Prediction of Light Gas Composition in Coal Devolatilization - Energy

May 6, 2009 - At a much higher level of complexity are the generalized devolatilization models such as tar formation models, species evolution/functio...
1 downloads 15 Views 1MB Size
Energy & Fuels 2009, 23, 3063–3067

3063

Prediction of Light Gas Composition in Coal Devolatilization Ravichandra S Jupudi,*,† Vladimir Zamansky,‡ and Thomas H. Fletcher§ GE Global Research, 122, EPIP, Phase 2, Hoodi Village, Whitefield Road, Bangalore, India 560066, GE Global Research, IrVine, CA, USA, and Department of Chemical Engineering, Brigham Young UniVersity, ProVo, Utah, USA ReceiVed February 16, 2009

The chemical percolation devolatilization (CPD) model describes the devolatilization behavior of rapidly heated coal based on the chemical structure of the coal. It predicts the overall char, tar, and light gas yields. This paper presents an improved CPD model with improved capability for predicting light gas composition. This is achieved by incorporating a kinetic model that simulates the release of various light gas species from their respective sources/functional groups in coal. The improved CPD model is validated using experiments with a wire mesh reactor and published experimental observations.

1. Introduction Devolatilization, the first step in coal gasification, exerts its influence throughout the life of the coal particles from injection to burnout and is the step that is quite dependent on the properties of the coal. Composition and rate of evolution of individual gaseous species during devolatilization are important for defining product composition in coal gasification. Further, species composition is necessary to model the homogeneous reactions that happen subsequent to coal devolatilization. Incorporation of the capability to predict light gas composition in the established CPD model1 for coal devolatilization is the subject of this paper. Devolatilization models vary greatly in their complexity. The simplest of these are single-step and multiple-step methods that are purely empirical. At a slightly higher level of complexity are the distributed activation energy models (DAEM)2 that include standard deviation of the activation energy as an additional parameter. At a much higher level of complexity are the generalized devolatilization models such as tar formation models, species evolution/functional group models, and chemical network models that consider the evolution of gas species. These models are based on the descriptions of the coal structure and on the processes that the coal particle goes through during devolatilization. Examples of the tar formation models are the chemical percolation devolatilization (CPD)1 and FLASHCHAIN3 models, and an example of the chemical network model is the functional group-devolatilization vaporization crosslinking model (FG-DVC model).4 Since these models are based on fundamental processes, model predictions are applicable over a wide range of coal types and process conditions. The CPD model captures/considers all the physical features of FLASHCHAIN and FG-DVC models and takes only coal * To whom correspondence should be addressed. Phone: 91 80 4012 2447; e-mail: [email protected]. † GE Global Research, Bangalore, India. ‡ GE Global Research, Irvine, CA, USA. § Brigham Young University. (1) Grant, D. M.; Pugmire, R. J.; Fletcher, T. H.; Kerstein, A. R. Energy Fuels 1989, 3, 175–186. (2) Anthony, D. B.; Howard, J. B. AIChE J. 1976, 22, 625–656. (3) Niksa, S.; Kerstein, A. R. Energy Fuels 1991, 5, 647–665. (4) Solomon, P. R.; Hamblen, D. G.; Carangelo, R. M.; Serio, M. A.; Deshpande, G. V. Energy Fuels 1988, 2, 405–422.

structure coefficients (13C NMR measurements) as inputs without any adjustable constants. Further, it is computationally very economical. However, it involves only one rate equation that corresponds to the mixture of all the light gases. Hence, it predicts only the overall yield and not the composition of the light gases. A later CPD version5 included an empirical correlation to estimate the composition of the light gas evolved from any arbitrary unknown coal. The correlation links the light gas composition data of Solomon et al.6 and Chen7 with the extent of light gas release. A more accurate and reliable approach, however, would include individual rate equations for various light gas species that evolve during devolatilization. Such an approach was used in the FG-DVC model.4 The basic premise of this approach is that coal is viewed as an ensemble of functional groups that are organized into tightly bound aromatic ring clusters and are connected by bridges. Tar and light gas species are released by the thermal decomposition of these individual functional groups. The thermal decomposition of the functional groups into light gas species is incorporated into a kinetic model. The kinetics are expected to depend on the functional group or the nature of bond breaking. However, they are found to be relatively insensitive to coal rank for a good range of coals.8 The differences in the light gas composition for various coals, however, stems from the rank-dependent variations in the concentrations of the functional groups. Various functional groups and their corresponding kinetics are presented in Table 19 in which suffixes “extra loose,” “loose,” and “tight” represent various distinct sources/functional groups that lead to the release of a particular species. These suffixes indicate the nature of the binding between the functional group/source and the species, where loose would correspond to low binding energy and tight would correspond to high binging energy. The proportions of various functional groups depend on the coal (5) Genetti, D.; Fletcher, T. H.; Pugmire, R. J. Energy Fuels 1999, 13, 60–68. (6) Solomon, P. R.; Serio, M. A.; Carangelo, R. M.; Bassilakis, R.; Gravel, D.; Baillargeon, M.; Baudais, F.; Vail, G. Energy Fuels 1990, 4 (3), 319–333. (7) Chen, J. C.; Niksa, S. Energy Fuels 1992, 6 (3), 254–264. (8) Solomon, P. R.; Hamblen, D. G. Prog. Energy Combust. Sci. 1983, 9, 323–361. (9) Serio, M. A.; Hamblen, D. G.; Markham, J. R.; Solomon, P. R. Energy Fuels 1987, 1, 138–152.

