Kinetic Rates of Oxidation and Gasiﬁcation …...In this work, the kinetic rates of oxidation, and the oft-neglected CO2 and steam gasiﬁcation reactions for three pulverized coal

Paper # 070CO-0309 Topic: Coal and Biomass Combustion and Gasification

8th US National Combustion MeetingOrganized by the Western States Section of the Combustion Institute

and hosted by the University of UtahMay 19-22, 2013.

Kinetic Rates of Oxidation and Gasification Reactions of CoalChars Reacting in Oxy-Combustion Environments

Ethan S Hecht1 JoAnn S Lighty2 Christopher R Shaddix1

1Combustion Research Facility,Sandia National Laboratories, Livermore, CA

2Department of Chemical Engineering, Institute for Clean and Secure Energy,University of Utah, Salt Lake City, UT

In this work, the kinetic rates of oxidation, and the oft-neglected CO2 and steam gasification reactionsfor three pulverized coal chars were determined through experiments and analysis using a single-filmmodel amenable for inclusion in CFD simulations of boilers. The three coals considered in this workwere Black Thunder, Utah Skyline, and Illinois #6: one sub-bituminous and two high-volatile bitumi-nous coals. Size-classified char particles were pre-generated at a high heating rate and then reacted inan entrained flow reactor with oxygen concentrations ranging from 24-60%, steam concentrations from10-16%, and N2 and CO2 diluents. High oxygen concentrations were chosen to increase particle tem-peratures (from the exothermic oxidation reactions) and overcome the high activation energy barriersfor the gasification reactions. Temperatures of individual particles were measured, and burnout mea-surements were made on collected chars. Best-fit activation energy and pre-exponential factors for theoxidation, CO2 and steam gasification reactions were determined for each of the three chars. A transientsingle-film model demonstrates that the deduced kinetic rates effectively capture the experimentally ob-served trends, such as a higher temperature for the sub-bituminous char, the size dependence on peakchar temperature, and the differences between CO2 and N2 diluent atmospheres.

1 Introduction

Among the promising means of carbon capture and storage from coal-fired power plants is oxy-combustion. Oxy-combustion involves the separation of oxygen from air before introduction tothe boiler, and then combustion of the solid fuel with pure oxygen diluted with recycled flue gasto control the temperature. Recycle could occur before or after moisture removal, but the recyclestream (which is primarily CO2) is likely to contain some amounts of moisture [1]. After con-densing out the moisture, the oxy-combustion effluent is highly-concentrated CO2, which can bereadily condensed, transported, and stored. At the University of Utah, a DQMOM/LES formu-lation for heterogeneous chemistries has been developed [2] for which validation and verificationdata is necessary for predictive modeling of the oxy-combustion of coal. A predictive model withquantified uncertainty will allow the rapid development of new build and retrofit oxy-combustionboilers.

For a retrofit application, recirculation of the flue gas could cause CO2 levels in the furnace toapproach 60-70 vol-%., and water vapor levels of up to 25-35 vol-% in post-flame furnace gases.

1

8th US Combustion Meeting – Paper # 070CO-0309 Topic: Coal and Biomass Combustion and Gasification

Both CO2 and H2O have higher molar heat capacities than N2, and are radiantly active gas specieswhile N2 is not. These gas properties will cause obvious heat transfer differences between air-fired and oxy-combustion environments. In addition, oxygen diffuses more rapidly through N2

than CO2, but more rapidly through H2O than either N2 or CO2, which will change the masstransfer characteristics during oxy-combustion. Previous research has shown that the substitutionof CO2 for N2 delays coal ignition and reduces the rate of coal volatile consumption, for a givenfurnace temperature and oxygen concentration [3–5]. A reduced burning rate in a oxy-combustionenvironment has been attributed to the reduced rate of oxygen diffusion in the boundary layer[5–7].

In addition to these known effects, with elevated levels of CO2 and H2O, gasification reactions,which are traditionally neglected for air-fired combustion, may begin to impact the char combus-tion rate and temperature. There has been some evidence that gasification reactions enhance thecombustion rate under oxy-combustion conditions [8–10], but the high endothermicity of the gasi-fication reactions lowers the combustion temperature, complicating the effects of these reactions.In some recent modeling efforts, we found that when a char particle is reacting at a temperaturenear the gas temperature (i.e. when there is a low O2 concentration in the surrounding gas), theCO2 gasification reaction increases the overall consumption of solid carbon, but only by a smallamount. Alternatively, when the particle is reacting at a temperature considerably higher than theambient gas temperature (i.e. in a high O2 concentration environment), the CO2 gasification re-action decreases the overall carbon removal rate, again by a small amount [11]. With a differenttreatment of the surface area, we found that although both the CO2 and H2O gasification reactionsreduced the combustion temperatures, the char carbon removal rate slightly increased in a widerange of environments [12].

Our recent work demonstrated that under a wide range of environments, a single-film model can beaccurately used to predict pulverized coal particle combustion temperatures and burning rates [13].While gas-phase reactions only have a small influence on the combustion temperatures and burningrates, the CO2 and steam gasification reactions are important to accurately predict temperaturesin oxy- and conventional combustion environments. The heterogeneous reactions that must beconsidered are

2Cs + O2 → 2CO, (1)Cs + O2 → CO2, (2)

Cs + CO2 → 2CO, (3)Cs + H2O→ H2 + CO. (4)

In this analysis, we have focused on determining the kinetic reaction rates of these four reactions,for use as a single-film sub-model in a CFD code.

