55 dustin jones gasification

Published: December 15, 2011

r 2011 American Chemical Society 2362 dx.doi.org/10.1021/ie201713n | Ind. Eng. Chem. Res. 2012, 51, 2362–2375

ARTICLE

pubs.acs.org/IECR

Rigorous Kinetic Modeling and Optimization Study of a ModifiedClaus Unit for an Integrated Gasification Combined Cycle (IGCC)Power Plant with CO2 CaptureDustin Jones,†,‡ Debangsu Bhattacharyya,*,†,‡ Richard Turton,†,‡ and Stephen E. Zitney†

†U.S. Department of Energy, National Energy Technology Laboratory, Morgantown, West Virginia 26507, United States‡Department of Chemical Engineering, West Virginia University, Morgantown, West Virginia 26506, United States

ABSTRACT: The modified Claus process is one of the most common technologies for sulfur recovery from acid gas streams.Important design criteria for the Claus unit, when part of an Integrated Gasification Combined Cycle (IGCC) power plant, are theability to destroy ammonia completely and the ability to recover sulfur thoroughly from a relatively low purity acid gas streamwithout sacrificing flame stability. Because of these criteria, modifications to the conventional process are often required, resulting ina modified Claus process. For the studies discussed here, these modifications include the use of a 95% pure oxygen stream as theoxidant, a split flow configuration, and the preheating of the feeds with the intermediate pressure steam generated in the waste heatboiler (WHB). In the future, for IGCC plants with CO2 capture, the Claus unit must satisfy emission standards without sacrificingthe plant efficiency in the face of typical disturbances of an IGCC plant, such as rapid change in the feed flow rates due to load-following and wide changes in the feed composition because of changes in the coal feed to the gasifier. The Claus unit should beadequately designed and efficiently operated to satisfy these objectives. Even though the Claus process has been commercialized fordecades, most papers concerned with themodeling of the Claus process treat the key reactions as equilibrium reactions. Suchmodelsare validated bymanipulating the temperature approach to equilibrium for a set of steady-state operating data, but they are of limiteduse for dynamic studies. One of the objectives of this study is to develop a model that can be used for dynamic studies. In a Clausprocess, especially in the furnace and the WHB, many reactions may take place. In this work, a set of linearly independent reactionshas been identified, and kinetic models of the furnace flame and anoxic zones, WHB, and catalytic reactors have been developed. Tofacilitate the modeling of the Claus furnace, a four-stage method was devised so as to determine which set of linearly independentreactions would best describe the product distributions from available plant data. Various approaches are taken to derive the kineticrate expressions, which are either missing in the open literature or found to be inconsistent. A set of plant data is used for optimalestimation of the kinetic parameters. The final model agrees well with the published plant data. Using the developed kinetics modelsof the Claus reaction furnace, WHB, and catalytic stages, two optimization studies are carried out. The first study shows that thereexists an optimal steam pressure generated in theWHB that balances hydrogen yield, oxygen demand, and power generation. In thesecond study, it is shown that an optimal H2S/SO2 ratio exists that balances single-pass conversion, hydrogen yield, oxygen demand,and power generation. In addition, an operability study has been carried out to examine the operating envelope in which both theH2S/SO2 ratio and the adiabatic flame temperature can be controlled in the face of disturbances typical for the operation of an IGCCpower plant with CO2 capture. Impact of CO2 capture on the Claus process has also been discussed.

’ INTRODUCTION

Integrated gasification combined cycle (IGCC) is a promisingtechnology for generating clean, affordable, and secure power.However, a coal-fed IGCC power plant has lower net plantefficiency compared to a conventional natural gas combined cycle(NGCC) power plant. A NETL study shows that the net plantefficiency of a coal-fed IGCC plant with a General Electric Energy(GEE)-type gasifier is 38.2%, compared to 50.8% for NGCCplants.1 Moreover, the net plant efficiency of the IGCC plantfurther decreases to 32.5% when the CO2-capture option isconsidered. To make IGCC technology more competitive, effortsmust be exerted to improve its efficiency. Bhattacharyya et al.2

carried out anoptimization study for improving the overall efficiencyof an IGCC plant with CO2 capture. A slightly modified version ofthe plant layout considered in that study is shown in Figure 1. Thestudy considered a GEE-type entrained flow gasifier mainly becauseof its high carbon conversion rates and environmental advantages.2,3

Figure 1 shows that the shifted syngas from the gasifier goes to a dualstage Selexol solvent unit for acid gas removal (AGR). The stripperoff-gas from the Selexol unit, rich in hydrogen sulfide, is sent to theClaus unit, and carbon dioxide is sent for compression andsequestration. The off-gas from the sour water treatment unit, richin ammonia, is also sent to the Claus unit.

Environmental targets for the IGCC are shown in Table 1.1

Single pass conversion of hydrogen sulfide to sulfur in a dualstage Claus process is approximately 96%. To meet the environ-mental target on SO2 given in Table 1, 99.9% sulfur recovery isrequired. To accomplish this level of sulfur recovery, a hydro-genation unit is required.4 The tail gas from the Claus process is

Received: August 2, 2011Accepted: December 15, 2011Revised: December 14, 2011

2363 dx.doi.org/10.1021/ie201713n |Ind. Eng. Chem. Res. 2012, 51, 2362–2375

Industrial & Engineering Chemistry Research ARTICLE

treated in a hydrogenation unit, compressed from about 1 to 50atm, and recycled back to the AGR unit for recapture. In the AGRplant, a higher operating pressure results in a lower flow rate ofthe circulating solvent that decreases the refrigeration, pumping,reboiling, and equipment cost by increasing the partial pressuresof CO2 and H2S.

5 But the compression cost for recycling the tailgas from the Claus process increases. For improving the overallefficiency of the IGCC plant, this compression cost should beoptimized (reduced) without violating the environmental tar-gets. Other important design criteria for the Claus unit, when partof an IGCC power plant with CO2 capture, are the ability todestroy ammonia completely and recover sulfur thoroughly froma relatively low purity acid gas stream without sacrificing flamestability. A rigorous kinetic model is needed to design a plant thatcan satisfy these design criteria.

Several papers have been published on themodeling of the Clausprocess or specific unit operations of the Claus process.1,2,6�9

However, most of these modeling efforts rely on restricted equilib-riummodels to predict product distribution from the Claus furnace.One exception that could be found is the work ofNasto et al.6 Nastoet al. focused on the modeling of the waste heat boiler (WHB) witha reaction set consisting of H2S pyrolysis and COS formationreactions. However, considering only these reactions cannot fullydescribe theWHB, as these reactionswould not capture any changesin sulfur dioxide composition across the WHB, which is shown totake place from plant data.10

There have been several papers published on the kinetics ofthe Claus furnace and WHB that attempt to determine the

kinetics of the system.11�17 Despite this, there is still uncertaintyin reaction pathways and some reactions that lack rate expres-sions. In the work of Nasto et al.6 for modeling the WHB, thereactions considered were only the hydrogen sulfide pyrolysisand carbonyl sulfide formation reaction of carbon monoxidereacting with sulfur. Additionally, only the WHB was modeledand the feed composition and properties were taken as themeasurements of Sames et al.10 from the anoxic region of thefurnace. However, as these two reactions do not fully define thesystem, additional linearly independent reactions must be con-sidered to derive a consistent kinetic model of the Claus furnace.Little or no methodology is available in the open literature thataddresses the selection of such a kinetic model. To facilitate themodeling of the Claus furnace, a four-stage method has beenproposed, so as to determine which set of linearly independentreactions best describe the product distributions from availableplant data. Using this method, an optimal set of linearlyindependent reactions was determined and used to model theClaus furnace and WHB. Rate expressions were fitted when theywere not available within the open literature.

