratory scale, so a tool such as
numerical simulation, but not definitive, it is appropriate to give an idea of the performance of the technique. However, a
correct evaluation through the simulation depends heavily on a model suitable fluid that allows it to represent the behavior of
different conditions.

The combustion tube tests are highly useful as a tool for establishing a preliminary feasibility of developing an air injection
process at a specific site. It is possible to represent different mechanisms occurring in the in situ combustion process, giving
an idea of the process behavior. However, success of the technique in the laboratory does not mean success at the field level,
4 SPE 165429

but very likely if laboratory does not have good results in the field is not displayed. However, these tests are quite expensive
and are also associated with the attainment of a large amount of rock and fluid sample for the evaluation, so to get the results
of the test; you may initially using numerical simulation results for representing it obtained in the laboratory, and also raise
new scenarios test development.

A suitable simulation model must represent the most part, characteristics of the combustion tube tests. In the case of physical
dimensioning is used a grid withat least two cylindrical rings: one to represent the fluid-rock interaction and one for the
external coating. Figure 4 shows a simulation model for a combustion test tube (Rodriguez et al, 2011).

0.5 Ft

6 Ft

Figure 4. Simulation model for Combustion tube tests

It needs to define the properties tmicas for rock and fluid, as well as the outer ring of the model. Likewise, the petrophysical
properties are loaded to the model: porosity, permeabilities, etc. However, the phase behavior model who will focus more
attention due to the strong relationship between this and the development of the kinetic model.

Development of the phase behavior model

Previously, we highlighted the importance of having a good phase behavior model for simulating the in situ combustion
processes, because this was the basis of the kinetic model, which is responsible for representing all reactions occurring in the
process. Based on information gathered from PVT tests conducted at fluid in study, and using the right software, we proceed
to describe the development of model fluids. Once displayed known or reactions occurring in the process, it is necessary to
characterize the fluid, so the components are involved in the development of these.

Based on the composition given for the fluid, the first thing to do is adjust the saturation pressure, if the simulated value, do
not match with the reported, proceed to adjust using regression parameters such as the critical pressure, critical temperature
or molecular weight of the fraction or component that has uncertainty, the plus component in this case (Matinez & Vega,
2012). After matching the saturation pressure, is coming to group all the components to leave five pseudocomponents namely
N2, CO2, C1-C5, C6-C20 and C21-C36 +. This lumping can be done either on the basis of information gathered in literature
or observing the behavior of the K-values, lumping the similar components. Once the components are lumped, it is necessary
to check that the saturation pressure is matched, therefore runs the model and verifies the resulting saturation pressure. If this
setting is off, you need to regression of critical parameters and coefficients Omega A and B for achieve again setting this
property. Up to this point has been successfully matched the saturation pressure, now it need to match the properties of the
fluid obtained from PVT tests such as differential liberation or constant composition expansion (CCE and DL) and separator
tests. It is also necessary for the matching, have a function of viscosity in terms of pressure and temperature.
You start by loading the data for differential liberation test, these data are: oil volume factor, specific gravity of gas and oil
and GOR for three pressure levels. Subsequently load test separator that helps to adjust the behavior of Rs (gas in solution)
and oil API. A Constant Composition Expansion test is also loaded for adjustment of the oil density and the relative volume
of oil (ROV). It can be seen in figure 5 that the value of saturation pressure was no maladjusted (143.7 psia).
SPE 165429 5

Figure 5. Saturation pressure for the fluid.

In the first results can be demonstrated that the simulated properties are not consistent with the experimental results, so it is
necessary to perform a regression to adjust behavior. For this setting are selected tuning parameters in this case are: Volume
shift and the critical properties of the component that is uncertain, in this case one that includes the heaviest fraction. Other
parameters that may be useful to achieve the setting time are the molecular weights of each component, and the parameters
Omega A and B. It is clear that the latter should not be used simultaneously with the critical properties incur due to an error
by creating an endless cycle of calculation of these properties. The matched behavior for Relative Oil Volume (ROV) and gas
density are shown in the figures 6 and 7.

