geochemical modeling and experimental evaluation of high ... v92 pp216-23… · the interaction...
TRANSCRIPT
Fuel 92 (2012) 216–230
Contents lists available at ScienceDirect
Fuel
journal homepage: www.elsevier .com/locate / fuel
Geochemical modeling and experimental evaluation of high-pH floods: Impactof Water–Rock interactions in sandstone
Mahdi Kazempour, Eric Sundstrom, Vladimir Alvarado ⇑Department of Chemical and Petroleum Engineering, University of Wyoming, Laramie, WY 82071, USA
a r t i c l e i n f o
Article history:Received 27 March 2011Received in revised form 8 July 2011Accepted 14 July 2011Available online 28 July 2011
Keywords:Alkaline floodAnhydritepH bufferingOil recoveryChemical flooding
0016-2361/$ - see front matter � 2011 Elsevier Ltd. Adoi:10.1016/j.fuel.2011.07.022
⇑ Corresponding author.E-mail address: [email protected] (V. Alvarado)
a b s t r a c t
Injection of alkaline solutions in reservoir leads to mineral dissolution and precipitation, possibly result-ing in changes in permeability and porosity, and consequently altering solution pH. Accurate predictionof pH, alkali consumption and aqueous chemistry changes are required to design suitable chemicalblends in alkaline-polymer (AP) or alkaline-surfactant-polymer (ASP) flooding. Excessive consumptionof alkali can result in degradation of flood performance and lower than expected oil recovery. We reportstate-of-the-art geochemical simulation results for sandstone reservoir mineral assemblages and alkalisolutions (NaOH, Na2CO3, and NaBO2) employed in AP and ASP formulations. Single-phase high-pH core-floods were completed using Berea sandstone and reservoir samples to calibrate and validate geochem-ical simulations. Results show that rock-fluid interactions depend strongly on mineral type and amount,alkaline solution injection flowrate, and composition of the injected and formation water. Anhydrite, acommonly found calcium sulfate, significantly impacts pH buffering capacity, water chemistry and per-meability damage against conventional alkali agents in chemical flooding particularly for Na2CO3, but nosignificant pH buffering is observed during NaBO2 flooding. Experimental data and model results showthat the pH-buffering effect is maintained even after several pore volumes of alkaline solution areinjected, if a sufficient fraction of relevant minerals is present. The end consequence of this is insufficientalkalinity for reactions with the oil phase and the likely formation damage.
� 2011 Elsevier Ltd. All rights reserved.
1. Introduction
The role of an alkali agent in enhanced oil recovery methods hasbeen studied for more than four decades. A common claim is thatalkali aids oil mobilization by generating in situ soap [1–3], or bylowering IFT to ultra-low values in synergy with surfactants [4–6]. An alkaline agent can sequester divalent cations from the aque-ous phase enhancing the efficiency of surfactant partitioning andavoiding its precipitation [7]. Alkali agents can increase negativecharge density on rock surfaces, altering its wettability towardwater wetness in presence of crude oil [8]. Alkali can also reducethe adsorption of anionic surfactants on rock surfaces [7,9]. Theaforementioned roles of alkaline agents are attributed to their abil-ity to increase pH. One of the most important phenomena affectinghigh pH front propagation through a formation is alkali consump-tion or retention by rock during the injection period.
Ehrlich et al. [10] completed a series of static tests to measurealkali consumption in presence of different minerals at room tem-perature. They found that gypsum, a more stable form of anhydrite,is largely responsible for alkali consumption in contact with 1.25 N
ll rights reserved.
.
NaOH solution, more than quartz, dolomite, calcite, illite, kaoliniteand montmorillonite. Sydansk [11] showed that caustic solutionsin the form of a sodium-hydroxide solution strongly interactswith sandstone at elevated temperature. Sydansk concluded thatsodium hydroxide interacted with sandstone at elevated tempera-ture to promote: (a) significant dissolution of the more susceptiblesilicate minerals, predominantly clay and large-surface-area silicaminerals; (b) sandstone weight loss; (c) increased porosity; (d)propagation of significant concentrations of water-soluble silicates,including sodium orthosilicate; (e) in situ formation of new immo-bile aluminosilicate material; (f) changes in permeability; and (g)hydroxide ion consumption. Larrondo and Urness [12] performedcore flooding tests on unconsolidated sandstone with quartz, chert,feldspar, rock fragments and kaolinite as a clay using sodiumhydroxide, sodium orthosilicate and sodium methasilicate. Theyshowed that NaOH produced the greatest long-term consumption.They also concluded that loss of alkali possibly occurred throughion exchange and mineral dissolution reactions. Cheng [13] foundthat alkali consumption can be significantly decreased with the useof sodium carbonate compared to caustic and silicate alkali inpresence of Wilmington and Ottawa sand with dolomite.
Dezabala et al. [14] used a simple equilibrium chemical modelto present continuous, linear, alkali flooding of acid oil reservoirs.Jensen and Radke [15] used a local equilibrium chromatographic
M. Kazempour et al. / Fuel 92 (2012) 216–230 217
model to predict alkali propagation through Berea cores. Labridand Bazin [16] describe a model designed for calculating the trans-port of alkali in permeable media with either the assumption thatalkali consumption was determined by thermodynamics or by theinterference of kinetic process. They considered quartz and kaolin-ite as the only minerals initially in the system and analcime as aproduct of rock-alkali interaction. Soler [17] used Global ImplicitMulticomponent Reactive Transport (GIMRT) code to simulatethe interaction between hyper-alkaline solution and fracture marlat 25 �C. Soler used different kinetics for mineral dissolution andprecipitations, but there was no experimental data for comparisonor testing the efficiency of the model.
