CFD simulation of entrained-flow coal gasification: Coal particle density/sizefraction effects

Powder Technology 203 (2010) 98–108

Contents lists available at ScienceDirect

Powder Technology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p ow t e c

CFD simulation of entrained-flow coal gasification: Coal particle density/size fraction effects Andrew Slezak a,b, John M. Kuhlman a,b,⁎, Lawrence J. Shadle a, James Spenik a,c, Shaoping Shi a,d a

National Energy Technology Laboratory, Morgantown, WV 26507-0880, United States West Virginia University, Mechanical and Aerospace Eng. Dept. Morgantown, WV 26506-6106, United States REM Engineering Services, Morgantown, WV 26505, United States d Ansys, Inc., Morgantown, WV 26505, United States b c

a r t i c l e

i n f o

Article history: Received 21 August 2009 Accepted 18 February 2010 Available online 8 April 2010 Keywords: Coal gasification Entrained-flow reactor CFD Discrete Phase Method DPM Particle–wall interactions Coal size/density fractions

a b s t r a c t Computational Fluid Dynamics (CFD) simulation of commercial-scale two-stage upflow and single-stage downflow entrained-flow gasifiers was conducted to study effects of simulating both the coal particle density and size variations. A previously-developed gasification CFD model was modified to account for coal particle density and size distributions as produced from a typical rod mill. Postprocessing tools were developed for analysis of particle–wall impact properties. For the two-stage upflow gasifier, three different simulations are presented: two (Case 1 and Case 2) used the same devolatilization and char conversion models from the literature, while Case 3 used a different devolatilization model. The Case 1 and Case 3 solutions used average properties of a Pittsburgh #8 seam coal (d = 108 μm, SG = 1.373), while Case 2 was obtained by injecting and tracking all of the series of 28 different coal particle density and size mass fractions obtained by colleagues at PSU as a part of the current work, for this same coal. Simulations using the two devolatilization models (Case 1 and Case 3) were generally in reasonable agreement. Differences were observed between the single-density solution and the density/size partitioned solution (Case 1 and Case 2). The density/size partitioned solution predicted nominally 10% less CO and over 5% more H2 by volume in the product gas stream. Particle residence times and trajectories differed between these two solutions for the larger density/size fractions. Fixed carbon conversion was 4.3% higher for the partitioned solution. Particle–wall impact velocities did not vary greatly. Grid independence studies for the two-stage upflow gasifier geometry showed that the grid used in the comparison studies was adequate for predicting exit gas composition and wall impact velocities. Validation studies using experimental data for the Pittsburgh #8 coal from the SRI International pressurized coal flow reactor (PCFR) at 30 atmospheres indicated adequate agreement for gasification and combustion cases, but poor agreement for a pyrolysis case. Simulation of a single-stage downflow gasifier yielded an exit gas composition that was in reasonable agreement with published data. © 2010 Elsevier B.V. All rights reserved.

1. Introduction The present study has been undertaken to demonstrate capability to predict the performance of entrained-flow gasifiers using current state-of-the-art Computational Fluid Dynamics (CFD) software, and more specifically to predict differences within the gasifier in behaviors (e.g., particle residence times, char conversion, and trajectories) of various size and density fractions of the pulverized coal fuel produced from a typical rod mill. This effort is a part of a larger study, the Coal Partitioning Project, or CPP, which is being supported by the DOE National Energy Technology Laboratory (NETL). In the CPP project, a Pittsburgh #8 seam coal from the Bailey mine in southwestern ⁎ Corresponding author. National Energy Technology Laboratory, Morgantown, WV 26507-0880, United States. Tel.: +1 304 293 3180; fax: +1 304 293 6689. E-mail address: [email protected] (J.M. Kuhlman). 0032-5910/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.powtec.2010.03.029

Pennsylvania has been ground using a METSO rod mill to specifications typical of slurry-fed entrained-flow gasifiers and separated into four density fractions, and then each of these density fractions has been further classified into seven diameter ranges [1,2]. The parent coal and each of the 28 resulting density/size cuts have been characterized by an ultimate and proximate analysis, a maceral analysis, and a mineral analysis of the ash from each cut [1,2]. While a density separation is known to separate hydrogen-rich, liptinite macerals in the lighter fractions from denser wood-derived vitrinite, and denser still mineral impregnated inertinite macerals, there was little evidence that this was the case in this vitrinite-rich coal. The density separation however did produce a separation in the mineral matter such that the lighter fractions consisted of fine clays imbedded within the coal particles, the denser fractions had subsequently higher concentrations of clays and heavier pyrite minerals, while the densest fraction was essentially individual minerals with some organic

A. Slezak et al. / Powder Technology 203 (2010) 98–108

99

Table 1 Chronological summary of previous coal gasification CFD studies. Ref # Authors

Year

[12]

1999 CFX 4.2

[13] [14] [15] [16] [7] [17] [8] [9] [18] [19] [20] [21] [22] [23] [3] [24] [25] [26] [27]

Bouma et al. Chen et al. Shanam et al. Fletcher et al. Chen et al. Chen et al. Tominaga et al. Chen et al. Liu et al. Bockelie et al. Liu and Kojima Liu and Kojima Roucis et al. Watanabe et al. Guenther and Breault Shi et al. Tominaga et al. Lu et al. Forstner et al. Syred et al.