There have been no studies found within the open literaturethat discuss the optimal operation of the Claus plant as part of anIGCC power plant. It is generally accepted that optimal opera-tion occurs when the H2S/SO2 ratio in the feed to the tail gastreatment unit is as close as possible to 2, so as to maximize thesingle-pass conversion of hydrogen sulfide.18,19 However, in theoperation of an IGCC power plant, there are other considera-tions in addition to single-pass conversion. The Claus processyields hydrogen and raises considerable amounts of steamthrough the recovery of waste heat. Additionally, as these studiesconsider the use of a highly enriched oxygen stream (95 mol %)from an air separation unit as the oxidant to the furnace,20 thecost associated with that oxygen production should also beconsidered. Therefore, the effect of oxygen flow and steampressure on these variables was examined.

Little information could be found in the open literature on thetopic of operational limits for the Claus process. It was desired to

Figure 1. Block flow diagram of IGCC power plant with carbon capture.2

Table 1. Environmental Targets of IGCC1

pollutant environmental target

NOx 15 ppmv (dry) @ 15% O2

SO2 0.0128 lb/MMBTU

particulate matter 0.0071 lb/MMBTU

mercury 90% capture



determine the limiting conditions for maintaining both adiabaticflame temperature and H2S/SO2 ratio without violating safetyconstraints.Configuration andKineticModeling of theModified Claus

Process.As shown in Figure 2, the process begins with the Clausreaction furnace, which is also referred to as the thermal stage. Inthe reaction furnace, the hydrogen sulfide contained in the acidgases is combusted to form sulfur dioxide via reaction 1a. This gasthen passes through the WHB to generate intermediate-pressure(IP) steam to be used for the heating duties required in theprocess. The process gas is then further cooled to condense thesulfur formed during the thermal stage. Approximately 60% ofthe inlet sulfur gets converted in the thermal stage.18 The gas isthen reheated with the steam generated in the WHB and sent tothe first of two catalytic stages where hydrogen sulfide and sulfurdioxide react in a 2:1 ratio to form sulfur and water via reaction1b. The overall sulfur chemistry of the Claus process can beshown as reaction 1c.

H2S þ 32O2 f SO2 þ H2O ð1aÞ

2H2S þ SO2 T 2H2O þ 3nSn ð1bÞ

H2S þ 12O2 f

3nSn þ H2O ð1cÞ

Reactions 1a�1c represent the overall chemistry of the Clausprocess; however, the actual chemistry is much more compli-cated. One of the major purposes of this paper is the determina-tion of the most likely pathways for the formation anddestruction of species important in the Claus process.For operations with acid gases whose hydrogen sulfide com-

position is greater than 50 mol %, a straight-through configura-tion is generally chosen.18 However, a split-flow configuration isgenerally required to address the challenges that can arise whenhydrogen sulfide concentrations in the acid gases are less than∼50 mol %. In the split-flow configuration, a fraction of the acidgas bypasses the burner so that the adiabatic flame temperature isgreater than approximately 930 �C to ensure flame stability.18,21

In addition, a split-flow configuration is also sometimes requiredwhen processing a sour water stripper gas that has a highconcentration of ammonia. To prevent the plugging of down-stream reactors caused by ammonia slip through the furnace,high furnace temperatures are required to ensure the completedestruction of ammonia.18 Because a limit exists on the amountof acid gas that can be bypassed, other options are often required

to ensure sufficiently high flame temperatures.18 The optionsconsidered here are the preheating of all feed streams and the useof an oxygen-enriched stream instead of air. Because of thesemodifications to the conventional Claus process, this process is alsoreferred to as a “modified Claus process.” The process gas that exitsthe waste heat boiler passes to a cooler to condense and separate thesulfur. The process gas is then reheated and sent to the first of twocatalytic reactorswhere hydrogen sulfide and sulfur dioxide react in a2:1 ratio to form sulfur and water. After reacting in the catalyticreactor, the gas is then cooled to condense and remove sulfur andsent to the second catalytic stage. After the second stage, the tail gasis sent to a hydrogenation unit. This unit converts all remainingsulfur species to hydrogen sulfide. This gas is then cooled to removethe water and compressed and recycled to the Selexol unit.Flame Zone. A zoned approach was taken in the modeling of

the Claus furnace. A simplified diagram of the zoned approach isshown in Figure 2. The first zone is an adiabatic flame zone that ismodeled as a plug flow reactor. The linearly independentreactions considered for this zone are the oxidation reactionsshown in Table 2. Accurately predicting the relative extents ofreactions occurring in the flame zone is important for predictingthe adiabatic flame temperature and the consumption of oxygen.Ensuring that no oxygen slips into the anoxic region is also animportant safety related issue because of the possibility of aflashback in the bypassed acid gas.18

Anoxic Zone and WHB. A wide number of possible reactionsthat may occur in the anoxic and WHB section of the Clausfurnace are reported in the literature.12,24 For this work, a set oflinearly independent reactions that would best describe theproduct distribution at the outlet of the WHB was desired. Inthe open literature, little or no results have been presented onmethodologies to select such a set of linearly independentreactions. In this paper, a four-stage method was devised todetermine which set of linearly independent reactions would bestdescribe the product distributions from available plant data. Thefirst stage of this method is to determine the maximum numberof linearly independent reactions. The first stage is based on thework of Mah and Aris.25 Let us consider that there are S reactivechemical species consisting of N atomic species in a system. In adata set p, ifmi represents the amount of atomic species i, yj is thenumber of chemical species j ,and bij represents the number ofatoms i in species j, then

m ¼ By ð2ÞThen, the maximum number of linearly independent reactions is

Rmax ¼ S� rankðΒÞ ð3Þ

Figure 2. Simplified flowsheet of Claus and hydrogenation units.



If a S� SmatrixΔY exp is constructed from the experimental data,where each column represents a particular experimental data setsuch that Δyjp

exp = yjpexp � yjp

exp,in where yjpexp,in represents the

number of the chemical species j at the inlet in the data set p, thenthe number of reactions actively taking place, R, is given by rank(ΔY exp). However, one may not have S sets of experimental data.In addition, determination of rank is an involved process in thepresence of experimental error, especially for large systems. Inthis case, a qualitative approach can be taken based on theliterature data and/or experimental observations. This point willbe further clarified as the method is applied to the Claus plantmodel. Once the independent number of reactions, R, is deter-mined, the important step is to determine a consistent set ofreactions that can address the experimental data. In the existingliterature, this determination is based mainly on a qualitativeargument relying on an inspection of the experimental data. Inthe discussion that follows, a quantitative method is proposedthat systematically addresses the issue of the correct set ofreactions, and this method is then applied to the Claus plant.In the second stage, the best possible prediction of the experi-mental data is determined in the presence of experimental error/noise. If the extent of a reaction r in a data set p is given by ζrp, andαjr represents the stoichiometric coefficient of chemical species jin reaction r, then

ΔY ¼ αζ ð4ÞNote that ζ is unknown and some of the elements of α may beknown from experimental studies and/or literature data but, ingeneral, the matrix α is either partially or completely unknown.However, because of experimental errors

