Gasification Reaction Pathways of Condensable Hydrocarbons

Apr 26, 2016 - In addition, UCS that are considered to contain primary tar are divided into four subgroups, to encompass two levels of reactivity with...
1 downloads 10 Views 2MB Size
Download PDF
Article pubs.acs.org/EF

Gasification Reaction Pathways of Condensable Hydrocarbons Mikael Israelsson* and Henrik Thunman Department of Energy and Environment, Division of Energy Technology, Chalmers University of Technology, SE-412 96 Gothenburg, Sweden S Supporting Information *

ABSTRACT: Dual fluidized bed (DFB) gasification of biomass generates numerous species with large differences in size and boiling point. At the heavier (tar) end, the produced species range from benzene to coronene. In this work, a method for elucidating the pathways of tar evolution is applied to previously presented measurements that satisfy the carbon balance of the Chalmers 2−4-MW DFB gasifier. In addition to quantifying the cold gas and tar, the measurements yield information regarding the amount and C, O, H composition of unknown condensable species (UCS). The reaction pathways were identified by means of fitting a model to the performed measurements. The employed solver varies freely the reaction rate coefficients of three global reactions (mimicking dissociation and reactions with hydrogen and steam) per modeled group, as well as the carbon distribution coefficients within a predefined reaction scheme. The mature tar (excluding primary tar) spectrum is divided into phenols, furans, benzene, naphthalene, pyrene, and one-, two-, and three-ring aromatic components. In addition, UCS that are considered to contain primary tar are divided into four subgroups, to encompass two levels of reactivity with varying composition. Ultimately, the solver converges, yielding a reaction scheme that is based on the findings of earlier works and that describes the creation of mature tar from UCS. Furthermore, the importance of individual reaction routes is discerned for the pertinent measurements. Thus, it is demonstrated that the maturation of secondary tar species (e.g., toluene and phenol) is not in itself sufficient to describe the formation of the tar spectrum.

1. INTRODUCTION Indirect biomass gasification is a thermochemical, primary conversion process that can be applied for biofuel production, whereby biomass is converted into a mixture of combustible components,1 often referred to as “raw gas”. A large fraction of the generated gas consists of cold gas (or permanent gas) species, such as hydrogen and methane, although condensable hydrocarbons are also formed during the pyrolysis of the fuel. Initially, these condensable species (CS) comprise mainly primary tar, which contains a high fraction of oxygen, and are very reactive at higher temperatures.2 As the primary tar reacts, it forms additional cold gas, as well as increasingly stable secondary and tertiary tar species, which are more aromatic in nature and, in the case of tertiary tar, heavier components with higher boiling points. The high boiling points of the heavier tar species are associated with increased complexity of the gasifier operation, as the tar condenses on cold surfaces, eventually causing extensive fouling and blockages in the downstream equipment. In addition to the problems related to fouling, unreacted tar represents a loss of efficiency.3 Therefore, a comprehensive understanding of tar formation and maturation is needed to minimize or otherwise control its formation. Several studies have been presented on various reaction schemes for tar, describing its formation from the primary species generated during pyrolysis,4−6 as well as the reformation of already mature tar.7−9 Unfortunately, many of these studies have been based on incomplete mass balances or have relied on species that are not measured, making validation difficult. A global reaction mechanism that includes the formation, subsequent maturation, and decomposition of tar, preferably based entirely on measurable species, would bring together all the previously mentioned reaction schemes. The © XXXX American Chemical Society

aim of the present study was to construct an expanded (yet simplified) reaction scheme based on the previously presented measurements,10 for which the carbon mass balance of the gasifier is fulfilled. The measurements were carried out in the Chalmers 2−4MW dual fluidized bed (DFB) gasifier operated using silica sand as the bed material, wood pellets as the fuel, and steam as the fluidizing agent. The investigated parameters were temperature, level of fluidization, and residence time at constant steam-to-fuel ratio.10 The measured raw gas spectrum was divided into two parts: (1) the cold gas and (2) CS. Part of the CS was further categorized as SPA tar, which contains all detectable species in the boiling-point range between benzene and coronene, as measured using the solid-phase adsorption (SPA) method.11−13 The remainder of the CS was denoted as unknown condensable species (UCS),3,10 which contain all condensable species that are not detected using the SPA method and are considered to consist of primary and secondary tar species, as well as intermittent species. The amount of carbon present as UCS is determined by comparing the carbon yield of the cold gas and SPA tar to the total carbon yield of the raw gas flow, as described in subsequent sections. The purpose of the present work was to describe how the reformation of the intermediate and primary tar species affect the final composition of the tar, and to identify the key parameters for describing the evolution of CS. This was done by means of identifying the predominant conversion routes for the CS by fitting a model that describes the evolution of the tar, Received: March 3, 2016 Revised: April 22, 2016

A

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

implies that the additional reactivity mainly affected the conversion of the primary tar species, resulting in a higher yield of cold gas species and lower levels of tar species being formed. Therefore, the reaction patterns of the primary species must be fully elucidated to achieve a clear understanding of the formation of secondary and tertiary tar species. In summary, for the modeling of primary conversion systems, the inclusion of primary tar in the mechanism is crucial for describing the creation of the mature tar spectrum. In contrast, for secondary systems that contain no primary tar in the inlet stream, the inclusion of primary tar in the mechanism is not necessary. In addition to the various employed reaction schemes, several different reaction expressions have been proposed.7−9,17,18 The types of reactions used for tar include thermal dissociation reactions that are dependent on the level of the species of interest,7 as well as hydrogenation, polymerization, and heterogeneous reactions.5,17,19 For experimentally determined reaction kinetics, the reaction rate expressions are defined as a function of reactant concentration (C) and temperature (T):

starting from the UCS yield obtained during the pyrolysis of wood pellets.14 The present work was performed to ensure that relevant species and conversion routes can be implemented in future models.