Code used

1999 – 2000 Fluent 5.4 2000 CFX 4 2000 – 2000 – 2000 Fluent 2001 2001 Modified Ψ–ω 2002 GLACIER 2004 – 2004 – 2004 CDadapco 2004 CFX 4

Method used

Configurations(s) simulated

Eulerian– Axisymmetric entrained-flow Eulerian gasification simulator DPM 200 T/day 2-stage HYCOL gasifier DPM N. Dakota EERC transport gasifier DPM 1 MW entrained-flow biomass gasifier DPM 200 T/day 2-stage HYCOL gasifier DPM 200 T/day 2-stage HYCOL gasifier DPM 2 T/day and 50 T/day 2-stage HYCOL gasifiers DPM 200 T/day 2-stage HYCOL gasifier DPM Bench-scale KIER entrained-flow gasifier DPM 2-Stage upflow and 1-stage downflow gasifiers DPM Pilot-scale 2-stage HYCOL gasifier DPM pIlot-scale 2-stage HYCOL gasifier – Pilot-scale gasifier DPM

2 T/day entrainedflow gasifier 2005 MFIX Eulerian– Transport gasifier Eulerian 2006 Fluent 6.1 DPM Commercial 2-stage, upflow entrained-flow gasifier 2006 Fluent and DPM 2 T/day and 50 T/day RESORT 2-stage HYCOL gasifiers 2008 – DPM Lime-injected fluidized-bed coal gasifier 2006 Fluent 6.1 DPM Biomass-fueled, fixed-bed furnace 2007 Fluent 6.1 DPM 500 kV coal-fired furnace

Table 2 Proximate and ultimate analyses of Pittsburgh #8 coal from the Bailey mine [1,2]. Component

Pittsburgh #8 coal

Fixed carbon Volatiles Ash Moisture C H O N S Cl

57.69 33.52 7.79 1 76.83 5.49 3.91 1.4 1.46 0

Fig. 1. Commercial-scale, two-stage, upflow, slurry-fed, entrained-flow gasifier geometry and grid.

impurities [2]. The CPP project objectives are to evaluate the influence of these compositional differences in coal particles on the partitioning of inorganics produced during gasification into flyash and slag. As such it is important to understand the differences in the trajectories of these various particles and their relative tendencies to strike the wall. The main focus of the present CFD simulations is a commercialscale, two-stage, upflow, slurry-fed, entrained-flow gasifier geometry previously studied by Shi et al. [3]. The Fluent user-defined function (UDF) developed by Shi to simulate gasification has been extended herein to allow the feeding of (28) multiple coal fractions into the gasifier, each at the appropriate mass flow rate and chemical composition. Results have been obtained at a typical operating condition using the devolatilization chemistry model of Kobayashi et al. [4] and the shrinking core char conversion models of Wen and Chaung [5]. Results are compared for one solution for the parent coal at the measured Sauter-mean diameter of 108 μm and average density of 1373 kg/m3 for the parent coal, and for a second solution for the injection of all 28 density/size cuts, each at the appropriate injection mass flow rate, and each with its own volatile chemical composition and ultimate yield as governed by the measured ultimate and proximate analysis for that size and density fraction, as given by the Kobayashi et al. devolatilization model. Also, a solution using the devolatilization model from PC Coal Lab6 has been compared with the first solution. 2. Background Some previous CFD gasifier simulations have investigated the behavior of a range of coal particle size fractions (e.g., [7–9]), but none to date have included various density fractions. This is at least partly

Table 3 Composite Bailey coal specific gravity (SG) and particle size (PS) [1,2]. Gravity fraction

BSG1PS0

BSG2PS0

BSG3PS0

BSG4PS0

Avg. specific gravity wt.%

1.2a 47.84

1.45 47.57

2.1 3.46

3.3a 1.12

Size class

PS1

PS2

PS3

PS4

PS5

PS6

PS7

Avg. diameter, μm BSG1 wt.% BSG2 wt.% BSG3 wt.% BSG4 wt.%

800a 5.31 3.91 16.15 8.69

512 10.85 9.53 11.5 5.93

318 31.15 21.66 23.38 17.5

181 13.36 9.15 10.71 8.31

128 11.57 7.38 9.12 8.65

90 10.79 6.99 8.79 13.49

50a 16.97 41.39 20.4 37.43

Composite SG = 1.373; composite Sauter-mean diameter = 108 μm. a Assumed value.

100

A. Slezak et al. / Powder Technology 203 (2010) 98–108

Table 4 Assumed model conditions for present solutions.

