A Pilot-Scale Gasifier Freeboard Equipped with Catalytic Filter Candles for Particulate Abatement and Tar Conversion: 3D-CFD Simulations and Experimental Tests

Abstract

This work deals with the catalytic steam reforming of raw syngas to increase the efficiency of coupling gasification with downstream processes (such as fuel cells and catalytic chemical syntheses) by producing high-temperature, ready-to-use syngas without cooling it for cleaning and conditioning. Such a combination is considered a key point for the future exploitation of syngas produced by steam gasification of biogenic solid fuel. The design and construction of an integrated gasification and gas conditioning system were proposed approximately 20 years ago; however, they still require further in-depth study for practical applications. A 3D model of the freeboard of a pilot-scale, fluidized bed gasification plant equipped with catalytic ceramic candles was used to investigate the optimal operating conditions for in situ syngas upgrading. The global kinetic parameters for methane and tar reforming reactions were determined experimentally. A fluidized bed gasification reactor (~5 kWth) equipped with a 45 cm long segment of a fully commercial filter candle in its freeboard was used for a series of tests at different temperatures. Using a computational fluid dynamics (CFD) description, the relevant parameters for apparent kinetic equations were obtained in the frame of a first-order reaction model to describe the steam reforming of key tar species. As a further step, a CFD model of the freeboard of a 100 kWth gasification plant, equipped with six catalytic ceramic candles, was developed in ANSYS FLUENT^®. The composition of the syngas input into the gasifier freeboard was obtained from experimental results based on the pilot-scale plant. Simulations showed tar catalytic conversions of 80% for toluene and 41% for naphthalene, still insufficient compared to the threshold limits required for operating solid oxide fuel cells (SOFCs). An overly low freeboard temperature level was identified as the bottleneck for enhancing gas catalytic conversions, so further simulations were performed by injecting an auxiliary stream of O₂/steam (50/50 wt.%) through a series of nozzles at different heights. The best simulation results were obtained when the O₂/steam stream was fed entirely at the bottom of the freeboard, achieving temperatures high enough to achieve a tar content below the safe operating conditions for SOFCs, with minimal loss of hydrogen content or LHV in the fuel gas.

1. Introduction

Interest in renewable energy sources (RESs) is steadily increasing due to concerns about climate change. The main purposes of using RESs are to limit greenhouse gas (GHG) emissions, diversify energy supply, and reduce the dependence on fossil fuel markets [1]. Of considerable interest are RESs related to the use of biomass, i.e., the biodegradable fraction of products, waste, and residues of biological origin derived from agriculture (including vegetal and animal substances), forestry, and related industries such as fishery and aquaculture, as well as the biodegradable fraction of industrial and municipal waste, as defined by European Directive 2009/28/EC [2,3].

In recent decades, interest in biomass and waste gasification has increased due to the growing attention to sustainable energy [4,5]. Gasification is the thermochemical conversion of an organic material into a valuable gaseous product, syngas, and a solid product, char [6,7], obtained by adding air/oxygen/steam to the solid fuel. Syngas contains carbon monoxide (CO), hydrogen (H₂), methane (CH₄), and carbon dioxide (CO₂), but also light hydrocarbons, such as ethane and propane, and heavier hydrocarbons, such as tar, which condense at temperatures between 250 and 300 °C. Raw syngas includes minor but significant amounts of unwanted impurities, known as contaminants. Syngas contaminants consist of tars, nitrogen-based compounds (NH₃, HCN, etc.) [8], sulfur-based compounds (H₂S, COS, etc.) [9], hydrogen halides (HCl, HF, etc.), and metals in traces (Na, K, etc.) [10]. Cleaning raw syngas is an essential step before its use in downstream applications. For this reason, research has recently paid particular attention to syngas cleaning and conditioning to reduce contaminants below tolerable limits [7,11].

The technologies for removing contaminants from raw syngas are classified according to the gas temperature exiting the clean-up device: hot (T > 300 °C), cold (T < ~100 °C), and warm gas cleaning technologies [12]. Cold gas cleaning techniques are relatively well established and highly effective, although they often generate wastewater streams and suffer from energy inefficiencies, mainly due to losses of the sensible heat of syngas and organic matter contained in the separated tar. Most of these techniques are based on the use of wet scrubbers [13]. Hot gas cleaning technologies are interesting because they avoid gas stream cooling and reheating and allow for the conversion of tar into lighter chemical compounds [14].

Among hot gas cleaning technologies, the addition of a catalyst to the gasifier bed inventory was first considered. However, due to the low activity demonstrated by iron-based catalysts (such as olivine or olivine enriched with iron), the utilization of a toxic component would have contaminated the product gas. As a consequence, a technical solution was found to avoid nickel dispersion in the product gas by retaining a small, fixed bed of catalyst inside the body of a commercial ceramic filter candle. This concept has been studied extensively experimentally and proven at the plant scale [15]. Catalytic filter candles installed in the freeboard of fluidized bed gasifiers enable high-temperature syngas cleaning by simultaneously removing particulate matter and promoting the steam reforming of tars and methane. This is achieved through the incorporation of a Ni-based catalyst within the filter structure, enabling in situ conversion without heat loss. The main advantages of this concept, unlike low-temperature gas cleaning systems, are the avoidance of tar condensation, an increase in gas yield, and an operating temperature closer to the gasification temperature [16].

As is known, this is the optimal temperature range for using a nickel catalyst. Many different catalysts of this type have been developed, studied, and applied industrially. They differ mainly in the structure, morphology, and composition of the support, as well as the percentage of nickel and its distribution within the fine particles of the structure, aimed at reducing carbon deposition [15]. These differences influence the kinetics of reforming reactions when considering the various compounds, ranging from methane to heavy cyclic hydrocarbons [17]. With reference to the standard operation for nickel catalysts, their application to the conditioning of gasification product gas is characterized by a large excess of steam, as required by gasification, in relation to the tar to be reformed, so that the temperature and concentration ratio helps avoid carbon deposition. These considerations point to the need to develop a reliable kinetic model for such specific operating conditions.