2. THEORY The maturation of tar described by Elliot et al.2 is depicted in Figure 1. Generally, as the temperature is increased, the

Figure 1. Tar maturation scheme. (Reproduced from ref 2. Copyright 1998, American Chemical Society, Washington, DC, USA).

composition moves from a relatively high oxygen-to-carbon ratio to zero and the tar takes on a more aromatic nature. While Figure 1 does not present a complete reaction scheme, the general transition of mass is an important element that forms the basis of the present work. Condensable species were thoroughly investigated in previous studies,2,5−9 resulting in several reaction schemes being proposed. The various schemes differ in terms of their complexity, ranging from describing the evolution of predominant species as a function of temperature2 to establishing reaction schemes that contain several tar species, whereby stable species are formed from the decomposition of less-stable species.5,6,9 Furthermore, the schemes differ, with regard to which process they describe. For example, Corella et al.7 and others7−9 describe the evolution of an already mature tar spectrum that contains stable species, such as benzene, while Font Palma5 employs a scheme in which primary tar is formed from lignin and progressively matures into secondary and tertiary tar. The models that have been designed for already-matured gas mixtures can be implemented to describe systems in which primary tar no longer exist and the CS can be quantified using only, e.g., the SPA method. These systems include secondarymeasure equipment and the freeboard following a pyrolysis step of sufficient severity to convert all the primary tar. However, in constructing this type of model, one must ensure that the investigated system does not contain significant amounts of primary tar species, which are not sufficiently quantified using the SPA method and may act as hidden source terms for the species of interest. Failure to ensure negligible levels of primary tar species may result in unrealistic reaction pathways, wherein secondary tar species must be formed from tertiary species to satisfy the mass balance of the system. When describing systems that operate at lower degrees of severity, such as noncatalytic low-to-medium temperature gasification, the conversion and measurement of primary tar species becomes important. Previous studies15,16 have shown that, depending on the means of implementation, the use of an active bed material in indirect gasification can strongly affect both the amount and composition of tar. For example, Larsson et al.15 have shown that, for low levels of fluidization, replacing part of the bed material with ilmenite reduces the yields of all the tar species, to approximately the same extent. Presumably, if this added layer of reactivity affected the tar at a later stage of maturation, it would have shifted its composition toward higher levels of maturity (according to Figure 1), resulting in lessoxygenated tar with a higher fraction of tertiary species. This

r = f (C , T )

(1)

In the case of controlled, dedicated experiments, the determined rate expressions can be highly accurate, while the kinetic data derived from larger systems with unknown fluctuations will have associated uncertainties. Consequently, uncertainties related to, for example, the temperature of the system will be reflected in the derived rate coefficients, turning them into lumps that encompass several unknown parameters in order to ensure a satisfactory global reaction rate that is proportional to the perceived temperature. The present work is limited to proposing a general reaction scheme for CS, and the obtained measurements are used to derive these types of lumped kinetic coefficients. Consequently, the derived rate coefficients are referenced as “relative rate coefficients”, to underline the fact that they are limited to describing the dynamics of the investigated system in proportion to the perceived process parameters. The reactivity of each modeled group or species (i) is described through three global reaction rates that cover thermal dissociation and reactions with steam and hydrogen: ⎡ ⎛ k 2,H O, i ⎞⎤ ⎡ ⎛ k 2, T , i ⎞⎤ x 2 ri = − Cik1, T , i exp⎢− ⎜ ⎟⎥ ⎟⎥ − CiC H2Ok1,H2O, i exp⎢− ⎜ ⎣ ⎝ T ⎠⎦ ⎣⎢ ⎝ T ⎠⎥⎦ ⎡ ⎛ k 2,H , i ⎞⎤ 2 − CiC Hy 2k1,H2, i exp⎢ − ⎜ ⎟⎥ ⎢⎣ ⎝ T ⎠⎥⎦

(2) 3

where C denotes the concentration [mol/m ]; x and y are the reaction orders of steam and hydrogen, respectively; and the constants k1 and k2 represent the relative rate coefficients. The use of the three global reactions takes into account the effects of the temperature, steam, and hydrogen concentrations. This, together with the use of relative rate coefficients, makes the model sufficiently flexible to describe the behavior of the studied system. Regardless of which individual types of reactions control the maturation of tar, the sum total of the proposed reactions for any species should generate a reasonable overall reaction rate. Additional factors that affect the reaction rate expressions, such as the effect of bed material aging, the extent of gasparticle interactions, and the catalytic effects of ash components such as alkali,20,21 are probably needed to explain fully the reaction rate of any given specie. While evaluation of the effects B