Devolatilization Char conversion Coal particle(s) Cp equations k equations

Case 1

Case 2

Case 3

Kobayashi Wen/Chaung d = 108 μm Polynomial Power law

Kobayashi Wen/Chaung 28 CPP size/ρ cuts Polynomial Power law

PC Coal Laba Wen/Chaung d = 108 μm Polynomial Power law

a Recommended one point tuning of this model using experimental entrained-flow data has not yet been completed.

Table 5 Comparison of average temperatures, FC conversion, and exit gas compositions (by vol.).

Tavg (K) stage 1 Tavg (K) stage 2 Overall FC conversion Specie CO H2 CH4 CO2 H2O O2 N2 H2S Ar

Case 1

Case 2

Case 3

2522.4 K 1817.7 K 85.5%

2574.2 K 1805.4 K 89.8%

2775 K 1601 K 83.1%

0.6489 0.2764 0.0410 0.0038 0.0094 0.0000 0.0035 0.0063 0.0106

0.5491 0.3290 0.0445 0.0416 0.0106 0.0000 0.0076 0.0072 0.0104

0.6266 0.2360 0.0016 0.0189 0.0000 0.0000 0.0000 0.0045 0.0110

because commercial CFD software such as Fluent typically has the built-in capability to allow use of a user-selected particle size pdf represented by several different diameter classes, and the individual particle trajectories and burn out behaviors of each size class may then be investigated individually. However, this built-in procedure requires that all particles have a single diameter distribution, and also have the same density value and chemistry models. The study by Cloke et al. [10] experimentally investigated the variation in coal properties versus particle diameter for ten different world coals; these coals were selected to give a wide range of properties. The coals used in this study were not density-separated. The experimental work by Wu et al. [11] studied the composition of coarse and fine slags that were sampled for a Chinese commercial single-stage, downflow, slurry-fed entrained-flow gasifier unit. The coarse slag was collected from the outlet of the lock hopper beneath the main gasifier unit,

while the fine slag was collected from the fine slag removal filter for the water streams in the quench chamber and the downstream gas scrubber/convective cooler. Previous CFD simulation studies of coal gasification [3,7–9,12–27] are briefly summarized in Table 1; none of this previous work has simulated variations in particle densities. Our present simulations have focused on the same gasifier geometry, CFD mesh, and chemistry models as in Shi et al. [3], but have utilized the ultimate and proximate analysis data for the Pittsburgh #8 seam coal from the Bailey mine that is under study in the current CPP, again at an operating pressure of 2.84 MPa. The proximate and ultimate analysis data for the parent Bailey coal are presented in Table 2. For the current simulations, the computed Pittsburgh #8 coal Sauter-mean diameter of 108 μm and average specific gravity of 1.373 for the composite coal have been used; see Table 3. For comparison purposes, the simulation by Shi et al. using Illinois #6 coal assumed a specific gravity of 1.4 and an average particle diameter of 30 μm. (This average particle diameter is believed to be too small to represent conditions in a slurry-fed entrained-flow gasifier.) For the present simulations using the Bailey coal, the same total coal mass flow rate as used by Shi et al. [3] has been used in both of the present simulations (28.04 kg/s), with the same split between the first and second stages (78%/22%). However, because of the higher carbon content of this Pittsburgh #8 coal relative to the Illinois #6 (86.08% versus 80.51%), the stage 1 inlet carrier gas (consisting of 94.4% O2) was scaled up from 21.61 kg/s to 23.17 kg/s, and the slurry water flow rates were scaled in the same proportion (from 11.62 kg/s to 12.31 kg/s). This was done to maintain the same (O/C) ratio of 1.44, and (steam/C) ratio of 0.34 as was used by Shi et al. for Illinois #6 coal. Further details of the present case setup conditions and solution process have been given in the thesis by Slezak [28]. 3. Results Details of the present Fluent CFD solutions using the Discrete Phase Method (DPM) for the commercial-scale two-stage upflow entrained-flow gasifier geometry previously studied by Shi et al. [3] at an operating pressure of 2.84 MPa are summarized and compared herein. Additional comparisons have been presented by Slezak [28]. Shi et al. [3] used the devolatilization chemistry model of Kobayashi et al. [4] and the char conversion models of Wen and Chuang [5]. The Case 1 simulation is for the parent coal at a Sauter-mean diameter of 108 μm and the mean density of 1373 kg/m3 (Table 3), while the Case 2 simulation is for the injection of all 28 density/size cuts, each at the

Fig. 2. Computed gas phase temperature field: Fluent simulation of upflow entrained-flow gasifier; p = 2.84 MPa.