Studies on catalytic ceramic candles in terms of gas flow field, particle filtration, and tar reforming under different operating conditions have been conducted using 2D-CFD and 3D-CFD simulations [18,19] with different computing models, respectively. The numerical results confirmed the outputs of experiments carried out in the laboratory (2D-CFD) and on pilot-scale (3D-CFD) fluidized bed gasifiers with catalytic ceramic filter candles inserted into the respective reactor freeboard. In fact, these candles allow for the complete removal of the particulate content by means of their anisotropic porous filter structure; furthermore, thanks to the Ni catalyst impregnated in the filter body, they act as an efficient means to convert tar (<200 mg/Nm³) into lighter products upon exit from the gasifier [20]. To bypass the technical problems related to Ni impregnation directly on the porous body of ceramic filter candles and to make the overall process more feasible in practice, commercial ceramic candles for particulate abatement at elevated temperatures can be partially filled with catalytic pellets in their inner empty space for steam reforming of hydrocarbons [21].

This work aimed to simulate the behavior of ceramic candles filled with a catalyst and installed directly inside the freeboard of a pilot-scale gasifier.

Numerous studies have demonstrated the potential of CFD to investigate gas–solid flows involving chemical reactions. The CFD approach has proven to be a reliable and flexible tool for analyzing the impact of operating conditions, feedstock properties, and reactor geometry on gasification process performance [22,23]. Furthermore, recent developments in multi-scale CFD methodologies have enabled a more comprehensive understanding of coupled hydrodynamic and kinetic phenomena, particularly relevant in systems characterized by high levels of heterogeneity and reaction complexity [24]. These strategies bridge the gap between detailed chemical kinetics and macroscopic flow fields, allowing for an improved representation of gas–solid interactions, species transport, and reaction mechanisms. Such modeling frameworks have been applied to simulate biomass gasification under various operating conditions and reactor designs [25]. These contributions confirm the relevance of CFD modeling as both an advanced simulation framework and a powerful tool for the design and optimization of next-generation gasification systems. The integration of validated numerical models with experimental data offers a solid basis for scaling up promising configurations and improving process robustness in industrial applications.

The results of 3D simulations reported in previous studies have already demonstrated the effectiveness of the ceramic filter candle system for particulate removal [18] and tar reduction through the use of quadrilobar catalyst particles [19]. In that case, the inner empty volume of the candle was completely filled by catalytic pellets. The use of 3D simulations highlighted the need for oxygen injections into the freeboard to increase the operating temperature and improve tar conversion. Although the results showed a significant enhancement of conversion efficiency with a final tar content below 1 g/Nm³, this concentration level is still too high for some syngas applications.

In particular, tar represents a potential issue in combined heat and power (CHP) plants that integrate biomass gasification and solid oxide fuel cells (SOFCs) [26]. Such plants offer the opportunity to generate sustainable and environmentally friendly energy; however, the gas contaminants present in the syngas (tars, sulfur, alkali, and halogen metals) could limit the performance of SOFCs and reduce their lifetime [27,28]. Tar leads to carbon formation and deposition on the nickel particles in the anode of SOFCs [29]. The rate of catalytic reactions leading to coke formation is faster than that of carbon gasification and reforming reactions [30]. Based on the rate of reactions reported in the BLAZE research project results [31], aromatic compounds can be classified into fast- and slow-converting tars, represented by toluene and naphthalene, respectively. In particular, the so-called slow tars tend to accumulate easily in the anode of SOFCs, decreasing the conversion of other hydrocarbons and damaging the microstructure of the anode [32]. The tolerable limits for toluene and naphthalene obtained in the BLAZE project highlight the need to further reduce the tar content with respect to the results reported in the work of Savuto et al. [19].

Unlike the previous work, this study used the results of preliminary gasification tests with a pilot-scale plant to set up the simulation. Additionally, to enhance tar conversion, a different catalyst (pelletized small cylinder, h = d = 3 mm) was considered for the candles. In this case, to reduce the pressure drop through the catalytic bed, the inner empty space of the candle was only partially filled in an annular region, leaving a central empty hole for the cleaned gas flux. A bench-scale gasifier was utilized to conduct tests under various operating conditions and derive the kinetic parameters of the reactions for the new catalyst and operating conditions, which were then incorporated into the model. An axisymmetric CFD model of the freeboard region of the bench-scale gasifier was developed using ANSYS FLUENT^® v18.2 software. Syngas compositions obtained from tests without a catalyst were utilized as input data for the freeboard region, while the kinetic constants were determined by comparing the simulation results with experimental data obtained at a lab scale, using a candle filled with the catalyst and installed in the reactor’s freeboard. In particular, all relevant chemical reactions characterizing the behavior of the catalytic filter candle were identified (Section 2.2.3). A first-order apparent kinetic model was assumed for these reactions. Utilizing experimental data on input/output concentrations at different temperatures around the filter candle, the optimal set of numerical values for activation energies and pre-exponential factors was obtained through the 2D-CFD model of the bench-scale freeboard gasifier. In this study, three different temperatures were considered, specifically 1023, 1073, and 1123 K, included within the typical operating range for circulating and more conventional fluidized bed steam gasification [15]. As mentioned in the literature and confirmed by a lengthy series of experimental studies conducted by this research group, nickel catalysts offer sufficient conversion efficiency and are characterized by a relatively low cost [33,34].

Upon completion of the kinetic study, the freeboard of the 100 kWth fluidized bed gasifier was simulated using FLUENT^® software to identify the optimal operating conditions and configurations for achieving syngas with a tar content within the limits acceptable for syngas utilization, as mentioned above.

2. Materials and Methods

2.1. Gasification Setup

2.1.1. Bench-Scale Apparatus

Figure 1 shows the bench-scale apparatus used for the gasification tests aimed at determining the catalyst kinetic parameters. A detailed description of the system and test procedure is reported in a previous work [35]. It consisted of a fluidized bed reactor (ID = 0.10 m) heated by an electric oven. The reactor was equipped with a filter candle filled with commercial nickel catalyst pellets, inserted into the freeboard of the gasifier to allow for hot gas conditioning and cleaning. The filter candle was provided by PALL Schumacher GmbH (Mengen, Germany). Notably, the filter candle used in the bench-scale apparatus was a segment of reduced-length commercial candle inserted into the freeboard of the pilot-scale gasifier. More specifically, the candle’s full length was 1.5 m, while the segment inserted into the bench-scale gasifier was 0.44 m. OD (60 mm) and ID (40 mm) were the same, so that the cross-section of the region occupied by the catalyst remained equal to allow the use of the same catalyst in both systems, both in nature (Ni catalyst) and size (cylindrical pellets with a height and diameter of 3 mm).