In this research, we are interested in determining how rock-fluid interactions affect pH and aqueous-phase chemistry in chem-ical enhanced oil recovery (EOR) that include alkaline agents. Sincehigh-pH conditions favor EOR performance by mechanisms such assaponification and by affecting rheological behavior of polymers[18], understanding the so-called alkali consumption is vital to asustainable alkaline flood in any of its modalities. A systematic ap-proach is utilized to investigate the effect of anhydrite on pH buf-fering capacity during alkaline floods. We test two differentsandstones, Wyoming’s Minnelusa reservoir and Berea cores usingthree different alkaline agents: NaOH and Na2CO3, which are themost commonly used alkalis in EOR processes, and NaBO2, whichhas been proposed as a promising alkali in harsh environment[19]. We perform all the experiments at reservoir temperatureand moderate confining pressure, and unlike other studies, a com-prehensive chemical analysis of samples at effluent is completedand reported to track ions concentrations and other related infor-mation. In addition to chemical analysis, thin sections and X-raydiffraction (XRD) tests are obtained. Subsequently this informationis used for building a reactive transport model using a state-of-the-art geochemical simulator (Geochemist’s Workbench [20]). Oncecalibrated, this base model is used to predict the results of othertests to evaluate its robustness. After successful evaluation, thebase model is used to predict the high pH penetration length inpresence of anhydrite in forward (field) simulation cases. For thiscase, a 1-D radial case was examined.
2. Materials and methods
2.1. Experimental
The cores are vacuum-saturated with a 30,000 ppm NaCl solu-tion and aged at 60 �C for one week. The cores dimensions are1.500 � 300 (diameter � length). The core flooding experiments areconducted at 200 psi of confining pressure. The physical propertiesof these cores are tabulated in Table 1. The results of XRD test forMinnelusa and Berea cores are shown in Figs. 1 and 2. The flow ratefor the core flooding experiments is set at 0.5 ml/min for Bereacores and 0.2 ml/min for Minnelusa cores, due to lower permeabil-ity and porosity of the latter. The flooding sequence is as follows.First, 5 pore volumes of the 30,000 ppm NaCl solution are injected,followed by 20 pore volumes of an alkali solution, having one shut-in event, after 15 pore volumes of alkali for 15 h. Three alkaline
Table 1The properties of the cores.
Core Type Porosity (%) Pe
BC-Y-13 Berea 19.9 20C-10 Minnelusa 8.73
BC-Y-11 Berea 19.92 20C-12 Minnelusa 9.26
BC-Z-05 Berea 20.9 17C-7 Minnelusa 17.6 7
solutions were prepared at 1 wt% NaOH, 1 wt% Na2CO3, and1 wt% NaBO2 in 3 wt% NaCl brine with the resulting initial pH at25 �C: 13.13, 11.15 and 10.36 respectively. For each alkaline solu-tion, two different cores were used to test the effect of anhydriteon results. The cores are flushed with brine at the end of the alka-line flood for additional 5 pore volumes. Pressure drop is recordedto track the permeability changes during each test. Effluent sam-ples are collected in vials using an automated fraction collector.The vials are used to determine the effluent samples pH, chartingthe change in pH as a function of pore volumes injected. The efflu-ent samples are also analyzed for anion and cation concentrationsto determine the dissolution of minerals within the core from alka-li flooding. The cores are dried, cut and thin sections are made foranalysis.
2.2. Numerical simulation
The Geochemist’s Workbench Professional 8.0 (GWB) [20] isused as a numerical simulator. GWB can model reactive transportfluid flow in porous media under a variety of conditions. The modelfor activity coefficient calculation is the B-dot equation, whichworks better for intermediate and high salinity solutions [21].The standard thermodynamic database was modified to includethermodynamic properties of sodium carbonate and sodium meta-borate from the thermo.com.v8.r6+ and thermo_phrqpitz dat-abases. For radial cases, rapid changes in velocity around thewellbore require increasing spatial resolution by adding morenodes particularly around the wellbore in order to provide moreaccurate solutions. To find the most efficient balance betweenmodel node density and computational time, the number of gridblocks was increased until no significant changes in predictionswere observed. Calibration of the simulation model is based onthe experiment done on Minnelusa rock using sodium hydroxideat 60 �C.
The base model contains five minerals: quartz, dolomite, anhy-drite, kaolinite and k-feldspar. The choice of minerals in the model,particularly anhydrite, is determined through X-ray diffraction(XRD) on Minnelusa and Berea core samples (Figs. 1 and 2). Asshown, Minnelusa cores contain anhydrite, but Berea does not.For all the initial minerals in the system, different dissolution ki-netic rates are considered. Except anhydrite, other minerals havepH dependent dissolution rate kinetics. Due to the high tempera-ture condition, Arrhenius type of dissolution kinetic rate (1) is usedfor each mineral (Tables 2 and 3):
dmi
dt¼ AiK
�i e
�EaiRT
� �ð1�XiÞani
Hþð1Þ
where dmidt is dissolution kinetic rate in mole
sec ; Ai is the surface area incm2, K�i is the modified Arrhenius pre-exponential factor inmole=cm2 s; Eai is the activation energy in J
mole, R is the universalgas constant in J
mole K, T is temperature in K, Xi is saturation ratio,aHþ is activity of hydrogen ion, ni is an empirical parameter and i de-notes each mineral (i = 1,2, . . . ,5).
During the injection of alkali, secondary minerals, e.g. calcite,portlandite and brucite, can be produced if certain conditions are
rmeabilityair (md) Size Injected alkali
3.9 1.500 � 300 NaOH3.4 1.500 � 300
6.6 1.500 � 300 Na2CO3
6.0 1.500 � 300
9 1.500 � 300 NaBO2
1 1.500 � 300
Degrees 2-Theta
Cou
nts/
Min
ute
Fig. 1. XRD spectrum of Berea sample used in this study.
Degrees 2-Theta
Cou
nts/
Min
ute
Fig. 2. XRD spectrum of Minnelusa sample used in this study.
218 M. Kazempour et al. / Fuel 92 (2012) 216–230
met. Our strategy for history matching (or calibration) keeps thedissolution rate constant (here pre-exponential factor and activa-tion energy) of each mineral the same as the reported value [22],
changing their initial volumes and their specific surface areas.The ultimate initial volume and specific surface area of each min-eral in the base model are shown in Table 4. For predicting the re-
Table 2Information on minerals reaction status in the model [21].
Mineral Reactioncondition
Dissolution kinetic rateconstant (K) at 25 �C(mole/(m2 s)) at high pH
Activationenergy(Ea) (J/mole)
Quartz Kinetic 10�16.29 108,366Anhydrite Kinetic 10�3.19 14,300Dolomite Kinetic 10�5.11 34,800Kaolinite Kinetic 10�17.05 17,900K-feldspar Kinetic 10�21.20 94,100
Table 3Used dissolution kinetic rate formula for each mineral at high temperature.