ΔYexp ¼ αζ þ ε ð5Þwhere each column in the error matrix ε represents the experi-mental errors in a data set. An optimization problem is thenformulated and solved to find the optimal estimate for theunknown elements in α and ζ

min ΔY �ΔY expk ks:t: ΔY ¼ αζ

Βα ¼ 0ð6Þ

where the second constraint is required to ensure the conserva-tion of atoms. Here, no bound on ζ is considered because theequilibrium reactions can yield negative values. More constraints

can be considered in this formulation. In addition, partial or fullinformation of α can be provided, if known. The minimum valueof the objective function achieved through this formulation is thebest possible solution that can be obtained in the presence ofexperimental errors. In the third stage of this approach, adatabase of the possible reactions is prepared on the basis ofthe open literature and experimental data. From this database, Rreactions are considered, thus forming a number of reactionstoichiometric matrices; that is, Q ,q = 1:Q. Then, the followingquadratic programming (QP) problem is solved:

min ΔY q �ΔY expk ks:t: ΔY q ¼ αqζq

ζij g 0, i ∈ ψ, j ∈ ω

ζij e 0, i ∈ ε, j ∈ χ

ð7Þ

whereψ,ω, ε, and χ are finite sets of indices, i∈ψ∪ ε are vectorswith w elements, and w e R and j ∈ ω ∪ χ are vectors withn elements and ne S. The bounds on elements of ζ are based onexperimental data and/or process knowledge. Here, the con-straint on the atom balance is not considered, as it is assumed thatit has been satisfied while creating the reaction database.If the relationship shown in eq 8 is satisfied by a stoichiometric

coefficient matrix αq, then that stoichiometric coefficient matrixis retained for the fourth stage of the analysis.

min ΔY q �ΔY expk k �min ΔY �ΔY expk k e tol ð8ÞHere, the tolerance, tol, can be low initially but can be relaxed at alater time if performance in the fourth stage is found to be poor.In this way, a list of stoichiometric coefficient matrices isgenerated, and they are ranked in ascending order based on their

)ΔYq � ΔY exp ) values. Note that the solution of this QPproblem is computationally inexpensive, but it can drasticallyreduce the computational load in the next stage. If there are Vreactions in the database, then the reaction sets extracted for

consideration are expected to be Q ,VR

!because, for

certain chemical species, the choices for reactions will be limited.This point will be further clarified while developing the model ofthe Claus process. The values ofΔY qwere generated without anyconsideration for the kinetics, thermodynamics, or hydrody-namics of the actual system. Therefore, these values may notbe physically realizable or feasible. Hence, in the fourth stage, the

Table 2. Reactions and Kinetics Occurring in the Flame Zonea

reaction reaction rate [mol/(cm3 s)] ref

H2S þ 32O2 f SO2 þ H2O r ¼ 14 exp

�11:0RT

� �PH2SP

1:5O2

11

NH3 þ 34O2 f

32H2O þ 1

2N2 r ¼ 4430 exp

�40:0RT

� �PNH3P

0:75O2

12

CH4 þ 2O2 f CO2 þ 2H2O r ¼ 6:7� 1012 exp�48:4RT

� �C0:2CH4

C1:3O2

22

H2 þ 12O2 f H2O r ¼ 1:08� 106 exp

�30:0RT

� �CH2CO2

23

CO þ 12O2 f CO2 r ¼ 3:98� 1014 exp

�40:0RT

� �C0:25O2

CCOC0:5H2O

22

aActivation energies are in kcal mol�1, partial pressures are in atm, and concentrations are in mol cm�3.



feasible instances of ΔYq, denoted by ΔYq,act, are identified byconsidering the configuration and design of the actual processsystem along with kinetic modeling of the reactor. Here, themissingkinetic parameters, if any, are estimated first. Then, the nonlinearprogramming (NLP) problem described in eq 9 is solved:

minx∈Rn

ΔY q, act �ΔY expk ks:t: hðxÞ ¼ 0

υðxÞ e 0xL e x e xU

ð9Þ

where the nonlinear equality constraints are due to the processmodel that includes the kinetic expressions for those reactionsdetermined by the choice of αq. The inequality constraints mayappear based on the process system. The decision variables wouldbe the kinetic parameters of one ormore reactions that are boundedby the limits of the reported experimental errors. There can beadditional decision variables, for example, some parameters such asheat transfer coefficient, heat loss, etc. that may not be knownexactly. The lead coefficient matrix from stage three is examinedfirst. If its correspondingΔYq,act is lower than the nextΔYq in the listfrom stage three, a feasible reaction set is found. Otherwise, thesearch continues. It can take several hours of simulation time tosolve the above optimization problem for each of the coefficientmatrices, especially if any of the following occur: the number ofdecision variables is large; there are large plug flow reactorsconsidered, each with multiple reactions; or a sequential modularapproach is used to solve the simulation. The list prepared in stagethree can significantly reduce the number of reaction sets consideredin the fourth stage.For the kinetic modeling of themodifiedClaus process, all four

stages of the approach were applied. In the first stage, it was foundthat, Rmax = S � rank(B) = 13 � 5 = 8. A total of only fiveexperimental data sets were found in the open literature.8

However, based on the open literature, the number of reactionsthat are reported to take place for this system is more thaneight;13,14,17,24 therefore, it was assumed that R = Rmax. Thesecond stage optimization was carried out in the numericalcomputing software package MATLAB. In this optimizationformulation, the entire matrix α is unknown, along with ζ.

It was found that min )ΔY � ΔY exp ) = 3.53 . In the third stage,a list of possible reactions from the open literature wereconsidered; they are shown as reactions 10a�10l.12,24

H2S þ 12SO2 S

34S2 þ H2O ð10aÞ

H2S þ SO2 þ H2 S S2 þ 2H2O ð10bÞ

H2S S H2 þ 12S2 ð10cÞ

H2 þ 12SO2 S

14S2 þ H2O ð10dÞ

CH4 þ H2O f CO þ 3H2 ð10eÞ

CO2 þ H2 S CO þ H2O ð10fÞ

CO þ 12S2 S COS ð10gÞ

CO þ H2S S H2 þ COS ð10hÞ

NH3 f12N2 þ 3

2H2 ð10iÞ

NH3 þ 34SO2 f

38S2 þ 3

2H2O þ 1

2N2 ð10jÞ

3S2 S S6 ð10kÞ

4S2 S S8 ð10lÞFrom these reactions, a number of reaction sets, each of which arecomprised of four reactions, are considered and are shown inTable 3. Only four reactions are needed for this examination, asthere is no methane or ammonia, reducing S by two. Also, theplant data from Sames et al.10 considers all sulfur species as S,reducing the number of chemical species by another two. Itshould be noted that it was assumed that all sulfur existed as S2 forthe studies discussed in this section. The bounds on the elementsof ζ are based on plant data from Sames et al.8 and thethermodynamics of the problem. The saturation temperatureof water in the WHB is 488 K, and the process gas exittemperature is about 527 K. In addition, the rate of reaction willdecrease with a decrease in temperature, and the reactions are notvery exothermic. Therefore, the decrease in the temperature ofthe process gas along the length of the WHB can be consideredmonotonic. If δ is some tolerance, and if

��ΔQSj¼ 1

Cαjr

j

ΔðKeqÞr

�� e δ ð11Þ

then a constraint is imposed that the rth reaction could onlyproceed in the direction of the equilibrium shift. The aboverelation was computed from the experimental data for all theequilibrium reactions, assuming a linear temperature and com-position profile. However, if the above relation was not true for agiven reaction, the constraint was not imposed. The left-handside of inequality 11 gave values ranging from 0.53 for reaction 10c

Table 3. Reaction Sets and Objective Function Values

reaction sets considered min )ΔY q � ΔY exp ) min )ΔY q,act � ΔY exp )