The length of the candle was chosen to keep the filtration velocity similar (almost equal) in both the lab- and pilot-scale equipment. As a result, the conditions adopted for steam reforming reactions of the raw syngas produced by each gasifier were fully comparable, given the corresponding temperatures and syngas compositions.

The main operating conditions are summarized in

Table 1. Summary of the gasification test operating conditions.

Test	#1	#2	#3
Temperature (K)	1023	1073	1123
Biomass flow rate (g/min)	11	11	11
Steam/biomass	0.52	0.52	0.52

.

The chosen temperatures enable the determination of kinetic parameters within a temperature range commonly used for biomass steam gasification.

2.1.2. Pilot-Scale Apparatus

A complete description of the plant is reported in a previous work [14]. It consisted of a dual fluidized bed reactor, which integrates an internal cylindrical combustor (OD = 0.22 m) fed with air, surrounded by a coaxial gasification reactor (OD = 0.44 m) fed with steam. This configuration was designed to process up to 20 kg/h of biomass, corresponding to a thermal capacity of 100 kWth.

The freeboard section of the gasification chamber (with an overall height of 2.1 m) held six cylindrical filtration candles made of alumina (OD = 0.06 m, ID = 0.04 m; length = 1.5 m) suspended from the top. As for the bench-scale system, the ceramic candles were partially filled with Ni-based catalyst pellets for in situ cleaning and conditioning of the syngas [36].

2.2. Development of the Models

2.2.1. 2D-CFD Model of the Bench-Scale Reactor Freeboard

The results of the gasification tests were used to tune the kinetic parameters of the reactions involved in the process, covering a wide range of species concentrations and temperature values, as determined by different gasification conditions. For this reason, an axisymmetric 2D control volume replicating the bench-scale freeboard was defined in a previous study and utilized in this work. A complete description of this model is reported in [19]. By means of ANSYS-FLUENT^® (ANSYS Inc., Canonsburg, PA, USA), the steam reforming reactions described in Section 2.2.3 were implemented and simulated to reproduce the experimentally obtained flow rates and gas compositions. The control volume and its associated mesh are shown in Figure 2. ANSYS-FLUENT^® version 18.2 was used.

As a boundary condition, the wall temperature in the model was varied to obtain the measured operating temperatures in the catalytic candle (1023, 1073, and 1123 K, as reported in Table 1).

An additional boundary condition was the inlet superficial velocity of the gas inside the simulated volume; this value, set equal to 0.27 m/s, implies a filtration velocity of the gas through the candle of 90 m/h and a residence time in the catalytic zone of 0.33 s.

The filtration velocity adopted in the experimental tests and included in the CFD model corresponded to typical values recommended by the manufacturer of the catalytic candles [37].

The input composition of the syngas (

Table 2. Input composition of the syngas obtained from tests using a noncatalytic candle in the bench-scale reactor.

	Mass Fraction
H2	1.30 × 10−2
CO	1.30 × 10−1
CO2	1.49 × 10−1
CH4	2.31 × 10−2
H2O	1.15 × 10−1
C6H6	2.87 × 10−3
C7H8	1.50 × 10−3
C10H8	1.29 × 10−3

) was derived from previous tests carried out on the same gasification test rig with a noncatalytic candle in the reactor [21].

2.2.2. 3D-CFD Model of the Pilot Plant Freeboard

The final objective of this work was the simulation of the freeboard region of the gasification section of a pilot-scale dual bubbling fluidized bed reactor.

Figure 3 shows the 3D model of the freeboard region used for the simulations. A more accurate description is reported in a previous work [19].

The mass and momentum balances, along with the chemical species and energy conservation equations, were solved for the gas phase in the freeboard volume, in the solid region of the filter, and in the catalytic filter bed inside the candle.

A second-order upwind method was used to solve the mass and momentum balances and chemical species equations. Meanwhile, a third-order method was used for the energy equation, with a residual value set to 1 × 10⁻⁷. Regarding the balances on benzene, toluene, and naphthalene, the residual values were set to 1 × 10⁻⁶.

A summary of the governing equations applied for the three simulated regions is reported in

Table 3. Summary of key governing equations used in the 3D model.

Freeboard Region	Equation
$\nabla ·\left(\rho \overrightarrow{v}\right)=0$	(1)
$\nabla ·\left(\rho \overrightarrow{v}\overrightarrow{v}\right)=-\nabla p+\nabla ·\left(\stackrel{̿}{{\tau }_g}\right)$	(2)
$\nabla ·\left(\overrightarrow{v}\left(\rho h\right)\right)=\nabla ·\left[\left(\stackrel{̿}{{\tau }_g}·\overrightarrow{v}\right)+k_g\nabla T\right]+S_h$	(3)
$\nabla ·\left(\overrightarrow{v}\rho y_i\right)=\nabla ·\left(D_i\nabla y_i\right)+S_i$	(4)
Porous structure of the filter candle and bed of catalyst pellets
$\nabla ·\left({\epsilon }_c\rho \overrightarrow{v}\overrightarrow{v}\right)=-{\epsilon }_c\nabla p+{\epsilon }_c\nabla ·\left(\stackrel{̿}{{\tau }_g}\right)-\left({\displaystyle \frac{\mu }{\alpha }}+{\displaystyle \frac{C\rho }{2}}\left|\overrightarrow{v}\right|\right)\overrightarrow{v}$	(5)
$\nabla ·\left({\epsilon }_c\overrightarrow{v}\left(\rho h\right)\right)=\nabla ·\left[k_{eff}\nabla T\right]+{\displaystyle \sum _j}{r}_j{\rho }_{cat}\left(1-{\epsilon }_c\right)\Delta H_{j0}$	(6)
$\nabla ·\left({\epsilon }_c\overrightarrow{v}\rho y_i\right)=\nabla ·\left({\epsilon }_c\rho D_i\nabla y_i\right)+{\displaystyle \sum _j}{v}_{ij}{r}_j{\rho }_{cat}\left(1-{\epsilon }_c\right)M_i$	(7)
Insulation region (only the energy equation was solved for this zone)
$\nabla ·\left(k_{ins}\nabla T\right)=0$	(8)

where $i$ and $j$ indicate each of the chemical species and the reaction, respectively; r_j is the reaction velocity, set to zero in the porous structure of the candle wall (where reactions can be considered negligible) and is different from zero only in the catalyst bed.

