Society of Petroleum Engineers

SPE 25257

Cleaning Up Spilled Gasoline With Steam: Compo~itional SimulationsA.E. Adenekan, Exxon Production Research Co., and T.W. Patzek, * U. of California

'SPE Member

Copyright 1993, Society of Petroleum Engineers, Inc.

This paper was prepared for presentation at the 121h SPE Symposium on Reservoir Simulation held in New Orleans, LA, U.S.A., February 28-March 3, 1993.

This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper,as presented, have not been reviewed by the Society of Petroleum Engineers and are sUbject to correctlo~ by the author(s). The.ma~enal, ~s presen~ed: does not necessanly reflectany position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE me~tmgs are subject to publicatIon revIew by Edlton~1 Comm~ttees of the SocIetyof Petroleum Engineers. Permission to copy is restricted to an abstract of not more than 300 words. lJIustrallOns ~ay not be copied. The abstract should contam conspIcuous acknowledg­ment of where and by whom the paper is presented. Write Publications Manager, SPE, P.O. Box 833836, RIchardson, TX 75083-3836, U.S.A. Telex, 163245 SPEUT.

Abstract

A finite-difference compositional simulator has beendeveloped and tested at D.C. Berkeley to model theflow of mixtures of Nonaqueous Phase Liquids(NAPLs) through the air zone and into aquifers. Thesimulator has been successfully used to history-match asteam injection pilot at a 'Clean Site' near the LawrenceLivermore National Laboratory in California -- a testsite for the Gasoline Spill Area (GSA) cleanup pilotplannned for early '93. Because of its multicomponentcapabilities, the simulator has been used to calculate(a) production rates of individual gasoline componentsin the GSA to size treatment facilities, (b) areal andvertical distribution of gasoline after the first cycle ofsteam injection, and (c) steam injection rate that limitsgrowth of the steam zone beyond the cleanup area. Ithas been shown that gasoline present in the penneablesands and gravel layers can be successfully recoveredby injecting steam into these layers in a 7-spot pattem.For the conditions asswned in the model, it will takeless than 16 days to recover nearly all of the gasoline inthe sands and gravel layers. By that time, themaximwn aqueous concentrations of all hydrocarboncomponents in these layers will have dropped to lessthan 0.01 mg/I. The results show that vaporization,followed by bulk movement of the vapor to theproduction well is the dominant recovery mechanism.In tenns of time required for cleanup, model results aremost sensitive to penneability of the mediwn. Otherparameters, such as the relative penneabilities and thenwnber of components, also affect the outcome, but toa lesser extent.

261

Introduction

A compositional simulator M2NOTS (MulticomponentNon-isothermal Organics Transport Simulator) hasbeen developed [3,4] at D.C. Berkeley to model theflow of mixtures of Nonaqueous Phase Liquids(NAPLs) through the vadose zone and into aquifers.The NAPLs can have arbitrary densities, boilingtemperatures, and viscosities. The M2NOTS simulatoris a major extension of an existing two-phase (water &gas), two-component (H20 & air), and nonisothermalsimulator TOUGH2, developed at the LawrenceBerkeley Laboratory [1,2] for geothermal applications.The M2NOTS integrated finite-difference model is (a)three-dimensional; (b) fully implicit; (c) three-phase(aqueous, gaseous, and NAPL); (d) non-isothennal ifneeded; and (e) multicomponent (water, air, and anynwnber of hydrocarbon components). Thecompositional and multiphase part of M2NOTS has thefollowing features: (a) each component of every phasein a grid block may partition into every other phase; (b)each phase in a grid block may appear or disappear; (c)the appearance or disappearance of a phase isestablished by a multicomponent, isothennal flashcalculation performed at each iteration; and (d) therelative penneabilities are calculated from generalizedpower law equations for two-phase flow and the StoneII model for three-phase flow. The M2NOTS wellmodel can handle multiple fluid or heat injection wellson pressure or rate constraints, and fluid producers withvariable pwnp levels and on deliverability or pwnppressure constraints.

2 Cleaning up Spilled Gasoline with Steam: Compositional Simulations SPE25257

In the last twenty years, sophisticated thennal andcompositional simulators have been developed tomodel oil reservoirs and it appears that these codesmay also be used to study NAPL contaminationproblems. However, usually this is not the casebecause oil and NAPL transport and recovery aredominated by different mechanisms. For example,reservoir engineers are generally not interested in therelatively insignificant quantity of oil dissolved in thewater phase; neither is diffusion considered important.On the other hand, dissolution and transport of organiccompounds in the water phase, and diffusion of organicvapors in the gas phase may be quite important inNAPL contamination studies. Because oil reservoirsare generally deep and confined, oil industry simulatorsusually assume no-flux boundaries for a modeleddomain. In subsurface contamination problems, we aregenerally dealing with shallow systems, and are usuallyinterested in evaluating the exchange of contaminantsbetween the subsurface and the atmosphere. Yetanother example of how the differences in emphasisenter into code tonnulation is the treatment ofappearance and disappearance of phases. Most oilreservoir codes assume that the oil phase cannotcompletely disappear from a grid block. This isjustifiable because most crude oils contain heavy,nonvolatile components. In contrast, the goal of manyremediation efforts is to completely remove a NAPL.Therefore, codes used to study transport of NAPLsneed to be more flexible in dealing with the appearanceand disappearance of phases. For example, theM2NOTS model has a sophisticated algorithm forchoosing the optimal sets of unknowns for all sevenpossible phase combinations (water, NAPL, Gas, Water+ NAPL, Water + Gas, Gas + NAPL, Water + NAPL +Gas) and no component is required to have a 'master'phase to calculate its partitioning.

