Volume Issue 0 2000 [Doi 10.2118%2F63230-Ms] Serhat, Akin_ Kok, Mustafa_ Suat, Bagci_ Ozgen, Karacan -- [Society of Petroleum Engineers SPE Annual Technical Conference and Exhibition

7/28/2019 Volume Issue 0 2000 [Doi 10.2118%2F63230-Ms] Serhat, Akin_ Kok, Mustafa_ Suat, Bagci_ Ozgen, Karacan -- [Soc…

http://slidepdf.com/reader/full/volume-issue-0-2000-doi-1021182f63230-ms-serhat-akin-kok-mustafa-suat 1/12

Copyright 2000, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the 2000 SPE Annual Technical Conference andExhibition held in Dallas, Texas, 1–4 October 2000.

This paper was selected for presentation by an SPE Program Committee following review ofinformation contained in an abstract submitted by the author(s). Contents of the paper, aspresented, have not been reviewed by the Society of Petroleum Engineers and are subject tocorrection by the author(s). The material, as presented, does not necessarily reflect anyposition of the Society of Petroleum Engineers, its officers, or members. Papers presented atSPE meetings are subject to publication review by Editorial Committees of the Society ofPetroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paperfor commercial purposes without the written consent of the Society of Petroleum Engineers isprohibited. Permission to reproduce in print is restricted to an abstract of not more than 300words; illustrations may not be copied. The abstract must contain conspicuousacknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

AbstractIn situ combustion is a thermal recovery technique whereenergy is generated by a combustion front that is propagatedalong the reservoir by air injection. Most of the previouslyconducted studies report thermal and fluid dynamics aspects of the process. Modeling in situ combustion process requiresextensive knowledge of reservoir data as well as reactionkinetics data. Unfortunately, limited kinetic data are availableon the rates and the nature of partial oxidation reactions andthe high-temperature combustion reactions of crude oils andtheir saturate, aromatic, resin, and asphaltene (SARA)fractions. Moreover, the impact of such data on the modelingof the in situ combustion process has not been investigatedthoroughly. Thus, we modeled in situ combustionexperiments conducted on a 3D semi-scaled physical modelthat represents one fourth of a repeated five spot pattern. In allexperiments a vertical injector is employed whereas, bothvertical and horizontal producers have been installed torecover two different crude oils (heavy and medium). Severallocations for the producers have been tried while keeping the

length of the wells constant: vertical injector-vertical producer,vertical injector-horizontal side producer, and vertical injector-horizontal diagonal producer. In these experiments diagonalproducers performed better than the others. We first simulatedthe experiments by incorporating a kinetic model that is basedon grouping the products of cracking into six pseudocomponents as heavy oil, medium oil, light oil, two non-condensable gases and coke using a commercial thermalsimulator (CMG’s STARS). Four chemical reactions wereconsidered: cracking of heavy oil to light oil and coke, heavyoil burning, light oil burning, and coke burning. Most of the

experiments were history matched successfully with theexception of ones where a diagonal horizontal producer wasused. We then repeated the simulations using SARA kineticparameters. We observed that all matches were somewhatimproved. We finally present a discussion of application of the models to field scale.

IntroductionIn-situ combustion is an important enhanced oil recoveryprocess that has been studied extensively the past 45 years.This process has been considered particularly applicable forin-situ recovery of medium and heavy oil reservoirs. In in-situcombustion, heat is generated within the reservoir by ignitingthe formation oil and then propagating a combustion frontthrough the oil reservoir. The fuel necessary to sustain thecombustion front is supplied by the heavy residual material or“coke” that deposits on the sand grains during distillation,thermal cracking, pyrolysis etc. of the crude oil ahead of thecombustion front.

Sweep efficiency during in-situ combustion is one of themost important process parameters, but which has not beenextensively evaluated and is least understood. Most laboratoryinvestigations are conducted in combustion tubes, whichessentially use a vertical well arrangement and which, of course, because of their basically one-dimensional geometry,cannot provide information on either areal or vertical sweep.Information on the combustion sweep efficiency is veryimportant for comparing process variations and also forpredicting performance. 3-D scaled physical models canprovide much valuable insight into the areal and verticalsweep processes and stability of the combustion front over arange of operating conditions. Results from such experiments

may be used in conjunction with those from combustion tubetests to predict performance in the field, and also to validatenumerical simulator models.

Accuracy of field simulation results for in situ combustionis not well established in the literature. Often, one starts witha reasonable history match of laboratory in situ combustiondata. This matching procedure requires data on the the ratesand the nature of partial oxidation reactions, the high-temperature combustion reactions of crude oils and theirsaturate, aromatic, resin, and asphaltene (SARA) fractions.Laboratory data are usually matched using 1-D process

SPE 63230

Oxidation of Heavy Oil and Their SARA Fractions: Its Role in Modeling In-SituCombustionSerhat Akin, Mustafa V. Kok, Suat Bagci, Middle East Technical University, and Ozgen Karacan, PennState



2 AKIN, KOK, BAGCI, KARACAN SPE 63230

description with grid block sizes of the order of centimeters.However, field simulation requires 3-D process descriptionwith grid block sizes of the order of tens of meters.

This paper presents experimental data conducted withvarious injection-production well couples (vertical injector-horizontal side producer, vertical injector-diagonal horizontalproducer, vertical injector-vertical producer) in a 3-D modelusing two different heavy crudes. In these experimentshorizontal side producers performed better than the others.We first simulated the experiments by incorporating a kineticmodel that is based on grouping the products of cracking intopseudo components as heavy oil, medium oil, light oil, twonon-condensable gases and coke using a commercial thermalsimulator (CMG’s STARS). Several different chemicalreactions were considered: cracking of heavy oil to light oiland coke, heavy oil burning, light oil burning, and cokeburning. Most of the experiments were history matchedsuccessfully with the exception of ones where a diagonalhorizontal producer was used. We then repeated thesimulations using SARA kinetic parameters that were obtainedusing thermogravimetric (TG) and differential thermalanalysis (DTA) equipment with the hope of improving thehistory matches. We observed that all matches weresomewhat improved. We finally present a discussion of application of the models to field scale.

