Development of a Spouted Bed Reactor for Chemical Looping Combustion

Chemical looping combustion (CLC) has emerged as an attractive alternative for carbon dioxide (CO₂) capture, where a near-pure CO₂ stream is produced without the use of oxygen obtained from air separation units. Use of biomass or coal/biomass blends is particularly attractive in CLC, because capture and storage of the produced CO₂ would result in “negative” carbon emissions.

In CLC, a solid oxygen carrier (OC) is used to bring oxygen to the fuel to convert it to a pure CO₂ stream. The solid is then regenerated separately using air. CLC technology is expected to be more cost-effective than cryogenic or membrane methods for air separation. The majority of the CLC units worldwide use the configuration of two interconnected fluidized bed reactors, one being the fuel reactor and the other, the air reactor [1,2]. Interconnected fluidized bed systems can provide good solids/gas contacting for oxygen uptake and release, as well as enable efficient segregation of the solids from the gas streams using cyclones. However, there are two problems with such systems—insufficient time at temperature for high fuel conversions, and OC attrition, necessitating its recharge/replacement [3,4]. When a solid fuel such as coal or biomass is used, complete conversion of the fuel is a challenge, but is of paramount importance because the primary motivation of CLC is high CO₂ capture efficiency. Incomplete conversion of the fuel in the fuel reactor results in a slip of carbon that is sent to the air reactor where it is oxidized, thus forming CO₂ in the wrong reactor. The “optional” char separation step prior to transfer of the OC to the air reactor becomes necessary and leads to a more complex system. As coal undergoes pyrolysis, char, volatiles, and tars are formed [5]. The volatiles and tars can react with the oxygen carrier, but the chars do not directly react with the oxygen carrier. The char is mostly converted to CO and H₂ via gasification. Reaction rates for this conversion are slow and limited by how high a temperature the fuel reactor can be operated at without bed agglomeration, and the residence time that can be provided for the char [6].

Reported carbon conversion efficiencies for fluidized bed fuel reactors range from 26% to 89%, depending on the reactor configuration and size as well as oxygen carrier used [7–9]. In Ref. [7], the authors demonstrated a 1 megawatt (thermal) pilot plant using ilmenite as oxygen carrier and attained 40% carbon conversion in the fuel reactor of a dual fluidized bed system. In terms of gas conversion, researchers have reported conversions of 84–91.5%, depending on the oxygen carrier to gas ratio, reactor configuration, and oxygen carrier used [7,8]. In Ref. [8], the authors reported gas conversion at best 84% for ilmenite in a 100 kW interconnected circulating fluidized bed (CFB). The proposed benefit of a spouted fluid bed (SB) reactor is to increase conversion efficiency due to effective mixing of char and oxygen carrier while employing similar or less gas flow compared to bubbling fluidized beds [9,10]. It would be ideal to limit the amount of steam required for fluidization to only the amount required for conversion of the char. An area of concern for bubbling beds is large fluidizing gas requirements, including reacting steam for char conversion, which can be exacerbated by tendency of gas bypassing to occur via the gas bubbles, preventing good contact with the char/OC [11]. The use of dissimilar solids in bubbling beds may lead to inhomogeneous solid composition along bed height due to differences in fluidization behavior of the solids [12]. The less dense char can accumulate at the top of the bed, thereby decreasing the contact between gasification products and the oxygen carrier.

To overcome the above challenges, for example, researchers have used dual-stage moving beds [13]. Coal is introduced between the two beds; the top section allows contacting between the volatile components and OC and conversion to CO₂ and H₂O; the lower section provides sufficient residence time for converting the char. However, in such moving bed designs, the particle sizes for the OC used are in the 1–5 mm range [14,15]. Such large particles are prone to attrition (loss represents operating cost penalty), difficult to transport within system reliably, and contribute to high pressure drop in the oxidizer section (designed as a fluid bed) because of the high air flow rates required [3,6].

There is a need for an improved gas–solid contacting configuration to achieve:

• High fuel conversion by increasing the char residence time

• Low OC attrition

• Lower gas flow requirement in comparison to bubbling and circulating fluidized beds

• Low gas bypassing

What are also needed are validated computational fluid dynamics (CFD) modeling tools that will simulate the multiscale, multiphase combustion process in such equipment. This predictive capability can allow evaluation of novel design configurations that would be unrealistic to verify experimentally at the large scale. This paper provides results of a combined experimental and CFD study to develop a spouted fluid bed reactor for CLC. This paper will delineate several key features of the spouted fluid bed reactor that provide advantages over current reducer configurations.