2 Methods

Three coals were chosen for this work: a representative high volatile bituminous coal from theIllinois #6 seam, a low-sulfur western bituminous coal from the Utah Skyline mine, and BlackThunder sub-bituminous coal from the Powder River Basin (PRB). A proximate and ultimate anal-ysis of these coals is shown in Table 1. The PRB coal has the lowest fixed carbon, the highest

2


Table 1: Proximate and Ultimate analyses for the three project coals

Coal Black Thunder Utah Skyline Illinois #6

Moisture 23.69 3.18 9.64 wt-%, as receivedAsh 4.94 8.83 7.99 wt-%, as received

Volatile Matter 33.36 38.6 36.78 wt-%, as receivedFixed Carbon 38.01 49.39 45.58 wt-%, as received

C 75.27 80.24 78.51 wt-%, dry, ash freeH 5.03 5.75 5.49 wt-%, dry, ash freeN 1.09 1.61 1.36 wt-%, dry, ash freeS 0.32 0.60 4.83 wt-%, dry, ash free

O (by diff) 18.29 11.80 9.81 wt-%, dry, ash free

HHV 12720 14327 14080 BTU/lb., dry, ash free

moisture content, and the lowest heating value. These characteristics are expected since the PRBcoal is sub-bituminous and of lower rank than other two bituminous coals. The two bituminouscoals are similar in composition and heating value, although the Illinois #6 coal has a higher sulfurcontent.

These coals were first devolatilized using a drop tube furnace at the University of Utah. The furnacewas operated at 1200 ◦C while flowing 15 SLPM nitrogen with approximately 1 vol-% oxygen toburn off any tars that might have been generated during devolatilization. The furnace has an IDof 5 cm and a heated length (of temperatures between 1000 and 1200 ◦C) that is about 60 cmlong. This led to a residence time of about 0.9 sec. Chars were collected and then sieved into 6narrow size bins (53-63 µm, 63-75 µm, 75-90 µm, 90-106 µm, 106-125 µm, and 125-150 µm).Interestingly, the majority of PRB char was smaller than the parent coal (likely through fracturingand loss of volatiles), the Utah Skyline char was roughly the same size as the parent coal, whilethe Illinois #6 chars seemed to swell during devolatilization.

Chars in the narrowly sieved size bins were then fed into Sandia’s optical entrained flow reactor(EFR) that has been described previously [14]. A flat-flame Hencken burner is used to generatea high-temperature, well characterized environment into which the char particles are fed. As theparticles flow upwards in the reactor, optical diagnostics simultaneously measure size, velocity,and temperature of individual particles. Twelve different environments were used in this study:two at 24 vol-% O2, with 14 vol-% H2O, and a balance N2 or CO2; four at 36 vol-% O2, with10 and 14 vol-% H2O, and a balance of N2 or CO2; and 6 at 60 vol-% O2, with 10, 14 and16 vol-% H2O, and a balance N2 or CO2. Significantly oxygen-enriched conditions were chosenso that oxidation reactions would increase the combustion temperature, and thereby the rate ofthe gasification reactions to measurable levels. Intuition might suggest that at high temperaturesreactions would proceed very fast leading to diffusion limitations, but Murphy and Shaddix foundthat char particles react under increasing kinetic control at higher oxygen concentrations because ofthe higher oxygen flux available to the particle at these elevated oxygen concentrations[14]. Whennitrogen was used as a diluent, the Hencken burner was operated at an adiabatic flame temperatureof 1750 K, and with the CO2 diluent, at 1850 K. A thermocouple was used to measure the reactor

3


temperatures, and these conditions produced similar temperature profiles that all peaked around1650 K. Temperature, velocity and size measurements were made at 3-7 heights for around 100particles in each of the size bins in all of the environments.

In addition to the optical measurements, under several conditions, chars were also collected, andburnout measurements were made off-line in a TGA. By assuming that the ash in the chars doesnot gasify, the extent of reaction can be determined as a function of residence time, by comparingthe percentage of fixed carbon in the chars. To measure the fixed carbon using the TGA, theASTM standard for proximate analysis of coal was followed [15], neglecting the lid placementand removal and the minimum sample size. A typical sample would have an initial mass of 10 mg,and the instrument (TA Instruments Q600) is sensitive to 0.1 µg. The procedure involved rampingthe sample to 107 ◦C under a nitrogen atmosphere and soaking at that temperature to removeany moisture that might have accumulated during storage and collection of the sample. This wasfollowed by a temperature ramp at 30 K/min and soak for 7 mins at 900 ◦C to remove any remainingvolatiles. The temperature was then ramped down to 600 ◦C, the atmosphere was switched to air,and the temperature was ramped back to 750 ◦C at 3 K/min to react the remaining carbon in thesample. When the weight change was zero, the TGA was cooled back to room temperature. It wasassumed that all of the fixed carbon was removed in the air environment portion of the TGA run,and that only ash remained at the end of this procedure.

In several cases, repeat burnout measurements with samples chosen from one set of collectionexperiments in the EFR were made to estimate errors. Errors in these measurements can be at-tributed to various sources. As the chars were reacting in the EFR, changes in gas delivery tothe Henken burner may have caused variations in the gas temperatures and velocity profiles seenby the particles. The potential for particle-particle interactions (such as radiative heat transfer,disrupted boundary layer, etc.) in the EFR was also increased during collection, since the feedrate was increased during the collection experiments. During collection, variations in suction mayhave caused non-iso-kinetic measurements to be made, effectively changing the residence timeof the collected particles. The TGA measurements require the operator to tare the crucible, re-move it from the balance, fill it with sample, and replace it on the balance. As the balance isa lever-arm, any slight discrepancies in crucible placement can cause significant errors in massmeasurements. Finally, the collected samples may not have been homogeneous, and the selectedsample for burnout measurement on the TGA may not have been representative of the collectedsample.

Since there were three chars in six size bins that reacted in twelve different environments, it was notfeasible to collect chars and make burnout measurements on all of the different variables explored.Therefore, for a given coal char, chars that experienced variations in oxygen concentration or steamconcentration, or size were collected, or for a given environment and size, variations in coal charswere collected.

Quantitative analysis of reaction kinetics is difficult, owing to particle-to-particle variability inreactivity, and the largely unknown kinetic mechanisms. In this work, the single-film model de-scribed by Hecht et al. [13] was fit to the data. The four reaction steps shown by Rxns. 1–4 wereconsidered, and the reaction orders were each first order with respect to the reactant (O2, CO2,and H2O). Reaction 2 was set, based on the kinetic parameters in Rxn. 1 by the relationship givenby Tognotti et al. [18]. Therefore the pre-exponential factor for Rxn. 2 is 0.02 times the pre-

4


exponential of Rxn. 1, and the activation energy is 3070R J/mol lower than the activation energyof Rxn. 1. With these constraints, the six parameters used to fit the data were the activation energiesand pre-exponential factors for Rxns. 1, 3, and 4. The activation energies for these reactions werefurther constrained, based on the literature review described by Hecht et al. [12]. The activationenergy for oxidation to CO (Rxn. 1) was constrained to lie between 150–190 kJ/mol, 230–270J/mol for CO2 gasification (Rxn. 3), and 190–270 for H2O gasification (Rxn. 4).