[38,39,40].

The equation used to calculate the source (sink) terms $S_i$ and $S_h$, the effective thermal conductivity $\left(k_{eff}\right)$, the mass diffusivity ($D_i)$, and the boundary condition to solve the energy equation for the insulation region are reported in the Supplementary Materials.

The catalyst and ceramic candles were modeled using the chemical–physical properties of alumina (density = 3939 kg/m³; specific heat ($c_p$) and thermal conductivity ($\lambda$) calculated according to Equations (9) and (10)). The piecewise polynomial expression was developed according to NIST property data for advanced materials [41].

$$c_p \left[{\displaystyle \raisebox{1ex}{$J$}\!\left/ \!\raisebox{-1ex}{$kg·K$}\right.}\right]=731.04+1.2119·T-0.0007·T^2$$

(9)

$$\lambda \left[{\displaystyle \raisebox{1ex}{$W$}\!\left/ \!\raisebox{-1ex}{$m·K$}\right.}\right]=34.274-0.0644·T+0.00004·T^2$$

(10)

Moreover, both the catalyst and the ceramic candles were defined by their porosity and permeability. For the latter, the value derived in the work of Rapagnà et al. was used [42]. According to the manufacturer’s data, the ceramic candle porosity (${\epsilon }_{filter candle}$) was set to 0.3 [37], while for the catalyst, a porosity (${\epsilon }_{cat}$) equal to 0.7 was measured.

It should be stressed that in the bench-scale model, a segment of reduced length of full-scale ceramic candle modeled for the pilot-scale gasifier was used. The catalyst and its assembly were the same with corresponding raw syngas face velocity and gas hourly space velocity.

As a result, the same material properties were used for the bench- and pilot-scale models.

Moreover, the reactor insulation, made using an Insulfrax LT 128 ceramic blanket (thickness equal to 0.2 m, Unifrax, Tonawanda, NY, USA), was modeled as a solid region characterized by its thermal conductivity. The average thermal conductivity value $\left(k_{ins}\right)$ of Insulfrax LT 128 ceramic blankets [43] was used to obtain the expression as a function of temperature (Equation (11)).

$$k_{ins} \left[{\displaystyle \raisebox{1ex}{$W$}\!\left/ \!\raisebox{-1ex}{$m·K$}\right.}\right]=2.05\times {10}^{-7}{T}^2+9.11\times {10}^{-5}T+5.26\times {10}^{-2}$$

(11)

In addition, the properties of the syngas were included in the model as follows:

• The specific heat was computed using a weighted average of the specific heat of its components.
• The thermal conductivity and viscosity were expressed as polynomial functions of temperature.

The 3D-CFD numerical model was implemented in ANSYS-FLUENT^® software using the kinetic expressions derived from simulations of the experimental tests using the bench-scale reactor. As outlined in Section 2.2.1, the kinetic expressions encompass a range of concentrations and temperatures resulting from different gasification conditions. This ensured the reliability of the predictions for the pilot-scale gasifier.

2.2.3. Catalyst Bed Chemical Reaction Rates

The reactions considered for both models developed in ANSYS-FLUENT^® software were water gas shift (R1) and the methane (R2) and tar (R3, R4, R5) steam reforming taking place in the catalyst bed.

$$\mathrm{R}1:\mathrm{C}\mathrm{O}+{\mathrm{H}}_2\mathrm{O}\rightleftarrows {\mathrm{H}}_2+{\mathrm{C}\mathrm{O}}_2\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}$$

$${\mathrm{R}2:\mathrm{C}\mathrm{H}}_4+{\mathrm{H}}_2\mathrm{O}\rightleftarrows 3{\mathrm{H}}_2+\mathrm{C}\mathrm{O}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}$$

$$\mathrm{R}3:{\mathrm{C}}_6{\mathrm{H}}_6+6{\mathrm{H}}_2\mathrm{O}\rightleftarrows 9{\mathrm{H}}_2+6\mathrm{C}\mathrm{O}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}$$

$${\mathrm{R}4:\mathrm{C}}_7{\mathrm{H}}_8+7{\mathrm{H}}_2\mathrm{O}\rightleftarrows 11{\mathrm{H}}_2+7\mathrm{C}\mathrm{O}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}$$

$${\mathrm{R}5:\mathrm{C}}_{10}{\mathrm{H}}_8+10{\mathrm{H}}_2\mathrm{O}\rightleftarrows 14{\mathrm{H}}_2+10\mathrm{C}\mathrm{O}$$

For the water gas shift reaction (R1) and methane steam reforming (R2), the model of Numaguchi and Kikuchi [44] with the kinetic parameters of de Smet [45] and the equilibrium constants of Hou and Huges [46] was used as the reference kinetic model.

$$r_{R1}=k_{R1}{\displaystyle \frac{(p_{CO}{p}_{H2O}-p_{C{O}_2}{p_{H_2}}^3/K_{eq,R1})}{p_{H_{2}O}}}$$

(12)

$$r_{R2}=k_{R2}{\displaystyle \frac{(p_{C{H}_4}{p}_{H2O}-p_{CO}{p_{H_2}}^3/K_{eq,R2})}{{p_{H_{2}O}}^{1.596}}}$$

(13)

The kinetics chosen for tar steam reforming reactions was an apparent first order, as follows:

$$r_{R3}=k_{R3}{C}_{C_6{H}_6}$$

(14)

$$r_{R4}=k_{R4}{C}_{C_7{H}_8}$$

(15)

$$r_{R5}=k_{R5}{C}_{C_{10}{H}_8}$$

(16)