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels of catalytic components is outside the scope of the present investigation, earlier and parallel studies have indicated that it is an important parameter for the evolution of tar.22 Therefore, if the levels of catalytic components are constant throughout the measurements, its effects will be included in the various k-values of eq 2, although variations in the levels of catalytic activity will not be taken into consideration. It should be mentioned that online measurement systems for the quantification of alkali species have been presented previously.23 2.1. Mass Balance. Since the employed mass balance and the measurements included in this paper are described elsewhere,3,10 only those aspects of particular relevance will be summarized here. The measurements were performed using two parallel systems: (1) a gas conditioning system for the cold gas and SPA tar analyses,12 and (2) a high-temperature reactor (HTR)3 for complete quantification of the total carbon flow of the raw gas and the oxygen transport. The HTR is essentially a heated tube that is maintained at 1700 °C and through which part of the raw gas is transported. During this transport, all the species present are thermally converted into H2O and CO, H2, and CO2, all of which are readily quantifiable in a microgas chromatograph24 (micro-GC). During operation of the gasifier, a known flow of helium is added to the fluidization steam prior to it entering the gasifier, to create an even distribution throughout the bed and freeboard. This flow is used to determine the yields of all the measured species, as helium is quantified together with both the cold gas species of the gas conditioning system and the gas generated by the HTR. Once the yields of all the species are determined, the total molar flow of carbon in the cold gas and SPA tar can be compared to that in the gas exiting the HTR. This comparison means that the carbon yield of the UCS can be determined together with its CHmin ratio, which signifies the lowest possible ratio of hydrogen to carbon atoms in the UCS, i.e., when no oxygen is present in the molecules.3 The CHmin ratio varies between measurements, as a function of varying levels of conversion, which complicates the determination of a representative molecule. Consequently, the UCS is described as a mixture of two pseudospecies, namely, butane C4H10 (B) and formic acid CH2O2 (F), thereby permitting the existence of a wide range of CHmin values for the UCS. Since the individual CHmin ratios of B and F are 2.5 and −2.0, respectively, this approach reflects not only changes in the total carbon content of the UCS, but also changes in C, O, and H composition.

Table 1. Summary of Process Parameters for the Chalmers 2−4-MW DFB Tavg. bed [°C]

Tavg. gas [°C]

fuel feed [kg/h]

steam-to-fuel ratio [kg/kgdaf]

residence time [s]

823 804 786 811 811 813 809 801 799

792 774 758 783 781 779 773 773 772

294 294 295 295 294 295 252 338 369

0.87 0.87 0.86 0.98 0.87 0.75 0.86 0.86 0.85

4.72 4.90 5.07 4.52 4.86 5.19 5.61 4.31 4.01

of the fluidization steam) affects the residence time and the steam-tofuel ratio, as well as the level of particles in the freeboard. Similarly, the residence time was varied by adjusting the level of fluidization at a constant steam-to-fuel ratio, which was achieved by matching an increase in the steam supply with an appropriate increase in the fuel feed. Therefore, changing the residence time affects the level of particles in the freeboard in the same way as changes in the fluidization level. The raw gas is assumed to maintain a constant temperature throughout the gasifier which, combined with the assumption of constant volume flow, allows the residence time to be estimated. Presumably, both the temperature and volumetric flow rate will vary in the freeboard, because of reactions, heat transfer, and mixing. However, the current measurements do not allow the combined effect of these parameters to be determined. Nevertheless, assuming that the errors in temperature and residence time are similar for all cases, the generated data are sufficient to match the purposes of the present work. The collected experimental data points were compared to the same starting point for the purpose of deriving the reaction pathways. The employed starting point was adopted from Neves et al.14 and was derived for pyrolysis of a similar fuel (wood pellets) at 600 °C. The main assumption linked to the usage of this measurement as the starting point is that the gas composition of all the performed measurements at one point in time resembles that of the pyrolysis time point. Consequently, all the modeled cases originate from the point of pyrolysis at 600 °C and evolve over a timespan that reflects the difference in residence time between the pyrolysis point (set at 3 s)14 and the gasification points. 3.2. Theoretical Section. Earlier studies of pyrolysis and gasification have shown that the tar spectrum consists of several hundred different species during its evolution from primary to mature tar.25 Since not all these species can be included in the model, they are divided into different groups and represented by a model compound. The chosen model compounds are usually the most abundant species found in each group. As an example, phenol is often used to represent either phenolic species or all tar species that contain oxygen. A mature tar spectrum from primary tar components has been modeled in previous publications.5,6 In these models, the yield of primary tar is calculated based on either the lignin fraction of the fuel5 or its elemental composition and the devolatilization temperature.6 The representative primary tar species used include oxygenated hydrocarbons, such as acetol (C3H6O2), catechol (C6H6O2), and vanillin (C8H8O3), which are converted to secondary tar species, such as toluene, phenols, and intermediate species. In the model of FuentesCano et al.,6 all these species contribute to the subsequent formation of naphthalene and benzene, which, in turn, form the larger tertiary tar (pyrene) and soot. In the model of Font Palma,5 benzene is formed from catechol or the conversion of naphthalene, and all the heavier tar species are formed through polymerization reactions that involve catechol and cyclopentadiene, which is generated during the conversion of phenol. With the exception of the formation of benzene from naphthalene, neither of the two models allows reactions that