Initially, we validated [3,5] the M2NOTS simulatorwith a laboratory, one-dimensional, 'two-component(benzene-toluene), water and steamflood data set [6].We also used M2NOTS to history-match a shallowsteam injection pilot at a 'Clean Site' near the LawrenceLivennore National Laboratory [3,7] -- a test for theGasoline Spill Area (GSA) cleanup project [7] to bestarted in early '93. Apart from the GSA, the CleanSite is perhaps the most extensively imaged anddescribed subsurface volume in the world and,therefore, matching its response was a crucial test ofthe simulator. In the Clean Site simulation study, wehave achieved good agreement with the field data, andmatched (a) the injection and production rates, (b) thetemperature histories at two monitoring wells, and (c)the steam breakthrough time at the production well.

262

Because the M2NOTS simulator is compositional, wehave used it to evaluate how effectively can we removegasoline spilled at the GSA site by injecting steam. Inparticular, we have calculated (a) production rates ofindividual gasoline components to design treatmentfacilities), (b) areal and vertical distribution of gasolineafter the first cycle of steam injection, and (c) steaminjection rate that limits growth of the steam zonebeyond the cleanup area.

Simulations of Gasoline Spill Area

Soil and groundwater contamination at the LawrenceLivennore National Laboratory Gasoline Spill Area(GSA) resulted from leaking underground storagetanks. The size of spill is not known and estimatesrange from 10,000 to over 100,000 gallons. It isknown, however, that the gasoline was released from1952 to 1979 and the leaking tanks were removed in1980. Extensive field studies have been perfonned atthis site over the last five years [8,9] to evaluate theextent of the contamination, and obtain anwlderstanding of the hydrogeology of the site.

The GSA area is underlain by an unconsolidated,heterogeneous deposit of interbedded gravels, sands,silts and clays, as well as mixtures containing varyingproportions of these soil types. Average staticgroundwater level at the site is about 30 meters belowthe ground surface. Below a depth of 40 meters, thereis a continuous, low penneability silts and clays layerthat is at least 9 meters thick in most places. Chemicalanalyses of soil and groundwater samples show thatthis layer is not contaminated and, therefore, we havelimited our simulations to the soils above it. Thebottom silts and clays unit is overlain by a unifonnlythick (about 3 meters) layer of sands and gravels,designated as the Lower Aquifer. The Lower Aquiferin tum is overlain by another fairly continuous layer ofsilts and clays that is 3 to 5 meters thick at mostlocations. This unit is referred to as the Aquitard.Another sands and gravels unit, called the UpperAquifer, extends above the Aquitard and is about 7.5meters thick in the central section of the site. Thelocation of the water table at the GSA is such that thelower 2 meters the Upper Aquifer is saturated.Overlying the Upper Aquifer is an 18-meter layer ofsilts and clays containing scattered zones of sandy andgravelly materials.

Non-aqueous phase gasoline is present both above andbelow the water table at the GSA. A 7.5-meter rise inthe regional water table that occurred after the spill has

SPE25257 A. E. ADENEKAN and T. W. PATZEK 3

caused some gasoline to be trapped below the currentwater table. Data shows that most of the gasolinetrapped below the water table is at the interfacebetween the Lower Aquifer and the Aquitard.Contours of benzene concentrations in soil samplestaken from the Lower Aquifer [13] are shown in Fig. 1.A significant portion of the gasoline trapped in theunsaturated zone has been removed by vacuwnextraction operations that have been carried out in thelast few years. In addition, biodegradation seems to besignificant in the unsaturated zone with low residualgasoline saturation [7]. However, neither vacuumextraction nor biodegradation has significantly changedthe conditions immediately above and below the watertable. Thus, it has been decided [7] that steaminjection coupled with vacuwn extraction and electricalresistance heating will be used to recover the residualgasoline near and below the water table. Electricalheating will be used to heat up the Aquitard which,because of its low permeability, is not expected to bepercolated by steam. Current plans call for sixinjection wells around the periphery of the spill andone production well in the center of the spill (Figs. 1and 2). Steam will be injected into the Upper andLower Aquifers while the Aquitard is electricallyheated.

Figs. 3 and 4 show the vertical and plan view of thefinite-difference mesh chosen for this study. We haveidealized the proposed arrangement of injectors andproducer as a seven-spot pattern and asswned that flowand transport within the pattern possess symmetry.Therefore, simulation of only one-twelfth of the patternis required. The model consists of four horizontallayers: the Upper Aquifer, the Aquitard, and two layersthat represent the Lower Aquifer. The Lower Aquiferhas been divided into two layers to account for thedifference in current gasoline saturations. The modellayer representing the high gasoline saturation zone ofthe Lower Aquifer is 0.6 meter thick. The entire modelconsists of 506 grid blocks. The thick silts and clayslayer above the Upper Aquifer and the one below theLower Aquifer are the over- and underburden in themodel. This is supported by results of the Clean Sitesimulations that showed minimal percolation of steaminto the low permeability wlits that were included inthat model [3,7].

Table 1 lists the soil properties in the Base Run. Thepermeability is based on the results of several pumpingtests that have been performed at the site and ranked[13] as poor, fair, good, or excellent. Only the top twocategories were used in obtaining tlle current estimatesof permeability, but the results were scattered

263

nevertheless. The more reliable ones yieldedpermeabilities between 0.5 and 40 darcies for the sandsand gravels, with more tllall half of the data pointsfalling between 1 and 7 darcies. Hence the value of 5darcies was used in the Base RWI. The silts and clayswere assigned permeability of 10 millidarcy. Thevertical to horizontal permeability ratio was set to one­tenth.

Field-measured relationships were not available forrelative permeability and capillary pressure functions.Therefore, the Stone II model was used to calculate thethree-phase relative permeabilities, while the gas­NAPL and NAPL-water capillary pressures werecalculated from the modified van Genuchten'sfunctions [10] . The parameters appearip.g in thesefunctions have been assigned values based on literaturedata published for similar soil types.