Literature SurveyBinder et al 1 conducted combustion experiments in severalmodels of massive, unconsolidated reservoirs containing 2100cp. oil and with properties generally lending themselves tofavorable burning behavior. These experiments indicate thatrecovery levels of 60-70 % pore volume might be obtainedeconomically with a well spacing-sand thickness ratio of 3 andthick, shale-free sand. Experiments with a higher spacing-thickness ratio of 7, both with and without restrictions tovertical permeability, indicated that performance is adverselyaffected by thinner sands and by the presence of discontinuouspermeability barriers simulating interbedded shale. Garon et.al2 developed a facility for simulating fireflooding processes in3-D model. Results from 3-D model tests provided basicunderstanding of the mechanisms of in-situ combustion innonhomogeneous reservoirs. In a reservoir heated from belowby steam injection into a noncommunicating bottom waterzone, a fireflood moves rapidly through the heated layer withpoor vertical displacement if the air injection rates are toohigh. Fireflooding at relatively low air injection rates was veryeffective in a reservoir with a simulated heated fracture at themidplane. Much more oil was swept from above the fracturethan from below, and little or no oil was recovered frombeyond the areal extent of the fracture.

Garon et. al 3 conducted three dimensional (3D) scaledmodel experiments to investigate the volumetric sweep duringfireflooding. The effect of oxygen vs. air injection,water/oxygen ratios, injection rates and crude oil parameterson sweep efficiency and performance of the fireflood wereevaluated. Results indicate that the sweep of a fireflood was

similar for both oxygen and air combustion; water injectionresulted in a small decrease in the sweep of the fireflood; wetcombustion required less oxygen or air and increased the oilrecovery and recovery rate; fireflooding a medium-gravitycrude oil reservoir resulted in a larger sweep than a heavy oilreservoir; and higher injection rates improved the sweepefficiency.

Greaves et. al 4 used a semi-scaled 3-D physical model toinvestigate the process of dry forward in situ combustion usingdifferent configurations of horizontal producer wells. A higheroverall rate of oil production and higher recovery of OOIP wasachieved with both single and double horizontal wellconfigurations compared to that achieved with a single verticalproducer well. Cumulative oil recovery increased with thenumber of horizontal producer wells, but the net recovery perwell decreased compared to the single horizontal well case.The tendency for gas override was greatly reduced by theaction of the horizontal producer well, so that the verticalsweep efficiency of the combustion front was considerablyimproved compared to that for the single vertical producerwell.

Coates et. al 5 studied a new in-situ combustion strategy,top down process under detailed laboratory study. The time toreach ignition after commencing air injection was highlydependent on the degree of pre-heating. After a steam processin an Athabasca formation, there was enough fuel remaining toinitiate and sustain an in-situ combustion process. Wetcombustion showed the potential to increase production rate if it was commenced before pack has become depleted. The topdown in-situ combustion process produced bitumen with alower viscosity than the original native bitumen. The processresulted the stable propagation of a combustion front from thetop to the bottom of a reservoir, exploiting gravity drainage of the mobilized oil to a lower horizontal well.

Greaves and Al-Shamali 6 conducted a series of dry in-situcombustion experiments, and a single wet combustionexperiment on heavy Wolf Lake crude oil using a rectangular3-D combustion cell. High oil recoveries were achieved duringdry combustion, ranging from 64.3 to 72.3 % OOIP, and 78.8% OOIP during wet combustion. Gas override condition wasnot a major problem using the horizontal producer well indirect line drive. The volumetric sweep efficiencies calculatedfrom the vertical and horizontal temperature profiles weregenerally in good agreement with the measured oil recoveryvalues, indicating almost complete recovery of oil from theswept regions of the sandpack.

Greaves and Mahgoub 7 investigated low-pressure airinjection/in-situ combustion in waterflooded light oilreservoirs in a 3-D combustion cell using a horizontalproducer well in line drive. Oil recovery varied from 63.9 to85 % OOIP with fuel consumption varying from 3.5 to 4.5 %OOIP. These values indicate that improved oil recovery fromwaterflooded light oil reservoirs using in-situ combustion isnot limited by any apparent low fuel availability, but rather theintrinsic ability to sustain a stable combustion front at asufficiently high temperature.



SPE 63230 OXIDATION OF HEAVY OIL AND THEIR SARA FRACTIONS: ITS ROLE IN MODELING IN-SITU COMBUSTION 3

Greaves et. al 8 used a one-quarter of a 5-spot physicalmodel to carry out a series of low pressure air injection/in-situcombustion tests on light and medium heavy crude oils toinvestigate the effect of reservoir heterogeneity on oilrecovery. It was found that a horizontal producer well does notact as a high permeability streak, but actually increases andbrings forward production.

Although kinetic studies on both oil samples and oil-rock mixtures had been conducted there are almost no, to ourknowledge, or few studies governing the behavior and kineticsof crude oil SARA fractions under oxidizing environment.Verkocy and Jha 9 performed thermogravimetric anddifferential thermogravimetric (TG/DTG) experiments onheavy oils and cores. They have estimated kinetic and thermo-chemical data for low temperature oxidation, cracking, cokingand combustion reaction in cores and oils. Yoshiki andPhilips 10 used DTA and TG/DTG at high temperatures andpressures and concluded that, both low temperature oxidationand high temperature oxidation rates increased with pressure,as did the exothermicity of each. Bae 11 studied the effect of pressure and observed that the results were oil specific, but ingeneral increase in pressure causes the low temperature heatgeneration to increase. Drici and Vossoughi 12 applieddifferential scanning calorimetry (DSC) and TG/DTG to crudeoil combustion in the presence and absence of metal oxides.They observed that in the presence of a large surface area suchas silica, the surface reactions are predominant and unaffectedby the small amount of metal oxide present. Del Bianco et al 13