10c, 10d, 10f, 10g 3.80 6.26

10b, 10c, 10f, 10g 3.80 5.71

10b, 10f, 10g, 10h 3.80 6.89

10a, 10b, 10f, 10g 3.82 5.98

10b, 10d, 10f, 10g 4.87 6.26

10a, 10d, 10f, 10g 5.02 6.32

10d, 10f, 10g, 10h 5.02 7.47

10a, 10b, 10f, 10h 6.98

10b, 10c, 10f, 10h 6.98

10c, 10d, 10f, 10h 6.98

10a, 10d, 10f, 10h 7.37

10b, 10d, 10f, 10h 9.01

10a, 10c, 10f, 10h 34.60

10a, 10f, 10g, 10h 34.60

10a, 10c, 10f, 10g 34.76



to 1� 10�21 for reaction 10h. Neglecting reaction 10c, the averagevalue for the remaining reactions was approximately 10�3. Using avalue of δ = 10�2, the constraint regarding the direction of theequilibrium shift was not considered for reaction 10c. The constraintwas imposed for rest of the reactions. Therefore, the extents ofreaction for reactions 10a, 10f, and 10h were constrained to valuesless than zero. Reactions 10b, 10d, and 10g were constrained tovalues greater than zero. A list of all the possible linearly independentcombinations of reactions from the list above is shown inTable 3. Asthe plant data showed no methane or ammonia entering the WHBand all sulfur species were considered as S, reactions 10e and10i�10l were not considered. The reaction sets are arranged inTable 3 in ascending order on the basis of the minimum values ofthe objective functions, as explained previously. By using a value oftol = 1.5, only the first seven reactions are considered in the fourthstage analysis, but the values for all possible reaction sets are listed inTable 3 for completeness.In the fourth stage, the WHB was modeled in Aspen Plus as an

ideal, multitube, plug flow reactor with constant coolant tem-perature and constant heat transfer coefficient. The optimizationproblem in eq 9 was solved by using the Aspen Plus’s optimiza-tion toolbox. The decision variables were the pre-exponentialfactors and activation energies of the reactions considered,shown in Table 3. In addition, the heat transfer coefficient ofthe WHB was also considered to be a decision variable. An initialestimate of the heat transfer coefficient was calculated using theGnielinski correlation26 assuming smooth pipes, eq 12. An initialestimate was found by assuming that all resistance to heat transferwas on the tube side. The limits for the heat transfer coefficientswere considered to be(20% of the calculated values from eq 12.No bounds were given on the pre-exponential factors of anyreaction. For reactions 10c,16 10f,13 10g,27 and 10h,27 activationenergies were allowed to vary between the error bounds given inthe papers in which they were reported. For reactions 10b and10d, which have no rate expressions available within the openliterature, no bounds were provided for the activation energy. Itwas also assumed that reactions 10b and 10d had an elementary

form. For reaction 10e, rate parameters from Zhang et al.28 wereused. The equilibrium constants used for reactions 10b, 10d, and10h are derived from the equilibrium constants used for reactions10a,15 10c,16 and 10g14 to ensure consistency between allreactions. The results of these optimizations are shown inTable 3. Each reaction set in Table 3 was optimized until areaction set was found whose value of min )ΔY q,act � ΔY exp )

was lower than the min )ΔY q� ΔY exp ) of the next reaction set.The reaction set consisting of reactions 10b, 10c, 10f, and 10g isfound to be the best set, as seen in Table 3. A comparison ofequilibrium predictions and plant data are shown in Figure 3. Theresults of this optimization for this best reaction set are shown inFigure 4, and the respective errors are given in Table 4. As shownin Figures 3 and 4, the proposed kinetic model of the furnace andWHB provides superior predictions to a simple equilibriumapproach. Rate equations found from the optimization of reac-tion set consisting of reactions 10b, 10c, 10f, and 10g are shownin Table 5. It should be noted that the degree of constraint doesnot appear to affect the reaction set that is selected. As anexample, the reaction set consisting of reactions 10b, 10d, 10f,and 10g is the least constrained reaction set that is considered.However, this reaction set ranked only fifth out of the seven setsexamined. Additionally, the activation energy for reaction 10cthat was used for themodeling of the furnace, throughout the restof this paper, did not change from the value reported byHawboldt et al.,16 even though the bounds on the activationenergy were set equal to those reported in the paper.16

NuD ¼

ð0:79 lnðReDÞ � 1:64Þ�2

8

!PrðReD � 1000Þ

1þ12:7ð0:79 lnðReDÞ � 1:64Þ�2

8

!1=2

ðPr2=3 � 1Þ

ð12ÞKinetic Modeling of Ammonia Destruction Reactions.The

experimental data sets from Sames et al.10 that were used above

Figure 3. Equilibrium predictions vs plant data from the outlet of the WHB using reaction set 10b, 10c, 10f, and 10g, where dotted lines show(20% error.



did not contain ammonia in its feed. However, completeammonia conversion is one of the critical aspects of a Clausprocess design whenever it is present in the feed to the Clausprocess. In this section, the appropriate reaction(s) and its(their) kinetic parameters to capture the ammonia conversionprecisely are sought. However, the two ammonia conversionreactions, reactions 10i and 10j, in the set above are not linearlyindependent in the reacting system. The work of Clark et al.17 hasshown that the reaction of ammonia with sulfur dioxide, reaction10j, is more rapid than the pyrolysis of ammonia, reaction 10i,which they have also shown to be inhibited by the presence ofwater and hydrogen sulfide, two components that are in highconcentrations in the Claus furnace. Therefore, reaction 10j waschosen, as Clark et al.17 have shown that this reaction is initiatedat lower temperatures and that reaction 10i is strongly inhibitedby water and hydrogen sulfide. The missing kinetic parametersfor reaction 10j, an ammonia destruction reaction, are esti-mated first. For deriving a rate expression for reaction 10j, thepublished data from Clark et al.17 are used. As these experimentsare carried out in a heated tubular reactor, an ideal plug flowreactor is used to model the reactor. For reaction 10j, the form ofthe rate expression is assumed to be that given in eq 13 whereconcentrations are in kmol/m3.

�rNH3 ¼ A1 exp�Ea1RT

� �Cm1NH3

Cn1SO2

ð13Þ

An optimal set of parameters for this rate expression were foundusing the Aspen Plus’s Datafit toolbox. The optimization algo-rithm in the toolbox attempts to minimize a sum of squares errorfunction by manipulating specified decision variables. The reac-tion rate found is given as eq 14 and the comparison of thesimulated ammonia conversion and experimental ammoniaconversion is shown in Figure 5.

�rNH3

kmolm3s

� �¼ 22860 exp

�27:5 kcalmol

� �RT

0BBB@

1CCCAC0:25

NH3C0:5SO2

ð14Þ

Catalytic Reactors. Unlike the reactions taking place in thereaction furnace, the reactions occurring in the catalytic reactorshave been more thoroughly studied.7,29,30 The catalytic reactorsconvert the hydrogen sulfide and sulfur dioxide into sulfur andwater. In addition, COS formed in the WHB is converted tohydrogen sulfide by hydrolysis. Reactions and rate expressionsused for modeling the catalytic reactors are shown in Table 6.From Figure 2, it can be seen that any COS in the syngas will

pass through the AGR and will be vented to atmosphere afterpassing through the combustion turbine and HRSG. In addition,any COS produced in the Claus unit that slips through the Clausprocess and hydrogenation unit will likewise be recycled to theAGR and be vented to the atmosphere. This represents a majorsource of sulfur emissions from the Claus plant. Therefore, thefirst reactor is generally run at higher temperatures than theproceeding reactors to ensure COS hydrolysis, even though thisdecreases the conversion of hydrogen sulfide by the Clausreaction due to equilibrium. For this study, it was assumed thatthese reactions are kinetically controlled. The sizing of thereactors is based on heuristics from Tong et al.7

Hydrogenation Unit. Inevitably, some sulfur species will notbe converted to sulfur in the Claus unit. To mitigate sulfuremissions to the atmosphere, it is common to consider a tail-gas

Figure 4. Model predictions vs plant data from the outlet of the WHB using reaction set 10b, 10c, 10f, and 10g, where dotted lines show(20% error.