The initial composition of the NAPL used in the modelis based on the results of a laboratory analysis of asample of the gasoline found at the site [13]. Becausethis analysis does not provide sufficient breakdown ofthe gasoline into its constituents (percentageconcentrations are given for groups of compowlds suchas paraffins), it was necessary to use additionalliterature data on the composition of regular gasolines[11,12]. Because the computer CPU and storagerequirements limit the number of components that Callbe included in a simulation model, threepseudocomponents were chosen to represent the spilledgasoline, benzene, p-xylene, alld n-deCalle. Themixture consisted of 2.2 mole % of benzene, 74.1 % ofp-xylene, and 23.7% of n-decane. Despite its lowmole fraction, benzene was included as a separatecomponent because of its high solubility in waterrelative to other gasoline components and itsimportance as an environmental contaminant; p-xylenerepresented the C7 - C9 hydrocarbons, while n-deCallethe C10 - C12 hydrocarbons. In choosing pseudocomponents to represent the gasoline, other factors thatmight affect its transport were also taken intoconsideration [3]. Among these were the boiling pointdistribution curve, aqueous solubility, and viscosity.

Initial condition for the steam injection simulations wasthat of gravity-capillary equilibrium. The initial NAPLdistribution (Fig. 5) reflected roughly the data fromchemical analyses of soil and growldwater samples[13]. The total amount of NAPL initially in-placewithin the element of symmetry was 3,500 gallons(42,000 gallons within the entire 7-Spot pattern). Asindicated earlier, no-flow conditions were imposed atthe upper and lower boundaries, although heat losses to

4 Cleaning up Spilled Gasoline with Steam: Compositional Simulations SPE25257

the confining layers were allowed. Ambienttemperature and equilibrium hydrostatic pressure werespecified at the outer vertical boundary of the model.

Steam was injected into both the Upper and LowerAquifer. The constant injection pressure was specifiedas 3.5 x 105 Pa and the quality of the injected steamwas 90%. The production well was open to all layersand it produced on deliverability against a bottomholepressure of 0.6 x 105 Pa, i.e., a pump vacuum of about0.4 atmospheres. Each steam injection simulation wasallowed to proceed until some time after steambreakthrough at the producer.

Results of the Base Run

Figs. 6-8 show the simulated recovery rates of benzene,p-xylene, and n-decane in the gas phase, the aqueousphase, and as part of a separate hydrocarbon (NAPL)phase. Recovery rates of all hydrocarbon componentsin both the gas and NAPL phase show two distinctpeaks. Recovery rates of benzene and p-xylene in thewater phase also show peaks, but they are lesspronounced. These peak rates coincide with steambreakthrough at the producer in model layers thatcontained some NAPL initially. The first peakcorresponds to steam breakthrough in the UpperAquifer while the second one to breakthrough in theLower Aquifer. Steam breakthrough at the produceroccurs earlier in the Upper Aquifer than in the LowerAquifer for two reasons: (i) steam is injected into theUpper Aquifer at a higher rate because the fonnationfluid pressure is lower than in the Lower Aquifer wherethe additional hydrostatic head adds to the pressure;and (ii) initially, the Upper Aquifer is only partiallysaturated with water and gas mobility is higher. Steambreakthrough occurs in the Upper Aquifer after about 7days of steam injection while it takes about 12 days forbreakthrough to occur in the Lower Aquifer.

The rate of recovery of each component as part of ahydrocarbon liquid phase is significant only for a shorttime interval immediately preceding steambreakthrough. Although the initial NAPL saturation isbelow residual everywhere, evaporation of hydrocarboncomponents into the steam zone and their condensationjust ahead of the steam condensation front cause thelocal NAPL saturation to rise above residual and theNAPL be mobile. The locally higher NAPL saturationahead of the steam condensation front is referred to as aNAPL bank. Another important feature of the resultsshown here is the overlap between the time intervalduring which there is significant recovery in the NAPL

264

phase and the time interval of significant recovery inthe gas phase, especially in the Lower Aquifer. Thisimplies that even in an initially water saturated aquifer,the gas phase may extend beyond the steam zone. Twomechanisms are responsible for the emergence of a gasphase ahead of the steam condensation front: (i) thedevelopment of a heated zone downstream from thisfront; and (ii) enhanced vaporization or even boiling ofthe more volatile NAPL components in the heatedzone. Figs. 6-7 show that, as one might expect, theprominence of this feature increases with volatility ofthe organic compound -- it is more pronounced for themore volatile benzene than for p-xylene, and it isalmost absent for n-decane.

Cumulative recoveries of the hydrocarbon componentsare shown in Figs. 9-11 and Fig. 12 shows thefractional recovery of hydrocarbon components fromLayers 1 and 3 within each phase. As indicated earlier,initial conditions in Layers 1 and 3 are different. Layer3 was initially saturated with water, whereas Layer 1was only partially saturated. Moreover, NAPL wasinitially present in Layer 1 only in the immediate areaaround the producer, that is the initial average NAPLsaturation in Layer 1 was significantly less than that inLayer 3. This is why these two layers responddifferently to steam injection. With the exception ofsome benzene recovered in the aqueous phase, almostall contaminant recovery from Layer 1 is in the gasphase. Because this layer has a high initial gassaturation, gas is able to move toward the producersoon after injection starts, evaporating hydrocarboncomponents along the way. This continues throughoutthe injection period. Another reason for the dominanceof gas phase recovery is that the total volume of NAPLinitially present in Layer 1 is insufficient to sustain amobile NAPL bank. Therefore, even when a NAPLbank fonns, it eventually evaporates into the mobilegas phase recovered at the producer. The situation isdifferent for the initially water-saturated Layer 3.Recovery in the gas phase accounts for 55%, 84%, and91 % respectively, of the total benzene, p-xylene, and n­dec~U1e recovered from Layer 3. Recovery in the waterphase accounts for 22% of the total benzene and 2% ofthe total p-xylene recovered. Because of its lowaqueous solubility, no significant amowlt of n-decaneis recovered in the water phase. Twenty-three percentof the benzene, 14% of the p-xylene, and 9% of the n­decane recovered from Layer 3 are recovered as NAPL.These results are consistent with the most importantinterphase mass transfer characteristics of thecomponents: the saturated vapor pressures and aqueoussolubilities. Since the initial NAPL saturation is lessthan residual everywhere, recovery in the NAPL phase

SPE25257 A. E. ADENEKAN and T. W. PATZEK 5

must be preceded by the fonnation of a NAPL bank asdescribed above. The NAPL bank is richer in the morevolatile components because distillation favors suchenrichment. Therefore fractional recovery as NAPLshould be highest for the most volatile compowlds stillpresent in the gasoline.