used a vacuum residue of crude for thermal cracking in orderto define a kinetic scheme and to calculate the kineticparameters of thermal cracking of petroleum residues. Theyconcluded that distillate production can be described as asimple first-order reaction while coke formation seems to bethe consequence of consecutive reactions, in particular thoseinvolving asphaltenes. Yoshida et al 14 investigated the thermalbehavior of coal derived asphaltenes by TG/DTG. In theirwork, they conclude that weight loss is rapid from 300 to 500oC and is slow above 500 oC. They also state that theasphaltene that comparatively low in molecular weight showsgreater weight loss. Ciajolo and Barbella 15 usedthermogravimetric techniques to investigate the pyrolysis andoxidation of some heavy fuel oils and their separate paraffinic,aromatic, polar and asphaltene fractions. They also found thatthe thermal behavior of fuel oil can be interpreted in terms of low temperature phase involving the volatilization of paraffinic and aromatic fractions, and a high temperaturephase in which the polar and asphaltene fractions pyrolyse andleave a particulate carbon residue. Ranjbar and Pusch 16

studied the effect of oil composition, characterized on thebasis of light hydrocarbons, resin and asphaltene contents andthe pyrolysis kinetics of the oil. The results indicate that thecolloidal composition of oil as well as the transferability andheat transfer characteristics of the pyrolysis medium has apronounced influence on fuel formation and composition. Aliand Saleem 17 investigated the asphaltenes precipitated fromcrude oils by thermogravimetric analysis and pyrolysis-GC

analysis. Under severe pyrolysis conditions 98-100% of asphaltenes are converted to the products. The evaluation of methane and other normal alkanes from all the asphaltenesunder mild pyrolysis conditions indicates that theseasphaltenes contain thermally labile alkyl groups on theperiphery of these asphaltenes. Kok 18 characterized thcombustion properties of two heavy crude oils by DSC andTG/DTG. On combustion with air, three different reactionregions were identified, known as low temperature oxidation,fuel deposition and high temperature oxidation. Heat valuesand reaction parameters are also obtained from DSC andTG/DTG experiments.

Experimental Equipment and ProcedureIn the experimental work, steel 3-D in-situ combustionexperimental equipment with dimensions of 40x40x15 cm thatis comprised of three major components (fluid injection, fluidproduction, and measuring and control system) was used (Fig1). Combustion experiments were conducted by employingfive different well configurations shown in Fig 2. Perforatedstainless steel tubing of 8 mm diameter was used as injectionand production wells. The details of the experimentalequipment are given elsewhere 19.

Figure1. Schematic drawing of the experimental setup.

Figure 2. Schematic drawing of experimental well configurations.




The SARA fractions of the crude oils were obtained byusing thermo gravimetric analysis (TG/DTG) equipmentdeveloped by Dupont. TG/DTG has the capability of measuring the weight loss either as a function of temperatureor time in a varied but controlled atmosphere. Prior to theexperiments TG/DTG system was calibrated with calciumoxalate monohydrate for temperature readings and silver wasused in order to correct for buoyancy effects.

Experimental ProcedureThe physical properties of the porous media and the operatingconditions (Table 1) were kept constant in each experiment.Prior to ignition, the model packed with crushed limestonewas heated with the aid of band heater until the temperatureinside the model reached to 50 °C that was presumed to be thereservoir temperature. During this period nitrogen wascontinuously injected to prevent low temperature oxidation.After pre-heating, igniter was switched on to heat thecombustion cell to 500 °C with the aid of a temperatureprogrammer. At this instant, nitrogen flow ended and airinjection commenced.

Table 1. Experimental packing data and operating conditions.Raman rm1 rm 2 rm 3 Rm 4 rm 5 rm 6 rm 7

Config. vi-vp vi-hp-d vi-hp-r vi-hp-l vi-2hp vi-vp vi-hp-d

K, D 9.8 9.8 9.8 9.8 9.8 9.8 9.8φ, % 38 38 38 38 38 38 38So,% 73.5 75 75 75 71.0 72.2 73Sw,% 25 25 25 25 25 25 25Sg, % 1.5 0.0 0.0 3.99 2.82 2.0 5.64Voil, cc 6705 6850 6850 6477 6583 6663 6326Vwater , cc 2280 2280 2280 2280 2280 2280 2280

PInj, Kpa 300 294 290 250 210 205 230QAir, m

3 /m 2-hr 8.35 8.45 4.61 2.40 2.88 1.68 1.84Q Inj.m

3 /hr 0.626 0.634 0.348 0.180 0.216 0.126 0.138

B.Kozluca bk1 bk2 bk3 bk4

Config. vi-vp vi-hp-d vi-hp-r vi-hp-l

K, D 9.8 9.8 9.8 9.8φ, % 38 38 38 38So,% 69.36 75 75 75Sw,% 25 25 25 25Sg, % 0.0 0.0 0.0 0.0Voil, cc 6850 6850 6850 6850V

water, cc 2280 2280 2280 2280

PInj, Kpa 200 120 178 248QAir, m

3 /m 2-hr 14.44 9.97 7.57 2.76Q Inj.m

3 /hr 1.080 0.748 0.568 0.207

Temperature distribution inside the model was registeredevery 10 minutes during each experiment. The otherparameters that were recorded during the experiment were airinjection pressure and rate, volume of produced gas,production pressures and, oil and water productions. Producedgas samples were fed to gas chromatography in every twenty

minutes for compositional analysis of gas. The combustionruns were terminated as soon as the front arrived at thethermocouple nearest to the producer.