Fits were found to median optically measured particle temperatures for a size-binned char, in anenvironment, at a given residence time. With measurements at 3-7 heights in each environment,there were about 330 measurements for each char type. A ‘particle’ was created for each of theseroughly 100 optical measurements, reacting at the median temperature at a given residence time.Due to significant uncertainties in the optically measured particle sizes, the diameter of each ofthese ‘particles’ was set to the mid-point of the fed size-bin (regardless of residence time or degreeof burnout at which the temperatures were measured). Then, for a given set of kinetic parameters,the ‘particle’ temperature was set, and the species conservation equations were solved to determinethe concentration of reactants at the particle surface, and the kinetic reaction rates. The error for the‘particle’ was then calculated as the error in the thermal energy balance equation. The appropriatekinetic parameters were determined by minimizing the least-squared error for these ‘particles’through MINPACK’s lmdif and lmdir subroutines, as implemented in SciPy [19].

With kinetic fits, the burnout characteristics of the chars were calculated. A thermal inertia termwas added to the energy balance described by Hecht et al. [12] (using the same nomenclature), sothat it becomes

N ′′C,phC,p − εσ(T 4p − T 4

w

)− ρcprp

3

dT

dt=

ngas∑i=1

N ′′i hi,p +λ

rp

[κ

eκ − 1

](Tp − T∞) , (5)

where cp is the heat capacity of the char (assumed to be constant for all the chars at 2000 J/kg-K). The dT/dt term was evaluated as (T n − T n−1)/∆t, where superscript n indicates the currenttime-step, at which all other temperatures and properties were evaluated. A pseudo-steady statewas assumed, where the particle followed the single-film model at any point in time. Burnoutprogressed following the scheme proposed by Hurt et al. [20]. For a given time, the single-filmmodel was solved (including the inertial term), calculating the particle temperature, mole fractionsof reactants at the surface, and the molar burning rate, NC. The molar burning rate was converted toa mass rate (dm/dt) through the molecular weight of carbon, then the updated mass of the particleis given by

m =4

3ρπr3p +

dm

dtdt. (6)

Conversion was calculated asX = 1−m/m0 wherem0 is the initial mass of the particle. The massfraction of carbon was related to the conversion through the relation xC = (X − XC,0)/(1 − X)where XC,0 is the initial mass fraction of carbon (relative to carbon and ash) in the char parti-cle. The mass of carbon was calculated by mC = xCm (similarly the initial mass of carbon wascalculated by using the initial mass fraction) and then the density of carbon was updated by

ρC = ρC,0

(mC

mC,0

)α. (7)

5


In this equation, α is the mode of burning parameter, assumed to be the constant value of 0.25, assuggested by Mitchell et al. [21], that describes the fraction of carbon removed from the internalsurfaces of the char to the fraction removed from the surface. For α = 1, the density of the particlewould continually decrease, but the diameter would remain fixed, and for α = 0, the density wouldremain constant while the particle diameter would shrink as burnout progressed. After the densityof carbon was calculated, the particle density was updated through the expression

1

ρ=xCρC

+1− xCρa

, (8)

where ρa is the density of the ash (assumed to be 2500 kg/m3). In these simulations, the initialdensity of carbon was assumed to be 400 kg/m3 for all three of the chars, and the initial apparentparticle density was calculated through Eq. 8, using the initial weight fraction of carbon in eachchar. Finally, with the updated density, the diameter of the particle was also updated through therelation

dp =

(6m

πρ

)1/3

. (9)

Using these relationships, the particle solution was calculated from its initial injection into theEFR through heat-up, reaction, and burnout. It should be noted that many of the char structuralparameters, such as the surface area, porosity and tortuosity were assumed to remain fixed asburnout progressed in this model.

3 Results and discussion

3.1 Experimentally measured temperatures

The gas temperatures were measured in each of the different environments using at 0.001” typeR thermocouple. Because of the small diameter, radiation losses were minimized. Nonetheless,these losses were accounted for by an energy balance on the thermocouple wire. For a cylinder incross-flow at low Reynolds number, the Nusselt number was found through the relation given byHilpert [22], Nu = 0.989Re0.33Pr1/3, where Re is the Reynolds number (based on the diameter ofthe thermocouple and the gas velocity) and Pr is the Prandtl number. Convection is balanced byradiation and the gas temperature will satisfy the equation

σε(T 4g − T 4

w) =Nuk

dTC(Tg − TTC), (10)

where σ is the Stefan-Boltzmann constant, ε is the emissivity of the thermocouple wire (assumedto be 0.2), Tw is the wall temperature (assumed to be 500 K), k is the thermal conductivity ofthe gas, dTC is the diameter of the thermocouple wire, TTC is the temperature measured by thethermocouple, and Tg is the gas temperature.

Gas temperatures (corrected thermocouple measurements) are shown in Fig. 1. Differences be-tween the gas and thermocouple temperatures were around 20 K. There was a heat up of the inertgas that fed the particles in the first two inches of the furnace. This was followed by a peak tem-perature of around 1700 K in all but a few of the environments, and then a nearly linear drop-off of

6


0 2 4 6 8 10 12 14 16 18

height above burner (in)

1100

1200

1300

1400

1500

1600

1700

1800

gas

tem

pera

ture

(K)

24% O2 , 14% H2O, CO2

36% O2 , 10% H2O, CO2

36% O2 , 14% H2O, CO2

60% O2 , 10% H2O, CO2

60% O2 , 14% H2O, CO2

60% O2 , 16% H2O, CO2

24% O2 , 14% H2O, N2

36% O2 , 10% H2O, N2

36% O2 , 14% H2O, N2

60% O2 , 10% H2O, N2

60% O2 , 14% H2O, N2

60% O2 , 16% H2O, N2

Figure 1: Gas temperature measurements in the different environments.

around 400 K over the length of the furnace. Environments that showed the largest deviation fromthe mean temperature profile had 60 vol-% O2 in the bulk gas. Some of the hypodermic burnerfuel tubes would clog with oxidized metal under this condition, and maintaining a flat flame profilewas difficult. This likely led to deviations from the ideal temperature profile (which was the same,based on the adiabatic flame temperatures, for all of the conditions).