Mineral Calculated K� ¼ Kðat 25 �CÞexp �Ea
R�298:15ð Þ(moles/cm2 s)
Applied dissolutionkinetic rate (mole/s)
Quartz 0.001 [21] dm1dt ¼ A1K�1eð
�Ea1RT Þð1�X1Þa�0:5
Hþ
Anhydrite 2.07 � 10�5dm2dt ¼ A2K�2e
�Ea2RT
� �ð1�X2Þ
Dolomite 9.71 � 10�4dm3dt ¼ A3K�3e
�Ea3RT
� �ð1�X3Þa0:5
Hþ
Kaolinite 1.22 � 10�18dm4dt ¼ A4K�4e
�Ea4RT
� �ð1�X4Þa�0:472
Hþ
K-feldspar 1.93 � 10�9dm5dt ¼ A5K�5e
�Ea5RT
� �ð1�X5Þa�0:823
Hþ
Table 4Final values of adjusted parameters in base model.
Mineral Mineral vol. (%) Specific surface area (cm2/gr)
Quartz 80 400Anhydrite 3 400Dolomite 2.267 400Kaolinite 3 400K-feldspar 3 400
10−3
100
102
al s
atur
atio
n (Q
/K)
Anhydrite Portlandite Brucite
Saturation line
0 5 10 15 20 2510−6
10−5
10−4
10−3
10−2
10−1
100
PV
Ca2+
SO42−
0 5 10 15 20 25456789
1011121314
PV
pH
Inlet pH
Shut−in for 15 hrs
Brine injection
Brine injection
NaOH injection
NaOH injection(a)
(b)
(c)
Com
pone
nts
in fl
uid
(mol
al)
M. Kazempour et al. / Fuel 92 (2012) 216–230 219
sults of those tests completed on Berea, the base model is changedby dropping anhydrite from mineral assemblage, keeping otherparameters unchanged. We have included modeling results of al-kali flood at elevated temperature to evaluate the robustness ofthe model under reservoir conditions.
0 5 10 15 20 2510−9
10−6
PV
Min
er
NaOH injection Brine injection
Fig. 3. (a) Inlet and effluent pH during NaOH flooding through Minnelusa core; (b)Chemical composition of effluent; (c) Mineral saturation status of possiblesecondary minerals of collected samples.
3. Results and discussion
3.1. Results of high pH flood in Minnelusa sandstone
3.1.1. Results of NaOH floodUpon injection of this alkaline solution, 0.5 pH unit difference
between inlet and outlet pH was observed. This is an indicationof weak pH buffering capacity of this rock during NaOH flood(Fig. 3a). Only small amounts of aluminum, silica and potassiumwere produced during this experiment, which might have beendue to the low dissolution rate of quartz and aluminum–silicateminerals in this rock. First, during brine injection the molaramounts of produced calcium and sulfate were equal, which wasinterpreted as ions originating from the same source mineral (pre-sumed here to be anhydrite). After initiating the alkaline flood,these ions were no longer produced in equal molar fractions(Fig. 3b), which provided good evidence for significant precipita-tion of secondary calcium minerals such as portlandite (calciumhydroxide) mainly and calcite (calcium carbonate), to some extent,in this high pH environment. Also, there was no significant amountof magnesium in effluent samples, which was probably caused bybrucite (magnesium hydroxide) precipitation (Fig. 3c). Saturationratios of different minerals, in this study, were calculated usingGSS module of GWB for each collected sample at effluent. It isimportant to remark that by injecting NaOH through this type of
rock, the amount of calcium in the aqueous phase drops should,which should favor chemical EOR processes, because the phasebehavior of components in chemical blends, such as polymersand surfactants, improves with decreased calcium concentration.In contrast, the amount of sulfate increases drastically, which isunfavorable because of its side effect on typical anionic surfactantsas well as for its ability to form harmful deposits in the presence ofsome certain cations. Sulfate precipitates such as barium andstrontium sulfate are common oilfield scales that can cause a sig-nificant reduction in permeability [23].
3.1.2. Results of Na2CO3 floodIn this case, the inlet-to-outlet pH difference during alkali injec-
tion at about 2 pH units is a sign of strong pH buffering capacity ofthis rock to Na2CO3 flooding (Fig. 4a). Like sodium hydroxide, no
6789
1011121314
Inlet pH
NaBO injection
Brine injection
Shut−in for 15 hrs
(a)
pH
220 M. Kazempour et al. / Fuel 92 (2012) 216–230
significant amounts of aluminum, silica and potassium were pro-duced (Fig. 4b). Unlike NaOH, the chance of portlandite and bruciteformation is very low here because, as expected, the pH is lower(Fig. 4c). Because of the improbable portlandite precipitation, thepossible sink for calcium becomes calcite precipitation due to thehigh concentration of carbonate ion coming from the injected alka-li plus the present calcium ions in the solution coming from anhy-drite. Similar to NaOH, Na2CO3 sequesters calcium and it increasessignificantly the amount of sulfate in the water phase.
0 5 10 15 20 2545
100
PV
(b)
3.1.3. Results of NaBO2 floodFig. 5a shows that the pH buffering capacity of the rock during
NaBO2 flooding, as reflected by the inlet-to-outlet 0.35 pH differ-ence. Unlike NaOH and Na2CO3, after initiating NaBO2 injection,
0 5 10 15 20 25 30456789
1011121314
PV
pH
Inlet pH
Shut−in for 15 hrs
Na CO injection
Brine injection
0 5 10 15 20 2510−6
10−5
10−4
10−3
10−2
10−1
100 Al3+
Ca2+
Mg2+
K+
Na+
Cl−
SO42−
SiO2
Ca
SO
Na CO injection
Brine injection
0 5 10 15 20 2510−9
10−6
10−3
100
103
PV
Min
eral
sat
urat
ion
(Q/K
)
Anhydrite Portlandite Brucite
Saturation line
Na2CO3 injection
Brine injection
PV
(a)
(b)
(c)
Com
pone
nts
in fl
uid
(mol
al)
Fig. 4. (a) Inlet and effluent measured pH during Na2CO3 flood through Minnelusacore; (b) Chemical composition of effluent; (c) Mineral saturation status of possiblesecondary minerals of collected samples.