10.1021/ef9001346 CCC: $40.75  2009 American Chemical Society Published on Web 05/06/2009

3064

Energy & Fuels, Vol. 23, 2009

Jupudi et al.

Figure 1. Chemical reaction scheme in the CPD model.1 Table 1. Kinetic Rate Coefficients of Various Light Gas Species (Taken from ref 9) gas CO2 extra loose CO2 loose CO2 tight H2O loose H2O tight CO ether loose CO ether tight HCN loose HCN tight NH3 CHx aliphatics CH4 extra loose CH4 loose CH4 tight H aromatic CO extra tight

primary functional group source carboxyl carboxyl carboxyl hydroxyl hydroxyl ether O ether O

H(al) methoxy methyl methyl H(ar) ether O

A (s-1)

E/R (K)

0.56 × 1015 0.65 × 1017 0.11 × 1016 0.22 × 1019 0.17 × 1014 0.14 × 1019 0.15 × 1016 0.17 × 1014 0.69 × 1013 0.12 × 1013 0.84 × 1015 0.84 × 1015 0.75 × 1014 0.34 × 1012 0.10 × 1015 0.20 × 1014

30 000 ( 1500 33 850 ( 1500 38 315 ( 2000 30 000 ( 1500 32 700 ( 1500 40 000 ( 6000 40 500 ( 1500 30 000 ( 1500 42 500 ( 4750 27 300 ( 3000 30 000 ( 1500 30 000 ( 1500 30 000 ( 2000 30 000 ( 2000 40 500 ( 6000 45 500 ( 1500

rank.9 For example, aryl ether linkages are a source of CO and they are categorized as CO-tight. Aryl ether linkages are found in all ranks of coal. However, CO-loose appears in the low rank coals only. Similarly, evolution of other species such as CO2, H2O, HCN, aliphatic hydrocarbons (CHx), hydrogen (H aromatic), and Sulfur (S organic) are also sensitive to coal rank. In this paper, the kinetic model that relates the functional groups to the light gas composition is incorporated in the CPD model. This improvement adds the capability of accurate prediction of light gas composition to the CPD model. Mathematical formulation corresponding to the improved model is presented in Section 2. Validation of the improved CPD model is presented in Section 3. 2. Mathematical Formulation A detailed description and mathematical formulation of the CPD model is available in ref 1The CPD model employs percolation statistics to describe the generation of light gas/tar precursors of finite size based on the number of cleaved labile bonds in the infinite coal lattice. The model includes treatment of vapor-liquid equilibrium and a cross-linking mechanism. Coal-independent kinetic parameters are employed, and coaldependent chemical structure coefficients are taken directly from 13 C NMR measurements or by using correlations5 developed from 13C NMR measurements of several coals. Figure 1 shows the simple reaction scheme in the original CPD model.1 The reaction starts with the cleaving of a chemical bond in a labile bridge (l ) to form a highly reactive bridge intermediate (l *).The reactive bridge intermediate may either be (i) released as a light gas (g2) with the concurrent relinking of the two associated sites within the reaction cage to give a stable or charred bridge (c) or (ii) the bridge material may be stabilized to produce side chains (δ) that may eventually convert into light gas (g1) fragments through a subsequent slower reaction. The kinetic expressions corresponding to this reaction mechanism are dl ) -kbl dt

(1)

dl* ) kbl - (kδ + kc)l * dt

(2)

Figure 2. Van Krevelen diagram used to compute functional group parameters for an unknown coal by interpolation.