A simulation was run using COMSOL Multiphysics R© to determine the developing velocity profilein the Sandia reactor at 300 K. The calculated velocity profile at the reactor centerline was correctedby the mean gas temperature at the gas centerline, and the resulting velocity profle was comparedto the mean optically measured particle velocities. Velocities predicted by the corrected developingflow profile, when scaled up by 7% agreed well with the optically measured particle velocities. Thescaled, temperature corrected, developing flow profile was taken to be the particle velocities (it isthe same for all particles), and the residence time in the reactor was calculated by integrating theheight/velocity ratio. This approach neglects particle slip, which is expected to be small for theparticle sizes considered in these experiments. The overall residence time in the 18 inch reactorwas calculated to be 171 ms.

As particles were reacting, temperature measurements were made at 3-7 heights of roughly 100char particles. As shown in Fig. 2, there was significant spread in the particle temperatures, causedpartially by measurement error (estimated at ±20 K), but mostly by different reactivities, andvariations in size of individual char particles. Because of this spread, the mean and median val-ues measured for temperature were not always equivalent, with the median value generally beinghigher than the mean temperature. As the residence time in the reactor increased, there was a heatup of the particles, and then they tended to react at a fairly steady temperature, if not decreasingslightly at longer residence times as the gas temperature decreased. To compare all of the charsreacting in all of the environments, the maximum median temperature (regardless of the residencetime) are plotted in Fig. 3. These temperatures are not a direct measurement of reactivity, but ingeneral, higher temperatures for a given particle size correlate with higher combustion reactivity.

The more reactive sub-bituminous PRB char burned at a higher temperature than either of the twobituminous chars, which burned at about the same temperatures in all of the different environments.There was some dependence of temperature on particle size. For the Black Thunder char, the tem-perature peaked around 80 µm, while the Utah Skyline and Illinois #6 char temperatures increased(although they appear to level off) up to the largest particle size. A trade off exists between heat

7


20 30 40 50 60 70 80

time (ms)

1600

1700

1800

1900

2000

2100

2200

tem

pera

ture

(K)

90-106µm, CO2

14%H2O, 36%O2

106-125µm, N2

14%H2O, 24%O2

Figure 2: Utah Skyline char temperatures. Legend describes fed particle size, diluent gas (balanceother than O2 and H2O), and O2 and H2O vol-%. Circles are mean temperatures, and shaded regionin the background are the normalized number density of particles within 20 K bins, showing thespread of the data.

losses and available reactive surface area, leading to a dependence of temperature on particle size.

At both 24 and 36% O2, there is an obvious influence of diluent on particle temperature. Particlescombusted about 100 K hotter with the N2 diluent than with the CO2 diluent. This is consistentwith other experimental observations [5, 6]. However, at 60% O2, all of the curves seem to col-lapse, and there are no obvious trends in temperature either due to diluent or steam concentration.In addition to larger scatter in the data due to the poor Henken burner flames in the 60% O2 envi-ronments, there was also significantly lower diluent concentrations. Only 24-30% of the gas wasdiluent in these flames, as compared to 58-64% diluent in the 36% O2 environment. Since theparticle temperature difference between N2 and CO2 atmospheres is attributed to diffusion differ-ences through the diluent and the endothermic CO2 gasification reaction, it is not too surprisingthat the difference decreases when there is very little diluent and a greater oxygen concentration.Independence of temperature from diluent may also be evidence of improved kinetic control inhighly oxygen-enriched environments, as Murphy and Shaddix observed [14].

In the environments where the steam concentration was varied, there is no obvious trend of com-bustion temperature with steam concentration. An increase in the diffusivity of oxygen throughsteam would cause the exothermic oxidation reaction to proceed at a higher rate, and an increasedconcentration of H2O would cause the endothermic steam gasification reaction to proceed at ahigher rate. These two effects are in competition to limit the temperature change in these environ-ments, but lead to a greater consumption rate of carbon because of the higher concentrations ofreactants at the particle surface. As described in Hecht et al. [11, 12], little temperature change,but an increase in carbon consumption as the steam concentration increases is predicted when thegasification reactions are considered.

3.2 Burnout measurements

The measured burnout characteristics are shown in Fig. 4. There is a lot of scatter in the burnoutdata, as confirmed by the large error bars on the data when experiments were repeated. Nonethe-less, there are several noteworthy points shown in Fig. 4. First, the particles do not seem to havelost mass until at least 20 ms have elapsed. Some initial time was required to heat the gas and char

8


60 80 100 120 1401850

1900

1950

2000

2050

2100

2150

Black Thunder, 24%O2

2000

2050

2100

2150

2200

2250

2300

2350

2400

tem

pera

ture

(K)


2200

2300

2400

2500

2600

2700

2800


60 80 100 120 140

mean fed size (µm)

Utah Skyline, 24%O2

Utah Skyline, 36%O2

Utah Skyline, 60%O2

60 80 100 120 140

Illinois #6, 24%O2

Illinois #6, 36%O2

Illinois #6, 60%O2

10% H2O, CO2

14% H2O, CO2

16% H2O, CO2

10% H2O, N2

14% H2O, N2

16% H2O, N2

Figure 3: Char temperatures as a function of fed char diameter. Closed circles with long dashesdenote CO2 as a diluent while open circles and short dashes denote N2 as a diluent. Colors indicatedifferent steam concentrations, and the oxygen concentration varies with the vertical frames, asshown on the graph.

particles to a high enough temperature that the char oxidation reactions proceeded at a relevantrate. The particle temperatures lagged the gas temperatures, due to thermal inertia, but eventually,the exothermic oxidation reaction added enough energy to the thermal energy balance that the charburning was sustained. Second, the char burning rate was greatest in 60% O2, followed by 36%O2 and then 24% O2. Finally, we again note that minor variations in steam concentration had verylittle effect on the burning rate of the chars, as evidenced by the slopes of the fixed carbon fractionsbeing roughly equal, regardless of the steam concentration.