3. METHODOLOGY 3.1. Experimental Section. The experimental conditions for the measurements performed in the Chalmers 2−4-MW DFB are summarized here. The gasifier performance was evaluated for three different temperatures, levels of fluidization, and gas residence times in the freeboard, resulting in nine experimental setpoints being collected over a period of 3 days. The employed operational parameters (Table 1) were, with the exception of the investigated parameter, maintained as constant as possible throughout the different measurements. The presented values for the average raw gas temperature differ somewhat from the previously published values,10 because a distant temperature measurement was excluded to obtain a more representative value of Tavg. As a result, Tavg was determined as the average of the mean bed temperature and the outlet raw gas temperature. This, together with minor changes in the calculations of the molar flow rate, slightly affected the determined residence time. Changes in the bed material temperature mainly affect the raw gas temperature and, to a lesser extent, the residence time. The level of fluidization (i.e., the flow rate C

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels generate smaller aromatic species from larger species, as reported previously.7,9,26,27 In the present work, the SPA-tar spectrum was complemented by additional model compounds, to describe the behaviors of furans, naphthalene, and species that contain three aromatic rings more accurately. In addition, UCS were used to describe primary tar and intermediate species that can affect the formation of SPA-tar. Consequently, the SPA-tar and UCS were categorized into groups, according to size and level of maturation. The SPA-tar was divided into the following eight groups (with the representative species in parentheses): phenolic species (phenol); furans (2,3-benzofuran); benzene, one aromatic ring species with branches (toluene); naphthalene, two aromatic ring species with branches (acenaphthylene); three aromatic ring species (phenanthrene); and four or more aromatic ring species (pyrene). The UCS were divided into the abovementioned pseudo-components of formic acid and butane (F and B), and the secondary species F′ and B′ were employed to facilitate two levels of reactivity of the UCS. In other words, while some UCS will form SPA-tar and cold gas directly, some will generate secondary UCS (UCS′), which, in turn, can form SPA-tar and gas. Therefore, the experimental data for UCS are assigned as secondary UCS, while the primary species are considered to attain the measured UCS levels twice as fast as the secondary UCS (i.e., in half the residence time). The composition of the CS at the starting point is based on the pyrolysis measurements described in the Experimental Section. Here, the amount of SPA-tar is assumed to be zero, while all carbon found in the CS is designated as UCS, resulting in yields for F and B of 8.0 and 2.8 mol/kgdaf, fuel, respectively. Furthermore, as all the cases exhibit different volumetric flow rates and temperatures, the starting concentrations of UCS (in mol/m3) will vary. The actual starting yields of F and B most likely varies for the different cases, due to, e.g., the varying temperature of the gasifier. However, considering the uncertainties related to other parameters, such as temperature and residence time, assuming constant yields of F and B is acceptable for the purpose of this work. The proposed reaction scheme (presented in the Results section) is based on the findings reported previously2,5−8,10 and includes the contribution of the determined UCS. Thus, the formation of SPA-tar can be described using a known parameter, while satisfying the mass balance of the system. The scheme should comply with the order of maturation described by Milne et al.,2 whereby the CS evolve from oxygenated to aromatic species. Furthermore, it should take into account the reactions associated with mature tar,7,8 in which multiplering species can either dissociate into smaller aromatic species or form even larger species, approaching soot. Consistent with previous studies,4−6 the UCS are considered to form simple tar species, such as phenols, toluene, and UCS′, which, in turn, can form more-complex species. Furthermore, as the UCS′ are considered to contain species, such as cyclopentadiene, which are known to partake in polymerization reactions that yield polyaromatic species with and without functional groups,4,5,8,28 the scheme also accounts for the formation of mature components from UCS′. Nevertheless, these reaction routes are somewhat suppressed in favor of the remaining routes when deriving the reaction scheme, so as not to undermine the concept of tar evolution. The main assumptions concerning the proposed scheme are as follows:

computational time, assuming that they are identical for the reactions of all the groups, and initial values for the solver are generated. Furthermore, this offers an efficient means for assessing whether the measurements follow sufficiently uniform trends to allow a good fit. It is well-known from previous experience with the examined system that aging of the bed material and other unquantified factors can drastically alter the gasifier performance.16 This means that measurement points that do not comply with the major trends must be excluded and subsequently compared with the final model to assess the severity of the deviation. The proposed reaction scheme was expanded and refined throughout this work to obtain a satisfactory model, and it currently contains 35 unknown distribution factors, each having a value in the range of 0−1, which describes the fraction of reacted carbon that is being transferred via a specific route. For species with several outgoing distribution routes, the sum of the distribution factors must also be a value between 0 and 1, to avoid creating additional carbon. In addition to the unknown distribution factors, the employed relative reaction rates described in eq 2 contribute with six unknown constants per reactive species, resulting in a total of 107 unknown constants. The solver was constructed to generate randomly distribution factors and relative reaction rate coefficients, within a predetermined window of variability, to simulate the gasifier as a plug flow reactor and calculate the error determined in relation to the measured points. Following the simulation of a significant number of different cases, the window of variability is centered on the case that gives the lowest error, and the cycle is repeated. Eventually, the most promising case must be assessed on the basis of additional demands, to ensure that an acceptable solution is available. This assessment involves ensuring that sufficient mass is transferred through the system to allow adequate formation of heavier species, and that the system is neither exceedingly fast nor slow. The resulting model, which describes the tar evolution, should be capable of presenting realistic trends for the amount and composition of the tar in the studied system. However, because of the uncertainties related to the temperature and residence time, as well as those associated with describing the gasifier as a plug flow reactor, the predicted rate of change within the system will be approximate. Consequently, the present work should not be viewed as a kinetic study or as a construction of a general model, but rather as an attempt to map the routes by which secondary and tertiary tar species are formed during the maturation of the tar.