For Equations (14)–(16), similarly to what has been shown to happen in the literature for reaction R1 and R2, the kinetic parameter k is expressed as a function of temperature by means of the Arrhenius dependency.

$$k_{Ri}={ A}_{Ri}{e}^{-{\displaystyle \raisebox{1ex}{$E_{Ri}$}\!\left/ \!\raisebox{-1ex}{$RT$}\right.}}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}i=1,2,3,4,5$$

(17)

2.2.4. Combustion Reactions and Modeling

As is well known, tar steam reforming reaction kinetics are accelerated by increasing temperature, as long as the threshold limit for catalyst deactivation is not approached. Previous work on the same model has shown that the injection of a small amount of oxygen in the freeboard allows heating the syngas before it enters the candle through the combustion of a slight fraction of the syngas itself. The same solution is also simulated in this work by considering the following reactions:

$$\mathrm{R}6: {\mathrm{C}\mathrm{H}}_4+1.5{\mathrm{O}}_2\rightleftarrows 3{\mathrm{H}}_2\mathrm{O}+\mathrm{C}\mathrm{O}$$

$$\mathrm{R}7:\mathrm{C}\mathrm{O}+0.5{\mathrm{O}}_2\rightleftarrows {\mathrm{C}\mathrm{O}}_2\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}$$

$${\mathrm{R}8:\mathrm{H}}_2+0.5{\mathrm{O}}_2\rightleftarrows {\mathrm{H}}_2\mathrm{O}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}$$

Combustion reactions of heavy hydrocarbons were neglected in this study due to their low concentration. Moreover, despite the potential for temperatures to exceed 1000 °C due to combustion in the freeboard, the naturally localized temperature peaks within the fluid are considered to have little effect on tar cracking. Therefore, these reactions were also neglected. In any case, these assumptions lead to more conservative model results.

To simulate syngas combustion, a finite rate/eddy dissipation model was used. Most fuels exhibit rapid combustion, where the reaction rate is controlled by both Arrhenius-type kinetics (finite rate) and turbulent mixing. In non-premixed flames, such as those analyzed in this study, turbulence gradually blends the fuel and oxidizer into reaction zones where combustion proceeds rapidly. The eddy dissipation model proposed by Magnussen and Hjertager [47] was utilized to evaluate the influence of turbulent mixing.

In the adopted finite-rate/eddy dissipation model, calculations were performed for both the Arrhenius-type finite rate (r_j,KIN), and the eddy dissipation reaction rate (r_j,EDM). The effective reaction rate r_j was determined as the minimum of r_j,KIN and r_j,EDM.

The kinetic model developed by Dryer and Westbrook [48] was adopted for CH₄ and CO combustion, and that of Marinov et al. [49] for H₂ combustion.

The results obtained using these kinetic models are in good agreement with those obtained using a more detailed reaction mechanism under hot and dilute oxidation conditions [50].

Table 4. Combustion mechanisms and kinetic rate data *.

	A (1/s)	β	E/R (K)	Reaction Orders
R6 [48]	5.03 × 1011	0	24,056	${\left[{\mathbf{C}\mathbf{H}}_{\mathbf{4}}\right]}^{\mathbf{0.7}}{\left[{\mathbf{O}}_{\mathbf{2}}\right]}^{\mathbf{0.8}}$
R7 [48]	2.24 × 1012	0	20,484	$\left[\mathbf{C}\mathbf{O}\right]{\left[{\mathbf{O}}_{\mathbf{2}}\right]}^{\mathbf{0.25}}{\left[{\mathbf{H}}_{\mathbf{2}}\mathbf{O}\right]}^{\mathbf{0.5}}$
R8 [49]	5.69 × 1011	0	17,609	$\left[{\mathbf{H}}_{\mathbf{2}}\right]{\left[{\mathbf{O}}_{\mathbf{2}}\right]}^{\mathbf{0.5}}$

* Units in kmol, m³, K, s.

shows the parameter values used in the numerical simulations.

The Arrhenius-type finite rate of the combustion reactions has the following expression:

$$\begin{gathered} r_{j,KIN}={ T}^{\beta j}{A}_j{e}^{-Ej/RT}\prod _i{c}_i^{\eta i} \\ i=H_2, CO, C{O}_2,H_{2}O,O_2 \\ j=reactions R6-R8 \end{gathered}$$

(18)

where β_j and ηi are empirical constants; A_j and E_j are the pre-exponential factor and activation energy of the reaction j, respectively; and c_k is the concentration of the species k.

Given that the flow rate of injected O₂ was extremely low, according to the eddy dissipation model (EDM) of Magnussen and Hjertager, the rate of reaction j was given by

$$r_{j,EDM}={ C}_{EDM}\rho {\displaystyle \frac{\epsilon }{k}}\left({\displaystyle \frac{y_{O2}}{{\nu }_{O2j}{M}_{w,O2}}}\right)$$

(19)

where C_EDM is an empirical constant equal to 4.0.

As already mentioned, the minimum between r_j,KIN and r_j,EDM was chosen as the net reaction rate r_j.

Due to the potential for the temperature to exceed 1000 °C during combustion, even in localized hot spots, radiation must be accounted for. Therefore, a discrete transfer radiation model (DRTM) was selected, as detailed by Carvalho et al. [51]. Ilbas [52] demonstrated that this model yields satisfactory results for the non-premixed combustion of a composite fuel (CH₄ and H₂) when compared with experimental values, particularly for geometries similar to those examined here and using the same turbulence model (k-ε) adopted in this study. As suggested by Ilbas, the absorption coefficient for the composite fuel in the radiation model was set at 0.50 m⁻¹, while the scattering coefficient used was 0.01 m⁻¹.

3. Results and Discussion

3.1. Gasification Tests

Table 5. Summary of the main gasification test results and data from the work of Savuto et al. [21].