3.50 CCE Calc. - After Regression

3.00
Relative Volume

2.50

2.00

1.50

1.00

0.50
0 1000 2000 3000 4000 5000 6000
Pressure (psia)
ROV Psat Exp. ROV

Figure 6. Relative oil volume

6 SPE 165429

Figure 7. Gas density

Once matched these properties, it is necessary to adjust the viscosity, which have different tuning parameters, work
independently. In fact setting for this tool, when not initially achieved should be checked that the viscosity adjustment
parameters are not reaching the limits. If so, they should be modified to allow the matching. Figure 8 shows the viscosity
behavior of the set.

Figure 8. Matching for oil viscosity

In general, the procedure would be:

1. Match the oil composition

2. Identification of the properties and experimental tests
3. Load on suitable simulator
4. Evaluate initial fluid behavior in terms of its components
SPE 165429 7

5. Match the saturation pressure having all components

6. Lump the comopnents into a number of defined pseudocomponents established by any of the above criteria
7. Check and adjust saturation pressure once grouped components
8. Match the behavior and PVT properties
9. Match the viscosity behavior

Matched the phase behavior model properties, this is migrated to a thermal simulator precisely to the cylindrical grid for
representing the combustion tube tests. Once migrated, be careful that not been modified fluid properties. Figure 9 Display
currently behavior for the oil volume factor and viscosity.

Figure 9. Oil volume Factor and oil viscosity.

8 SPE 165429

Development of the kinetic model

The kinetic model is that part of the material balance that represents the speed and extent of transformation of different
components involved in chemical reactions, based on Arrhenius equations (Rodrguez et al, 2012). In the study developed by
Ordoez et al (2012), it was presented a workflow for the construction of a simulation model for in situ combustion (figure
10). As shown, obtaining a kinetic model is essential for a correct representation of the process. This model should contain
the reactions to occur in the process, as well as kinetic parameters that feed: activation energy, reaction enthalpy and
frequency factors.

Figure 10. General scheme for defining the design basis for the field pilot

During development of the phase behavior model, it must identify the number of pseudocomponents to use, they should
contain at least: one for the light fraction, another for heavy fraction, one representing the gas, and additionally the gaseous
phase representing no hydrocarbon gas. Each must define its molecular weight properly, so that follow represent PVT
properties, and must not alter the original saturation envelope is made once and matched adter the lumping. After establishing
the pseudo-component description, equation-of-state modeling is then performed to determine component properties such as
viscosity, density, and k-values. It is important that the modelling is carried out compositionally, to represent the change of
the fluid and each of its components when it is subjected to different temperature and pressure conditions. In reviewing the
behavior is observed, these data are consistent with the experimental, so the model adequately represents fluid properties
(Rodriguez et al,2012). Now the next step is to add a set of reactions that represents the process of combustion. For this case
it is used an addition reaction or low temperature (LTO) (1) a pyrolysis or cracking (intermediate temeperatura ITO) (2) and
two bond scission reaction or high temperature (HTO) (3 & 4). These reactions are stated above and it is clear that in their
formulation, they must be developed in terms of the same pseudocomponents characterized in that fluid.

Before adding the reactions, it is necessary to check whether the model already have all the components, both reactants and
products, if any of them is missing, it must add and set the stage to which they belong. It can also define which phases will be
burned during the process, being here the two heavier fractions of hydrocarbons. Reactions that were raised in this case are:
SPE 165429 9

c5 c 20 AO2 BC 21 C 30 (1)

C 21 C 30 Cc5 c 20 DCoque (2)

c 21 c30 EO2 FCoque (3)

GO2 coque JH 2 O CO2 (4)

Where A, B, C, D, E, F, G, J are the stoichiometric coefficients previously calculated to balance the reactions. These
coefficients are calculated taking into account the molecular weight defined for each pseudocomponent and in turn the
composition that was used when developing the phase behavior model.

Simulation Events and History Matching

The next step is adding the simulation events: inert gas injection (nitrogen in this case), air injection and the injection and
production constraints. Likewise, it needs to have the history files for temperature profiles, gas composition and the fluids
accumulated productions. Thus, it is possible to compare the response that is throwing the simulation model against
experimental data y verificar que el comportamiento del modelo de fluidos es adecuado para representar el proceso. Figures
11-13 show the temperature profiles, gas composition and cumulative productions respectively.