4. RESULTS During initial attempts to construct a general set of reaction expressions, it was noticed that the experimental points collected while varying the residence time (Table 1) did not comply with the remaining points. Therefore, these points were excluded from the numeric solver, and were subsequently used for comparison with the obtained model during the later stages of this work. The two exponents x and y in eq 2 were, as described above, determined by solving analytically the individual reaction expressions for UCS and phenol. The resulting values for x and y were ∼2 and ∼1, respectively, and were used to describe the reactivities of all the species, with the exception of primary F for which an x-value of 2.3 was obtained. Presumably, the actual values for x and y may vary between different species, but it is unlikely that such variations would be large enough to alter the findings of this study. It was similarly concluded that a reaction order of one should be used for all of the CS. The evolution of the various groups of CS, as described by the obtained reaction parameters, is shown in Figures 2−7 for the three cases that describe the situations at high (792 °C), medium (774 °C), and low (758 °C) temperatures (Table 1). The lines indicate the modeled trends, while the points represent the measured values. In all of the figures, the triangles

(1) all CS are initially found as primary UCS; (2) SPA-tar is formed via reactions of UCS, UCS′, or other SPA-tar species with a carbon yield lower than unity; and (3) oxygen cannot be added to the CS; it can only be transferred from species that already contain oxygen (UCS, phenols, and furans). In the proposed reaction scheme, UCS are formed only during the primary pyrolysis and are not affected by any other tar species. This allows the conversion reactions of UCS, UCS′, and phenol to be solved analytically, provided that they are solved in sequence, since UCS′ is dependent only on the conversion of UCS, while phenol is dependent on the conversion of both UCS and UCS′. As a result, the exponents x and y in eq 2 can be obtained in a reasonably short D

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 2. Evolutionary profiles of the primary UCS at high, medium, and low temperatures, indicated as black, red, and blue lines, respectively.

Figure 5. Evolutionary profiles of toluene and phenanthrene for high, medium, and low temperatures, indicated as black, red, and blue lines, respectively.

Figure 3. Evolutionary profiles of the secondary UCS for high, medium, and low temperatures, indicated as black, red, and blue lines, respectively.

Figure 6. Evolutionary profiles of benzene and naphthalene for high, medium, and low temperatures, indicated as black, red, and blue lines, respectively.

Figure 4. Evolutionary profiles of phenol and 2,3-benzofuran for high, medium, and low temperatures, indicated as black, red, and blue lines, respectively.

Figure 7. Evolutionary profiles of acenaphthylene and pyrene for high, medium, and low temperatures, indicated as black, red, and blue lines, respectively.

correspond to the dashed lines and the circles correspond to the solid lines. As previously mentioned, it is assumed that all the CS is present as primary UCS at the starting point, and that all the other species are formed through its decomposition. Overall, the derived parameters provide a reasonably good fit to the provided data, which means that the derived scheme ably

depicts the general trends of the measurements. Similar curves for the measurements obtained at different levels of fluidization are comparable to those obtained at the medium temperature. The main effects of altering the level of fluidization are alterations of the residence time and the concentrations of the tar species, with the latter being due to dilution by the steam. E

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels The derived relative rate coefficients, which result in the best fit, are summarized in Appendix A, which is given in the Supporting Information. The derived reaction scheme and distribution factors are presented in Figure 8 and Table 2, respectively. In Figure 8, all

without employing reaction routes 13−20. This implies that the contribution of UCS′, either as a direct reaction or as part of the polymerization reactions, is of great importance for both the creation and maturation of the SPA-tar. However, as shown in Table 2, the current model does not differentiate between F and B, as long as a sufficient mass is transported to the heavier species. The obtained model was used to simulate the various cases for a residence time of 300 s, to assess the trends in the extrapolated data. Figures 9 and 10 depict the evolution of the

Figure 8. Proposed reaction scheme, with numbers denoting the distribution factors (summarized in Table 2) and with line thickness representing relative significance.

of the reaction routes involving UCS and UCS′ are assigned two distribution factors, with the odd numbers referring to F and the even numbers referring to B for all of the routes. The thickness of the lines that denote the different routes indicate how much carbon was transported via a specific route during the first 5 s of the reactions. The dashed lines all transported 5 times less mass than a solid line of equal thickness. Note that, although the thick lines represent significant routes of mass transfer, the thin and dashed lines are less certain and arise from the solver and reaction scheme used. Since all of the routes represent global reaction mechanisms, higher levels of resolution are needed to describe properly the transport of lower amounts of mass between the various groups. Nevertheless, upon investigation of the less-significant routes, it was found that reactions regarded as unlikely, such as the formation of furan from phenol, agreed well with the findings of earlier studies.2,29 The green lines in Figure 8 indicate routes of mass transfer that proceed directly from UCS′ to much more mature tar species, because of the previously mentioned polymerization of species, such as cyclopentadiene. In addition to the mass transfer that occurs between the different tar species, the reactions yield various amounts of cold gas species, although these routes are not included in the figure. Interestingly, it was not possible to transport enough mass to the tertiary tar species

Figure 9. Normalized carbon yields of the modeled species, as a function of severity.