0 5 10 15 20 2510−6
10−5
10−4
10−3
10−2
10−1
Al Ca Cl K Mg Na SO
SiO
SO
Ca NaBO injection Brine injection
0 5 10 15 20 2510−10
10−6
10−3
100
102
PV
Anhydrite Portlandite Brucite Saturation line
NaBO injection Brine injection
PV
(c)
Min
eral
sat
urat
ion
(Q/K
)C
ompo
nent
s in
flui
d (m
olal
)
Fig. 5. (a) Inlet and measured pH at effluent during NaBO2 flood through Minnelusacore; (b) Chemical composition of effluent; (c) Mineral saturation status of somepossible secondary minerals of collected samples.
no additional molar fraction of sulfate was produced (Fig. 5b) andboth calcium and sulfate ions were produced at almost equal molarfractions during the injection period. This latter result is attributedto the lack of effective calcium sink under test conditions. Fig. 5cshows all the collected samples were under-saturated with respectto shown minerals, but the saturation ratio of brucite is close to one,showing that brucite might precipitate. If this is the case, calcitemight form as well due to the high concentration of calcium inaqueous solution and intensified dolomite dissolution, which cansupply required carbonate ion. By having no strong evidence of sec-ondary mineral precipitation and pH buffering, NaBO2 is found inprinciple to be a good candidate for alkali flooding under this coreflooding condition. However, EW need to consider the reactionroutes and upscaling. For instance, longer-term bottle test demon-strate that this alkali can and will precipitate in a matter of weeks.
Berea
M. Kazempour et al. / Fuel 92 (2012) 216–230 221
3.1.4. Formation damage in the presence of anhydriteTo track possible permeability damage of Minnelusa cores un-
der different alkali floods due to rock-fluid interactions, the pres-sure drop was continuously recorded during each flood, asshown in Fig. 6. Changes in inlet-to-outlet pressure drop were usedas a permeability damage indicator. If a significant increase in pres-sure drop upon injection of an alkali occurs, this is taken as a signof formation damage. It was observed that among the utilized alka-lis, NaOH caused the most severe permeability damage in presenceof anhydrite. Low-permeability rock is more prompt to permeabil-ity damage, but when compared to Na2CO3 results in similar rocksamples, the observed damage is much higher, which clearly evi-dence secondary minerals precipitation. No significant permeabil-ity damage was observed during NaBO2 flooding.
3.2. Results of high pH flood in Berea sandstone
3.2.1. Results of NaOH, Na2CO3 and NaBO2 floodsAs shown in Fig. 7, there is no sign of significant pH buffering by
Berea rock in these tests for the entire time period of flooding.There is a small pH drop after shut-in for Na2CO3 and NaBO2 that
0 5 10 15 20 25 30 350
10
20
30
40
50
60
70 Minnelusa
Inj. PV
Pre
ssur
e dr
op (p
si)
NaOH with Kair=3.4 md
Na2CO3 with Kair=6 md
NaBO2 with Kair=71 md
Shut−in for 15 hrs
Brine Inj. Brine Inj. Alkali Inj.
Fig. 6. Pressure drop along the core during the injection of brine & different alkalisolutions through Minnelusa core.
Fig. 7. Inlet and measured pH at effluen
might be a sign of possible long-tem alkali consumption. Producedamounts of silica, potassium and aluminum concentrations werehigher for NaOH than for other alkalis, probably caused by higherdissolution rate of quartz and aluminum–silicate clays and miner-als under higher pH condition at elevated temperature. Very lowconcentration of calcium and magnesium were produced in allthe experiments of this part attributed to the lack of anhydrite inthe system and lower contribution of dolomite dissolution usingBerea rock samples.
3.2.2. Pressure dropPressure drop data of this set is depicted in Fig. 8. Unlike the re-
sults of previous test on Minnelusa, none of utilized alkali resultedin severe permeability damage in Berea core. The equivalent datafrom NaOH, Na2CO3 and NaBO2 floods were somewhat noisy, butserved to establish that the permeability damage was not signifi-cant, when compared to the former case. The latter can be attrib-uted to the fact that no secondary minerals were formed during
t for different alkali floods in Berea.
0 5 10 15 20 25 30 350
0.5
1
1.5
2
2.5
3
3.5
Inj. PV
Pre
ssur
e dr
op (p
si)
NaOH with Kair=204 md
Na2CO3 with Kair=207 md
NaBO2 with Kair=179 md
Shut−in
Brine Inj. Alkali injection Brine Inj.
Fig. 8. Pressure drop along the core during brine & different alkaline floods throughBerea core.
222 M. Kazempour et al. / Fuel 92 (2012) 216–230
alkali injection through Berea core to plug some pore throats caus-ing permeability reduction.
3.3. Geochemical modeling results
3.3.1. Predicted results of calibrated geochemical base modelThe predicted results of base model are plotted in Fig. 9. There is
a good agreement for pH and cations concentration betweenpredicted and measured values. The discrepancy shown between
0 5 10 15 20 252
4
6
8
10
12
14
Inj. PV
pH
0 5 10 15 20 250
20
40
60
80
100
Inj. PV
SiO
2 con
c. (p
pm)
0 5 10 15 20 250
10
20
30
40
Inj. PV
Mg2+
con
c. (p
pm)
Measured Predicted
Measured Predicted
Measured Predicted
Shut-in
Brineinjection
Alkaliinjection
Fig. 9. Comparison between predicted and measured pH and also pred
Table 5Trace element analysis (in Minnelusa sample (C13)).
Element Ti Sr Ba Zr Co MnConc. (ppm) 397.6 331.2 59.4 53.5 36.4 29.4
predicted and measured sulfate concentration might be explainedas follows: barium and strontium present in this Minnelusa corecan react with sulfate to form barium (Ba) and strontium (Sr) sul-fates, two well-known insoluble precipitates. This was examinedby X-ray Fluorescence (XRF) to determine the amount of trace ele-ments in one Minnelusa core sample (Table 5). As shown, bariumand strontium are indeed found in this rock. Although the reportedvalues are not sufficient to cause this much difference, their exis-tence in this type of sandstone can contribute to consume sulfate.