A. Slezak et al. / Powder Technology 203 (2010) 98–108

101

Fig. 3. Computed gas velocity magnitude: Fluent simulation of upflow entrained-flow gasifier; p = 2.84 MPa.

appropriate injection mass flow rate (Table 3), and each with its own volatile chemical composition and proximate analysis [1,2]. A third simulation (Case 3) has also been obtained using a devolatilization model developed using secondary pyrolysis results from the PC Coal Lab software developed by Niksa [6], and the char conversion models of Wen and Chaung. The gasifier geometry and the grid used in the simulations are shown in Fig. 1, while the assumed model conditions for the Case 1, Case 2, and Case 3 runs are summarized in Table 4. The computed average gas temperature in the first stage for Case 2 is about 50 K hotter than results for Case 1 (Table 5), while the Case 2 second stage average temperature is about 12 K cooler than for Case 1; these are modest differences. However, the computed exit gas species volume fractions for these two solutions differ (Table 5). In particular, the exit CO volume percentage is 10% lower for the partitioned Case 2 solution than for Case 1, while the exit H2 volume fraction increased by over 5%. Computed exit volume fraction of CH4 increased by about 0.5%, while the exit CO2 volume percent increased by about 4%. The other gas species volume percentages are quite low, and are similar for both solutions. The Case 3 solution (PC Coal Lab devolatilization model) has exit gas volume fractions that compare more closely with

the Case 1 solution (Kobayashi devolatilization model, as used in Shi et al. [3]). However, the average temperatures differ dramatically from those of Case 1, being approximately 250 K hotter in stage 1 and 215 K cooler in stage 2. This is believed to be due to the different assumed volatile composition for the two models, where PC Coal Lab predicts that the predominant secondary pyrolysis product is soot, which was not produced at all in the Kobayashi model. Soot was allowed to react in the gas phase with oxygen and rapidly consumes any available O2 in the first stage leading to higher temperatures, and gasifies with the available steam in the second stage. The overall predicted fixed carbon conversion is 85.5% for Case 1, 90% for Case 2, and 83.1% for Case 3. Average temperature contours for Cases 1 and 3 appear nearly symmetric in the first stage, while for Case 2 there is an asymmetry in the temperature distribution in the first stage of the gasifier (Fig. 2B). This may be because the heaviest particles become trapped in the gasifier first stage. Mean velocity contours (Fig. 3) for all three cases look more nearly the same than do the temperature contours, although asymmetry in the velocity distribution is visible in the first stage for Case 2 in Fig. 3B. The CO mole fraction distributions for Case 1

Fig. 4. Computed CO species mole fraction: Fluent simulation of upflow entrained-flow gasifier; p = 2.84 MPa.

102

A. Slezak et al. / Powder Technology 203 (2010) 98–108

Fig. 5. Computed H2 species mole fraction: Fluent simulation of upflow entrained-flow gasifier; p = 2.84 MPa.

and Case 3 show higher concentrations in stage 2 as well as in the vicinity of the slag tap relative to levels for Case 2; compare Fig. 4A and C to Fig. 4B. This is consistent with the CO exit volume fractions shown in Table 5. Again, asymmetry is observed in the spatial contours in

stage 1 for the Case 2 solution (Fig. 4B). Conversely, the H2 mole fraction is higher in stage 2 for the Case 2 solution; see Fig. 5. Predicted overall fixed carbon conversion is compared for the Case 1 and Case 2 solutions for each particle size, for the SG1, SG2, and SG3

Fig. 6. Fixed carbon conversion for Cases 1 and 2.

A. Slezak et al. / Powder Technology 203 (2010) 98–108

103

Fig. 7. Average particle residence time for Cases 1 and 2.

density cuts in Fig. 6. Trends are quite similar for the two solutions, with nearly 100% conversion predicted for the three smallest particle diameter ranges (PS7, PS6, and PS5, or 50, 90, and 128 μm), and lower carbon conversion predicted for the larger particle sizes. For example, between 60% (for SG3) and 40% (for SG1) conversion is predicted for the largest particles (PS1, or 800 μm). Case 1 conversion levels were always lower than those using the individual density fractions (Case

2) which is consistent with the lower temperatures for the averaged density fractions used for Case 1. The computed average particle residence times are compared for these two solutions in Fig. 7. Residence times for the three smallest particle size classes (up to 128 μm diameter) are approximately 3 s, with residence times increasing for the larger particles, especially for the Case 2 solution. There are significant differences in the predicted

Fig. 8. Impact velocity normal to the wall versus axial location in second stage for density class SG1 for Cases 1 and 2. Particles injected from stage 2 inlet.