9


36% O2

106-125µm, 14% H2O, N2

90-106µm, 10% H2O, N2

90-106µm, 14% H2O, CO2

90-106µm, 14% H2O, N2

90-106µm, 10% H2O, CO2

36% O2

106-125µm, 14% H2O, N2

90-106µm, 10% H2O, N2

90-106µm, 14% H2O, CO2

90-106µm, 14% H2O, N2

90-106µm, 10% H2O, CO2

0 10 20 30 40 50 60 70 80

36% O2

90-106µm, 14% H2O, CO2

90-106µm, 10% H2O, CO2 1200

1300

1400

1500

1600

1700

1800 Gas

Temperature

(K)

36% O2

90-106µm, 14% H2O, CO2

90-106µm, 10% H2O, CO2

Illinois #6

60% O2

125-150µm, 10% H2O, N2

106-125µm, 16% H2O, CO2

125-150µm, 14% H2O, N2

125-150µm, 16% H2O, N2

106-125µm, 10% H2O, CO2 1200

1300

1400

1500

1600

1700

1800

60% O2

125-150µm, 10% H2O, N2

106-125µm, 16% H2O, CO2

125-150µm, 14% H2O, N2

125-150µm, 16% H2O, N2

106-125µm, 10% H2O, CO2

0

20

40

60

80

100Black Thunder

60% O2

90-106µm, 14% H2O, CO2

60% O2

90-106µm, 14% H2O, CO2

0

20

40

60

80

100

fixed

carb

on(%

)

36% O2

90-106µm, 14% H2O, CO2

75-90µm, 14% H2O, CO2

75-90µm, 14% H2O, N2

36% O2

90-106µm, 14% H2O, CO2

75-90µm, 14% H2O, CO2

75-90µm, 14% H2O, N2

0 10 20 30 40 50 60 70 80

residence time (ms)

24% O2

106-125µm, 14% H2O, N2 1200

1300

1400

1500

1600

1700

1800

24% O2

106-125µm, 14% H2O, N2

0 10 20 30 40 50 60 70 800

20

40

60

80

100

24% O2

90-106µm, 14% H2O, CO2

24% O2

90-106µm, 14% H2O, CO2

Utah Skyline

60% O2

106-125µm, 14% H2O, N2

60% O2

106-125µm, 14% H2O, N2

Figure 4: Complete set of burnout data as a function of residence time in the reactor. Faint dashedlines in the background are the gas temperatures for the different environments. Symbol types andsizes denote differences in fed char size, and colors denote differences in steam concentration.Open symbols are used for N2 diluent, whereas closed symbols are CO2 diluent results. Error barsindicate the standard deviation, for those cases where experiments were repeated.

10


Table 2: Best fit kinetic parameters for the project chars. Units of A are m/s, and units of E arekJ/mol.

Black Thundersub-bituminous

bituminous (Utah Sky-line and Illinois #6)

Reaction A E A E

2C + O2 → 2CO 2.22× 104 169 5.00× 103 152C + O2 → CO2 4.44× 102 144 1.00× 102 126C + CO2 → 2CO 2.58× 104 251 1.00× 103 255C + H2O→ CO + H2 2.39× 104 217 5.00× 103 192

3.3 Kinetic fits

Optically measured temperatures were used to find kinetic fits for each of the chars, as describedin Section 2. The slope of the error field was not very steep, and the solutions found were sensitiveto the initial guess. Initially, solutions were found starting from the ‘best-guess’ parameters de-scribed by Hecht et al. [12, 13]. After convergence, the optimization routine was restarted from thelast solution, until a stable set of kinetic parameters was found. While this optimization procedureproduced kinetic parameters that were able to match the experimental data in Fig. 3 reasonablywell for the Black Thunder char, the temperature predictions for the two bituminous chars weresignificantly lower than the experimentally measured values. This solution was likely a local ratherthan a global minimum. Since the two bituminous chars reacted at similar temperatures, the best-fit activation energies found in the original fit for the Utah Skyline char were set, since they bothtrended to similar values, and the pre-exponential factors were manually adjusted to achieve bettertemperature predictions. Optimized parameters are shown in Table 2. For the oxidation reactions,the activation energy found for the two bituminous chars was near the lower bound of 150 kJ/molfor oxidation to CO and significantly lower than the activation energy found for the sub-bituminouschar. The lower activation energy is a lower barrier for a reaction to occur, ‘turning on’ the reac-tion at a lower temperature. However, pre-exponential factors for the oxidation reaction of thebituminous chars were also nearly an order of magnitude lower than the pre-exponential factorsfor the sub-bituminous char, meaning the reaction proceeds at a lower rate at higher temperatures.Reaction rate coefficients for the Black Thunder and bituminous chars are plotted as a functionof temperature in Fig. 5. The oxidation rate coefficients are the same value around 1200 K, andfaster for the Black Thunder char at higher temperatures. The bituminous chars had an order ofmagnitude lower CO2 gasification rate coefficient than the sub-bituminous char, and very similarH2O gasification rate coefficients. Although the rates appear similar in Fig. 5 (other than the CO2

gasification rate), the temperature predictions were different.

Maximum combustion temperatures are compared to the experimental data in Fig. 6. Many of theexperimentally observed trends in temperature are also apparent in the simulations, although themagnitudes of the temperatures show some disagreement. All of the char particles were predictedto react at higher temperatures when N2 was the diluent as compared to CO2, and the differencesin temperatures between the two diluents were similar for both the experimental results and thesimulations, perhaps with the exception of the reactions in 60% O2. The maximum measured

11


10−510−410−310−210−1

100101102

krx

n1

(m/s

)

Black ThunderBituminous

10−510−410−310−210−1

100101102

krx

n2

(m/s

)

10−510−410−310−210−1

100101102

krx

n3

(m/s

)

1000 1500 2000 2500 3000

T (K)

10−510−410−310−210−1

100101102

krx

n4

(m/s

)

10−5

10−4

10−3

10−2

10−1

100

101

102

k(m

/s)

Black Thunder

rxn 1rxn 2rxn 3rxn 4

1000 1500 2000 2500 3000

T (K)

10−5

10−4

10−3

10−2

10−1

100

101

102

k(m

/s)

Bituminous

Figure 5: Reaction rate coefficients using the kinetic parameters shown in Table 2.

temperatures peaked at a similar size in the simulations and the measured data. Finally, the pre-dicted temperatures for the sub-bituminous char were higher than the predicted temperatures of thebituminous chars, in agreement with the measurements.