Table 4. Percentage Differences between Model Predictionsand Plant Data

data set 1 2 3 4 5

H2S �10.5 �6.4 27.6 �7.2 6.0

SO2 �15.8 2.5 7.7 �18.1 0.8

H2 2.5 20.1 �39.3 �20.5 6.0

CO �19.6 �3.8 �32.4 �15.4 �30.4

COS 0.8 11.5 2.3 0.9 �11.6

S 15.0 10.8 �2.7 5.2 12.6



treatment unit. In this study, a hydrogenation unit is consideredimmediately following the Claus plant. The hydrogenation unit’sprimary function is to convert all remaining sulfur species in thetail gas of the Claus unit back to hydrogen sulfide so that it may berecovered in the AGR. No rate expressions or experimental datafor the reactions taking place in the hydrogenation unit could befound within the open literature. However, Massie31 claims thatreactions will convert all of the sulfur compounds to hydrogensulfide. For these reasons, the hydrogenation reactor is modeledas an equilibrium reactor comprised of reactions 15a�15e.

CO þ H2O T CO2 þ H2 ð15aÞ

H2S T H2 þ 12S2 ð15bÞ

H2S T H2 þ 16S6 ð15cÞ

H2S T H2 þ 18S8 ð15dÞ

3H2 þ SO2 T H2S þ 2H2O ð15eÞIt should be noted that it is unclear from the literature whetherreaction 15a is catalyzed in the hydrogenation reactor. To clarify

the importance of reaction 15a in the hydrogenation reactor, anexamination of its effect on the performance of the hydrogena-tion reactor was undertaken. As shown in Table 7, the effect ofreaction 15a is quite significant on the performance of thehydrogenation reactor. If reaction 15a is catalyzed in the hydro-genation reactor, the hydrogenation of the remaining sulfurspecies present in the tail gas is nearly complete and requiresno externally supplied source of hydrogen. In fact, the Claus andhydrogenation units are net hydrogen producers. However, ifreaction 15a is not catalyzed, nearly all of the hydrogen producedin the furnace and WHB is consumed and a substantial quantityof the sulfur dioxide and elemental sulfur species still remain.Therefore, to attain the high level of sulfur capture required, anexternal source of hydrogen would be required to ensure thecomplete hydrogenation of all remaining sulfur species. Becauseof the need for an externally supplied source of hydrogen, theClaus and hydrogenation units will be hydrogen consumers.From the literature, it appears that a cobalt molybdenum based

catalyst is often used in the hydrogenation reactor.31,32 From thework of Massie,31 it also appears that reaction 15a is not catalyzed.However, Rameshni32 reports that the cobalt molybdenum catalystdoes catalyze reaction 15a. Traditionally, the catalyst used for thesour shift reactors is a cobalt molybdenum based catalyst.2 As theoperating temperature of the hydrogenation reactor in this work(about 558 K) is higher than the inlet temperature of the sour shiftreactors (about 490 K in the work of Bhattacharyya et al.2), it isanticipated that reaction 15a will take place in the hydrogenator. Forthis reason, reaction 15a is modeled in this work.Base Case Model Development of the Claus plant in an

IGCC plant with CO2 capture. With the developed kineticmodels of the Claus reaction furnace, catalytic reactors, andequilibrium models of the hydrogenation unit, operations of theClaus unit in an IGCC plant were examined. The base case feedsand compositions were based on ref 2 and are given in Table 8.The sizing of the reaction furnace is based upon the size of thefurnace reported by Sames et al.,10 so that the base case residencetime for this case is approximately the same as for the furnace inSames et al.10 The WHB was sized based upon the informationgiven by Nasato et al.6 The tube diameter was taken as 2 in., andthe number of tubes was adjusted to give the same mass flux asreported by Nasato et al.6 As shown in Figure 6, the heat transfercoefficient of the WHB has a strong effect on the productdistribution in the WHB. To capture the effects caused bychanges in flow, composition, and temperature, a calculatorblock was used to determine the heat transfer coefficient of the

Figure 5. Comparison of simulated versus experimental ammoniaconversion.

Table 5. Reactions and Reaction Rates Used in Further Modelinga

reaction reaction rate [kmol/(m3 s)]

H2S þ SO2 þ H2 T S2 þ 2H2Or ¼ 3:583� 107exp

� 26:0RT

� �PH2SPSO2PH2 �

1Keq

PS2P2H2O

!

H2S T12S2 þ H2 r ¼ 9:169� 105 exp

� 45:0RT

� �PH2SP

0:5S2 � 1

KeqPS2PH2

!

CO2 þ H2 T CO þ H2Or ¼ 1:515� 1012 exp

� 60:3RT

� �C0:5H2CCO2 �

1Keq

CCOCH2O

C0:5H2

!

CO þ 12S2 T COS r ¼ 1:0242� 106 exp

� 13:5RT

� �CCOCS2 �

1Keq

C0:5S2 CCOS

!

aActivation energies are in kcal mol�1, partial pressures are in atm, and concentrations are in kmol m�3.



WHB using the Gnielinski correlation, eq 12. It was assumed thatall resistance to heat transfer was from the process gas side andthat the heat transfer coefficient was constant along the length ofthe WHB. The heat transfer coefficient was calculated from theinlet properties of the process gas. This gave a heat transfercoefficient that was approximately 49 W m�2 s�1 in most of thecases investigated and is marked with the dotted line in Figure 6.Figures 7 and 8 show the composition and temperature profilesof the WHB when the oxygen flow is manipulated to give anH2S/SO2 ratio of 2 at the outlet of the WHB, for the base casefeed flows, compositions, and temperatures. These figures showthat the reactions get quenched within about 40% of the WHBlength, after approximately 0.29 s, as the temperature decreasesbelow 873 K. The heat transfer coefficient can vary along thelength of the WHB. The maximum variation is found to be about40%. However, the majority of the composition changes occur-ring in the WHB occur at the entrance, where the heat transfercoefficient is calculated in this work. To keep the optimizationstudy tractable, the heat transfer coefficient is assumed to beconstant along the length of the WHB

’RESULTS AND DISCUSSIONS

Little discussion is available within the open literature aboutthe optimal operation or limiting operational conditions of aClaus unit as part of an IGCC power plant with carbon capture.Using the developed kinetic model, two optimization studies are

Table 7. Effect of Reaction 15a on Exit Flow Rates [kmolh�1] from the Hydrogenation Reactor

H2S SO2 Svap H2

with reaction 15a 7.28 7.43 � 10�12 9.55 � 10�3 17.7

without reaction 15a 6.08 0.754 0.451 9.73 � 10�4

Table 8. Base Case Feeds and Compositions for the ClausFurnace

AGR off-gas sour water stripper gas

temp. [�C] 48.9 119

pressure [atm] 2.51 2.38

molar flow [kmol h�1] 388 87.2

Composition [mol %]

H2S 41.0 27.3

CO2 44.8 19.3

NH3 2.0 36.1

N2 8.1 1.5

CO 0.0 2.4

Ar 0.0 1.5

H2O 4.0 0.0

H2 0.0 11.6

Table 6. Reactions and Rates Considered in Catalytic Reactorsa

reaction reaction rate [mol/(s kgcat)] ref

H2S þ 12SO2 T H2O þ 3

16S8

r ¼ 5364 exp�7:35RT

� � PH2SP0:5SO2

� 1Keq

PH2OP0:1875S8

1 þ 1:140 exp�0:6RT

� �PH2O

� �2

0BBBB@

1CCCCA

29

COS þ H2O T CO2 þ H2S

r ¼ 8:258� 10 � 4 exp13:84RT

� �PCOSPH2O

1 þ 1:141� 10 � 5 exp19:9RT

� �PH2O

0BBB@

1CCCA

30

aActivation energies are in kcal mol�1, and partial pressures are in atm.