Figure 10. Normalized carbon yields of the modeled species, as a function of severity.

Table 2. Derived Carbon Distribution Factors (According to Figure 8) route

carbon dist factor

route

carbon dist factor

route

carbon dist factor

route

carbon dist factor

1 2 3 4 5 6 7 8 9 10 11

0.50 0.90 0 0.051 0.017 0 0.066 0 0.0093 0.0016 0.15

12 13 14 15 16 17 18 19 20 21 22

0.0021 0.15 0 0.075 0 0 0.10 0 0.022 0.33 0.11

23 24 25 26 27 28 29 30 31 32 33

0.84 0.11 0.11 0.43 0.17 0.57 0.28 0.53 0.19 0.22 0.58

34 35 36 37

0.81 0.31 0.50 0.059

F

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

12, respectively, together with the measured values. While the general trend of the modeled data reflects that of the measurements, the predicted behavior greatly underestimates the conversion of UCS. This, combined with additional shortcomings associated with the expressions for the remaining reactions, affects all of the remaining species in the raw gas. In conclusion, additional parameters are affecting the reactivity within the gasifier, and they must be identified to allow a clear description of the reaction rates.

modeled CS as a function of increasing severity for the medium-temperature (774 °C) case and includes measurements from experiments in which the temperature and fluidization levels were varied. The normalized carbon yield is determined as the fraction of the carbon present as CS at the starting point in relation to that determined by measurement, and the severity at a certain point is proportional to this ratio. The curve that represents the normalized CS yield in the two figures describes its correlation with severity [expressed as −log(CCS/CCS,start)] and decreases from unity as the severity increases, which allows easy determination of the level of severity for any point. The obtained yields of all the species increase as long as significant levels of UCS are present in the raw gas. The point at which the SPA-tar intersects the CS curve denotes complete depletion of the UCS. After this point, the yields of benzene, naphthalene, and pyrene continue to increase with increasing severity, at the expense of the less-stable species, until they are also depleted. The derived model was applied to the measurements that describe the effects of varying the residence times, to investigate the observed deviations. The predicted evolutionary profiles for the primary and secondary UCS are shown in Figures 11 and

5. DISCUSSION The proposed reaction scheme and model provide a satisfactory description of the observed trends during the maturation of biomass-derived tar. Furthermore, while a wider range of experimental points would have been desirable, the current data are sufficient to support the proposed reaction scheme and to show how SPA-tar can be formed from measurable (albeit unknown) CS. These results are in good agreement with those of Cypres,4 who showed that the formation of aromatic tar during coal pyrolysis encompassed the formation and dehydrogenation of cyclo-olefins, as well as the decomposition of phenol. The alternative to allowing the formation of SPA-tar directly from UCS′ is shown in Figure 13, which depicts the

Figure 11. Evolutionary profiles of the primary UCS for long, medium, and short residence times, indicated as black, red, and blue lines, respectively.

Figure 13. Evolutionary profiles of phenol and toluene for high, medium, and low temperatures, indicated as black, red, and blue lines, respectively, when reaction routes 13−20 are excluded.

profiles of phenol and toluene when reaction routes 13−20 are excluded from the model. Excluding these routes from the model means that all of the mass required to form the heavier tar species must be transported through toluene. This, in turn, results in an unreasonably high reaction rate of toluene and shifts its maxima to significantly shorter residence times. Phenol is not affected to the same extent, and even if one was to increase transport through phenol, this would do little to diminish the observed behavior of toluene. Setting the maximum yield of toluene at a residence time that is shorter than those of all the measurements does not give results that are in agreement with the trends evident in Figure 9, which indicates that most of the measurements have not yet reached the maximum toluene yield. Therefore, it is concluded that a reaction scheme that does not include at least some variation of routes 13−20 will become exceedingly fast and fail to describe certain species, such as phenol and toluene. The thickness of each line in Figure 8 denotes the amount of carbon being transported via a specific route during the initial 5 s of the reaction. As shown in Figure 3, the yield of UCS is

Figure 12. Evolutionary profiles of the secondary UCS for long, medium, and short residence times, indicated as black, red, and blue lines, respectively. G