0 5 10 15 20 25
200
400
600
800
1000
1200
Inj. PV
Ca2+
con
c. (p
pm)
0 5 10 15 20 250
2
4
6
8
10
Inj. PV
Al3+
con
c. (p
pm)
0 5 10 15 20 25
2000
4000
6000
8000
10000
Inj. PV
SO
42- c
onc.
(ppm
)
Measured Predicted
Measured Predicted
Measured Predicted
icted and measured different ions concentrations for base model.
Ce Cr As Rb Nd V Ta Ni19.5 13.5 12.1 9.3 9.1 4.8 4.8 4.3
M. Kazempour et al. / Fuel 92 (2012) 216–230 223
A second possibility is gypsum formation (or replacement of anhy-drite by gypsum) which can reduce sulfate concentration in aque-ous phase. Both possibilities can lead to the formation of sulfateprecipitates that create conditions that explain the observed per-meability damage of Minnelusa core during NaOH flood. Regard-less to the type of sulfate sink, this does not affect theconcentration of calcium in aqueous phase because there is alwaysa balance between anhydrite dissolution and portlandite deposi-tion. The presence of sulfate sink can affect anhydrite dissolutionand portlandite precipitation without modifying the calcium con-centration. Due to the important role of portlandite precipitationin the aforementioned mechanisms, XRD and scanning electronmicroscope (SEM) tests were run on C-10 sample after NaOH flood-ing to prove its formation. The results (both XRD and SEM) (Fig. 10)clearly shows formation of portlandite during NaOH flood inMinelusa cores.
3.3.2. Predictability of base modelThe based model was used to predict the results of other floods
(including remaining Minnelusa and Berea samples). It should bepointed out that for each evaluation test only the initial volume
Fig. 10. The results of XRD (upper) and SEM (low
of quartz was changed to consider porosity changes in differentcores keeping other parameters unchanged. Quartz was chosen be-cause the results have the lowest sensitivity with respect to quartzinitial volume among present minerals. Fig. 11 shows a comparisonbetween measured and predicted pH values for Na2CO3 flood inMinnelusa. As depicted, the predictive model shows high pH buf-fering capacity by rock in tests similar to measured data. However,at some points, the measured pH data are lower than the predictedvalues and their trend is not linear, showing more complex buffer-ing mechanisms, which were included in this test during Na2CO3.Despite the small discrepancy between measured and predictedpH values, there is a reasonable match between predicted ions con-centration and experimental data in this complex system. Unlikethe predicted sulfate concentration in the base model, there is aclose match between predicted and measured sulfate. Similar tothe results from NaBO2 flood experiment in Minnelusa, no pH buf-fering was predicted by the model either (Fig. 12) and the differ-ence between measured and predicted pH values representsuncertainties involved in thermodynamic databases (Table 6). Inthe case of alkali flood in Berea, in which anhydrite is not present,our model did not predict any strong alkali buffering by this rock
er) tests on C-10 sample after NaOH flood.
0 5 10 15 20 25 302
4
6
8
10
12
14
Inj. PV
pH
0 5 10 15 20 25100
200
300
400
500
600
700
800
900
1000
1100
1200
Inj. PV
Ca2+
con
c. (p
pm)
0 5 10 15 20 250
5
10
15
20
25
30
35
40
Inj. PV
Mg2+
con
c. (p
pm)
0 5 10 15 20 25
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
11000
Inj. PV
SO
42- c
onc.
(ppm
)
Measured Predicted
Measured Predicted
Measured Predicted
Measured Predicted
Alkali injection Brine injection
Shut-in
-Measured inlet pH = 11.09 -Predicted inlet pH = 11.23
Fig. 11. Comparison between predicted and measured pH and also predicted and measured different ions during Na2CO3 flood in Minnelusa.
224 M. Kazempour et al. / Fuel 92 (2012) 216–230
and it is consistent with experimental data (Fig. 13). In this case,small discrepancy between measured and predicted values statesdifferent degree of uncertainty for different materials in these ther-modynamic databases. For instance, the error for NaOH is smallerthan that for Na2CO3 and they both exhibit smaller error than forNaBO2 under this study conditions (Table 6).
3.4. Forward geochemical simulation under high pH flood
In this step, the calibrated base model was used to completeforward simulations. In this forward simulation, a 1D radial modelwas built to see how far high pH conditions extent into the reser-voir during different alkaline injections. To build a 1D radial case,grid blocks with higher density resolution around the wellborewere generated. The advantage of this technique, which includesfiner nodes around the wellbore and coarser ones away from it,is that it avoids numerical dispersion and captures abrupt changesaround the wellbore. The chemistry of formation water was set tomimic water chemistry in Minnelusa sandstone core at the end of3% NaCl flooding.
3.4.1. Forward simulation (1D radial case)The forward simulation in 1D radial case has one injector and
one producer. The properties of this model are summarized inTable 7. This case was selected to show the importance of non-linear velocity profile around the wellbore. pH profile (at 60 �C)and the effect of anhydrite dissolution and secondary minerals pre-cipitation after one pore volume injection of 1 wt% NaOH in 3 wt%NaCl brine are plotted in Fig. 14. As plotted, by flooding the forma-tion with NaOH, anhydrite is dissolved and replaced with portlan-dite and other minor minerals and the net volume changes of theseminerals leads to create more porosity around the wellbore. Also,this graph shows that the high inlet pH is only attainable justaround the wellbore and the rest of the field remains at lowerpH than the inlet, which is unfavorable for the chemical process.Forward simulations were also completed for the case of Na2CO3,as shown in Fig. 15. Notice that for this alkali, severe pH bufferingoccurs and calcite precipitation is prevalent. Sodium carbonatecannot produce significant alkalinity as long as anhydrite is pres-ent, as the insert shows. This particular alkali is unlikely to be agood candidate for chemical floods, as can be concluded from this
0 5 10 15 20 252
4
6
8
10
12
14
Inj. PV
pH
Measured at effluent Predicted at effluentMeasured inlet pHPredicted inlet pH
Shut-in
Brine injection Alkali injection
Fig. 12. Comparison between predicted and measured pH during NaBO2 flood through Minnelusa core.