Test	#1	#2	#3	[21]
Temperature (K)	1023	1073	1123	1033
Catalyst	YES	YES	YES	NO
S/B	0.52	0.51	0.52	0.52
Dry gas yield (Nm3/kg)	1.80 ± 0.06	1.82 ± 0.08	1.84 ± 0.07	1.10
H2 (vol.% dry)	55.7 ± 2.4%	55.6 ± 1.6%	55.7 ± 2.0%	40.6%
CO (vol.% dry)	26.1 ± 0.9%	28.2 ± 1.1%	31.6 ± 1.2%	29.2%
CO2 (vol.% dry)	16.4 ± 0.5%	14.3 ± 0.7%	11.1 ± 0.4%	21.2%
CH4 (vol.% dry)	1.9 ± 0.1%	1.9 ± 0.1%	1.7 ± 0.1%	9.0%
H2O conv (%)	40.8 ± 0.4%	43.7 ± 0.4%	56.2 ± 0.6%	24.0%
Carbon conv (%)	83.8%	86.1%	89.7%	-
Tar content (mg/Nm3)	2102 ± 10.1	986 ± 9.8	342 ± 6.2	3276
Benzene (mg/Nm3)	2301 ± 17.7	1384 ± 19.7	679 ± 17.1	2439
Toluene (mg/Nm3)	765 ± 5.8	291 ± 3.2	73 ± 1.3	1711
Naphthalene (mg/Nm3)	661 ± 5.3	338 ± 9.3	162 ± 8.6	1095

reports a summary of the results of the experimental gasification tests of this work. Moreover, data obtained by Savuto et al. [21] without a catalytic candle in the freeboard are also included in Table 5 and used as input for the composition of the raw syngas entering the freeboard.

The data presented in Table 5 indicate that the presence of the catalyst significantly enhanced the gasification performance and altered the composition of the syngas due to improved water gas shift and steam reforming reactions. Tests #1, #2, and #3, conducted with a catalyst, showed higher dry gas yields (1.45, 1.49, and 1.57 Nm³/kg, respectively) compared to the test without a catalyst (1.10 Nm³/kg). The catalyst also increased the hydrogen content, maintaining it around 55.7% (vol.% dry) across different temperatures.

The carbon monoxide content increased with temperature in the presence of the catalyst, reaching up to 31.6% in test #3. Conversely, the carbon dioxide content decreased with temperature in the catalyzed tests, from 16.4% to 11.1%, while the test without the catalyst recorded a higher CO₂ content of 21.2%. Methane reforming resulted in a relatively low concentration in the presence of the catalyst, around 1.7–1.9%.

Water and carbon conversion rates were markedly improved with the catalyst (up to 56.2% and 89.7%, respectively). Additionally, the presence of the catalyst greatly reduced the benzene, toluene, and naphthalene concentrations, which were substantially higher in the uncatalyzed test.

Temperature variations further influenced the results. As the temperature in the catalyzed tests increased from 1023 K to 1123 K, there was a clear trend of increasing dry gas yield and H₂ content, along with decreasing tar and hydrocarbon contents. This demonstrates that higher temperatures, coupled with a catalyst, optimize the gasification process, enhancing syngas quality and reducing undesirable by-products.

3.2. 2D-CFD Model

The key chemical compounds and the relevant reactions in the system under examination were selected. By means of a trial-and-error procedure, the simulation resulted in the following choice for the overall kinetic model to assure good agreement with the experimental results.

As far as the water gas shift (reaction R1), the literature kinetic model and the values of constants were kept equal to those proposed in the literature. The model mentioned above was adopted [44,45].

For methane reforming (reaction R2), only the value of k_R2 was modified to agree with experimental data, as reported in Equation (13).

Finally, for the steam reforming of key tar components under conditions of very low concentration and very large steam excess, an apparent first-order kinetic model was adopted (reactions R3 to R5), and the respective constants are reported in Equations (14) through (16). It is worth mentioning that in all systems considered in this work, the concentration of H₂S was always very low (of the order of 50 ppm or less), so that the influence of sulfur on the kinetic results was negligible [53,54].

According to Equation (17), plotting the logarithm of k as a function of 1/RT showed the trends reported in Figure 4 and Figure 5.

From the data in Figure 4 and Figure 5, the following functions were derived, allowing the calculation of the new values of the pre-exponential factor and activation energy:

$$ln{k}_{R2}=-98756 x+0.44$$

(20)

$$ln{k}_{R3}=-184125 x+24.41$$

(21)

$$ln{k}_{R4}=-224926 x+29.86$$

(22)

$$ln{k}_{R5}=-147180 x+20.63$$

(23)

where $x={\displaystyle \frac{1}{RT}}$.

From the above functions, it was possible to calculate the averaged new values of the pre-exponential factor (A) and activation energy (E) for each reaction.

Inserting the new parameter values into the corresponding kinetic relations allowed us to conduct all simulations at different operating temperatures.

It is worth stressing that the apparent kinetic model was developed using experimental syngas compositions obtained at three different gasification temperatures and dense bed raw product gas in the freeboard without a catalytic filter candle. The last data were obtained at the lowest temperature level investigated (750 °C), where the tar content in the raw syngas was certainly the highest. As a result, the conditions adopted to define the kinetic constants were somewhat conservative.

Figure 6 compares the curves as functions of temperature, obtained by FLUENT^® numerical integration using the inputs in

Table 6. Averaged new values of the pre-exponential factor (A) and activation energy (E) for each reaction.

Reaction	A (s−1)	E (J/kmol)
R2	1.55	9.88 × 104
R3	3.99 × 10	1.84 × 105
R4	9.32 × 1012	2.25 × 105
R5	9.06 × 108	1.47 × 105

, with the corresponding tar content obtained experimentally.

The graph below highlights that the simulation results are in good agreement with the relative experimental data (with a maximum percentage difference of 3%). Therefore, the kinetic model can be considered validated and completely defined for use in 3D simulations of the pilot gasifier.

3.3. 3D-CFD Simulation Results of the Pilot Gasifier Freeboard

The 3D-CFD model was then used to simulate tar conversion in the freeboard of the 100 kWth gasifier with six commercial candles filled with the catalyst. Based on the considerations outlined above (see Section 2.1 and Section 2.2.2), the kinetic expressions derived from the experimental results gathered using the bench-scale reactor were fully applicable to the pilot-scale reactor due to the complete similarity of the operating conditions. Specifically, the dependence of the catalytic process within the candles on chemical species concentrations and temperature was expected to be consistent across both reactors.