104

A. Slezak et al. / Powder Technology 203 (2010) 98–108

Fig. 9. Impact velocity normal to the wall versus axial location in second stage for density class SG4 for Cases 1 and 2. Particles injected from stage 2 inlet.

residence times, with Case 2 consistently predicting larger values. Including accurate density variations, and the individual inclusion and treatment of the larger size fractions (especially PS1, PS2, and PS3) while iterating the DPM solution for Case 2 results in the longer predicted residence times, which are captured in the Case 1 solution only for the highest density and largest particle size fraction (SG4PS1; see Fig. 7D). The computed average wall impact velocity distributions are quite similar for the Case 1 and Case 2 solutions; see Figs. 8 and 9 for examples for particle injection from the stage 2 injection location only. Additional examples have been given by Slezak [28]. (Slurry injection velocities were set to the same value of 50 m/s as used by Shi et al. [3] at all three injection locations.) Also, the wall impact velocity is largest near the stage 2 injection location, and then decreases significantly (to between 1 and 0.5 m/s). There also is an increase in impact velocity near the exit, due to the contraction just prior to the exit. Impact velocities near the second stage injector are largest for the larger particles (PS1; 800 μm). There is a trend for the heavier particles to have higher wall impact velocities (SG4; Fig. 9). Also, impact velocities for second stage injection tend to be higher than for particles that were injected into the gasifier first stage [28]. An example of measured particle–wall impact probability results is shown in Fig. 10 along the length of the second stage. This depicts the probability of particle–wall collision for injection of all 28 density/size fractions into both the first and second stages assuming that all

particles stick to the wall at their first impact with the wall. Impact probability is highest near the second stage injection location. The Case 1 solution predicts 10% of all particles never impact a wall, while Case 2 predicts only 5% do not impact the wall. Again, further details have been given by Slezak [28]. Fig. 11 displays contours of computed impact particle mass flux for particles injected from both stage 1 and stage 2 at the Sauter-mean diameter and composite average density of the Bailey coal. If one assumes that 100% of the particles impact the wall and 80% of the carbon has been converted prior to impact, then the average mass flux impacting the second stage wall from this rough estimate is approximately equal to the area-averaged value obtained from the data shown in Fig. 11. The mass flux is highest almost directly across from the second stage slurry injection port, as one would expect. Significant differences have also been observed for the computed particle trajectories, for different particle initial density and diameter. The dramatic differences in particle trajectories for the extreme ends of the particle size-density spectrum are highlighted in Fig. 12, where the computed particle trajectories until the initial particle impact with the gasifier wall are shown for second stage injections of the lightest, smallest cut (SG1PS7) and the heaviest, largest cut (SG4PS1). Impacts for the SG1PS7 cut appear to be distributed over the entire gasifier second stage wall, while the SG4PS1 particles all impact the wall almost directly opposite the injection location. Also, the particle–wall

Fig. 10. Impact probability versus second stage distance for 28 size/density fractions. Particles injected from both stage 1 and stage 2.

Fig. 11. Contours of mass flux (kg/m2 s) for Sauter-mean diameter and composite average density. Particles impact once and then stick.

A. Slezak et al. / Powder Technology 203 (2010) 98–108

105

Fig. 12. Particle trajectories until first wall impact for second stage injection of: A) cut SG1PS7 (SG = 1.2, d = 50 μm), and B) cut SG4PS1 (SG = 3.3, d = 800 μm).

impact velocities are much larger for the SG4PS1 cut, being as large as 10–14 m/s, as compared to average values of 1–3 m/s for the SG1PS7 cut; see Figs. 8B and 9A. These trajectory and wall impact differences would significantly influence the location and rate of slag formation, which then would feed back into a secondary influence on the local gas phase properties (temperature; gas composition). Grid independence checks for the present geometry have been reported in the thesis by Slezak [28]. Simulation results for the setups for Case 1 and Case 2 were run to 10,000 iterations for increasing grid densities of 12,256 cells (the present mesh), 20,061 cells, 28,264 cells, and 40,112 cells. Centerplane contours of gas species mole fractions, average gas temperatures, and wall temperatures were presented by Slezak [28], along with the second stage particle–wall impact normal velocities, and some localized visible differences were observed in some regions. However, no clear trends could be discerned for these variations versus the grid density. As an overview, the computed exit gas mole fractions are compared in Fig. 13 for Case 1 and in Fig. 14 for Case 2. Similar plots are given by Slezak [28] for the gasifier stage 1 and stage 2 average temperatures. Only modest differences are observed in these major gas species and average temperatures, so the present mesh (12,256 cells) is judged to be adequate to predict these overall gasifier characteristics. Also, particle–wall impact velocities generally did not change significantly as the grid density was increased. However, variations were observed in the predicted spatial

Fig. 13. Case 1 exit gas compositions for 12,256, 20,061, 28,264, and 40,112 grid elements.