Table 6Predicted and measured pH values for initial alkaline solutions.
Alkali solution T = 25 �C T = 60 �C
Measured pH Calculated pH by GWB Difference Predicted by GWB
1 wt% NaOH + 3 wt% NaCl 13.07 13.13 �0.06 12.151 wt% Na2CO3 + 3 wt% NaCl 11.09 11.23 �0.14 10.661 wt% NaBO2 + 3 wt% NaCl 10.36 10.92 �0.55 10.37
0 5 10 15 20 252
4
6
8
10
12
14
Inj. PV
pH
Berea
Measured pH for NaOH
Predicted pH for NaOH
Measured pH for Na2CO
3
Predicted pH for Na2CO
3
Measured pH for NaBO2
Predicted pH for NaBO2
Alkali injection
Brine injection
Shut-in for NaOH & NaBO2
Shut-in for Na2CO
3
Fig. 13. Comparison between predicted and measured pH during different alkali floods through Berea cores.
M. Kazempour et al. / Fuel 92 (2012) 216–230 225
evaluation under single-phase flow conditions. However, thismight change in the presence of crude oil. Similar to the othertwo alkalis, a high pH value is reachable just around the wellbore
after one pore volume injection of 1 wt% NaBO2 (Fig. 16) too. As ob-served, unlike NaOH and Na2CO3, NaBO2 injection does not formsignificant amount of secondary minerals and the dominant
Table 71D radial case properties.
Case Injected fluid Injected pH at 60 �C Injection time (days) Flow rate (m3/day) Field size /(%) H (m) Mineral (vol.%)
1D radial 1 wt% + NaOH 3 wt% NaCl 12.15 365 263.6 Rw = 12.7 cm 20 5 Q: 68.733, D: 2.267A: 3
1D radial 1 wt% NaBO2 + 3 wt% NaCl 10.37 Re = 175 m K: 3K–f: 3
Fig. 14. pH profile, formed secondary minerals and anhydrite status after 1 pore volume injection of NaOH versus radial distance from wellbore.
0 20 40 60 80 100 120 140 160 1807.5
8
8.5
9
9.5
10
10.5
11
X position (m)
pH, t
= 3
65.2
day
s
0 20 40 60 80 100 120 140 160 1801e–7
1e–6
1e–5
1e–4
.001
.01
.1
1
10
X position (m)
Som
e m
iner
als
(vol
ume
%),
t = 3
65.2
day
s
AnhydriteCalcite
Fig. 15. pH profile, formed secondary minerals and anhydrite status after 1 pore volume injection of Na2CO3 versus radial distance from wellbore.
226 M. Kazempour et al. / Fuel 92 (2012) 216–230
Fig. 16. pH profile, formed secondary minerals and anhydrite status after 1 pore volume injection of NaBO2 versus radial distance from wellbore.
Fig. 17. A thin section sample of Minnelusa core under non-polarized and crosspolarized light.
M. Kazempour et al. / Fuel 92 (2012) 216–230 227
mechanism for porosity increase around the wellbore is dissolu-tion of anhydrite. It should be pointed out that compared to NaBO2,NaOH has larger high-pH penetration length into the formation,which might be more desirable in enhanced oil recovery; NaOHcan cause more side problems than NaBO2, such as propagationof a front which is more enriched in sulfate and also permeabilityreduction.
3.5. Thin section analysis
To gain insight into what goes on in these alkaline floods, sev-eral thin sections were taken from the inlet and outlet sectionsof rock plugs used in the experiment. In this study, both plainand polarized lights were used for better detection of anhydritecrystals in Minnelusa’s sample. It is known that sulfate cements,i.e. gypsum, anhydrite, and barite, are present in some clastic sys-tems, e.g. sandstones, particularly those deposited in evaporiticenvironments, where gypsum is precipitated as a cement at shal-low burial depth. Late-stage anhydrite may precipitate at greaterburial depths if an adequate source of SO�2
4 ions is available in sub-surface brines [24]. Gypsum and anhydrite crystals take a varietyof forms including single crystals, radial clusters of crystals, andnumerous types of complex twinned crystals. The crystallographicinformation can be used to detect gypsum and anhydrite in thinsections, because their crystalline features allow one to differenti-ate them from other minerals present. Gypsum crystals can beidentified by characteristic lath-like crystal shapes, weak birefrin-gence and low reliefs. Anhydrite is distinguished from gypsum bya higher relief and stronger birefringence. The results of thin sec-tion analysis for one of the Minnelusa samples (before flooding)under non-polarized and cross polarized light is shown in Fig. 17.Presence of anhydrite particles and their widespread distributionare noticeable in the micrographs. Fig. 18 shows the porosity dif-ference between inlet and outlet cross sections of the cores usedfor NaOH and Na2CO3 respectively. As observed, dissolution ishigher than precipitation at inlets, yielding porosity values larger
Fig. 18. Comparison between porosity values of inlet and outlet face of Minnelusa cores afte NaOH (upper one) and Na2CO3 (lower one) floods.
228 M. Kazempour et al. / Fuel 92 (2012) 216–230
at inlet than at the outlet. Also, the micrographs show a reductionin porosity during the NaOH flood is more significant than that ofNa2CO3 flood (although this is only qualitative comparison), whichis consistent with what was found before in this study.