Predictions for temperatures in 60% O2 were high, and for 24% O2 were low, but overall, thesekinetics captured the experimentally observed trends. Further manual adjustment of the kinetic fitscould be made, but manual trial and error is a tedious process. The least-squares optimization pro-cedure compared steady-state temperatures to the experimentally measured median temperaturesas a function of time. Rather than using the transient data (i.e. particle temperatures as a func-tion of height), a more appropriate data set for optimization may have been the maximum mediantemperatures as a function of fed particle size (as is compared in Fig. 6). Alternatively, rather thanthe comparing temperatures predicted by the steady-state model, the transient model may havebeen more appropriate for making data comparisons. However, running on a PC, the iteration timeof the transient model for each particle was around 3-5 minutes, and an optimization procedurerequiring transient model results as a function of kinetic parameters would take an unreasonableamount of time. Future optimization routines may also focus on a subset of the data, as the fits arenaturally skewed due to the abundance of data at 60% O2 (where there are 3 steam concentrations),compared to the lower oxygen concentration data.

12


60 80 100 120 1401800

1850

1900

1950

2000

2050

2100

2150


2000

2050

2100

2150

2200

2250

2300

2350

2400

2450

tem

pera

ture

(K)


2200

2400

2600

2800

3000

3200


60 80 100 120 140

mean fed size (µm)

Utah Skyline, 24%O2

Utah Skyline, 36%O2

Utah Skyline, 60%O2

60 80 100 120 140

Illinois #6, 24%O2

Illinois #6, 36%O2

Illinois #6, 60%O2

10% H2O, CO2

14% H2O, CO2

16% H2O, CO2

10% H2O, N2

14% H2O, N2

16% H2O, N2

Figure 6: Simulated and experimentally measured char temperatures as a function of fed char di-ameter. Simulations are shown by the lines, while the data of Fig. 3 are plotted as the symbols forcomparison. Symbol styles and colors are the same as in Fig. 3.

3.4 Parametric Analyses

With the kinetics in hand, comparisons can be made between the temperatures and burnout char-acteristics across the parameters varied in these studies. First, experimental and simulated burnouttrends and temperatures for the three chars are shown in Fig. 7. As expected due to their ranks,the sub-bituminous Black Thunder char reacted at the highest temperatures and burned out in theshortest residence time. With the same kinetic parameters, the simulated temperatures and burnoutcharacteristics of the Utah Skyline and Illinois #6 chars were very similar. The chars all startedwith a different percentage of fixed carbon, due to different ash contents of the chars, as listed inTable 1. Black Thunder had the lowest (moisture free) ash content, followed by the Illinois #6, andthe Utah Skyline had the highest ash content. Fairly good agreement between the experiments andsimulations are shown. The particles took more time to heat up in the simulations than experimen-

13


0 20 40 60 80

time (ms)

0

20

40

60

80

100

Fix

edC

(%)

1600

1800

2000

2200

2400

T(K

)

90-106µm, 36% O2,14% H2O, CO2

BTUSI6

Figure 7: Temperature and burnout characteristics of the three chars used in this study. Sym-bols/density plots are experimental data, and lines are simulations. The fed size and environmentwere held constant at the conditions shown above the plot. BT is Black Thunder, US is Utah Skyline,and I6 is Illinois #6.

0 20 40 60 80

time (ms)

0

20

40

60

80

100

Fix

edC

(%)

16001800200022002400260028003000

T(K

)

Black Thunder,90-106µm, 14% H2O, CO2

0 20 40 60 80

time (ms)

Utah Skyline,106-125µm, 14% H2O, N2

24%O2

36%O2

60%O2

Figure 8: Temperature and burnout characteristics of two chars used while holding all but the oxy-gen concentration constant. Symbols/density plots are experimental data, and lines are simula-tions.

tally, this is likely due to the large uncertainty in the heat capacity of the chars, and the fixed valueof this parameter. There is evidence that the heat capacity is a function of char type, temperatureand ash content [24, 25], which was not taken into consideration in this analysis. Also, as will bediscussed later, the size of the particle has an impact on both the time required to heat up, andthe length of time the particles react. These simulations were only for the mean fed particle size,but by modeling both the smallest sized particles in the bin and the largest, the times the particlesare at the high temperatures would shift to both longer and shorter residence times. As discussedearlier, there is large uncertainty in the burnout measurements, making comparisons to the modelchallenging.

Figure 8 shows the temperature and burnout results for two chars where the only parameter variedwas the oxygen concentration in the bulk gas. Both of the chars reacted at the highest temperatureand burned out in the shortest time with the highest oxygen concentration in the bulk gas. Anincreased concentration of oxygen in the bulk gas caused a larger flux of oxygen to the particle

14


0 20 40 60 80

time (ms)

0

20

40

60

80

100

Fix

edC

(%)

1600

1800

2000

2200

2400

T(K

)

Black Thunder,36% O2, 14% H2O, CO2

0 20 40 60 80

time (ms)

Utah Skyline,36% O2, 14% H2O, N2

75-90µm90-106µm106-125µm

Figure 9: Temperature and burnout characteristics of two chars used while holding all but the fedparticle size constant. Symbols/density plots are experimental data, and lines are simulations.

surface, and higher oxidation rates. The oxidation reactions release heat, causing high particletemperatures. High temperatures caused the gasification reactions to proceed at faster rates (inaddition to the oxidation rate), consuming carbon at a higher rate and reducing the time requiredfor burnout. Longer burnout times are shown both experimentally and in the simulations, as theoxygen concentration in the bulk gas decreased.