Figure 11. Temperature profile for combustion tube model simulation

10 SPE 165429

Figure 12. Gas composition for combustion tube model simulation

Figure 13. Cumulative fluids production for combustion tube modelo simulation
SPE 165429 11

In the case of flue gas composition, it observes that the oxygen is close to zero, which means their consumption in developing
reactions. Now as the oxygen can be consumed in other reactions and not in the combustion, is necessary verifying the
concentration of the CO2. For a high temperature system or combustion, it is expected that the composition of CO2 in the
outlet is greater than 12%, which in this case. In the case of profiles of temperature, it is seen the progression of peaks,
showing that the front is moving in the reservoir

As can observer, it is possible to represent the historical performance of the test, although this is not always possible in a first
attempt, so the model must be adjusted to obtain an adequate representation of the process (Palma 2011). Among the
principals matching parameters are:

Relative permeability curves (gas-liquid, water-oil)

Properties of the outer ring in the simulation model
Kinetic parameters: frequency factor and activation energy (only when there is uncertainty about them)
The reaction enthalpy to be an intensive property can not be regarded as a matching parameter, except that it has not
been determined with a high degree of certainty.

Once the model is matched, it can be used in other models like the field-scale model, but it is necessary to check whether or
not there is a change in the properties, to be migrated to a model with much larger cells ( Rodriguez et Al, 2011).

Conclusions

The fluid must be represented in terms of these pseudocomponents that will be used in the reactions of combustion
model.

It is important to amatch the viscosity separately from other properties, because this has numerical dependence of other
parameters.

The temperature distribution, the advancing front and flue gas composition are indicative in order to infer whether there
is a good response of the fluid and the environment to the development process.

A greater number of tests and reported properties allow a better characterization of the fluid under study. It is at the
discretion of the engineer to establish in what order and what develops adjustment parameters used to represent
adequately the behavior.

Adknowledgments

The authors wish to thank and to the Colombian company Petrleos Ecopetrol s.a. for allowing the publication of this paper
and the society of engineers of petroleum (SPE) to provide the space and opportunity for presentation and publication.

References

1. BURGER, J. SORIEAU, P. Thermal Methods of oil Recovery. Texto, Editions Technip , Pars 1985.

2. SARATHI S. P. In situ combustion Handbook. Principles and Practices. BDM Petroleum Technologies. National Petroleum
Technology Office. Oklahoma. January 1999. pp 2-5

3. URSENBACH M,G. An overview of laboratory in situ combustion Behavior of Athabasca oil sands. University of Calgary. 1998

4. JOYA D, PALMA J. Determinacin de parmetros cinticos a partir de pruebas RTO para un proceso de combustin in situ, Tesis de
pregrado, UIS, 2012.

5. RODRIGUEZ E, ORDEZ A, COMAS J, BELGRAVE J, ALVAREZ C superando retos en el modelamiento numerico de

procesos de combustion in situ en yacimientos de crudo pesado. XIV Congreso Colombiano del Petrleo organizado
por ACIPET en Bogot D.C. Colombia, 22 - 25 de Noviembre de 2011.
12 SPE 165429

6. RODRGUEZ E, TRUJILLO M, BELGRAVE JOHN, COMAS J, A Framework for Consolidating Air Injection Experimental Data,
SPE 152048, 2012.

7. ORDEZ A, TRUJILLO M, RODRGUEZ E, BELGRAVE JOHN, BOTTIA H, LOBO A, COMAS J, FERNANDEZ F; ECO-
GSAI "gravity stable air injection" pilot test, an enhanced oil recovery opportunity to be implemented in Chichimene field
Colombia, SPE 154255, Tulsa, 2012.

8. MARTNEZ J, VEGA S; Anlisis y desarrollo de modelos de fluidos asociados a la simulacin numrica de procesos de combustin
in situ, Tesis de pregrado, Grupo de Investigacin de Recobro Mejorado, Universidad Industrial de Santander, 2012.

9. PALMA J, Procedimiento para el ajuste de variables involucradas en procesos de combustin in situ, Tesis de maestra, Grupo de
Investigacin de Recobro Mejorado, Universidad Industrial de Santander, 2011.