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels negligible at residence times of >5 s. Consequently, a figure that depicts the individual levels of importance of the reaction routes, for residence times in the range of 5−10 s would look considerably different, because routes 1−20 would be inactive. This would result in a scheme that describes the evolution of an already matured tar mixture.9 Throughout the present work, it has been assumed that the carbon distribution factors (summarized in Table 2) are constant, regardless of the reaction type, temperature, reactant concentration, or other process parameters. Allowing these constants to vary as a function of temperature and relevant concentrations would make the model more realistic and presumably more flexible. However, for the performed study, the impact of constant distribution coefficients is assumed to be low, given that the different cases are somewhat similar. The extended simulation indicates that, for high severity, the CS consist mainly of benzene, naphthalene, and pyrene. A similar behavior has been reported previously for benzene and naphthalene,9 whereas the values obtained for pyrene may be overestimated. No other measured species are dependent upon the destruction of pyrene to an extent that would induce a higher reaction rate in the solver. This could be remedied either by including soot formation in the reaction scheme or by observing the reactivity of the system over a wider range. However, neither of these alternatives is currently available, resulting in a less-reliable estimation of pyrene. The reported severity is determined by the total yield of CS, making it a measure of the combination of effects of operating conditions, residence time, and unknown parameters. As a result, the modeled curves in Figures 9 and 10 would look slightly different if they represented another case due to changes in the reactivities of the various species. This indicates the possibility of reducing the total yield of CS while minimizing the production of the more-stable species, such as benzene. Potential ways of accomplishing this are presumably best implemented as primary measures and include converting the various UCS into gas before they can form SPA tar. Thus, complications related to high tar levels could be partially avoided at an early stage. This would probably be beneficial, compared to implementing secondary measures to combat SPA tar, which is considerably more stable. In a comparison of the derived model with the measurements obtained from varying the residence time, the model was shown to underestimate the reactivity of the system significantly. Model deviations from the measured data are partially the result of incorrectly determined residence times and temperatures. However, additional factors, presumably related to aging of the bed material and gas-particle contacts in the freeboard, are needed to explain fully the changes in reactivity of the CS.20,21 Consequently, the proposed expressions for reactivity should include the impacts of catalytic species, such as the concentration of ash components or a measure of the catalytic surfaces in the freeboard. Interestingly, the experiments in which residence time was varied were the last to be performed in this study, which supports the notion that the observed change in reactivity relates to aging of the bed material. From the wider perspective, note that, if these effects are not taken into account, rate expressions will only be applicable to the type of system for which they were derived, provided that the system does not exhibit significant variability in the levels of catalytic species.

6. CONCLUSIONS The present work was performed to investigate the reaction routes needed to explain both the formation and decomposition of condensable species (CS) in biomass gasification. The investigation was based on measurement data for which the carbon mass balance of the system was satisfied. The main conclusions are listed below. A combination of previously suggested reaction schemes for the evolution of tar, at different stages of maturation, together with a comprehensive representation of all of the relevant CS enable a detailed investigation of tar behavior. It is shown that reaction mechanisms for low- and medium-temperature gasification require the inclusion of UCS to allow for increases in the yields of SPA-tar species. Furthermore, to attain the measured levels, a significant contribution of UCS′ that form stable aromatic species (such as benzene) directly is required, in addition to these species being formed from other SPA-tar species. This contribution is presumably attributable to species (such as cyclopentadiene) that are known to polymerize into stable aromatic compounds. The derived model was successfully implemented to describe the CS evolution for a majority of the investigated measurements. However, the derived rate expressions underestimated the reaction rate of the experimental points collected during the last day of the measurement campaign. This indicates that additional factors, such as bed material aging, gas-particle contacts, and mixing in the freeboard, can influence the reactivities of the CS. Consequently, current knowledge of the key parameters in gasification is inadequate to describe accurately the behaviors of the CS. Provided that these factors are insignificant or remain constant, a functioning model could be constructed that does not take them into consideration. Nevertheless, such a model would only be applicable to nearidentical systems, which motivates further studies of the behaviors and catalytic effects of ash components, and specifically of alkali compounds.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.energyfuels.6b00515. Appendix A: Derived relative reaction rates, as described by eq 2 (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*Tel.: +46 (0)31 772 52 54. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was performed within the Competency Center of the Svenskt Förgasningscentrum (SFC), in collaboration with Akademiska Hus, Valmet AB, E.ON AB, Göteborg Energi, and the Swedish Energy Agency.

■

REFERENCES

(1) Higman, C.; Van der Burgt, M. Gasification; Gulf Professional Publishing: Houston, TX, 2011. (2) Elliott, D. C. Relation of reaction time and temperature to chemical composition of pyrolysis oils; In Pyrolysis Oils from Biomass; H