The precipitation and column chromatographic methodsachieved separation of the crude oil into SARA fractions.Diluting the crude oil with forty volumes of n-hexaneseparated the asphaltene fractions of oils. The mixture wasshaken for about one hour and stored overnight in the dark.Filtration of this mixture was the next step to obtainasphaltene. The asphaltene fraction was washed with n-hexaneuntil no yellow color due to resins or oil was visible in thewash. The asphaltene fraction was dried in an oven at 70for 6 hours under helium atmosphere. To separate saturates,aromatics, and resins a combined alumna/silica column wasprepared. The column (60x1 cm ID) was slurry packed, thealumina was preheated for 6 hours at 300 °C and silica gel waspreheated for 6 hours at 130 °C. The height ratio of aluminaand silica was 1:1 in the column. The solvent (n-hexane)soluble compounds were separated into alkanes, aromatics andresins by chromatography using the prepared column. Thecarrier solvent for separation of alkanes was n-hexane. A 1:1mixture of n-hexane/benzene was used to elute aromatics fromthe column. Finally, 1:1 benzene/methanol was used to obtainresins from the column. The separation efficiency of saturatesfrom aromatics was checked by means of UV detectionExperiments (TG/DTG) were performed with a sample size of ~10 mg, at heating rate of 10 °C/min. Air flow rate through thesample pan was kept constant at 50 ml/min. in the temperaturerange of 20-600 °C. Experiments were performed twice forrepeatability.

Experimental Results

SARA fractions were separated from the crude oils both toinvestigate the behavior of the individual fractions in oxidizingenvironment and analyze the oxidation kinetics of thesefractions to be used in numerical simulations. TG/DTG curvesof crude oils and their fractions (Fig 3) are given to be able tofollow the behavior of the SARA constituents in oxidizingenvironment. The crudes were similar in properties to thoseused in in situ combustion experiments (Table 2). In B.Ramancrude, asphaltene fraction is the most resistant part of crudeoil. Asphaltene molecules are so heavy and resistant that,oxygen does not affect the asphaltenes until very hightemperatures are reached. There is almost no weight loss of asphaltenes due to distillation and low temperature oxidation

(LTO) reaction. LTO is very weak in asphaltenes and occurswith very little weight loss (2.1%). In derivative plots of weight loss (Fig 4) this behavior is shown in the derivativecurve, which is placid and smooth till MTO and shows a steepincrease in HTO showing a rapid reaction in that region forboth of the asphaltenes. This weight loss is probably due tothe visbreaking and pyrolysis of the asphaltenes. After thisslight reaction, middle temperature oxidation (MTO) startsnear 380 °C that creates a 32% weight loss for asphaltenes.Since these are the heaviest fractions in crude oil, they losealmost rest of their weight in high temperature oxidation




(HTO) region. This means that the overall oxidation of asphaltenes is slow with little production of higher molecularweight compounds to be burned.

Figure 3. TG/DTG curves for B.Raman crude and its SARAfractions.

Figure 4. Derivative plot for B.Raman crude and its SARAfractions.

Saturates are at the other extreme as far as oxidation of crude oil fractions are concerned. Saturates show a hugeweight loss (88.6%) until the end of LTO reaction that start at310 °C and their weight losses in this region are around12.6%. Saturates show very slow MTO reaction and a weak HTO of the fuel formed from the oxidation of paraffins inLTO. So, saturates do not contribute too much to HTOreaction, in oil phase.

Between these two extremes, asphaltenes and saturates,

there are aromatics and resins or polar materials. TG behaviorof the aromatic fraction is very similar to that of resins and thisobservation supports the hypothesis of the formation of resinsis due to the reaction between aromatics and oxygen. For theresin fraction of the crude oil, LTO region is between 320-370°C with a weight loss of 11%. It is also important to note thatthe behavior of resins in distillation and LTO region isdependent on the chemical structure of resins or the amount of waxy compounds, which is distilled at low temperatures asinferred from similar weight losses. The very same behavioris also observed for aromatics. They have almost the same

LTO region (330-390 °C), and also LTO starting temperatureis the same with resins. The calculation of kinetic data isbased on the formal kinetic equations:

( )nk dt d

α α −= 1 (1)

−= RT

E Ak exp (2)

Where α is the amount of sample undergoing reaction, n is torder of reaction, k is the specific rate constant, A is tArrhenius constant, E is the activation energy and R is the glaw constant.

TG/DTG data of the crude oils and their SARA fractionswere analyzed using Coats and Redfern 20 method. Thmethod has wide applications in analyzing the TG/DTG datain recent years and provides reliable results. In this study, fivedifferent reaction orders (0.5, 0.67, 1, 1.5 and 2) were assumedand correlation coefficients of each reaction were calculated.Here, the aim was to find out the highest correlationcoefficient of the reaction orders assumed. It was found thatthe reaction order of “ 1” has the highest value for all the crudeoils and fractions studied. The final form of the equation is,

( )( )

( ) RT E

E RT

E AR

T n

n

−

−

=

−−− − 2

1ln1

11ln 2

1

β α

(3)

A graph of ln [1-(1- α)(1-n) / (1-n)T 2] vs. 1/T yields a straightline of slope E/R for the correct value of the order of reaction.If the order of reaction is 1, equation 3 can be expressed as

( ) RT E

E RT

E AR

T −

−

=− 2

1ln1ln

ln 2 β α

(4)

Therefore E/R can then be obtained from the graph of ln [-ln(1- α) / T 2] vs. I/T. The kinetic analysis (activation energyand pre-exponential factor) on the SARA fractions of thesetwo crude oils by using the computing method explainedabove are given in Tables 2 and 3. These tables show thatasphaltenes are the strongest fractions toward oxidation. Forboth of the crudes, in all oxidation regimes, asphaltenes needthe highest amount of energy to be oxidized. This is due to theheavy nature of the asphaltene molecules. Unlike asphaltenes,saturates are the easiest oxidizable compound.