The mesh used for the simulation is shown in Figure 7 and was composed of 500,000 hexahedral cells. A sensitivity analysis performed by changing the number of cells (halving the characteristic size) did not significantly affect the results (differences were less than 2.5%). For this reason, the proposed mesh was considered to ensure the accuracy of results with a reasonable calculation time.

The input data for raw syngas entering the freeboard section of the gasifier used for the simulations were taken from the results of preliminary tests on the pilot-scale dual fluidized bed gasifier without the candles installed. The tests were carried out using a steam-to-biomass ratio equal to 0.6 and olivine as the bed material.

The tar species were grouped in lumps, marking the one-ring hydrocarbons as toluene and the two-ring tars as naphthalene; the most significant components in the respective category are shown in

Table 7. Input data for the CFD simulations without air injections.

Input Data
	Gasifier	Combustor
Gas flow rate (kg/h)	31	60
H2 (% vol. wet)	29
CO (% vol. wet)	19
CO2 (% vol. wet)	16	14
CH4 (% vol. wet)	8
H2O (% vol. wet)	21	14
N2 (% vol. wet)	7	70
C6H6 (g/Nm3)	5.3
C7H8 (+1-ring) (g/Nm3)	5.7
C10H8 (+2-ring) (g/Nm3)	4.6
T inlet (°C)	850	950

.

To verify the reliability of the model, a preliminary simulation was carried out assuming adiabatic conditions for the gasification reactor wall and no reactions. As expected under these conditions, the gas combustor temperature decreased, while the gas temperature of the gasifier increased. The temperature of the outlet gas from the combustor decreased by approximately 25 °C, while the gas outlet temperature of the gasifier increased by approximately 29 °C. The enthalpy balances of these two gas streams showed that the combustor lost ~450 W, which is exactly the sensible heat that was gained by the gas inside the gasifier. After this simulation, the boundary conditions were reset to those of interest, and the reactions were re-entered into the model.

Figure 8 shows the pressure drop due to the candles in the freeboard of the fluidized bed. As expected, the pressure drop was in the order of 40 mbar. This confirms that the model is also able to predict the pressure drop through the candles.

Figure 9 shows the temperature distribution inside the reactor as obtained by the simulation, while in Figure 10, the average temperature distribution at the surface of the candle versus its length is reported.

As expected, a temperature decrease occurred due to heat dispersion through the reactor wall and heat adsorption by endothermic steam reforming reactions.

In fact, from the mass balances, it was possible to determine the extent of the reactions. Knowing the specific enthalpy change associated with each individual reaction, it was possible to determine the contribution of the reactions, mainly endothermic, to the decrease in temperature. Finally, from the enthalpy balance, the contribution of thermal losses was quantified. The latter contributed to approximately half of the decrease in temperature.

This implies that the reaction temperature at different heights of the candle was not high enough to guarantee an optimal conversion of tar [17]. As shown in Figure 10, the temperature decreased from 1092 K (819 °C) to less than 1010 K (737 °C). Therefore, even with the inclusion of very thick insulation, the thermal dispersions were still too high, making it impossible to maintain a freeboard temperature close to that of the bed (850 °C). This effect is obviously much more pronounced for small-scale systems, like that studied in this work.

The gas composition obtained at the outlet of the gasifier is shown in

Table 8. Results of the 3D-CFD simulation for the input data from Table 7.

Output Data
H2 (% vol. wet)	31
CO (% vol. wet)	19
CO2 (% vol. wet)	16
CH4 (% vol. wet)	7
N2 (% vol. wet)	7
H2O (% vol. wet)	20
C6H6 (g/Nm3)	5.0
C7H8 (+1-ring) (g/Nm3)	1.17
C10H8 (+2-ring) (g/Nm3)	3.24

.

As far as the consistency of the model with pilot-scale results, a previous paper [19] presented experimental data showing that when a slipstream of raw syngas produced by the pilot gasifier was treated by a catalytic candle, a total tar concentration close to the values predicted by the model at 750 °C in the freeboard was obtained.

The tar conversion obtained in this simulation was higher than 80% for toluene (one ring); however, it was only 41% for naphthalene (two rings). Comparing the results with the acceptable concentration limits, naphthalene and toluene were still found to be too high; acceptable concentration values should be lower than 75 mg/Nm³ and 750 mg/Nm³ for naphthalene and toluene, respectively [32]. As previously discussed, this low conversion is likely related to the temperature drop along the candle, specifically the substantial temperature drop predicted near the reactor exit. It is thus confirmed that controlling temperature and reducing thermal losses are crucial issues in chemical reactor design for tar removal.

The reliability of the computational model to describe well-known experimental trends was further investigated by blowing a small flow rate of a mixture of steam and O₂ (50/50 weight%) into the freeboard of the reactor [55]. This allowed for the burning of a little of the fuel gas and increased the temperature enough to promote tar reforming reactions. A distributed oxygen feeding system was implemented in the model of the gasifier freeboard, made of four series of nozzles uniformly positioned around the reactor circular wall, each made of six nozzles located at different heights in the simulated volume (0.4 m, 0.8 m, 1.2 m, and 1.6 m from the bottom of the reactor freeboard). Figure 11 shows the volume simulated with the cylindrical nozzles for oxygen injection (the cylinder of the combustor and the insulation are omitted for clarity).

In this case, the stream of steam and oxygen was distributed evenly to the four circumferences, containing six nozzles each. Simulations were carried out with different O₂ flows to investigate the best solution in terms of tar conversion and gas composition. In particular, O₂ varied between 0.25 and 0.50 kg/h, which corresponds to 2% and 4% of the oxygen fed into the combustor. The results are shown in Figure 12 and Figure 13.

Figure 12 shows that the dry gas composition did not change dramatically. As expected, there was a decrease in hydrogen from 39% to 37–38% without and with O₂ injection, respectively. At the same time, CO increased from 24% to 26%.