For two chars, simulated and experimental temperature and burnout measurements are shown fortwo size bins while all other parameters are held constant in Fig. 9. Both the experimental mea-surements and the simulations show only a small effect of size on the peak temperatures reached.There was some dependence of temperature on size, as shown by the slopes in Fig. 6, but withina size bin, with a maximum variation in size of 25 µm, the effect on temperature was fairly mini-mal. However, even over the variation in initial particle sizes simulated here (around 16 µm), thesmaller particles heated up significantly faster, and the larger particles sustained the high tempera-tures for longer, which would account for the larger spread in time of the experimentally measuredhigh temperatures than was predicted by the simulations. The experimental range in fed size alsoaccounts for the greater spread in the measured particle temperatures towards the edges of the hotzone. Towards the beginning of the high temperature reaction zone, the larger particles were stillheating up, and at the longer residence times, the smaller particles were already cooling. Detectionof the cooler particles on the tail edge was somewhat suppressed due to a bias in the experimentallymeasured temperature to trigger off of larger particles. As expected, larger particles had a longerburnout time, as there was more mass to consume. This trend was observed experimentally andpredicted by the simulations.

The effect of the diluent on the combustion characteristics is shown in Fig. 10. The substitutionof CO2 for nitrogen suppressed the heat-up of the particle, reaction temperature and burnout. Thiseffect is partially attributed to the lower rate of diffusion of oxygen through CO2 as compared toN2, and partially attributed to the endothermic CO2 gasification reaction [12]. Both effects must beconsidered to properly model char combustion in both air-fired and oxy-fired conditions. The levelof agreement in both predicted char temperatures and burnout characteristics between the modeland simulations is quite good for the six conditions shown in this figure. At both higher and loweroxygen concentrations, there was more disagreement between the model and experiments with the

15


0 20 40 60 80

time (ms)

0

20

40

60

80

100

Fix

edC

(%)

1600

1800

2000

2200

2400

T(K

)

Black Thunder,75-90µm, 36% O2, 14% H2O

0 20 40 60 80

time (ms)

Utah Skyline,90-106µm, 36% O2, 14% H2O

0 20 40 60 80

time (ms)

Utah Skyline,90-106µm, 36% O2, 10% H2O

CO2

N2

Figure 10: Temperature and burnout characteristics while holding all but the diluent constant. Sym-bols/density plots are experimental data, and lines are simulations.

0 20 40 60 80

time (ms)

0

20

40

60

80

100

Fix

edC

(%)

16001800200022002400260028003000

T(K

)

Illinois #6,90-106µm, 36% O2, CO2

0 20 40 60 80

time (ms)

Illinois #6,106-125µm, 60% O2, CO2

0 20 40 60 80

time (ms)

Illinois #6,125-150µm, 60% O2, N2

10% H2O14% H2O16% H2O

Figure 11: Temperature and burnout characteristics while holding all but the bulk steam concentra-tion constant. Symbols/density plots are experimental data, and lines are simulations.

kinetic parameters used in this analysis, as can be seen in Figs. 6 and 8.

Finally, Fig. 11 shows the effect of the steam concentration in the bulk gas on the temperatureand burnout characteristics. While there was too much scatter experimentally to reveal any trends,the simulations can provide insight into the effects of H2O concentration. The predicted particletemperature and combustion rates are expected to increase slightly when the steam concentrationis increased [12]. In these simulations, however, the opposite trend was observed. A higher con-centration of steam in the bulk gas appears to suppresses the particle temperatures and lengthen theburnout time. The steam gasification reaction rate coefficients were significantly higher than theCO2 gasification reaction rate coefficients, as shown in Fig. 5. The endothermic steam gasificationsuppressed the particle temperature under higher steam concentrations, leading to this contradic-tory observation. However, the fits for the steam gasification reaction were likely too high, andinconsistent with literature data [12]. Nonetheless, these comparisons are valuable for highlightingthe complexities of this reaction system. In the CO2 diluent, increasing the steam concentrationdecreases the CO2 concentration, so the effect is to substitute one gasification reactant for the other.The diffusion rate of O2 through H2O is significantly higher than the diffusion of O2 through CO2,which could increase the oxidation reaction under wet-recycle oxy-combustion conditions. With

16


N2 as a diluent, there are also competing effects of the improved diffusion of oxygen throughsteam, and the increased steam gasification rate. These competing effects require an accurate de-termination of the reaction kinetics in this system.

4 Conclusions

Coals from the Illinois #6 seam, Utah Skyline, and Black Thunder mines were devolatilized toform chars. These chars were sieved into six narrow size bins, and fed into Sanda’s optical en-trained flow reactor. Twelve gaseous environments were chosen for study, spanning a wide rangeof conditions, from 24-60 vol-% O2, 10-16 vol-% H2O, with CO2 and N2 diluents. In all envi-ronments, the sub-bituminous PRB char burned at the highest temperature, with 80 µm particlesburning at the highest temperature. Both bituminous chars increased in combustion temperatureup to the largest mean fed char size of 140 µm. The substitution of CO2 for N2 lowered thecombustion temperature in the 24 and 36%O2 environments, due to the lower diffusivity of oxy-gen through CO2 as compared to N2, and the effect of the CO2 gasification reaction. At 60%O2,there was no discernible influence of combustion environment on combustion temperature, whichmay by a sign of improved kinetic control under highly oxygen-enriched combustion conditions.In all cases, there was no meaningful influence of steam concentration on temperature. Burnoutmeasurements on the chars had a large degree of uncertainty, due to experimental errors. Nonethe-less, faster burnout occurred as the oxygen concentration increased. The steam concentration alsoshowed little effect on the experimentally measured burnout characteristics. An initial heat-up timeof around 25-30 ms for the chars was apparent in the burnout measurements (as well as the opticalmeasurements).