Table-2 Properties of Crude OilsProperties Medium Crude Heavy Crudel

OAPI Gravity 26.12 14.95Viscosity (cp) 37 51935<C 15+ (%) 21.8 16.9Saturates (%) 45 18Aromatics (%) 29 31Resin (%) 14 22Asphaltene (%) 12 29Sulphur (%) 2.2 5.4Calorific Value(cal/gr)

11300 9870




5.00 10 .0 0 15 .0 0 20.00 2 5.00 30 .0 0

DISTANCE (cm)

5.00

10.00

15.00

20.00

25.00

30.00

D I S T A N C E ( c m )

PRODUCER

INJECTOR

5.00 10.00 15.00 20.00 25.00 30.00

DISTANCE (cm)

5.00

10.00

15.00

20.00

25.00

30.00


INJECTOR

PRODUCER

5.00 10.00 15.00 20.00 25.00 30.00

DISTANCE (cm)

5.00

10.00

15.00

20.00

25.00

30.00

D

I S T A N C E ( c m )

PRODUCER

INJECTOR

5.00 10.00 15.00 20.00 25.00 30.00

DISTANCE (cm)

5.00

10.00

15.00

20.00

25.00

30.00


PRODUCER

INJECTOR

Figure 5. Left- Temperature profiles at the center plane at t=95 and t=450 minutes forRM-1. Right- Temperature profiles at the center plane at t=100 and t=300 minutes forBK-1

Table-3 Kinetic Parameters of the Samples (ActivationEnergy, kJ/mol)

LTO MTO HTOSaturates 10.5 2.7 80.9

Medium Aromatics 74.4 27.9 121.4Crude Resins 77.6 37.3 136.3

Asphaltenes 131.8 130.6 226.8Saturates 18.9 7.3 136.6

Heavy Aromatics 103.1 41.9 152.7 Crude Resins 60.9 46.3 49.5

Asphaltenes 83.1 124.5 183.9

LTO MTO HTOSaturates 1.16E+05 6.26E+03 1.76E+09

Medium Aromatics 2.33E+06 7.32E+04 1.21E+08Crude Resins 7.69E+06 5.64E+04 1.99E+08

Asphaltenes 7.34E+07 6.11E+07 6.51E+13Saturates 1.27E+05 2.19E+04 7.43E+08

Heavy Aromatics 3.61E+07 7.50E+05 1.36E+09Crude Resins 9.14E+07 1.76E+06 5.74E+06

Asphaltenes 5.67E+05 2.43E+07 1.93E+11

Table 4 summarizes the experimental results obtained fromthe in-situ combustion tests. The temperature distributions attwo time levels for vertical injection-vertical productionscheme as shown in Figure 5 were discussed to make acomparison between the applications of different crudes with

different API gravity. Centre plane was observed to be muchwarmer than top and bottom planes in both runs. This isbecause the igniter was located near the centre plane. Theoccurrence of HTO at the vicinity of injection end in both runscan be verified by looking into reaction kinetics of Raman andB.Kozluca crudes mentioned earlier. The heat frontpropagation in RM1 was faster at the beginning and sloweddown in later time steps. The peak temperatures observed inRM1 were lower than those in BK1. Stabilized combustionfront was observed in RM1 whereas, no period of stabilizationwas apparent in BK1 due to severe bypassing.

Table 4. Overall summary of the combustion test results.Raman RM1 RM2 RM3 RM4 RM5 RM6 RM7TAverage peak °C 428 411 390 403 401 450 420VFront , cm/min 0.065 0.097 0.064 0.073 0.103 0.051 0.065Oxygen util, % 85.50 83.10 80.10 83.40 82.16 89.80 86.00Air requirement,m3(st)/m 3

1395 847 423 133 183 376 245

AOR, m 3(st)/m 3 2675 2313 1121 328 558 640 702AFR, m 3 /kg 11.32 11.47 13.03 11.04 10.84 12.50 10.66Fuel cons. rate,kg/m 3

123.2 73.82 32.42 12.04 17.05 30.08 22.90

Oil recovery,(% OOIP)

33.45 36.35 54.46 55.14 74.91 30.66 34.88




Table 4. ContinuedB.Kozluca BK1 BK2 BK3 BK4TAverage peak °C 458 460 445 465VFront , cm/min 0.0650 0.0820 0.0690 0.0612Oxygen util, % 90.65 90.80 92.21 95.06Air requirement,m3(st)/m 3

3930 1329 562 229

AOR, m 3(st)/m 3 4864 4234 1859 595AFR, m 3 /kg 12.86 10.65 10.43 10.92Fuel cons. rate,

kg/m3

305.45 124.67 5392 20.93

Oil recovery,(% OOIP)

25.95 28.40 45.03 43.22

However in RM6, which was a rerun of RM1 with a lowerair rate, the channeling occurrence was severe. The reductionof channeling may be related to the lower injection rate. Thesuperficial burning of the sand pack was also noted in thisexperiment. Therefore, it can be concluded that superficialburning of sand pack is usual in all vertical-vertical well

configurations, which leads to a poor vertical sweep in thisconfiguration.

Several experiments were conducted employing verticalinjector-horizontal producer configuration to recover bothcrudes. After the start of air injection, a uniform temperaturedistribution throughout the centre plane in both runs wasobserved. The instant of active combustion was obvious inRM3 and BK3 where vertical injector- horizontal right sideproducer was used at 247 minutes and 279 minutes,respectively. The front stabilized in both runs and proceededin the direction of the producer. The creation of isothermswere parallel to the producer in both runs which is suggestive

of distributed flow field at the centre plane as shown in Fig 6.Nevertheless, isotherms at the centre plane in BK3 spreaded tothe left boundary of the model. Early production of hot fluidsthat transport heat as a result of convection was noted. Heatwas removed from burned-out sand pack by vaporization of water behind the combustion front. The heat was depositedahead of the burning zone by condensation in cooler regions of the reservoir. The horizontal production well conveyed heatdownstream into the colder regions. However, anotherimportant factor was the conductive heat loss towards thewalls of the 3-D model because of the controlling of band

5.00 10.00 15.00 20.00 25.00 30.00

DISTANCE (cm)

5.00

10.00

15.00

20.00

25.00

30.00


PRODUCER

INJECTOR

5.00 10.00 15.00 20.00 25.00 30.00

DISTANCE (cm)

5.00

10.00

15.00

20.00

25.00

30.00


PRODUCER

INJECTOR

5.00 10.00 15.00 20.00 25.00 30.00

DISTANCE (cm)

5.00

10.00

15.00

20.00

25.00

30.00


PRODUCER

INJECTOR

5.00 10.00 15.00 20.00 25.00 30.00

DISTANCE, cm

5.00

10.00

15.00

20.00

25.00

30.00

D I S T A N C E

, c m

Figure 6. Left- Temperature profiles at the center plane at t=140 and t=615 minutes for RM-3. Right-Temperature profiles at the center plane at t=95 and t=485 minutes for BK-3




heaters. No indication of flow channel or bypass was noted.Since horizontal producers cause a larger areal contact, theycan drain fluid without creating a by passing flow path.