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Soltes, E. J., Milne, T. A., Eds.; ACS Symposium Series, Vol. 376; American Chemical Society: Washington, DC, 1988; Chapter 6, pp 55−65 (DOI: 10.1021/bk-1988-0376.ch006). (3) Israelsson, M.; Larsson, A.; Thunman, H. Online measurement of elemental yields, oxygen transport, condensable compounds, and heating values in gasification systems. Energy Fuels 2014, 28 (9), 5892−5901. (4) Cypres, R. Aromatic hydrocarbons formation during coal pyrolysis. Fuel Process. Technol. 1987, 15, 1−15. (5) Font Palma, C. Model for biomass gasification including tar formation and evolution. Energy Fuels 2013, 27 (5), 2693−2702. (6) Fuentes-Cano, D.; Gómez-Barea, A.; Nilsson, S.; Ollero, P. Kinetic modeling of tar and light hydrocarbons during the thermal conversion of biomass. Energy Fuels 2016, 30, 377. (7) Corella, J.; Caballero, M. A.; Aznar, M.-P.; Brage, C. Two advanced models for the kinetics of the variation of the tar composition in its catalytic elimination in biomass gasification. Ind. Eng. Chem. Res. 2003, 42 (13), 3001−3011. (8) Gerun, L.; Paraschiv, M.; Vijeu, R.; Bellettre, J.; Tazerout, M.; Gøbel, B.; Henriksen, U. Numerical investigation of the partial oxidation in a two-stage downdraft gasifier. Fuel 2008, 87 (7), 1383− 1393. (9) Nguyen, H. N. T.; Berguerand, N.; Thunman, H. Ind. Eng. Chem. Res. 2016, DOI: 10.1021/acs.iecr.6b00650. (10) Israelsson, M.; Berdugo Vilches, T.; Thunman, H. Conversion of Condensable Hydrocarbons in a Dual Fluidized Bed Biomass Gasifier. Energy Fuels 2015, 29 (10), 6465−6475. (11) Brage, C.; Yu, Q.; Chen, G.; Sjöström, K. Use of amino phase adsorbent for biomass tar sampling and separation. Fuel 1997, 76 (2), 137−142. (12) Israelsson, M.; Seemann, M.; Thunman, H. Assessment of the Solid-Phase Adsorption method for sampling biomass-derived tar in industrial environments. Energy Fuels 2013, 27 (12), 7569−7578. (13) Osipovs, S. Sampling of benzene in tar matrices from biomass gasification using two different solid-phase sorbents. Anal. Bioanal. Chem. 2008, 391 (4), 1409−1417. (14) Neves, D. d. S. F. d. Evaluation of thermochemical biomass conversion in fluidized bed; Thesis, University of Aveiro, Aveiro, Portugal, 2013. (15) Larsson, A.; Israelsson, M.; Lind, F.; Seemann, M.; Thunman, H. Using Ilmenite To Reduce the Tar Yield in a Dual Fluidized Bed Gasification System. Energy Fuels 2014, 28 (4), 2632−2644. (16) Berdugo Vilches, T.; Marinkovic, J.; Seemann, M.; Thunman, H. Comparing Active Bed Materials in a Dual Fluidized Bed Biomass GasifierOlivine, Bauxite, Sand and Ilmenite. Energy Fuels 2016, submitted for publication (DOI: 10.1021/acs.energyfuels.6b00327). (17) Jess, A. Mechanisms and kinetics of thermal reactions of aromatic hydrocarbons from pyrolysis of solid fuels. Fuel 1996, 75 (12), 1441−1448. (18) Morf, P.; Hasler, P.; Nussbaumer, T. Mechanisms and kinetics of homogeneous secondary reactions of tar from continuous pyrolysis of wood chips. Fuel 2002, 81 (7), 843−853. (19) Abdelouahed, L.; Authier, O.; Mauviel, G.; Corriou, J.-P.; Verdier, G.; Dufour, A. Detailed modeling of biomass gasification in dual fluidized bed reactors under Aspen Plus. Energy Fuels 2012, 26 (6), 3840−3855. (20) Elliott, D. C.; Baker, E. G. The effect of catalysis on woodgasification tar composition. Biomass 1986, 9 (3), 195−203. (21) Sutton, D.; Kelleher, B.; Ross, J. R. Review of literature on catalysts for biomass gasification. Fuel Process. Technol. 2001, 73 (3), 155−173. (22) Marinkovic, J.; Thunman, H.; Knutsson, P.; Seemann, M. Characteristics of olivine as a bed material in an indirect biomass gasifier. Chem. Eng. J. 2015, 279, 555−566. (23) Davidsson, K. O.; Engvall, K.; Hagström, M.; Korsgren, J. G.; Lönn, B.; Pettersson, J. B. A surface ionization instrument for on-line measurements of alkali metal components in combustion: Instrument description and applications. Energy Fuels 2002, 16 (6), 1369−1377.

(24) Larsson, A.; Seemann, M.; Neves, D.; Thunman, H. Evaluation of Performance of Industrial-Scale Dual Fluidized Bed Gasifiers Using the Chalmers 2−4-MWth Gasifier. Energy Fuels 2013, 27, 6665. (25) Evans, R. J.; Milne, T. A. Molecular characterization of the pyrolysis of biomass. Energy Fuels 1987, 1 (2), 123−137. (26) Bruinsma, O. S.; Geertsma, R. S.; Bank, P.; Moulijn, J. A. Gas phase pyrolysis of coal-related aromatic compounds in a coiled tube flow reactor: 1. Benzene and derivatives. Fuel 1988, 67 (3), 327−333. (27) Bruinsma, O. S.; Tromp, P. J.; de Sauvage Nolting, H. J.; Moulijn, J. A. Gas phase pyrolysis of coal-related aromatic compounds in a coiled tube flow reactor: 2. Heterocyclic compounds, their benzo and dibenzo derivatives. Fuel 1988, 67 (3), 334−340. (28) Dufour, A.; Masson, E.; Girods, P.; Rogaume, Y.; Zoulalian, A. Evolution of aromatic tar composition in relation to methane and ethylene from biomass pyrolysis-gasification. Energy Fuels 2011, 25 (9), 4182−4189. (29) Yang, Y.; Mulholland, J.; Akki, U. Formation of Furans by GasPhase Reactions of Chlorophenols. Symp. (Int.) Combust., [Proc.] 1998, 27, 1761−1768.

I

DOI: 10.1021/acs.energyfuels.6b00515 Energy Fuels XXXX, XXX, XXX−XXX