contours of gas specie volume fraction, for some species and in some localized regions. So, it is concluded that the finer grid should be used to obtain the most realistic overall simulations. However, this would result in significant run time increases and was not deemed necessary here when looking at the particle–wall collisions. Preliminary validation efforts have been summarized in the thesis by Slezak [28]. Radiant coal flow reactor (RCFR) experiments conducted by SRI International as a part of the CPP for a pyrolysis run, a combustion run, and a gasification run for a Bailey coal sample [29] were simulated in Fluent. Wen and Chaung's shrinking core char conversion model and the devolatilization model as implemented by Shi et al. [3] were used, with the same chemical kinetic rate parameters as have been used in the present commercial-scale gasifier simulations. The computational model of the SRI RCFR is shown in Fig. 15. The SG2PS5 fraction was chosen as the coal density/ size fraction for the validation study, because this fraction was the center point of the SRI test matrix. The computational grid used (Fig. 15) contains 91,584 hexahedral cells; no grid sensitivity studies were conducted for this SRI PCFR geometry The reactor is 1.2 m long, with a 12 mm ID. The coal was injected via the core flow inlet in Ar carrier gas, and any oxygen was introduced via the outer sheath flow. Steam was injected through the conical inlet shown in Fig. 15. Wall temperature boundary conditions were chosen to match the experimental setup at SRI as closely as possible. The inlet velocity of the coal particles was set at 24 cm/s. The

Fig. 14. Case 2 exit gas compositions for 12,256 and 40,112 elements.

106

A. Slezak et al. / Powder Technology 203 (2010) 98–108

Fig. 15. Computational model of SRI facility.

operating conditions for the three SRI runs chosen for this validation study are summarized in Table 6. Run 49 was a pyrolysis case, run 81 was a combustion run, and run 80 was a gasification run at the maximum steam and O2 levels. The computed centerline gas temperatures for all three runs are shown in Fig. 16. Results for the CO2, H2, and CO exit gas yields obtained from the CFD validation runs are compared to the SRI International experimental data for runs 80, and 81 in Figs. 17 and 18, respectively. The CFD predictions of the CO and H2 yields for the pyrolysis case (run 49; not presented) did not agree with the experimental data. These discrepancies are either due to an experimental O2 leak that was discovered after the run was completed, or to the fact that tar and soot were not included as pyrolysis species in the present devolatilization model. It is noted that the volatile distribution predicted in the CFD model [3] assumes that the product gases resulting from pyrolysis consist primarily of CH4, which is not found in the SRI pyrolysis results. The CO and H2 predictions are realistic for runs 80 and 81, but CO2 is underpredicted for both of these cases (by 40% of the measured value for run 80 and 34% for run 81). Fixed carbon conversion predicted by Fluent for the gasification case (run 80) was approximately 72%, versus 88.3% measured by SRI. For run 81 the computed fixed carbon conversion was 47%, versus a measured level of 72%. Additional Fluent CFD simulations have been obtained for a commercial-scale single-stage downflow entrained-flow gasifier,

Fig. 17. Comparison of exit gas yields for SRI run 80.

Fig. 18. Comparison of exit gas yields for SRI run 81.

Table 6 Operating conditions for SRI SG2PS5 density/size cut experimental runs [29]. Run #

Oven length (cm)

Coal loading (wt.%)

Oxygen loading (wt.%)

Steam loading (wt.%)

SR

49 80 81

15 120 120

1.5 2.10 1.88

0.0 3.22 2.88

0.0 9.0 0.0

0.0 0.68 0.68

Fig. 16. Computed centerline gas temperature for SRI validation runs.

Fig. 19. Commercial-scale, single-stage, downflow, slurry-fed, entrained-flow gasifier geometry.

A. Slezak et al. / Powder Technology 203 (2010) 98–108 Table 7 Proximate and ultimate analyses of Kentucky #9 coal. Component

Kentucky #9 coal

Fixed carbon Volatiles Ash Moisture C H O N S Cl

41.99 35.97 8.84 13.2 62.31 4.20 7.94 1.35 2.64 0

Fig. 20. Comparison of exit gas composition for commercial-scale downflow gasifier.

using the same chemistry models [4,5] and gasification UDF [3,28] as were used to simulate the two-stage upflow gasifier. The geometry and 7058-cell axisymmetric grid used are shown in Fig. 19. The present simulations were for a Kentucky #9 coal, to enable comparison with published performance data for the commercial unit [30]. Proximate and ultimate analysis data for this coal are given in Table 7. Coal slurry flow rate (27.8 kg/s coal and 11.0 kg/s H2O), gas flow rate (22.5 kg/s of 96% O2), and operating pressure (2.69 MPa) were selected to correspond to results given by Hornick and McDaniel [30]. Coal particle average diameter and specific gravity were set to d = 231 μm, SG = 1.373, with a Rosin–Rammler size distribution between 25 μm and 850 μm. The computed exit gas composition is compared with data from Hornick and McDaniel [30] in Fig. 20. Agreement between the CFD predictions and measured data are reasonable, except for underprediction of CO and overprediction of CO2 volume fractions. The computed fixed carbon conversion for this simulation is 95%,

107

compared to a measured value of 91%. Examples of the computed particle trajectories are shown in Fig. 21 for the lightest, smallest particles (SG1PS7; SG = 1.2; d = 50 μm) and the heaviest, largest particles (SG4PS1; SG = 3.3, d = 800 μm). The lighter, smaller particles tend to follow the gas flow and are recirculated in the gasifier, while the heavier, larger particles impact the wall nearer to the exit and do not recirculate. This is shown quantitatively in Fig. 22, where the percent of particles striking the wall is shown versus distance along the gasifier for the two particle size/density cuts.