The simulation results indicate that after 16 days ofsteam injection no free-phase gasoline is left in thepenneable layers. Aqueous phase concentrations ofbenzene in these layers become very low, less than0.01 mg/l everywhere. On the other hand, conditionsin the Aquitard do not change from what they wereinitially. There is some increase in temperature due tothennal conduction, up to about 70'C near the injector.Free-phase gasoline initially present in the Aquitardremains after the 16 days of steam injection. Asobserved at the Clean Site, the low penneability of theAquitard does not allow any significant amount ofsteam to enter and flow through it. Therefore, it hasbeen decided [7] that gasoline recovery from theAquitard will be achieved by resistance heating andvacuwn extraction. Resistance heating will be used todry up the Aquitard, creating a continuous andtherefore mobile gas phase. This aspect of theremediation plan is not included in the work presentedhere.

The rates of steam injection into the Upper- and LowerAquifer are shown in Fig. 13. As indicated earlier, theinjection is pressure-constrained. In both layers, therate of injection is highest at the beginning because thefonnation pressure is lowest at that time. As steaminjection continues, the fonnation pressure increases,and the difference between the wellbore pressure andthe fonnation pressure diminishes. This in turn leads toa decrease in the rate of steam injection. It should benoted that the cyclic variations or oscillations in theinjection rates seen in Fig. 13 are caused by the rathercoarse spatial discretization in the nwnericalsimulations. A comparison of Figs. 13a and 13b showsthat the cyclic variations are only significant forinjection into an initially water-saturated mediwn. Inmultiphase flow problems that also involve phasechange, calculated grid block pressures wldergo cyclicvariations as phase conditions in grid blocks change[1,3]. Consider how the steam condensation front ispropagated in the finite-difference grid within Layers 3and 4 (the Lower Aquifer). As water vapor enters agrid block that is just downstream from the steam front,it condenses, raising the temperature of this grid block.Eventually, this temperature reaches the boiling pointof water at the prevailing pressure and the entire gridblock makes a transition a two-phase condition. Water

265

vapor continues to enter the grid block, condensing inpart, and in part increasing gas saturation. Theenthalpy of condensation goes to increase thetemperature and pressure in the grid block, while steammobility increases with saturation. At some point thesteam begins to flow to the next downstream grid blockwhere the process is repeated. The pressure variationdescribed here propagates backward through the steamzone, causing the oscillation of the steam injection rateshown in Fig. 13. This oscillation is eventuallydampened out by the high compressibility of the steamzone between the injection well and the condensationfront.

The total rate of water production is shown in Fig. 14.During the first four days, almost all the water isproduced from the Lower Aquifer. The initial mobilityof water in the Upper Aquifer is low because this unitis unsaturated (the specified initial water saturation inthis layer is 0.2, the irreducible saturation is 0.1).During this period, steam injected into the UpperAquifer layer condenses, raising the water saturationthere. Water production from the Upper Aquiferbecomes significant after about four days of steaminjection, and increases until the condensation frontbreaks through at about 7 days. Following thisbreakthrough, steam injected into the Upper Aquifer isproduced with little condensation taking place.Because steam is less dense than water, the mass rateof production from the Upper Aquifer decreases. Theproduction rate drops again after about 12 days ofsteam injection. This coincides with the time ofbreakthrough in the Lower Aquifer, and is also due tothe difference in density between liquid water beingproduced prior to breakthrough and steam that isproduced after it.

The total mass of steam injected during the 16-daysimulation is 2200 m3 of cold water equivalent (CWE)for the entire 7-spot pattern. Seventy-five percent ofthis amount is injected into the Upper Aquifer and theremainder into the Lower Aquifer. In principle, it ispossible to reduce the mass of injected steam ifinjection into the Upper Aquifer is stopped soon aftersteam breakthrough there. The results of anothersimulation show that 1600 m3 (CWE) of steam wouldbe injected under that scenario. About 20% of the totalinjected heat is lost to the confining units.

Sensitivity Runs

A nwnber of sensitivity runs were made to evaluate theinfluence of several key parameters on the results