4. Discussion
The premises of an alkaline flood include the sustenance of highpH conditions with limited formation damage. Significant numberof experiments for screening purposes rely on the use of outcroprock such as Berea. The results shown here clearly indicate thatoutcrop rock would not represent a good analog for rocks contain-ing easily dissolvable minerals such as anhydrite, typically presentin evaporite (sandstone). As previously shown in Fig. 3, sodiumhydroxide is buffered as portlandite precipitates, but pH quicklyrises by continuously injecting the alkali. The observed pH wouldsustain conditions to saponify the crude oil and create surfactant.However, several drawbacks are the significant formation damage(as shown in Fig. 6) and the increase in sulfate concentration(Fig. 4b), especially for the NaOH injection. The former problemwould limit the use of sodium hydroxide in low-permeability for-mations, as secondary mineral precipitation is likely to cause sig-nificant formation damage, as clearly shown in Fig. 6, while thelatter would affect surfactant (synthetic and natural) phase behav-ior. This means that the choice of surfactant in laboratory screeningmust account for the geochemical conditions originating in therock. Traditional bottle tests for surfactant screening use simpleaqueous phase chemistry, i.e. typically sodium chloride solutions,which would not reflect in situ conditions. On the other hand,
the extreme buffering of sodium carbonate, as illustrated inFig. 4, seriously limits the ability of this alkali to provide conditionsfor saponification of acid crude oils (pH should be close to 11 to beeffective). This buffering effect is even more apparent at field scale,as shown in Fig. 15. Significant precipitation of secondary mineralsis also observed with sodium carbonate, althought the formationdamage was not as significant as in the case of sodium hydroxide.In addition, the use of this alkali would lead to increase in sulfateconcentration without the benefits provided by stronger alkalis(e.g. sodium hydroxide) such as eventually sustaining high pH. Fi-nally, sodium metaborate appears to fulfill at least one of the desir-able conditions for an alkali, namely limited formation damage.However, we should not neglect the uncertainties in understand-ing the long-term behavior of this alkali on one hand. It turnsout that our ability to predict this for sodium metaborate is notcompletely adequate and up to one pH unit may be associated inpredicting metaborate behavior in solution. On the other hand,longer-term observations in bottle tests clearly indicate precipita-tion. This signals that sodium metaborate may not be able to pro-vide alkaline conditions necessary for EOR processes requiringhigh-pH conditions in practice.
Upscaling of coreflooding experiments through simulation arequite revealing, as illustrated in Figs. 1,14–16. Portlandite is thedominant precipitant in sodium hydroxide injection, not observedin sodium metaborate flooding. Sodium carbonate leads to precip-itation of calcite, which can cause formation damage. In contrast,however, the amount of calcite precipitation for sodium metabo-rate is significantly higher than for sodium hydroxide and morecomparable to the sodium carbonate case. Also, high-pH frontsare more likely to propagate with sodium hydroxide as opposed
M. Kazempour et al. / Fuel 92 (2012) 216–230 229
to sodium metaborate. It is important to contrast the experimentalobservations, i.e. coreflooding results, with field predictions. De-spite the elevated pH that can be sustained in the cases of sodiumhydroxide and to a lesser extent with sodium metaborate, this isonly attainable after a couple of pore volumes in the lab and whatthe simulation indicates is that the high-pH front rapidly decayswith distance, particularly for sodium carbonate and sodium meta-borate. It might be misleading to extrapolate lab inferences to fieldsituations without proper modeling of field effects. The radial casessimulated here clearly indicate the impact of geometrical effects inthe field and further evidence of applicability or lack thereof of agiven alkali to field situations. In this sense we refer to lab resultsas limited predictors of field observations. However, at least for thesodium carbonate case, the poor expectations derived from core-floods are confirmed by modeling field-size cases.
Notice that the results reported here will be impacted by thepresence oil and the combined effects of the original rock wettabil-ity and the resulting one from the addition of alkali to chemicalblends. The presence of oil should limit access to the rock surface,particularly in oil-wet formations, and therefore will affect therock-fluid interaction. In order to draw conclusions regarding this,additional research must be carefully conducted, because as alkaliis consumed, the possible wettability alteration associated withthe injection of alkali will change. Additionally, some of the alkaliwill be consumed in neutralization reactions, some of which willgenerate desirable soaps, depending on the acid number and com-position of the oil. However, since excess alkali is usually injectedin chemical floods, this sink of alkali should dominate only after asubstantial fraction of the alkali has been consumed through fluid-mineral reactions. A detailed example of this in heavy oil can befound in reference [25].
The results regarding secondary mineral precipitation can be re-lated to large field pilot projects of alkali-surfactant-polymer (ASP)floods in which scales are a severe problem, such as Daqing in Chi-na, where some of the largest ASP pilots have been developed [26].Silicate scales have been found during NaOH-based ASP formula-tions [27], reason why silicate scale inhibitors have been used toprotect production infrastructure. However, carbonate scales havebeen found in conjunction with silicates [28,29]. Calcite has alsobeen found in the production system intermixed with silicates.The modeling strategy developed in this work can contribute tothe prediction of scale formation in downhole and surface facilities,but the extent of formation damage requires additional experi-mental data.
5. Conclusion
1. The impact of anhydrite as a pH buffering mineral in the rockand its contribution to changes in water chemistry during alkaliflood are strongly dependent on alkali used in the flood, i.e. themaximum pH buffering occurred when Na2CO3 solutions areinjected, while the minimum corresponds to NaBO2 flooding.However, this buffering is significant at field scale.
2. Small amounts of anhydrite can limit the maximum attainablepH upon injection of NaOH and more markedly Na2CO3, evenafter several pore volumes of the alkaline solution have beeninjected, although high pH is quickly attained after several vol-umes of NaOH are injected.
3. Conventional alkalis, namely NaOH and Na2CO3, reduce theamount of calcium in the aqueous phase in presence of anhy-drite, but they cannot completely sequester it. On the otherhand, these alkalis will promote an increase in the amount ofsulfate in solution to values three times larger or more thanthe values observed during waterflooding, which represents arisk to the success of any chemical flood.
4. Associated permeability damage by alkali flood can be intensi-fied by presence of anhydrite in mineral assemblage due tothe secondary minerals precipitation particularly for NaOH.
5. pH buffering capacity obtained in linear core flooding tests canbe more significant at field scale, as simulation results predict,indicating the importance of upscaling and forward simulation.
6. A robust geochemical model is a powerful tool that is essentialfor field application design of chemical flooding for which pH,water chemistry and optimum salinity are key factors, as labresults can turn out to be poor predictors of field behavior.
Acknowledgments
We would like to acknowledge the Enhanced Oil Recovery Insti-tute at the University of Wyoming for financial assistance. Theauthors thank the School of Energy Resources (SER) at the Univer-sity of Wyoming for financial assistance through the Anadarko Fel-lowship for Excellence in Energy Scholarship.
References
[1] Johnson Jr C. Status of caustic and emulsion methods. J Petrol Technol1976;28(1):85–92.