4. Conclusions Fluent CFD simulations using DPM particle tracking have been obtained for commercial-scale two-stage upflow and single-stage downflow entrained-flow gasifier geometries, and the solutions have been used to examine variability in the computed gas phase average exit conditions (temperature and species volume fractions), gas phase flow field contours, particle char conversion, particle residence time, and particle–wall impact velocity for the 28 varying particle density and size fractions under study in the current Coal Partitioning Project. For these density/size fractions, the density ranges between SG = 1.2 and SG = 3.3, while particle diameter ranges between d = 50 μm and d = 800 μm. For the two-stage upflow gasifier, the computed gas phase flow field, computed particle char conversion, and computed residence times for some coal fractions, each differ significantly between the simulation that used the parent coal properties (Case 1) and the solution that used properties for each of the 28 different Bailey coal density/size fractions that have been separated and characterized in the CPP project (Case 2). This indicates that in order to most accurately determine both the actual trajectories of coal particles for each density and size fraction and the gasifier flow field, it is necessary to also model the injections of each of these density/size cuts in the iterated DPM simulation. Significant differences have also been observed for the computed particle trajectories, as the particle initial density and diameter are varied. Solutions that were obtained using two different devolatilization models generally compared reasonably well. Grid independence studies showed the grid used in the comparison studies was adequate for predicting exit gas composition and wall impact velocities. Preliminary validation studies using experimental data for the Bailey coal by SRI International in their pressurized coal flow reactor at 30 atmospheres [29] indicated adequate agreement for gasification and combustion runs, but poor agreement for a pyrolysis case. Additional simulation of a single-stage downflow gasifier yielded an exit gas composition that was in reasonable agreement with published data.

Fig. 21. Particle trajectories for injection of: A) cut SG1PS7 (SG = 1.2, d = 50 μm), and B) cut SG4PS1 (SG = 3.3, d = 800 μm).

108

A. Slezak et al. / Powder Technology 203 (2010) 98–108

Fig. 22. Particle–wall impacts for lightest, smallest particles (SG1PS7; SG = 1.2, d = 50 μm) and heaviest, largest particles (SG4PS1; SG = 3.3, d = 800 μm).

Acknowledgements This work has been supported under the DOE Gasification Technology Program, funded by the National Energy Technology Laboratory in Morgantown, WV, as part of the University Research Initiative, through the “Collaboratory for Multiphase Flow Research,” RDS Contracts No. 41817M2318 and 41817M2100.

References [1] N. Soundarrajan, M. Klima, S.V. Pisupati, Coal Partitioning Project Monthly Progress Report, Project Progress Report, Pennsylvania State Univ, June 2007. [2] N. Soundarrajan, M. Klima, S.V. Pisupati, Coal Partitioning Project Monthly Progress Report, Project Progress Report, Pennsylvania State Univ, Jan 2008. [3] S.P. Shi, S.E. Zitney, M. Shahnam, M. Syamlal, W.A. Rogers, Modelling coal gasification with CFD and discrete phase method, Journal of the Energy Institute 79 (n4) (Dec 2006) 217–221. [4] H. Kobayashi, J.B. Howard, A.F. Sarofim, Coal Devolatilization at High Temperatures, 16th Symp. (International) on Combustion, The Combustion Institute, 1976. [5] C.Y. Wen, T.Z. Chaung, Entrainment coal gasification modeling, Ind. Eng. Chem. Process. Dev. 18 (1979) 684–695. [6] S. Niksa, PC Coal Lab Version 4.0 User Guide and Tutorial, Niksa Energy Assoc. LLC, Belmont, CA, 2006. [7] C. Chen, M. Horio, T. Kojima, Numerical simulation of entrained flow coal gasifiers. Part I: modeling of coal gasification in an entrained flow gasifier, Chemical Engineering Science 55 (2000) 3861–3874. [8] C. Chen, M. Horio, T. Kojima, Use of numerical modeling in the design and scale-up of entrained flow coal gasifiers, Fuel 80 (2001) 1513–1523. [9] X.J. Liu, W.R. Zhang, T.J. Park, Modelling coal gasification in an entrained flow gasifier, Combustion Theory and Modelling 5 (2001) 595–608. [10] M. Cloke, E. Lester, A. Belghazi, Characterization of the properties of size fractions from ten world coals and their chars produced in a drop-tube furnace, Fuel 81 (2002) 699–708. [11] T. Wu, M. Gong, E. Lester, F. Wang, Z. Zhou, Z. Yu, Characterization of residual carbon from entrained-bed coal water slurry gasifiers, Fuel 86 (2007) 972–982. [12] Bouma, P.H., de Goey, L.P.H., Tummers, M.J., and Kiel, J.H.A. (1999). “Numerical Modelling of an Entrained-Flow Gasification Simulator,” Computational Technologies for Fluid / Thermal / Structural / Chemical Systems with Industrial Applications—1999 (ASME Pressure Vessels and Piping Conf.), Boston, MA, Aug. 1–5. [13] C. Chen, T. Miyoshi, H. Kamiya, M. Horio, T. Kojima, On the scaling-up of a two stage air blown entrained flow coal gasifier, Canadian Journal of Chemical Engineering 77 (1999) 745–750. [14] M. Shahnam, M. Syamlal, D. Cicero, Numerical Modeling of Combustion and Gasification Processes using the Discrete Particle Method, International Joint Power Generation Conf, ASME Fuels & Combustion Technologies Div, Miami Beach, FL, 2000, July 23–26.