6 Cleaning up Spilled Gasoline with Steam: Compositional Simulations SPE25257

presented above. In Sensitivity Run #1 thepermeability of the sands and gravels units waschanged from 5 to 10 darcy. As one would expect (Fig.14), there is a nearly linear relationship between thetime of steam breakthrough and permeability of themedium. The relative permeability of the NAPL wasvaried in Sensitivity Runs #2 and #3. In Run #2, theexponents for the two-phase NAPL relativepermeabilities (Dow and nog) were changed from 1.5 to2.0. In Sensitivity Run #3 the residual NAPLsaturations (Sorw andl Sorg) were changed from 0.055to 0.085. Results of both simulations show that theNAPL relative permeability can significantly affect thequantity ofhydrocarbons recovered in the NAPL phase.Changes made here to the exponents and the residualNAPL saturation both have the effect of reducing theNAPL relative permeability. Therefore, the quantity ofhydrocarbons recovered as NAPL in both simulations issignificantly lower than that in the Base Run (Figs. 15­17). There is a corresponding increase in the quantityof hydrocarbons recovered in the gas phase. Theoverall recovery in all the phases shows very littlesensitivity to these parameters, although the time toachieve a certain level of recovery may be somewhatdifferent. Sensitivity Run #4 was designed to test thesignificance of the number of hydrocarbon componentsused to represent gasoline. Five hydrocarboncomponents were used in this run, compared to threeused in the Base Run. The components used torepresent gasoline in each of these two runs, and theirmole fractions are shown in Table 2. A comparison ofthe simulation results is presented in Fig. 18. In tennsof overall hydrocarbon recovery, there is no significantdifference between the two simulations. However,there are differences in the amount of hydrocarbonsrecovered in each phase. A greater quantity ofhydrocarbons is recovered in the gas and water phasesin the Base Run thatl in Run #4. On the other hand,more separate-phase hydrocarbon is recovered in Run#4 than in the Base Run. The introduction of n-octanein Run #4 to represent some of the componentsapproximated by p-xylene in the Base Run causes thesaturation of the NAPL bank to be higher since n­octane is more volatile than p-xylene. This in tumincreases separate-phase hydrocarbon recovery. Theintroduction of n-butylbenzene has the opposite effectsince it is less volatile than n-decane. The substitutionof n-octane for p-xylelle has more impact than that ofn-butylbenzene for n-decane. This is caused by arelatively higher differential of mole fractions and thesaturated vapor pressures for the n-octane/p-xylenepair. A similar argument holds for the amountrecovered in the water phase. The solubility of n­octane is much less thatl that of p-xylene, so that the

266

net effect is a lower recovery in the water phase in Run#4. Since the bulk of recovery in the gas phase(especially in an initially water-saturated medium)comes after what is recovered as NAPL, a higherquantity of the latter implies a lower recovery in thegas phase (Fig. 18b). As was the case for the CleanSite Model, the simulation results presented here showlittle sensitivity to the choice of capillary pressurefunctions. The dominant recovery mechanisms areinterphase mass transfer and viscous flow of the fluidphases.

Conclusions

A simulation model was developed to evaluate thepotential for using steam injection to clean up gasolinecontamination at the Gasoline Spill Area at LLNL.The results show that gasoline present in the permeablesands and gravel layers can be successfully recoveredby injecting steam into those layers in a 7-spot pattern.For the conditions assumed in the model, it will takeless than 16 days to recover nearly all of the gasoline inthe sands and gravel layers. By that time, themaximum aqueous concentrations of hydrocarboncomponents in these layers will have dropped to lessthan 0.01 mg/I.

The results show that vaporization, followed by bulkmovement of the vapor to the production well is thedominant recovery mechanism. In terms of timerequired for cleatlUp, model results are most sensitiveto permeability of the medium. Other parameters, suchas the relative permeabilities also affect the outcome,but to a lesser extent.

Although the simulations described in this paper arepredictive, in the sense that there is no historical datafor comparison, the reasonable and continuousdependence of the results on input data providesanother indication that the M2NOTS simulator is 'wellbehaved.'

References

1. Pruess, K., TOUGH User's Guide, NuclearRegulatory Commission, Report NUREG/CR­4645,1987.

2. Pruess, K., TOUGH2 -- A General PurposeNumerical Simulator for Multiphase Fluid andHeat Flow, Lawrence Berkeley Laboratory ReportLBL-29400, Berkeley, CA 1990.

SPE25257 A. E. ADENEKANand T. W. PATZEK 7

3. Adenekan, A. E., Numerical Modeling ofMultiphase Transport of Multicomponent OrganicContaminants and Heat in the Subsurface, Ph.D.Dissertation, University of California, Berkeley,1992.

4. Adenekan, A. E., Patzek, T. W. and Pruess, K,Modeling of Multiphase Transport ofMulticomponent Organic Contaminants and Heatin the Subsurface, I. Model Formulation, submittedto Water Resources Research, 1992.

5. Adenekan, A. E., Patzek, T. W. and Pruess, K,Modeling of Multiphase Transport ofMulticomponent Organic Contaminants and Heatin the Subsurface, II. Model Verification,submitted to Water Resources Research, 1992.

6. Hunt, J. R., Sitar, N. and Udell, K S., Nonaqueousphase liquid transport and cleanup, 2.Experimental Studies, Water Resources Research2481259-1269,1988.

7. Aines, R. (Editor), Dynamic UndergroundStripping Demonstration Project, Interim Report,Lawrence Livennore National Laboratory ReportNo. UCRL-ID-I09906, March, 1992.

Remediation Investigations Report for the LLNLLivermore Site, Lawrence Livermore NationalLaboratory, Livermore, CA., UCRL-AR-I0299,1990.

9. Nichols, E. M., M. D. Dresen, and J. E. Field,Proposal for Pilot Study at LLNL Building 403Gasoline Station Area, Lawrence LivermoreNational Laboratory Environmental RestorationSeries, UCAR-I0248, August 1988.

10. Parker, J. C., Lenhard, R. J., and Kappusamy, T.,A Parametric Model for Constitutive PropertiesGoverning Multiphase Fluid Flow in PorousMedia, Water Resources Research 23 4 618-624,1987.

11. Kreamer, D. K and K J. Stetzenbach,Development of a standard, pure-compound basegasoline mixture for use as a reference in field andlaboratory experiments, Groundwater MonitoringReview, Spring 1990.

12. Johnson, P. C., M. W. Kemblowski, and J. D.Colthart, Quantitative analysis for cleanup ofhydrocarbon-contaminated soils by in-situ soilventing, Groundwater, 28(3),413-429, 1990.