[2] Guo JX, Liu Q, Li MY, Wu ZL, Christy AA. The effect of alkali on crude oil/waterinterfacial properties and the stability of crude oil emulsions. Colloids SurfacesA – Physicochem Eng Aspects 2006;273(1–3):213–8.
[3] Li MY, Lin MQ, Wu ZL, Christy AA. The influence of NAOH on the stability ofparaffinic crude oil emulsion. Fuel 2005;84(2–3):183–7.
[4] Rudin J, Wasan DT. Mechanisms for lowering of interfacial-tension in alkaliacidic systems – effect of added surfactant. Ind Eng Chem Res 1992;31(8):1899–906.
[5] Nedjhioui M, Moulai-Mostefa N, Morsli A, Bensmaili A. Combined effects ofpolymer/surfactant/oil/alkali on physical chemical properties. Desalination2005;185(1–3):543–50.
[6] Liu Q, Dong MZ, Yue XG, Hou JR. Synergy of alkali and surfactant inemulsification of heavy oil in brine. Colloids Surfaces A – Physicochem EngAspects 2006;273(1–3):219–28.
[7] Hirasaki GJ, Miller CA, Puerto M. Recent advances in surfactant EOR. SPE115386, Presented at the SPE annual technical conference and exhibition,Denver, Colorado, USA; 2008.
[8] Ehrlich R, Hasiba H, Raimondi P. Alkaline waterflooding for wettabilityalteration-evaluating a potential field application. J Petrol Technol1974;26(12):1335–43.
[9] Wyatt K, Pitts MJ, Surkalo H. Mature waterfloods renew oil production byalkaline-surfactant-polymer flooding. SPE 78711, Presented at SPE easternregional meeting, Lexington, Kentucky, USA; 2002.
[10] Ehrlich R, Wygal RJ. Interrelation of crude-oil and rock properties withrecovery of oil by caustic waterflooding. Soc Petrol Eng J 1977;17(4):263–70.
[11] Sydansk RD. Elevated-temperature caustic sandstone interaction –implications for improving oil-recovery. Soc Petrol Eng J 1982;22(4):453–62.
[12] Larrondo L, Urness C. Laboratory evaluation of sodium hydroxide, sodiumorthosilicate, and sodium metasilicate as alkaline flooding agents for a westerncanada reservoir. SPE 13577, Presented at the SPE oilfield and geothermalchemistry symposium, Phoenix, Arizona, USA; 1985.
[13] Cheng K. Chemical consumption during alkaline flooding: a comparativeevaluation. SPE 14944, Presented at the SPE enhanced oil recovery symposium,Tulsa, Oklahoma, USA; 1986.
[14] Dezabala EF, Vislocky JM, Rubin E, Radke CJ. A chemical theory for linearalkaline flooding. Soc Petrol Eng J 1982;22(2):245–58.
[15] Jensen J, Radke C. Chromatographic transport of alkaline buffers throughreservoir rock. SPE Reservoir Eng 1988;3(3):849–56.
[16] Labrid J, Bazin B. Flow modeling of alkaline dissolution by a thermodynamic orby a kinetic approach. SPE Reservoir Eng 1993;8(2):151–9.
[17] Soler JM. Reactive transport modeling of the interaction between a high-phplume and a fractured marl: the case of wellenberg. Appl Geochem2003;18(10):1555–71.
[18] Kazempour M, Sundstrom E, Alvarado V. Effect of alkalinity on oil recoveryduring polymer floods in sandstone. SPE 141331, presented at the SPEinternational symposium on oilfield chemistry, The woodlands, Texas, USA;2011.
[19] Zhang J, Nguyen QP, Flaaten A, Pope GA. Mechanisms of enhanced naturalimbibition with novel chemicals. SPE Reservoir Eval Eng 2009;12(6):912–20.
[20] Bethke CM, Yeakel S. The geochemist’s workbench professional. Uinversity ofIllinois; 2009.
[21] Bethke CM. Geochemical and biogeochemical reaction modeling. 2nd ed. NewYork: Cambridge University Press; 2008.
[22] Palandri JL, Kharaka YK. A compilation of rate parameters for water-mineralinteraction kinetics for application to geochemical modeling. Tech. rep. USGeological Survey, Open File Report; 2004.
230 M. Kazempour et al. / Fuel 92 (2012) 216–230
[23] Todd A, Yuan M. Barium and strontium sulfate solid-solution scale formationat elevated temperatures. SPE Prod Eng 1992;7(1).
[24] Boggs JRS. Petrology of sedimentary rocks. 2nd ed. New York: CambridgeUniversity Press; 2009.
[25] Manrique E, Villaba M, Mendez Z, DerSarkissian J. The effect of crude oilcomposition on aqueous phase-rock interaction: implications on formationdamage in the enhanced recovery of heavy oil. SPE 27391, Presented at the SPEintl symposium on formation damage control, Lafayette, Lousiana; 1994.
[26] Wang Y, Liu J, Liu Y, Wang H, Chen G. Why does scale form in asp flood? how toprevent from it? – A case study of the technology and application of scalingmechanism and inhibition in asp flood pilot area of n � 1dx block in daqing.SPE 87469, Presented at the 6th international symposium on Oildfield Scale,Aberdeen, UK; 2004.
[27] Qing J, Bin Z, Ronglan Z, Zhongxi C, Yuchum Z. Development and application ofa silicate scale inhibitor for asp flooding production scale. SPE 74675,Presented at the SPE oilfield scale symposium, Aberdeen, UK; 2002.
[28] Gang C, He L, Gouchen S, Gouqing W, Zhongwen X, Huafeng R, et al. Technicalbreakthrough in pcp’s scaling issue of asp flooding in daqing oil field. SPE109165, Presented at the 2007 SPE annual technical conference and exhibition,Anaheim, California, USA; 2007.
[29] Jinling L, Tiande L, Jidong Y, Xiwen Z, Yan Z, Feng Y. Silicon containing scaleforming characteristics and how scaling impacts sucker rod pump in aspflooding. SPE 122966, Presented at the 2009 SPE Asia pacific oil and gasconference and exhibition, Jakarta, Indonesia; 2009.