A transient burnout model was used to simulate the chars reacting in the EFR. Kinetic fits to thetemperature data were determined. Both the activation energies and pre-exponentials determinedfor the oxidation and steam gasification reactions were much lower for the bituminous char thanthose parameters for the sub-bituminous char. The activation energy for the CO2 gasification re-action was similar for both the bituminous and sub-bituminous chars, but the pre-exponential tothe CO2 gasification reaction was much lower for the bituminous char than the pre-exponential forthe sub-bituminous char. Simulated temperatures showed similar characteristics as the measuredtemperatures. Further, the simulations showed a temperature dependence on steam concentrationthat was not discernible experimentally. The temperature was predicted to decrease as the steamconcentration increased. In addition, simulations were conducted holding all but one parameterconstant. As the char was varied, the Black Thunder reacted at higher temperature and burned outin less time than the two bituminous chars, which had similar characteristics. The reaction temper-atures increased and burnout times decreased as the oxygen concentration increased. Heat-up andburnout time were shorter for smaller chars, while peak temperature was not affected greatly byfed particle size. Char particles reacted at higher temperatures and had shorter burnout times whenN2 was the diluent than when CO2 was the diluent. The reaction temperature and burnout timesdecreased as the steam concentration increased, but the kinetics rates of the steam gasificationreaction were likely overestimated in the current analysis.

17


Acknowledgments

This material is based upon work supported by the Department of Energy under Award NumberDE-NT0005015, managed by David Lang and part of the University of Utah, Institute for Cleanand Secure Energy Clean Coal program. The authors are grateful for additional funding providedby the National Energy Technology Laboratory’s Power Systems Advanced Research Program,managed by Dr. Robert Romanosky, through Sandia. Support from Sandia’s Doctoral Study Pro-gram is also appreciated. Sandia National Laboratories is a multi-program laboratory managedand operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation,for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000. The views and opinions of authors expressed herein do not necessarily state orreflect those of the United States Government or any agency thereof.

References

[1] Y. Tan. Oxy-fuel combustion for power generation and carbon dioxide (CO2) capture, chapter Oxy-fuel powerplant operation, pages 35–53. Woodhead Publishing, 2011.

[2] Adel F. Sarofim, JoAnn S. Lighty, Philip J. Smith, Kevin J. Whitty, Edward Eyring, Asad Sahir, Milo Alvarez,Michael Hradisky, Chris Clayton, Gabor Konya, Richard Baracki, and Kerry Kelly. Chemical looping com-bustion reactions and systems task 5 topical report, utah clean coal program. Technical Report Task 5 TopicalReport, Utah Clean Coal Program, University of Utah, Salt Lake City, UT, 2011.

[3] Alejandro Molina and Christopher R. Shaddix. Proc. Combust. Inst., 31 (2007) 1905–1912.

[4] C. R. Shaddix and A. Molina. Proc. Combust. Inst., 32 (2009) 2091–2098.

[5] C. R. Shaddix and A. Molina. Oxy-fuel combustion for power generation and carbon dioxide (CO2) capture,chapter Ignition, flame stability, and char combustion in oxy-fuel combustion, pages 101–124. Woodhead Pub-lishing, 2011.

[6] C. R. Shaddix and A. Molina. Effect of O2 and high CO2 concentrations on PC char burning rates during oxy-fuelcombustion. Proceedings of the 34th International Technical Conference on Coal Utilization and Fuel Systems,June 1–5 2008.

[7] M. B. Toftegaard, J. Brix, P. A. Jensen, P. Glarborg, and A. D. Jensen. Prog. Energy Combust. Sci., 36 (2010)581–625.

[8] P. Heil, D. Toporov, H. Stadler, S. Tschunko, M. Forster, and R. Kneer. Fuel, 88 (2009) 1269–1274.

[9] R. K. Rathnam, L. K. Elliott, T. F. Wall, Y. Liu, and B. Moghtaderi. Fuel Process. Technol., 90 (2009) 797–802.

[10] D. Toporov, P. Bocian, P. Heil, A. Kellermann, H. Stadler, S. Tschunko, M. Forster, and R. Kneer. Combust.Flame, 155 (2008) 605–618.

[11] Ethan S. Hecht, Christopher R. Shaddix, Alejandro Molina, and Brian S. Haynes. Proc. Combust. Inst., 33(2011) 1699–1706.

[12] Ethan S. Hecht, Christopher R. Shaddix, Manfred Geier, Alejandro Molina, and Brian S. Haynes. Combust.Flame, 159 (2012) 3437 – 3447.

[13] Ethan S. Hecht, Christopher R. Shaddix, and JoAnn S. Lighty. Combust. Flame, in Press (2013) http://dx.doi.org/10.1016/j.combustflame.2013.02.015.

[14] Jeffrey J. Murphy and Christopher R. Shaddix. Combust. Flame, 144 (2006) 710–729.

[15] ASTM. Standard Test Methods for Proximate Analysis of Coal and Coke by Macro Thermogravimetric Analysis.Technical Report D7582-10, 2010.

18


[16] HF Stoeckli and J.P. Houriet. Carbon, 14 (1976) 253–256.

[17] H Marsh and T Siemieniewska. Fuel, Lond., 44 (1965) 355–367.

[18] L. Tognotti, J. P. Longwell, and A. F. Sarofim. Proc. Combust. Inst., 23 (1991) 1207–1213.

[19] Eric Jones, Travis Oliphant, Pearu Peterson, et al. SciPy: Open source scientific tools for Python, 2001–.http://www.scipy.org/.

[20] Robert Hurt, Jian-Kuan Sun, and Melissa Lunden. Combustion and Flame, 113 (1998) 181–197.

[21] Reginald E. Mitchell, RH Hurt, LL Baxter, and DR Hardesty. Compilation of Sandia Coal Char CombustionData and Kinetic Analyses. Technical report, 1992.

[22] R Hilpert. Forschung auf dem Gebiete des Ingenieurwesens, 4 (1933) 215–224.

[23] DB Anthony and JB Howard. AIChE Journal, 22 (2004) 625–656.

[24] F. Hanrot, D. Ablitzer, J.L. Houzelot, and M. Dirand. Fuel, 73 (1994) 305 – 309.

[25] Jerzy Tomeczek and Henryk Palugniok. Fuel, 75 (1996) 1089 – 1093.

19

Documents

Kinetic Rates of Oxidation and Gasiﬁcation …...In this work, the kinetic rates of oxidation, and the oft-neglected CO2 and steam gasiﬁcation reactions for three pulverized coal