13. Weiss Associates, Personal communication, 1992.8. Thorpe, R. K, W. F. Isherwood, M. D. Dresen,and C. P. Webster-Scholten (Editors), CERCLA

Table 1: Rock Properties used in the Base Run

PermeabilityPorositySoil grain densitySoil grain specific heat capacityDry media thermal conductivityLiquid saturated thermal conductivityResidual NAPL saturation, B orw = Borg

Irreducible water saturation, B wir

Residual gas saturation, Bgr

now = nog

nwng

k rwro

k rocw

k rgro

n

267

Sands and Gravels5.0 X 10 12 m 2

0.252650 kg/m3

720 J/kg·K0.50 W/m·K3.10 W/m·K

0.0550.100.011.52.01.20.81.01.05.0

40.040.00.0

Silts and Clays1.0 X 10 14 m2

0.252650 kg/m3

720 J/kg·K1.50 W/m·K3.10 W/m·K

0.0850.150.011.52.01.20.81.01.03.06.06.00.0

8 Cleaning up Spilled Gasoline with Steam: Compositional Simulations

Two-phase water-oil and oil-gas relative permeability functions:

k [Sw - Swir ] nw

rwro 1 - Sorw - Swir

SPE25257

krowk [ 1 - Sw - Sorw ] now

roew 1 - Sorw - Swir

[ ]n~

k1 - Swir - Sorg - Sg

roew S1 - Swir - org

[ ]

ng

_ k Sg - Sgr- rgro

1 - Swir - Sorg - Sgr

Two-phase water-oil and oil-gas capillary pressure functions:

Peow

Pego

f!w9 ((Sw_- Sm) -11m _ 1) lIn

a ow 1 Sm

(

-11m ) lInf!w9 (Sw +~o - Sm) _ 1ago 1 Sm

where m = 1 - lin.

Table 2: Pseudocomponents used to represent gasoline

Basle Run: 3 Pseudocomponents1. Benzene. Mole fraction = 0.022; represents Benzene~~. p-Xylene. Mole fraction = 0.741; represents all C.,-Cg hydrocarbons~:. n-Decane. Mole fraction = 0.237; represents all C lO- 0 12 hydrocarbons

SenBitivity Run #4: 5 Pseudocomponents1. Benzene. Mole fraction = 0.022; represents Benzene~:. p-Xylene. Mole fraction = 0.497; represents C7-C9 aromatics:3:. n-Octane. Mole fraction = 0.233; represents C.,-Cg paraffins, olefins, etc.4. n-butylbenzene. Mole fraction = 0.168; represents ClO-Cll aromaticsS. n-Decane. Mole fraction = 0.080; represents all C lO- C 12 hydrocarbons