At the beginning of all runs, the rate of frontal advancewas observed to be very low. This was probably due to thehigher fuel deposition at the beginning of the experiments as aresult of LTO and cracking. Since the frontal advance isdependent on the residual material burned per unit of sandcleaned, propagation of front is very slow at the beginning.Additionally, in a 3-D model the front is viable to change itsdirection of propagation during the courses of its travel to theproducers, which is one of the reasons of stagnancy in frontaladvance along the diagonal of the model at earlier time steps.

The average fuel consumption rates in vertical-verticalconfiguration for both crudes were the highest, which, in turn,resulted in the highest overall fuel consumption. This eventindicates the involvement of severe cracking and distillationall through these experiments, which enhance fuel deposition.The fuel consumption rate was greatly reduced in all runs withboth crudes by employing horizontal producers. For bothcrudes, air requirements were higher in vertical-verticalconfiguration, since higher fuel deposition takes place in thisconfiguration. The AOR for both crudes were highest invertical-vertical configuration.

Figures 7 and 8 represent the burned volume andcorresponding recovery of Raman and B.Kozluca crude oilrespectively, employing different well configurations. Withthe same burned volume it was observed that horizontalproducers recovered more oil than vertical producers. Theamount of oil recovered in vertical injector-right sidehorizontal producer (RM3) and vertical injector-left sidehorizontal producer (RM4) was almost the same, at around 55% of OOIP. The recoveries for the same well configurationsof the heavier crude were around 45 % of OOIP. Although,dual horizontal producers recoverd the highest amount of crude (72% of OOIP), the net recovery per producer waslower than that of a single producer placed alone at theboundary. The main reason for this is presumably the flowinterference between the twin producers due to the restricteddrainage area. The vertical-vertical configurations for bothcrudes exhibit poor performances. Early production noted inhorizontal wells remove oil without the creation of anyextensive mobile path in the colder region whilst, because of their flow geometry more volume should be burned to create aflow path between a vertical injector and a vertical producerwithin which the fluid can flow easily.

Simulation of Laboratory Combustion Test DataNumerical simulation of the experimental tests wereperformed using the CMG model STARS 21 by incorporating areaction model for in-situ combustion of used crude oils.Cartesian gridding with 12x12x5 gridblocks were used. Aconstant pressure producer and a constant rate injector locatedat the appropiate plane and location were used. The igniterwas modeled by constant addition of heat to the injectionblock.

0

10

20

30

40

50

60

70

80

0 200 400 600 800 1000Time (min)

O I l R e c o v e r y

( % O

O I P )

RM1(VI-VP)RM2(VI-HP(D))RM3(VI-HP(R))RM4(VI-HP(L))RM5(VI-2HP)RM6(VI-VP)RM7(VI-HP(D))

CRUDE OIL = RAMAN

Figure 7. Experimental Recovery Curves for Different WellConfigurations Using Raman Crude Oil.

0

5

10

15

20

25

3035

40

45

50

0 100 200 300 400 500 600 700 800

Time (min)

O i l R e c o v e r y

( % O

O I P )

BK1(VI-VP)BK2(VI-HP(D))BK3(VI-HP(R))BK4(VI-HP(L))

CRUDE OIL = B. KOZLUCA

Figure 8. Experimental Recovery Curves for Different WellConfigurations Using B.Kozluca Crude Oil.

We first modeled experiments using a kinetic model that isa consolidation of individual kinetic studies on thermalcracking, low temperature oxidation and high temperatureoxidation given in detail by Greaves et al. 22. In this modedescribing the oil component in terms of heavy, medium, andlight fractions with two non-condensible gases as well as asolid coke component followed a pseudo component approach.Seven chemical reactions were used. Two reactions dealt withcracking of heavy oil. Other reactions described burning of the heavy, medium, light oil, hydrocarbon gases, and the cokecomponents respectively. However, as reported in Kumar 23

was observed that only “heavy oil burning” reaction wasenough to describe the reaction kinetics for the heavy oils usedthroughout the experiments. Equilibrium K-values for themodel were estimated from the Table 2 in STARS manual.The numerical model incorporated external heater option toraise the temperature of the injector in the beginning. In orderto simulate adiabatic conditions, no external heat losses orgains were allowed.

The modeling efforts produced relatively satisfactoryresults especially in vertical injector-horizontal side producerconfigurations using both crude oils. Figures 9 and 10 showsample history matched experiments, RM-1 and BK-1 wherevertical injector-vertical producers for different crudes wereused. It can be observed that recovery and oil rate predictions




are very comparable. Similar successful matches wereobserved with vertical injector-horizontal side producerconfigurations (RM-3, BK-3) as shown in Figures 11 and 12.The pressure matches were somewhat poor because of the lowsampling rate of pressures during the experiments. A samplepressure match for BK-3 (vertical injector-vertical producer) isgiven in Figure 13. Recovery and pressure matches were thepoorest for the vertical injector-diagonal horizontal producerconfiguration (RM-2) as shown in Figures 14 and 15. Forboth crudes the numerical model severely underestimated theexperimental recovery especially after the ignition. Webelieve this problem is related to the definition of horizontalwells in the numerical model and further studies are required.Figure 16 gives temperature distribution at the center plane forthe experiment RM-1 (vertical injector-vertical producer).When compared with the experimental temperature profilespresented in Figure 3, it can be observed that modeltemperature profiles underestimated experimental ones.Although the simulation displays rapid channeling down thecommunication path observed in the experiment it does notduplicate the progression of a high temperature zone throughthe pack. This behavior was also observed in other injector-producer configurations.