Figure 13 shows, instead, that the tar conversion was enhanced thanks to O₂/steam injections. The toluene concentration was always lower than 750 mg/Nm³, and in particular, it was always lower than 100 mg/Nm³. Naphthalene conversion was also improved thanks to the O₂/steam injections, and its concentration was of the order of 100–200 mg/Nm³ and was really close to the fixed limit of 75 mg/Nm³ for 0.5 kg/h of oxygen injected. These results are obviously due to the combustion of a small portion of the fuel gas and a consequent increase in the reactor temperature. The next figure shows the average surface temperature of the candle at different candle heights for the case of 0.5 kg/h of oxygen injected.

As shown in Figure 14, the temperature on the external surface of the candle increased, and this explains the higher conversion of tar. In particular, the temperature varied between 1100 and 1260 K (827–987 °C) at the bottom and top of the candle, respectively. The temperature gradient was around 160 K, and the candle temperature in some parts also reached a very high temperature (>950 °C). These two effects could be an issue for the correct operation of the candle, which could be damaged; in fact, the operation temperature should not exceed 1223 K (950 °C). Furthermore, even if the candles can operate at high temperatures, being of ceramic material, they can suffer from thermal stresses induced by the temperature gradients.

For these reasons, new simulations were carried out, varying the distribution of oxygen injected at different heights. It is evident that most of the fuel gas had already percolated through the candle in the upper zone of the reactor, and it did not make sense to inject the same flow rate of oxygen as fed into the bottom of the reactor. The best distribution should be one assuring a temperature high enough to obtain an optimal tar conversion, preferably constant throughout the length of the candle. This can be achieved by increasing the oxygen injected into the bottom of the reactor and reducing that injected into the top of the reactor. In the new simulation runs, the total oxygen flowrate was always varied between 0.25 and 0.5 kg/h; however, 75% of this flow was injected into the nozzles located at 0.4 m from the bottom of the reactor (see Figure 11), while the remaining 25% was injected in the nozzles located at 0.8 m from the bottom of the reactor. No O₂ was injected into the other nozzles at 1.2 and 1.6 m (close to the top of the reactor). The results of these simulations are reported in Figure 15 and Figure 16.

Figure 15 shows that the gas composition was similar, as expected, to that obtained in the previous simulations. Additionally, in this case, there was a slight decrease in hydrogen from 39% to 38% and a slight increase in CO from 23% to 26%.

This new configuration further improved the tar conversion (Figure 16). In fact, in this case, the tar concentration was always lower than in the previous case at every O₂ flow rate (see Figure 13 for a comparison). Furthermore, the naphthalene concentration was lower than the fixed limit (75 mg/Nm³), even for a total oxygen flow rate of 0.38 kg/h.

In Figure 17, the average surface temperature distribution along the length of the candle is reported.

In this case, the temperature gradient was around 90 K, lower than in the previous case. Furthermore, the temperature was higher than 1123 K (850 °C) for most of the candle length, although always lower than the tolerable limit of 1223 K (950 °C).

Thus, by feeding the mixture of oxygen and steam only into the bottom part of the freeboard, better results were achieved in the simulations, both in terms of the temperature profile and tar conversion.

The injection of oxygen resulted in a decrease in the total outlet gas flow rate (Figure 18) and in the lower heating value (LHV) of the produced gas.

Similar observations hold for the LHVs of the gas leaving the freeboard (Figure 19).

The results show that feeding the oxygen/steam mixture with the second configuration produced a gas with a lower hydrogen content and a higher CO₂ concentration, which is also reflected in the lower calorific value of the resulting syngas. These observations indicate slightly reduced combustion intensity. Indeed, the average temperature was slightly lower when the injections were not distributed.

However, with the uneven distribution of the injections, the temperature increased more significantly in the lower regions of the candle, reaching an average temperature of 1123 K, leading to improved catalytic tar reforming.

In conclusion, the model proved to be highly reliable in capturing subtle variations in the results, making it particularly suitable for use during the design phase of industrial equipment. It can effectively determine optimal operating conditions from the perspective of tar reforming.

Finally, it is worth noting that under the optimum operating conditions, the temperatures reached by the catalytic candle fell within the same temperature range used for evaluating the kinetic parameters. This eliminated the need to extrapolate the chemical kinetic model over a broader interval, ensuring its robustness and accuracy.

4. Conclusions

A 3D-CFD model was used to investigate the behavior of a pilot fluidized bed steam gasifier (100 kWth) equipped with six catalytic filter candles in its freeboard. The simulations provided fundamental insights for optimizing the operating conditions of gasifiers.

The relevant kinetic parameters were obtained using a 2D-CFD model, developed using ANSYS FLUENT^®, applied to experimental tests conducted with a bench-scale gasifier. The simulation results demonstrated good agreement with the corresponding experimental data, validating the kinetic model for use in 3D simulations of the pilot gasifier. Future studies will focus on evaluating catalyst performance under extended operation times. Subsequently, several simulations were then conducted using a 3D-CFD model applied to a pilot-scale dual fluidized bed gasifier (100 kW). As expected, the operating temperature of the catalyst inside the ceramic candles in the freeboard was found to increase to minimize one- and two-ring tar compound concentrations to levels suitable for feeding syngas into solid oxide fuel cells. The simulations examined the optimum spatial and flow rate distributions of oxygen/steam injections along the gasifier freeboard to achieve a controlled temperature rise. Four circumferential locations, each with six injection points, were considered in the upper part of the reactor. Initially, evenly splitting the flow across all nozzles achieved the desired temperature increase. However, the resulting temperatures exceeded the maximum operating limit of the ceramic candles and created a temperature gradient of 160 K along their surface, potentially inducing thermal stress. To address these issues, 75% of the oxygen/steam stream was fed from the lowest nozzles and 25% from the circumferential nozzles just above them. This configuration successfully reduced tar concentrations below the limits allowed by fuel cells while maintaining an acceptable temperature gradient along the candles (90 K). The final design ensured temperatures between 1123 K (necessary for tar reforming reactions) and 1223 K (the maximum tolerance limit for the ceramic filter candles). The model showed that the injection of small amounts of oxygen into the freeboard, to increase its temperature, is useful for lowering the tar content to below 1 g/Nm³. This option will be tested experimentally in a coming research project with a full assembly of six candles in the pilot gasifier freeboard.