268

~

l:I1

~~~;I

~

~tI1~

tI:l

~tvVItvVI-.I

C.II. Noy••3/./82

\

\\ ...... W·216

\----x---

Aug.red borehole

locatlon uncenatn: Ilanl boring

Monitor/imaging .eU

Pre-Dynamic Stripping boring. andmonitor .eUa

St.am injection/eleclrical h••ting.ell

~ Steam .xUsclion well

+

o 20 4011

I I I I I

o

t

EXPLANATION

o078 B.nz~~ concentration in loil, parll. per m,H,on (ppm)

...@.1

'J Logorilhmic benzene concentration(). conlout Inlerval, ppm, approxima1elr

/'localed. dalhed where Inferred,queried where uncertain

N

'--;~I;"'--.

DRAFT

TEP-006 "J

_~J

+1O.OJCSB-IIO

nO)-lo.

•GSB-~O)

0.028-+-csw-n

GSW·4

-+-

0 0.002YEp·004

:\#,--,/ I.I®'

/ YEP-OOS \

,/ •. C40)-' ~ \

~4IJ I \I! Int<1 '<41.

I '- ._-;- cm.... 1-----

I, -~~

,

_ BUilding 41l1> kCSW~-20'eSD-004 'l"., + <J ------ "-

n. @CEW-'IO - ~I ~ 14, CSB-liIs ~

(;su-,I L:'OOI SVP _ -5.0 I "/+ / .(,P-OO) ............

eSD-t ~ I "' \

Isvu:' , \' 64 !LLCSW.11

• (;1-002, \' GSW40)-6 CSW-15 -'Y'1 \I H1'-ool® i. \.' ,SVB-CP-OI2" \. .'1" !\J / GSW·5

S.VU-GI'-OO4 SVIl-GP-UU....* // "",:,;.,"-'- ... HP.010" ;:»

I t:.., -+-. <i>. . ~ ..kSU

.6

' ' .do..GSW-)

eSD) . l:) T \ 'l"

\

Cl07' -+-t>. SVI.GP-<l06 \ '

• GSB-lIDO -+:~' CtO)·I· \ ,\\ \ ' . CSW.1b ~VB.GP-OO' ClO)-2 \ 1 -'22 \ (;SII.80'(> ~ S\·t:

P17-+- . td $,50" MW-508 I .-.-g.;-';--.

;/

CSW.6 \ (;'B.2 SVII-G\'·OH ( ~)\ ..,•• .' w:.,. J I ,_.,SVII-GI'U $"., ~GSD·4 ~to).) ,_':ILCSB-I06 I

~ - /2

CU~'7I1 2.0 ®II

YEP-OOI I ®lEP-007 (CSW'2

X \, I •• 25 8.60: •. en.•, •. ,nl__-x \ GSII-I02 . '.. TE:J'OOIX MW·20

X---- \ UOH'!' I ~ 4;--X'-- ~ '", ,_X· ~·m' ~.

, """" --(.;-----X --~x--- I'"- _ nr-002 ...I).\). -. --x·" - •., I X

O?.. ~ CSB-aOf /...... .1 /--- ,;'

_____ ..-'1""

/.

.>< <0.002TEp·OO)

®

X

NmCD

Maximum Benzene Soil Concenlralion· Lower Sleam Zone (pans per million) ....Fig. 1. Plan view of the Gasoline Spill Area.

\0

10 Cleaning up Spilled Gasoline with Steam: Compositional Simulations SPE25257

Vertical X-Section

~SteamInjector

iProducer Steam

Injector

WaterTable

Plan View

Depth (m)

Fig. 2. 7-spot pattern.

Ground Surface;dl1LSdtb.lISL2.l@L....L# )li!@£3.jjiUL!

Heat loss to overburden

24

#1

32

37

40

:::::::::::::::::::~cw~r ~: :

Heat loss to underburdenInjector

~ Water Table

Producer

CIIJ Sands and Gravel Silts and Clays

Fig. 3. Vertical cross-section of finite-difference grid.

270

til

~VlNVl-..l

;>tI1

I~8-;3

~'"tl

~tI1~

Initial NAPL Saturation

ill!l 0.05

[] 0.01

D 0.002

L/'

/'

<a) Layer 1

""Distance (m)

(b) Layer 2

,-

/,-

./I -Marks area

within actual ,;',J-spot

pattern

,-

.-

Injeclor 5(1

<;~':I:;'>

"'-Marks areawithin actual

40 J-spotpattern

:~. IProducer IllIeciOf 50 Dist~e (m)

(C) Layer 3

ProduCer injectOr 50 Dis~~e (m)

Producer A:J::i,

150100Distance (m)

Injector 50

Fig. 4. Plan view of finite-difference mesh.

-

/-

/-Marks area /

within actual ,//- 7-spot /''''

pattern,/

,/,/' /'-

/,

/ \

L1 \

............ 1 1 1 I ..- •

20

60

40

80

100

Producer

I\)........

Fig. 5. Initial distribution of gasoline.

--

12 Cleaning up Spilled Gasoline with Steam: Compositional Simulations

300 20

"

Gd "~

-250"~ ","0 " 15 "0- I -.2 200

(/)

I,

.0, :::::-.....- ,Q)

:~Q)

..- 1Uco 150 10a: : NAPL a:~ ~Q) 100

Q)

> >0 0() 5 ()Q) Q)

a: 50 a:

05 10 15 20

Fig. 6. Benzene recovery rates per 1/12 of 7-spot pattern.

10 1.2Gas

~9

~co 1.0 co"0 8 "0- -(/) (/).0 7 .0

0 0.8 00 6 00 0....- G.....-Q) 5 0.6 Q)

1U - ..-4 " NAPL co

a: • a:~

• :,--.. 0.4 ~"Q) 3 • II Q)

> I " , >0 2

I , I 0() Water

, , I ()Q)

, , I 0.2 Q)

a: , I I a:, I ,. ,0 -----------r ,

00 5 10 15 20

Fig. 7. P-xylene recovery rates per 1/12 of 7-spot pattern.9 0.4

~~

Gas ~8

I co"0 "0- 7 -(/)

! 0.3 (/)

.0 .0

0 6 00 \ 00

~0....- 5 ....-.....- .....-

Q) \ 0.2 Q)

1U 4 " 1U"a: " NAPL a:~ 3 " ~"Q) " ."~--" Q)

> " 0.1 >0 2 " ' I 0,

\)4() " , ()Q) " , Q)

a: " , a:Ii ,,

0 00 5 10 15 20

Days of Steam InjectionFig. 8. N-decane recovery rates per 1/12 of 7-spot pattern.

272

SPE25257

SPE25257 A. E. ADENEKAN and T. W. PATZEK 13

250,----r=============,-------------,

~Benzene·ln-Place = 363 Ibs

200

-en.0;::::.. 150

~(1)

>8 100(1)

a:

_--- All 3 Phases

_---Gas

................................ NAPL..,

50

5 10 15 20

Fig. 9. Cumulative recovery of benzene per 1/12 of 7-spot.

-en:Q

0.4 0oo....

0.3 '-""~

~0.2 8

(1)

a:0.1

8 -

10 rr.==;:~~=:;;==~===;=~~==::=;----------, 0.61L::lln~it::::ia::.1X~y~le~n:.::e~·in.:.:·P~I:::ac~e:-=.:....:.;16~6~8::.8::::Ib~s_---l. All 3 Phase

9 Gas................... NAPL 0.5

2

-en.0 7og 6....- 5~~ 4og 3a:

o l..-__iiiioiiO__I!IIIIl!L._--''- -'-,-----' 0

o 5 10 15 20

Fig. 10. Cumulative recovery of p-xylene per 1/12 of 7-spot.

5 r;===:::;:::====:=:::::============:::::::;-------j 0.20IInitial Decane·in-Place = 7156 Ibs

f...! NAPL

4-en:Qoo 3o....-~~ 2o(,)(1)

a: ............

..~== All 3 PhaseU G~ .":

....0.10 "'-'

~

~o(,)

0.05 ~

~U I -0 ..... .......-i'-------......:...................------10

o 5 10 15 20

Days of Steam Injection

Fig. 11. Cumulative recovery of n-decane per 1/12 of 7-spot.

273