0

500

1000

1500

2000

2500

0 100 200 300 400 500 600 700

Time, minutes

P r o

d u c e

d o

i l v o

l u m e ,

m l

0

10

20

30

40

50

60

70

80

90

O i l r a t e

, m

l / m i n

Figure 9. History Match of Experiment RM-1.

0

500

1000

1500

2000

2500

0 100 200 300 400 500 600 700

Time, minutes

P r o d u c e

d o

i l v o l u m e ,

m l

0

2

4

6

8

10

12

14

16

O i l r a t e

, m

l / m i n

Figure 10. History Match of Experiment BK-1.

0

500

1000

1500

2000

2500

3000

3500

4000

0 100 200 300 400 500 600 700 800

Time, minutes

P r o

d u c e

d o

i l v o

l u m e ,

m l

-20

0

20

40

60

80

100

120

140

160

180

Figure 11. History Match of RM-3.

0

500

1000

1500

2000

2500

3000

3500

4000

0 100 200 300 400 500 600 700 800

Time, minutes

P r o

d u c e

d o

i l v o l u m e ,

m l

0

5

10

15

20

25

30

Figure 12. History Match of BK-3.

The purpose of history matching the experimental tests isthat an acceptable reaction scheme and other operational

parameters can be obtained for field scale simulations that willultimately predict a possible process performance in the field.That’s why proper modeling of the experiments is extremelyimortant. Thus, in order to improve the matches we haveincorporated a kinetic model that was obtained using theobservations carried out using the thermogravimetricexperiments. Thus we described the heavy crude using itsSARA fractions and developed a new kinetic model based onthe experimental observations given in Tables 2 and 3. Whencompared with the previous model, the new model has anadditional oil fraction and the reactions are slightly different,(eg. reaction between aromatics and oxygen produce resins).It was observed that to achieve a successful run has been

difficult. Using the same fluid description along with theparameters for the SARA reactions given in Tables 2 and 3resulted in quick rise in the injection pressure. Materialbalance errors increased up to 2% in certain cases where adiagonal horizontal producer was utilized. By lowering thefrequency factors observed experimentally we were able todecrease run times and the pressures. At the end of theexperimental run times of around 500 minutes the SARAmodel achieved a cumulative production significantly higherthan that of a single component model as shown in Fig 17. Itshould be noted that the cumulative recovery was still lower




than the experimentally observed one. Similar to the pseudocomponent model, the temperature profiles for the SARAmodel showed that the temperatures were not as high as theexperimentally observed ones.

We have conducted several sensitivity runs to establishdifferences with the pseudo model, one component model, andthe SARA model. It was observed that the SARA model wassensitive to especially vertical grid size. It was also observedthat similar to the experimental observations air injection ratewas a critical parameter (see discussion on experiment RM 1and RM 6). On the other hand, the SARA model wasinsensitive to component and rock specific heats, and oilcomponent thermal expansion coefficient. We also observedlittle sensitivity to absolute permeability. On the other hand,K-values were found to be affecting the results more than thatof the thermal conductivity of overburden formation. As aconclusion we can say that the matches were somewhatimproved however, the usage of SARA fractions and theirkinetic parameters is not justified because of the experimentalcomplexity and time.

0

100

200

300

400

500

600

0 100 200 300 400 500 600 700 800

Time, minutes

P r e s s u r e ,

K p a

Figure 13. Differential Pressure Match of Experiment BK-3.

0

500

1000

1500

2000

2500

3000

0 100 200 300 400 500 600

Time, minutes

P r o

d u c e

d o

i l v o

l u m e ,

m l

0

10

20

30

40

50

60

70

80

O i l r a t e

, m l / m i n


0

20

40

60

80

100

120

140

160

180

0 100 200 300 400 500 600

Time, minutes

P r e s s u r e ,

K p a

Figure 15. Differential Pressure Match of Experiment RM-2.

Figure 16. Numerical temperature profiles at the center plane att=135 (top) and t=615 minutes (bottom) for RM-3.

Finally, we can conclude that the results obtained fromhistory matching experimental data cannot be used directly forfield scale simulation. The size of the grid blocks used in theexperimental history matching is compareble to the actual sizeof the combustion front. However, field scale grid blocks areseveral magnitudes larger. Moreover, in the numerical modelthat simulates the experimental data, temperature is theaverage of the entire grid block that adequately represents the




temperature in the reaction zone. It is inappropriate torepresent the peak combustion zone temperature in the fieldcase. A mathematical model that incorporates dynamic gridrefinement around the injector and producer as suggested byCoates et al 25 may be used to solve the aforementionedproblem.

0

500

1000

1500

2000

2500

0 100 200 300 400 500 600 700 800

Time, minutes

P r o

d u c e

d o

i l v o l u m e ,

m l

0

5

10

15

20

25

30

35

O i l r a t e

, m

l / m i n

SARA


ConclusionsWe present in situ combustion experiments where a verticalinjector is employed whereas, both vertical and horizontalproducers have been installed to recover two different crudeoils (heavy and medium) conducted on a 3D semi-scaledphysical model that represents one fourth of a repeated fivespot pattern using two crude oils: heavy and medium. Basedon the experimental analyses, the following conclusions canbe drawn:

1. Superficial burning of sand pack was usual in allvertical-vertical well configurations, which leaded topoor sweep efficiency in this configuration.

2. No indication of flow channel or bypass was notedduring vertical injector-horizontal producerconfigurations. Since horizontal producers cause alarger areal contact, they can drain fluid withoutcreating a by passing flow path. Therefore, it wasconcluded that the vertical injector-horizontal sideproducers performed better than the otherconfigurations.

3. There is almost no weight loss of asphaltenes due todistillation and LTO reaction. Asphaltenes are the

strongest fractions toward oxidation.4. Saturates show a huge weight loss till the end of LTOreaction. Unlike asphaltenes, saturates are the easiestoxidizable compound.