Figure 6. Sensitivity of hydrogen yield and oxygen demand to maintainH2S/SO2 ratio to heat transfer of WHB.

Figure 7. Composition profiles in the WHB (with a heat transfercoefficient of 49 W/m2 K).



presented that investigate the effect of the WHB steam pressureand H2S/SO2 ratio on the power output of the IGCC plant. Aninvestigation of the limiting operational conditions of the Clausunit is also examined.Optimal WHB Steam Pressure and H2S/SO2 Ratio to the

Tail Gas Treatment Unit. Equation 16 describes the effect theClaus unit has on the power requirements in the IGCC plant. Itshould be noted that eq 16 does not account for all powerrequirements and losses associated with the Claus unit in anIGCC plant with carbon capture. For example, it does notconsider the baseline power requirement for recovering thehydrogen sulfide in the AGR, power loss associated with burningthe carbon monoxide and hydrogen from the sour water shift inthe furnace rather than the gas turbine, etc. However, eq 16captures most of the trade-offs in the operation of a Claus unit.

PClaus ¼ � 11:72 _nO2

kmolh

� �

þ 47:96 _nH2

kmolh

� �� 61:19 _nH2S

kmolh

� �

þ Wtailcomp½kW� þ Wsteamturb½kW� ð16Þ

The first term of the function describes the power requirement inthe air separation unit (ASU) for producing the 95 mol % pureoxygen used as the oxidant for the Claus furnace. This powerrequirement is based on the power requirements of the main aircompressors and oxygen compressors of the ASU. As the Clausprocess yields hydrogen as a product, the second term isassociated with the power production possible by sending thegenerated hydrogen to the gas turbine (GT) and steam turbinesfor power production. This is a unique option for the Clausprocess as part of an IGCC. In more conventional applications ofthe Claus process, that is, the petroleum industry, not muchbenefit may be realized by recycling the hydrogen to the AGR, asit is usually burned in a furnace. However, in the IGCCapplication, there exists the possibility of efficiently recoveringthe heat of combustion of the hydrogen in the GT and steamturbines rather than burning it in the furnace to raise steam. Thepower production from hydrogen is based on the power produc-tion of the syngas expander, GT, and steam turbines. The third

term represents the power requirement for recovering theunconverted hydrogen sulfide from the tail gas recycle streamin the AGR. The power requirement of the AGR is based uponthe power requirements of the stripped gas compressor, leanSelexol pump, refrigeration duty in the solvent cooler, and thestripper reboiler.2 Using this approximation yields an expectedpower requirement of 87.41 kW kmol�1 H2S. It is anticipatedthat the true value will be lower than this because this value isbased largely on the recovery cost from the raw syngas. As thepartial pressure of both H2S and CO2 are higher in the tail gasthan in the syngas, it is anticipated that the incremental cost forrecovery from the tail gas will be lower. Comparing the partialpressure of carbon dioxide in the tail gas with that of the rawsyngas, it is expected that the power requirement will beapproximately 70% of value previously calculated, giving approxi-mately 61.19 kW kmol�1 H2S. Carbon dioxide is chosen as thescaling factor because, with the deep recovery of carbon dioxidethat is considered in the case of an IGCC plant with CO2 capture,the recovery of hydrogen sulfide is automatically satisfied. Thefourth term represents the power requirement to compress thetreated tail gas from approximately 1 atm to the operationpressure of the AGR, approximately 50 atm. This term isdependent upon the flow of tailgas from the hydrogenationreactor, which depends upon the operation of the Claus process.The fifth term represents the power production possible by usingthe remainder of the steam produced in the WHB to power steamturbines. For determining the amount of steam generated in theWHB, it was assumed that boiler feedwater’s temperature wasconstant at 411 K and that 5% of the total heating duty of theWHBis lost to the environment. After the steam is used to provide theheating duty for the Claus process and hydrogenation unit, theremainder is taken for power production in the steam turbines. Thenumbers used in eq 16 are derived from Bhattacharyya et al.2

Additionally, all sulfur condensers in the process generate steam at4.4 atm, which is also sent to the steam turbines.The first optimization study was done on the effect of the

pressure of the steam raised in the WHB on the objectivefunction given above. The results of this study are given inFigure 9. The pressure of the various stages of the steam turbine,to which the steam generated in the WHB can be sent, isconsidered to be fixed at the level set in the previous study fromour group.2 This results in discontinuities at 124.0, 96.4, and 59.5 atm.This study finds that there exists an optimal steam pressure

Figure 8. Temperature profile in WHB using a steam temperature of268 �C.

Figure 9. Effect of steam pressure on Claus power production.



of 61 atm that should be generated in the WHB. Generatingsteam at very high pressures has several negative impacts in aClaus plant. As hydrogen and sulfur dioxide are thermodynami-cally preferred at high temperatures, one would like to freezetheir composition at this higher temperature, where they arethermodynamically preferred. This means increasing the quenchrate by decreasing the temperature of the generated steam andthus the pressure. As shown in Figure 10, decreasing steampressure decreases the quench time, increasing hydrogen yieldand decreasing the quantity of oxygen required to maintain aspecified H2S/SO2 ratio. Also, at very high pressures, less powergeneration in the steam turbines is achieved. This is because lesssteam is generated in the WHB and a large portion of this highquality steam is used for the heating duties of the process.With the pressure of the steam generated fixed at 61 atm, an

optimization study was done to determine the optimal H2S/SO2

ratio to run the process in an IGCC application. The H2S/SO2

ratio is maintained by manipulating the oxygen flow to the Clausprocess. The results of this study are shown in Figures 11 and 12.Traditionally, the process is run at an H2S/SO2 ratio as close aspossible to 2 at the inlet of the tail gas treatment unit19 to

maximize the conversion of hydrogen sulfide. However, whenone considers the possibility of hydrogen production for powergeneration,6 the optimal ratio increases. Additionally, as anenriched oxygen stream is used for oxidation, there is anassociated cost for its production. Also important is that thoughhydrogen sulfide conversion is not particularly sensitive to theH2S/SO2 ratio beyond about 2, hydrogen flow is fairly sensitiveto changes in H2S/SO2 ratios above 2. For example, from a ratioof 2.1 to 6.4, the hydrogen sulfide flow increases by 0.597 kmol h�1;this corresponds to single pass conversion decreasing from 96.6%to 96.3%. However, over the same interval, hydrogen flowincreases by 13.1%. For these reasons, the IGCC power outputis maximized when the Claus unit operates at an H2S/SO2 ratioof approximately 0.8 at the inlet of the first catalytic reactor, 2.5 atthe inlet of the second catalytic reactor, and 4.3 at the inlet of thetail gas concentration unit. The corresponding H2S/SO2 ratios atthe inlet of the first and second catalytic reactors at varying ratiosat the inlet of the hydrogenation unit are provided in Figure 13.Operability Study. There are two important variables that

one would like to maintain while operating a Claus process: the

Figure 10. Effect of steam pressure on power production from steamturbines, hydrogen, and power requirement for oxygen production.