[15] D.F. Fletcher, B.S. Haynes, F.C. Christo, S.D. Joseph, A CFD based combustion model of an entrained flow biomass gasifier, Applied Mathematical Modelling 24 (n3) (March 2000) 165–182. [16] C. Chen, M. Horio, T. Kojima, Numerical simulation of entrained flow coal gasifiers. Part II: effects of operating conditions on gasifier performance, Chemical Engineering Science 55 (2000) 3875–3883. [17] H. Tominaga, T. Yamashita, T. Ando, N. Asahiro, Simulator development of entrained flow coal gasifiers at high temperature and high-pressure atmosphere, IFRF Combustion Journal, June 2002, Article No. 200004, http://www.industrial. combustion.ifrf.net/view_paper.html?paperId=14. [18] M. Bockelie, M. Denison, Z. Chen, T. Linjewile, C. Senior, A. Sarofim, N. Holt, CFD Modeling for Entrained Flow Gasifiers, Gasification Technologies Conference, San Francisco, CA, Oct. 27–30 2002. [19] H. Liu, T. Kojima, Theoretical study of coal gasification in a 50 ton/day HYCOL entrained flow gasifier. I. Effects of coal properties and implications, Energy and Fuels 18 (4) (July/Aug 2004) 908–912. [20] H. Liu, T. Kojima, Theoretical study of coal gasification in a 50 ton/day HYCOL entrained flow gasifier. II. Effects of operating conditions and comparison with pilot-scale experiments, Energy and Fuels 18 (4) (July/Aug 2004) 913–917. [21] J.B. Roucis, J.M. Kline, J.D. Smith, Advanced Modeling of Gasifiers: Gasification Process Modeling using Computational Fluid Dynamics (CFD), Gasification Technologies Conference, Washington, DC, Oct. 6 2004. [22] H. Watanabe, M. Otaka, J. Inumaru, Development of numerical simulation technique for coal gasification on entrained flow coal gasifier (modeling of coal gasification reaction and prediction of gasifier performance), Transactions of the Japan Society of Mechanical Engineers, Part B 70 (n695) (July 2004) 1856–1863. [23] C. Guenther, R. Breault, The Effect of Gas–Solids Dispersion on a Two Fluid Model of a Transport Gasifier, AIChE Annual Meeting and Fall Showcase, Cincinnati, OH, Oct. 30–Nov. 4 2005. [24] H. Tominaga, Y. Matsushita, H. Aoki, T. Yamashita, T. Akiyama, T. Miura, The development of entrained flow coal gasification simulator, Nihon Enerugi Gakkaishi/Journal of the Japan Institute of Energy 85 (n1) (Jan. 2006) 36–41. [25] J. Lu, L. Yu, X. Zhang, S. Zhang, W. Dai, Hydrogen production from a fluidized-bed coal gasifier with in situ fixation of CO2 Part I: numerical model, Chemical Engineering & Technology 31 (2008) 197–207. [26] M. Forstner, G. Hofmeister, M. Jooler, J. Dahl, M. Braun, S. Kleditzsch, R. Scharler, I. Obernberger, CFD simulation of ash deposit formation in fixed bed biomass furnaces and boilers, Progress in Computational Fluid Dynamics 6 (2006) 248–261. [27] N. Syred, K. Kuriawan, T. Griffiths, R. Ray, Development of fragmentation models for solid fuel combustion and gasification as subroutines for inclusion in CFD codes, Fuel 86 (2007) 2221–2231. [28] Slezak, A. A. (2008). “Modeling of Particle Trajectories of Coal Size and Density Fractions in a Gasifier,” M. S. Thesis, West Virginia Univ., Dept. of Mechanical and Aerospace Engineering, Morgantown, WV; URL: https://eidr.wvu.edu/etd/documentdata.eTD?documentid=6016. [29] D.J. Eckstrom, R. Malhotra, Coal Partitioning High Pressure Gasification Experiments, Project Progress Report, SRI International, Mar 2008. [30] Hornick, M. J. and McDaniel, J. E. (2002). “Tampa Electric Polk Power Station Integrated Gasification Combined Cycle Project,” Final Technical Report, U.S. DOE Coop. Agrmt.DE-FC-21-91MC27363.