5. A kinetic model based on SARA fractions in the crudeoil successfully modeled the experiments. However,due to experimental complexity, the use of the kineticmodel was not justified.

6. Numerical temperature profiles underestimatedexperimental temperature profiles.

7. Since the temperature is the average of the entire gridblock that adequately represents the temperature in thereaction zone. It is inappropriate to represent the peak combustion zone temperature in the field case. Amathematical model that incorporates dynamic gridrefinement around the injector and producer may beused to solve this problem.

References

1. Binder, G.G., Elzinga, E.R., Tarmy, B.L., and Wilmann,B.T., "Scaled -model Tests of In-Situ Combustion inMassive Unconsolidated Sands", Proceedings SeventhWorld Petroleum Congress, Mexico City, pp. 477-485,March 1967.

2. Garon, A.M., Geisbrecht, R.A., and Lowry, W.E. Jr.,"Scaled Model Experiments of Fire-flooding in Tar Sands",Journal of Petroleum Technology, pp.2158-2166, September1982.

3. Garon, A.M., Kumar, M., Lau, K.K. and Sherman, M., "ALaboratory Investigation of Sweep During Oxygen and Air

Fireflooding", SPE Reservoir Engineering, pp.565-574,November 1986.

4. Greaves, M., Tuwil, A.A., and Bagci, A.S., "HorizontalProducer Wells in In-Situ Combustion (ISC) Processes",Journal of Canadian Petroleum Technology, Vol. 32, No. 4,pp.58-67, April 1993.

5. Coates, R., Lorimer, S., and Ivory, J., "Experimental andNumerical Simulations of a Novel Top Down In-SituCombustion Process", SPE Paper 30295, Presented atInternational Heavy Oil Symposium, Calgary, Alberta, pp.487-498, June 1995.

6. Greaves, M., and Al-Shamali, O., "In Situ Combustion (ISC)Process Using Horizontal Wells", Journal of CanadianPetroleum Technology, Vol. 35, No. 4, pp.49-55, April1996.

7. Greaves, M., and Mahgoub, O., "3D Physical Model Studiesof Air Injection in a Light Oil Reservoir Using HorizontalWells", SPE Paper 37154, presented at SPE InternationalConference on Horizontal Well Technology, Calgary,Alberta, pp.951-962, November 1996.

8. Greaves, M. and Wilson, A., Al-honi, M., and Lockett, A.D.,"Improved Recovery of Light/Medium Heavy Oils inHeterogeneous Reservoirs Using Air Injection/In SituCombustion (ISC)", SPE Paper 35693, presented at WesternRegional Meeting, Anchorage, Alaska, pp.435-443, May1996.

9. Verkocy, J.; Kamal, N.J. J. Can. Pet. Technol .1986 , 47-5710. Yoshiki, K.S.; Phillips, C.R. Fuel 1985 , 64, 1591-159611. Bae, J.H.: “Characterization of Crude Oil for Fireflooding

Using Thermal Analysis Methods,” Soc. Pet. Eng. Journal June 1977 , 211-218

12. Drici, D.; SPE Reservoir Eng. 1977 , 591-59513. Del Bianco, A.; Panariti, M.A.; Beltrame, P.L.; Carniti, P.

Fuel 1993 ,72, 75-8514. Yoshida, R.; Takeda, S.; Teramoto, S.; Matsushita, T.;

Takeya, G. Fuel Process. Technol. 1984 , 9, 307-31315. Ciajolo, A. and Barbella, R.: “Pyrolysis and oxidation of

heavy oils and their fractions in a thermogravimetricapparatus” Fuel, Vol 63, pp 657-662, 1984.

16. Ranjbar, M. and Pusch, G. J.: “Pyrolysis and Combustionkinetics of crude oils, asphaltenes and resins in relation to




thermal recovery processes” Anal. Appl. Pyrolysis, 185,1991.

17. Ali, M.F. and Saleem, M.: “Pyrolysis and Combustionkinetics of crude oils, asphaltenes and resins in relation tothermal recovery processes” Fuel Sci. Technol. Int. No 9, pp461-468, 1991.

18. Kok, M.V. Thermochim. Acta 1993 , 214, 31519. Akin S., Bagci S. and Kok M.V.: “Dry Forward Combustion

with Diverse Well Configurations,” paper SPE 62551,Proceedings of the 2000 SPE/AAPG Western RegionalMeeting held in Long Beach, California, 19-23 June 2000.

20. Coats, A.; Redfern,R.; Nature 1964 , 201, 68-6921. Computer Modelling Group, "STARS Version 99 User's

Guide", Calgary Alberta, Canada, 1999.22. Greaves, M., Ron, S.R., and Rathbone, R.R., "Air Injection

Technique (LTO Process) for IOR from Light OilReservoirs:Oxidation Rate and Displacement Studies", SPEPaper 40062, presented at SPE/DOE Improved Oil RecoverySymposium, Tulsa, Oklahoma, pp.479-492, April 1998.

23. Kumar M.: “A Cross-Sectional Simulation of WestHeidelberg In-Situ Combustion Project,” SPE ReservoirEngineering, pp 46-54, Feb 1991.

24. Ambastha A.K. and Kumar M.: ”New Insights Into In-SituCombustion Simulation for Heavy Oil Reservoirs,” paperSPE 56543 presented at the 1999 SPE Annual TechnicalConference and Exhibition held in Houston, Oct. 3-6 1999.

25. Coates R., Lorimer, S., and Ivory J.: “Experimental andNumerical Simulations of a Novel Top Down In-SituCombustion Process,” paper SPE 30295, InternationalHeavy Oil Symposium held in Calgary, Alberta, Canada, pp487-498, 19-21 June 1995.

Documents

Volume Issue 0 2000 [Doi 10.2118%2F63230-Ms] Serhat, Akin_ Kok, Mustafa_ Suat, Bagci_ Ozgen, Karacan -- [Society of Petroleum Engineers SPE Annual Technical Conference and Exhibition