Figure 11. Effect of H2S/SO2 ratio at inlet to hydrogenation unit onpower production of the Claus unit.

Figure 12. Effect of H2S/SO2 ratio on hydrogen and hydrogen sulfideflow to the AGR.

Figure 13. Effect of H2S/SO2 ratio at inlet to hydrogenation unit onH2S/SO2 ratio at inlet of first and second catalytic reactor.



H2S/SO2 ratio and the adiabatic flame temperature. The adia-batic flame temperature is important because it affects both theflame stability and the level of destruction of ammonia. Main-taining the H2S/SO2 ratio is important for satisfying the sulfuryield in the catalytic reactors. A study was carried out investigat-ing the operating envelope in which the H2S/SO2 ratio andadiabatic flame temperature can be maintained by manipulatingthe oxygen flow rate and the furnace bypass flow in the face ofdisturbances that may be encountered in the operation of theClaus process as part of an IGCC. These disturbances includechanges in flow and composition of the feeds to the Clausprocess. For this study, the following assumptions were used: thesour water gas composition, given in Table 7, and the hydrogensulfide flow from the AGR are both constant. Hydrogen sulfideflow from the AGR was held constant at 159 kmol h�1. Aconstraint was considered to ensure that no oxygen can slip fromthe flame zone of the furnace to the anoxic region where thebypassed acid gas is injected. This was to avoid safety relatedissues concerned with flashback in the bypassed acid gas line. Asapproximately 78% of the sour water gas is combustible, theadiabatic flame temperature would always be greater than theminimum required flame temperature if sour water gas were theonly gas fed to the furnace. It was found that, for a given sourwater gas flow, there is a minimum purity requirement of the acidgas from the AGR and amaximum allowable acid gas bypass. Theresults of this study are shown in Figure 14. These results areobviously dependent on the composition of the sour water gas,hydrogen sulfide flow, and the H2S/SO2 ratio that is beingmaintained, but this study does show that there is a strongdependency of the minimum acid gas purity from the AGR andthe maximum allowable bypass on the flow of sour water gas. Asshown in Figure 14, as the flow of sour water gas increases, theminimum allowable acid gas purity required decreases. Also, asshown in Figure 15, as the sour water gas flow increases, theoxygen flow increases and the acid gas bypass required formaintaining the flame temperature increases. These results showthat given a sour water gas flow, the acid gas purity has a lowerbound, below which flame temperature is below the minimumvalue, and the acid gas bypass has an upper bound, above whichoxygen will slip to the anoxic region.

Impact of Carbon Capture on the Claus Process. Asmentioned before, the Claus unit considered here is part of anIGCC plant with CO2 capture. Due to the high partial pressure ofcarbon dioxide in the Selexol-based AGR unit,2 a relatively impureacid gas is obtained from the Selexol unit. CO2 concentration in thisacid gas is about 40%. This is very different than a typical Claus feedavailable in a refinery.10 It should also be noted that most of theresults available in the open literature for Claus processes relate torefinery operation. Because of this impure acid gas, the use of feedpreheating, enriched oxygen from an ASU as the oxidant, and anacid gas bypass are required to ensure sufficiently high furnacetemperatures. Ensuring the furnace is maintained above a criticaltemperature is necessary to ensure complete destruction of ammo-nia and methane, as well as ensuring a stable flame. Another strongimpact of the CO2 capture is that the high partial pressure of carbondioxide allows the hydrogenation unit to be operated without anexternal supply of hydrogen and allows the process to be a nethydrogen producer. This is due to the water gas shift reaction,reaction 15a, occurring in the hydrogenation reactor. Whether theClaus and hydrogenation units are net hydrogen consumers orproducers is dependent not only upon whether reaction 15a is or isnot catalyzed but also upon the process operation and feed gascompositions. From Rameshni,32 it is stated that reaction 15a canyield 70% of the hydrogen required in the hydrogenation reactor;however, this number is based upon some assumed feed gascomposition. The studies presented in this paper show thatsubstantial production of carbon monoxide from carbon dioxideoccurs in the furnace and WHB. This carbon monoxide, uponentering the hydrogenation unit yields hydrogen via reaction 15a.This hydrogen production, because of the high concentration ofcarbon monoxide, is greater than the hydrogen required forcomplete hydrogenation of the remaining sulfur species present inthe tail gas. Because of this, the Claus and hydrogenation units arenet hydrogen producers in this instance. Experimental data withsimilar feed composition would be very helpful to validate theseresults, but unfortunately such data are scarce in the open literature.

’CONCLUSION

To facilitate the effective modeling of a Claus furnace, a four-stage method has been proposed to determine an optimal set of

Figure 14. Effect of sour water gas flow on the limiting case of AGR gaspurity.

Figure 15. Effect of sour water gas flow on the limiting case of AGR gasbypass and oxygen flow.



linearly independent reactions that would best describe theproduct distributions from available plant data. The first stageinvolves the determination of the number of linearly indepen-dent reactions that will describe the reacting system. The secondstage determines the best possible fit that one can obtain in thepresence of experimental error/noise. The third stage requires adatabase of possible reactions acquired from process knowledgeor literature. A quadratic programming (QP) problem is thensolved, subject to constraints on the elements of the matrix ζbased upon thermodynamics and process knowledge. Solvingthis QP problem is computationally inexpensive and allows oneto rank the possible reaction sets. The fourth stage requiressolving a nonlinear programming problem for each possiblereaction set starting with the highest ranked reaction set fromthe third stage. This stage considers thermodynamics, kinetics,and any additional phenomena considered in the actual model.The solution of this NLP problem is computationally expensiveand can take hours to solve, especially using a sequential modularmethod. The minima of the objective function of this NLPproblem is compared against the results attained in the thirdstage. If it is found to be less than the minima of the next reactionset in the third stage, an optimal reaction set had been found, elsethe search continues. This method has been demonstrated on theClaus furnace and has shown that it significantly decreases thenumber of reaction sets to be considered in the fourth stage.

After developing the model of the Claus furnace, a kineticmodel of the entire modified Claus process has been developedto determine the effect of the unit’s operation on the poweroutput of an IGCC power plant with CO2 capture. With thedeveloped kinetic model, a sensitivity study was undertaken todetermine the optimal steam pressure of the WHB. At lowerpressures of steam generated in the WHB, more hydrogen isyielded from the process and less oxygen is required for main-taining a specified H2S/SO2 ratio; however, less power isgenerated from the excess steam not required for heating dutiesin the process. At higher pressures, less power is generated fromthe excess steam as less heat is recovered in the WHB and high-temperature high-pressure steam is largely used for heatingduties in the process. This study showed that there exists anoptimal steam pressure generated in the WHB that balanceshydrogen yield, oxygen demand, and power generation.

Additionally, with the developed kinetic model, a sensitivitystudy was undertaken to determine optimal H2S/SO2 ratio to thetail gas treatment unit. At H2S/SO2 ratios close to 2, single-passconversion is maximized, thus decreasing the cost associated withrecapture of hydrogen sulfide in the AGR. However, by operatingat H2S/SO2 ratios greater than 2, higher yields of hydrogen arepossible, which can be used for high efficiency power generation inthe gas turbine and steam turbines. This study showed that thereexists an optimal H2S/SO2 ratio that balances single pass conver-sion, hydrogen yield, oxygen demand, and power generation.