Figure 3. Comparison of predictions of CPD model and improved CPD model for Illinois No. 6 coal. Lines: model predictions; symbols: experimental results from ref 9.

where kb, kδ and kc are rate constants corresponding to bridge breaking, side chain formation, and char bridge formation. The steady state expression for l * becomes: l* =

kbl kδ + k c

(3)

and hence kckbl k bl dc ) k cl * = ) dt kδ + k c F+1

(4)

2Fkbl dδ ) 2kδl - kgδ = - k gδ (5) dt F+1 where F ) kδ/kc. Equation 5 corresponds to the overall side chains that will eventually lead to light gas formation. By considering several δ values corresponding to various light gas species as in the FG-DVC4 model, improved light gas composition prediction capability can be incorporated into the CPD model. Mathematical expression corresponding to various light gas species is taken as

[

2Fkbl dδi ) dt F+1

]

fgi

- kgiδi

17

∑ fg

(6)

j

j)1

where fgi is the functional group/gas specie source fraction and kgi is the kinetic rate coefficient for gas specie “i”, listed in Table 1. Functional group parameters are also necessary to calculate the initial population of side chains corresponding to each of the light gas species given by

Light Gas Composition in Coal DeVolatilization

δi,0 ) 2(1 - c0 - l0) ×

Energy & Fuels, Vol. 23, 2009 3065

fgi

(7)

17

∑ fg

j

j)1

gi gi di ) ) ) dimax gi,max gi(∞)

gi 2(1 - c0)

(11)

fgi 17

∑ fg

j

where subscript “0” denotes the initial value of the particular parameter. As in the FG-DVC4 model, functional group parameters for an unknown coal are obtained by a 2D interpolation method based on coal rank from the data corresponding to a library of coals.10The O/C and H/C molar ratios are used as indicators of coal rank. The elemental ratios of the library coals are used to form a 2D triangular mesh on a O/C vs H/C coalification diagram, called a Van Krevelen diagram, as shown in Figure 2. The three nearest nodes on the diagram form each triangle. For an unknown coal, the elemental composition determines the appropriate triangle. The functional group parameters of the unknown coal are obtained by interpolation from the parameters corresponding to the three nodes, as shown in Figure 2. The numerical details of the interpolation method are available in ref 10. From the values of assorted side chain populations, the amounts of various light gas species formed can be computed as gi ) [2(1 - l - c0) - δtot] ×

δi δtot

(8)

while the total amount of light gas formed is given by

j)1

Using eq 11 and the look-up table for the functional area under the normal curve, the activation energy is calculated. The improved CPD model can thus predict the light gas composition of any unknown coal. Various species that are predicted as part of the light gas composition are CO2, CO, CH4, H2O, H2, HCN, NH3, aliphatic hydrocarbons, and tar. Experimental validation of this improved CPD model is presented in the next section. 3. Validation of Improved CPD Model Predictions of the improved CPD model were first compared with published pyrolysis results for three well-known coals.9 The experimental results9 were generated in a heated tube reactor with a peak temperature of 1073 K. Particle-temperature history, which is an input to the CPD model, was extracted from ref 9 and coal structural parameters were obtained from ref 1. Species CHx represents the overall aliphatic hydrocarbons produced during devolatilization. In ref 9, a hydrocarbon-cracking model was employed to divide the CHx into paraffins, olefins, and acetylene. The hydrocarbon-cracking model is not included in the current work since the authors intend to develop a separate

17

g)

∑g

j

) [2(1 - l - c0) - δtot]

(9)

j)1

17 δj . where δtot ) ∑j)1 Species evolution is a result of breaking bonds that have a variety of activation energies. This distribution in activation energies could be handled by solving a large set of differential equations or by solving a complex distribution function involved in traditional distributed activation energy models.2 A highly computationally efficient alternative is the use of an effective activation energy (Ei) based on the mean value (E0), standard deviation (Vi), and extent of reaction to capture the spread of the Gaussian distribution of the activation energies.1 This methodology is employed in the CPD model, where the chemical reactions with distributed energies are viewed as progressing sequentially, with the low-activation-energy species reacting at lower temperatures, followed by the high-activation-energy species. Thus, the specific activation energy of these reactions is increased according to a normal distribution function as the reactions proceed. The value of the normalized probability function at different extents of reaction is taken as

di dimax

)

∫ √2πV 1

E

{ (

exp -∞

i

1 E - Ei 2 Vi

)} 2

dE

(10)

where di/di max represents the ratio of any distributed value to its maximum value. The right-hand side of eq 10 represents the fractional area under a normal curve for a particular value of activation energy. This is coded in the form of a look-up table. The left-hand side of eq 10 represents the extent of reaction, which for the release of particular species is given by (10) Zhao, Y.; Serio, M. A.; Bassilakis, R.; Solomon, P. R. TwentyFifth Symposium (Int.) on Combustion/The Combustion Institute 1994, 553– 560.