-~

en

~VINVI-....l

()

f.(JCl

fj

~....~Q..

[S'~

~

~

~~()

l'"a:[ens·a~~.

Layer 2

NAPL

. .

Water

.......

.............

..........

...........

Gas

.... ..................... . .

o 20 40 60 80 100

Percentage Recovered in Each Phase

Fig. 12. Relative hydrocarbons recoveries.

p-Xylene

p-Xylene

Benzene

Benzene

n-Decane

n-Oecane

20

(a)

1510

Upper Aquifer

5

1100

1000-~ 900"'0-C/) 800:B0 70000 600T'""-Q) 500~a: 400C0 300

t5200Q)

"c100

00

N......

300 I 4000~ ILower Aquifer (b)

3500

% 250 %"'0

"'0 en 3000en .0::Q -aa 200 a 2500a aa .......... ~

~ CD 2000CD iiiiii 150 a:a: c 1500C 00 nu ::J 1000CD "'0~ 100 e

a.. 500

50 " 00 5 10 15 20 0 5 10 15 20

Days of Steam Injection Days of Steam Injection

Fig. 13. Steam injection rates, Fig. 14. Water production rate.

15 ,r--------------------------------,

?>~

~8-;l

~

~tTl~

CI:l

tilNVINVI-...l

k = 5 darcy (Base Run)

k = 10 darcy

k = 5 darcy (Base Run)

k = 10 darcy

,-----------------.- Ir--------

~-----_._--------,...,/

,,,",,,

I,.,,I,

II,,,,,

I."""

15 r,--------------------------,

250 , ,

50

VI.0

~50Q)

8Q)0:100

200

..,,VI ~

,I

~ " ~:'.--_.

l,0

I,....I~

I~ ,~ I

I,£ 5 IIIIIIII

JI.

k = 5 darcy (Base Run)

k = 10 darcy

k = 5 darcy (Base Run)

k = 10 darcy

:------------------- /r--------I,

II,,.

_.."

1---------------_·_----: ~,..-------IIIIII,I,I,III,,

",,-'

J . ), I Io 5 10 15 20

Days of Steam Injection

(a) Recovery of all components in all phases

~Q)

~oQ) 50:

VI:Q 10

8,....~

I\).....(II

800

700

600VI.0=500~Q)

~ 4000Q)

0:300

200

100

o ! , ( t 0 '45"""': til I

o 5 10 15 20 0 5 10 15 20oars of Steam Injection Days of Steam Injection

(C) Recovery of a I components as NAPL (d) Recovery of all components in the water phase

Fig. 15. Sensitivity Run #1.

­VI

-0\

C/.l

~VINVI-.J

(1

i(JQ

.g-s.[

[S'o

;S.So~~(1

~'g'"&:[C/.ls·aa~.

2010 15

Days of Steam Injection

Exponent for NAPL Relative Perm. = 1.5(Base Run)

Exponent for NAPLRelative Perm. = 2.0

:/II

IIff,I,I,IIII

01 #o 5 ---.J

15 , I

Exponent for NAPL Relative Perm. = 1.5(Base Run) _

Exponent for NAPL Relative ,';;,.---Perm. = 2.0 /

I,I

I

_______ .",,1

~

~~ 50::

250 ,,----------------------------------,

(b) Recovery of all components in the gas phase

en::Q 10ooo.....-

200

50

en.D

~150

~o(J)

0::100

20

Co

"",'"r--_.......~

Exponent for NAPL Relative Perm. = 1.5(Base Run)

Exponent for NAPL RelativePerm. = 2.0

, .... ------------_.

Exponent for NAPL Relative Perm. = 1.5(Base Run)

Exponent for NAPLRelative Perm. = 2.0

I,I

Io ! !, I! 0 I............... ! I I

o 5 10 15 20 0 5 10 15 20Days of Steam Injection Days of Steam Injection

(C) Recovery of all components as NAPL (d) Recovery of all components in the water phase

Fig. 16. Sensitivity Run #2.

100

J },o 5 10 15

Days of Steam Injection

~(J)

~ 5o(J)

0::

15 "----------------------------,

(a) Recovery of all components in all phases

800 i i

en 10::Qooo.....-

600

700

en

i:~(J)

0:: 300

200

I\)......CJ)

(a) Recovery of all components in all phases

?>~

~8-;l

~

~~

CI:l

~VltvVl-...l

205 10 15Days of Steam Injection

o~ , I I

o

50

250 I 1

Residual NAPL Saturation = 0.055(Base Run)

Residual NAPLSaturation = 0.085 ~./'

<IIJ.'-

200

(b) Recovery of all components in the gas phase

Iii'.0=150~

~o~1oo

20

15 ,Residual NAPL Saturation = 0.055

(Base Run)

Residual NAPL

j'~ 10 ~Saturation = 0.085

00

IJ tI,,,

0 1 f20 0 5 10 15 20

Days of Steam Injection

15S

10 .Days of team Injection

5

5 10 15Days of Steam Injection

Residual NAPL Saturation = 0.055(Base Run)

Residual NAPLSaturation = 0.085

Residual NAPLSaturation = 0.085

I,,I,

II

IIII

,j ;/~--=-=-~_I

15 1.----------------------------,Residual NAPL Saturation = 0.055

(Base Run) /r.......,...--

II

II

II

- --",'

~Q)

~oQ) 5a:

,....-

Iii".0;; 10oo

I\)........ 800

700

600Iii'.0=500~Q)

~4oo0Q)

a: 300

200

100

00

(c) Recovery of all components as NAPL (d) Recovery of all components in the water phase

Fig. 17. Sensitivity Run #3.

..­-...l

­00

700

100

(")~

e.Jg.g~....~l:l-

[S'('D

:;:;~CI:let~(")

I::;:

[CI:lt::.

[~o't::(Il

,,,---------_.,.',...,

.........

5 Hydrocarbon Components

3 Hydrocarbon Components -----­(Base Run)200

250 I i

o I I [o 5 10 15 20

Days of Steam Injection(b) Recovery of all components in the gas phase

~

~~ 5a:

15 , ,5 Hydrocarbon Components

3 Hydrocarbon Components(Base Run)

50

'iii';§.150

~CD

~~1ooa:

'iii'::Q 10

8a.....~

- - - - 5 Hydrocarbon Components900 f- ,--------------- 3 Hydrocarbon

Components (Base Run)800

15 ,r-------------------------------,

(a) Recovery of all components in all phases

300

200

~

~~ 5a:

1000 rl----------------------------_

5 Hydrocarbon Components

3 Hydrocarbon Components(Base Run)

oI' Io 5 10 15 20

Days of Steam Injection

~10a8~

E=600

~~500o~400a:

N......Cl)

o! !,! I 1 0 ~- ! I r Io 5 10 15 20 0 5 10 15 20

Days of Steam Injection Days of Steam Injection

(C) Recovery of all components as NAPL (d) Recovery of all components in the water phase

Fig. 18. Sensitivity Run #4.

CI:l"'I:itr:lNVlNVl-.l