A study was undertaken to examine the operating envelope inwhich both H2S/SO2 ratio and adiabatic flame temperature canbe controlled in the face of typical disturbances possible in theoperation of an IGCC power plant with CO2 capture. It has beenshown here that, for a given flow of the sour water gas, there is aminimum H2S concentration in the acid gas stream from theAGR to maintain the desired H2S/SO2 ratio and the adiabaticflame temperature while ensuring no oxygen slips to the anoxicregion of the furnace.

This study shows that a Claus unit as part of an IGCC plantwith CO2 capture is a net hydrogen producer unlike the typical

Claus units as part of a refinery, which are hydrogen consumers.This happens because of a Claus feed that is very rich in CO2.The authors believe that experimental data considering a feed gasthat is similar in composition to the feed gas considered in thisstudy would be very beneficial for comparison with our resultsand further enhancement of the reaction kinetics, if needed.

’AUTHOR INFORMATION

Corresponding Author*Phone: 3042939335. E-mail: [email protected].

’ACKNOWLEDGMENT

This technical effort was performed in support of the NationalEnergy Technology Laboratory’s ongoing research in Processand Dynamic Systems Research under the RES contractDE-FE0004000.

’REFERENCES

(1) National Energy Technology Laboratory. Cost and PerformanceBaseline for Fossil Energy Power Plants Study, Volume 1: Bituminous coaland natural gas to electricity, DOE/NETL-2007/1281; Aug 2007;Available online: http://www.netl.doe.gov.

(2) Bhattacharyya, D.; Turton, R.; Zitney, S. Steady-state simulationand optimization of an integrated gasification combined cycle (IGCC)power plant with CO2 capture. Ind. Eng. Chem. Res. 2011, 50, 1674–1690.

(3) Encyclopedia of Energy Engineering and Technology; Capehart,B. L., Ed.; CRC Press: Boca Raton, FL, 2007; Vol. 2�3.

(4) Rameshni, M.; Street, R. PROClaus: The New Standard for ClausPerformanceWorleyParsons: 2001

(5) Jacobs Consultancy U.K. Impact of CO2 removal on coal gasifica-tion based fuel plants; Feb 2006.

(6) Nasto, L. V.; Karan, K.; Mehrotra, A. K.; Behie, L. A. Modelingreaction quench times in the waste heat boiler of a Claus plant. Ind. Eng.Chem. Res. 1994, 33, 4–13.

(7) Tong, S.; Dalla Lana, I. G.; Chuang, K. T. Effect of catalyst shapeon the hydrolysis of COS and CS2 in a simulated Claus converter. Ind.Eng. Chem. Res. 1997, 36, 4087–4093.

(8) Puchyr, D. M. J.; Mehrotra, A. K.; Behie, L. A.; Kalogerakis, N.Hydrodynamics and kinetic modeling of circulating fluidized bedreactors applied to a modified Claus Plant. Chem. Eng. Sci. 1996, 51,5251–5262.

(9) ZareNezhad, B; Hosseinpour, N. Evaluation of different alter-natives for increasing the reaction furnace temperature of Claus SRU bychemical equilibrium calculations. Appl. Therm. Eng. 2008, 28, 738–744.

(10) Sames, J. A.; Paskall, H. G.; Brown, D. M.; Chen, M. S. K.;Sulkowski, D.Field measurements of hydrogen production in an oxygen-enriched Claus furnace.Proceedings of Sulfur 1990 International Confer-ence, Cancun, Mexico, April 1�4, 1990.

(11) Hawboldt, K. A. Kinetic modelling of key reactions in themodified Claus plant front end furnace. Ph.D. Thesis, Department ofChemical and Petroleum Engineering, University of Calgary, Canada,1998.

(12) Monnery, W. D.; Hawboldt, K. A.; Pollock, A. E.; Svrcek, W. Y.Ammonia pyrolysis and oxidation in the Claus furnace. Ind. Eng. Chem.Res. 2001, 40, 144–151.

(13) Karan, K.; Mehrotra, A. K.; Behie, L. A. A high-temperatureexperimental and modeling study of homogeneous gas-phase COSreactions applied to Claus plants. Chem. Eng. Sci. 1999, 54, 2999–3006.

(14) Clark, P. D.; Dowling, N. I.; Huang, M.; Svrcek, W. Y.;Monnery, W. D. Mechanisms of CO and COS formation in the Clausfurnace. Ind. Eng. Chem. Res. 2001, 40, 497–508.



(15) Monnery, W. D.; Hawboldt, K. A.; Pollock, A.; Svrcek, W. Y.New experimental data and kinetic rate expression for the Clausreaction. Chem. Eng. Sci. 2000, 55, 5141–5148.(16) Hawboldt, K. A.; Monnery, W. D.; Svrcek, W. Y. New experi-

mental data and kinetic rate expression for H2S pyrolysis and re-association. Chem. Eng. Sci. 2000, 55, 957–966.(17) Clark, P. D.; Dowling, N. I.; Huang, M. Mechanisms of

ammonia destruction in the Claus furnace. Proceedings of the LauranceReid Gas Conditioning Conference 2001, 301–318.(18) Kohl, A. L.; Nielson, R. Gas Purification, 5th ed.; Gulf Professional

Publishing: Houston, TX, 1997.(19) Taggart, G. W. Optimize Claus control. Hydrocarbon Process.

1980, 133–138.(20) Jones, D.; Bhattacharyya, D.; Turton, R.; Zitney, S. E. Optimal

design and integration of an air separation unit (ASU) for an integratedgasification combined cycle (IGCC) power plant with CO2 capture. FuelProcess. Technol. 2011, 92, 1685–1695.(21) Sassi, M.; Gupta, A. K. Sulfur recovery from acid gas using the

Claus process and high tempeature air combustion (HiTAC) technology.Am. J. Environ. Sci. 2008, 5, 502–511.(22) Westbrook, C. K.; Dryer, F. L. Simplified mechanisms for the

oxidation of hydrocarbon fuels in flames. Combust. Sci. Technol. 1981,27, 31–43.(23) Peters, N. Premixed burning in diffusion flames—The Flame

Zone Model of Libby and Economos. Int. J. Heat Mass Transfer 1979,22, 691–703.(24) Clark, P Chemistry of the Claus furnace. Sulfur 2003, 285,

24–26.(25) Aris, R.; Mah, R. H. S. Independence of chemical reactions. Ind.

Eng. Chem. Fundam. 1963, 2, 90–94.(26) Gnielinski, V. Equations for heat and mass transfer in turbulent

pipe and channel flow. Int. Chem. Eng 1976, 16, 359–368.(27) Karan, K.; Mehrotra, A. K.; Behie, L. A. COS-forming reaction

between CO and sulfur: A high-temperature intrinsic kinetics study. Ind.Eng. Chem. Res. 1998, 37, 4609–4616.(28) Jia-yuan, Zhang; Jie-min, Zhou; Hong-jie, Yan Kinetic model

on coke oven gas with steam reforming. J. Cent. South Univ. Technol.2008, 15, 127–131.(29) Birkholz, R.; Behie, L.; Dalla Lana, I. Kinetic modeling of a

fluidized bed Claus plant. Can. J. Chem. Eng. 1987, 65, 778–784.(30) Tong, S.; Dalla Lana, I.; Chuang, K. Kinetic modeling of the

hydrolysis of carbonyl sulfide catalyzed by either titania or alumina. Can.J. Chem. Eng. 1993, 71, 392–400.(31) Massie, S. N.. Less fuel, less fire, less pollution using low

temperature tail gas catalyst and catalytic incineration in sulfur plants.Gas Processors Association Europe Spring Conference, 2006.(32) Rameshni, M. Selection Criteria for Claus Tail Gas Treating

Processes; WorleyParsons: 2010.

55 dustin jones gasification

Documents