Figure 4. Pyrolysis results of North Dakota Beulah Zap coal obtained with the improved CPD model. Solid lines: CPD model predictions; symbols: experimental results from ref 9.

3066

Energy & Fuels, Vol. 23, 2009

Jupudi et al.

Figure 6. Effect of heating rate on the pyrolysis yields of Illinois No. 6 coal. Solid lines: improved CPD model predictions; symbols: WMR experimental results.

Figure 7. Effect of peak temperature on the pyrolysis yields of Illinois No. 6 coal. Solid lines: improved CPD model predictions; symbols: WMR experimental results. Figure 5. Pyrolysis results of Montana Rosebud coal obtained with the improved CPD model. Solid lines: CPD model predictions; symbols: experimental results from ref 9.

homogeneous gas-phase reactions model.11 Figure 3 compares the predictions of the improved CPD model with the published version of the CPD model1 for Illinois No. 6 coal,9 and Figures 4 and 5 compare model predictions for Beulah Zap and Montana Rosebud coals with corresponding experimental data from 9. Predictions of the improved CPD model are in good agreement with the experimental observations. Some disagreement can be noted for H2O yields. This could be attributed to the difficulty in obtaining good data on H2O yields.9 In entrained flow gasifiers, the high heating rate, high temperature, and high pressures have a significant effect on devolatilization yields. A wire mesh reactor (WMR) is a good bench-scale experimental tool to study devolatilization at such extreme operating conditions. A high-temperature, high-pressure WMR was set up (as described elsewhere12) to validate the improved CPD model described in this paper. The effect of heating rate on the pyrolysis yields of Illinois No. 6 coal was studied in the WMR by subjecting coal samples to a peak temperature of 1173 K at different heating rates ranging from 5 to 5000 K/s. Similarly, the effect of peak temperature and pressure on pyrolysis yields were studied by subjecting coal samples to various temperature and pressures ranging from 300 to 1100 °C and from 1 to 50 bar, respectively. The experimental (11) Zamansky, V.; Ravichandra, J. S.; Zeng, C.; Eiteener, B.; Fletcher, T. H. International Pittsburgh Coal Conference: Pittsburgh, PA, 2008. (12) Zeng, C.; Chen, L.; Liu, G.; Li, W.; Huang, B.; Zhu, H.; Zhang, B.; Zamansky, V. ReV. Sci. Instrum. 2008, 79, 84–102.

Figure 8. Effect of pressure on the char and tar yields from Illinois No. 6 coal. Solid lines: improved CPD model predictions; symbols: WMR experimental results.

results and corresponding yields predicted by the improved CPD model are compared in Figures 6-8.Good agreement between the experimental observations and improved CPD model predictions establishes the capability of the CPD model in capturing the effect of heating rate, peak temperature, and pressure on pyrolysis yields. Additional validation of the improved CPD model was performed using the data of Xu and Tomita,13 who conducted devolatilization experiments on 17 coals in a Curie-point pyrolyzer, which heated samples to 1037 K at a heating rate of 3000 K/s. The carbon content of these coals ranged between 65.4 and 93.7 wt % on a dry, ash free basis. Figure 9 compares predictions of the improved CPD model with the experimental observations. While certain discrepancies exist in few individual (13) Xu, W.; Tomita, A. Fuel 1987, 66, 627–631.

Light Gas Composition in Coal DeVolatilization

Energy & Fuels, Vol. 23, 2009 3067

Figure 9. Pyrolysis results of the 17 coals studied by Xu and Tomita.13 Solid symbols: experimental measurements; open symbols: CPD model predictions. Lines are used for visualization purposes only.

cases, model predictions follow experimental observations qualitatively as well as quantitatively. The agreement is remarkable, since NMR input data for these coals are obtained by correlations,5 whereas the functional group parameters are obtained by data interpolation from the library coals. 4. Conclusions A capability for predicting light gas composition was incorporated into the existing CPD model. This ability was achieved by including rate equations corresponding to various light gas species. A 2D interpolation methodology was incorporated to compute the functional group parameters that are required as input to obtain the initial populations of the side chains corresponding to various light gas species. The improved

CPD model is validated by comparison with published experimental data. Modeling predictions quantitatively agree with experimental observations. Light gas compositions predicted using the improved CPD model can be used as input to various models on coal gasification. Acknowledgment. The work contained in this paper is funded by the Sustainable Energy Advanced Technology program of the General Electric Company (GE). Experiments with WMR were conducted at the GE Global Research at Shanghai by Lei Chen, Dr. Cai Zeng, and Dr. Gang Liu. Their support in providing experimental data is gratefully acknowledged. EF9001346