The spouted bed, as described is a variant of fluidization, which permits agitation of solid materials that are too coarse in size for good fluidization [16]. It also has several desirable characteristics that are tailored to chemical looping combustion as applied to solid fuels. In a spouted bed, gas enters at the bottom of the bed through a small orifice as a central jet and induces a systematic circulatory movement of the solid particles as illustrated in Fig. 1. A relatively small upward gas flow (background fluidization) in the annular region is also desirable [17] so that the material is kept in a fluidized state throughout and to avoid dead zones in the overall geometry. The advantages (long contact times for fuel conversion) associated with the dual moving bed reactor design of Ref. [13] can be captured in a spouted fluid bed system [18]. For example, the resulting char from coal devolatilization will have long residence time for its reaction with the OC because the solid flow in the annular region is a “slow” downward moving bed.

The OC–char mixture reaching the bottom of the annular section could be recycled for multiple trips around the reactor system, as would be the case in a CFB. However, a critical advantage of spouting beds over CFBs is a lack of cyclones in the spouting bed, which reduces particle attrition [19]. The recirculation of the solid material via the spout region has an additional benefit, in that any fuel ash particles (typically much smaller than OC particles), which tend to adhere to the OC, can be separated and entrained by the high velocity spout gas and removed from the reactor chamber. This minimizes the interaction between coal/biomass ash and OC.

Another advantage over CFBs is that the total flow rate required to fluidize particles in a spouted fluid bed is lower [10]; in some studies, significantly (approximately 36%) less gas flow (spout and annular) was required to circulate particles. Therefore, the choice of a spouted fluid bed for application to the oxidizer in the CLC process of Ref. [13] is particularly advantageous, as their operation was challenged by high air flow needs (large OC particles) for fluidization, resulting in high combustor gas exit O₂ concentrations (∼15%).

In summary, the spouted fluid bed configuration offers design simplicity and operational robustness compared to the moving, bubbling, or circulating fluid beds and can be applied to both the reducer and oxidizer portions of CLC. This research study targeted the development of a modeling tool for spouted fluid beds for application to chemical looping combustion and experimental validation in a laboratory-scale unit.

Multiphase Flow with Interphase eXchanges (MFIX), an open source code developed by the National Energy Technology Laboratory (NETL), was used to establish a nonreactive, room temperature spouted fluid bed reactor design [20]. The twin fluid model describes the motion of a mixture of gas and solids as two or more interpenetrating continuum phases using an Eulerian–Eulerian framework. This model is commonly used because in actual systems, it is difficult to describe the dynamics of individual particles. The twin fluid model is the most mature MFIX model and provides acceptable speed and accuracy [20]. With this model, neither the motion of individual particles nor the gas flow around each particle is calculated, but averaging procedures are applied over the local instantaneous conservative equations to produce averaged equations that can be solved [21]. To compensate for the reduced resolution, due to the averaging procedures, closure laws are required to deal with parameters and coefficients present in the averaged conservation equations. The closure laws provide correlations for viscous stress tensors, viscosities, pressures, and primarily the drag force. Correlations may be generated from empiricism or by applying a suitable theory such as the kinetic theory of granular flows [22].

The simulations served as a starting point, which were gradually adjusted for more complexity. The proposed reactor design had six important variables that could play a role in spouting behavior. The investigated factors are identified in Table 1.

A fractional factorial design was used to determine the effect of the draft tube geometry/position as well as spout (U_sp) and background (U_bg) velocities on the spouting behavior of the proposed reactor design. The high- and low-level values for the factorial design were constrained by an existing reactor that would eventually be converted into the spouted bed reactor. The ratios of U_sp/U_mf and U_bg/U_mf, where U_mf represents the minimum fluidization velocity of the bed material, were used to make dimensionless numbers that were used for hot flow analyses at a later stage. For the cold flow modeling, 2D axisymmetric simulations were set up using the Gidaspow drag law [23]. A structured grid of 7440 cells was used. The solid annular residence times from the simulations indicated similar trends as those observed in the experimental results. The slug velocity, that is, the velocity at which a parcel of solids is moving up through the draft tube, was not determined, but the material circulation rate recorded.

Initial simulations indicated material movement at U_sp/U_mf equal to ∼95. The results from the simulations were used to design and validate a laboratory test system as illustrated in Fig. 2. Thereafter, the experimental setup was modified for testing under high temperature reacting conditions between 1073 K and 1273 K (1472 and 1832 °F). The setup, represented in Fig. 2, was operated in either the spouted fluid bed or bubbling fluid bed regime to compare the performance attributes of each design.

Particle tracer experiments were set up to determine the mixing behavior and material circulation rate in the reactor. Silica filler was sized to attain a similar minimum fluidization velocity to that of the oxygen carrier. A layer of silica filler was placed near the top of the material bed and covered with oxygen carrier. The residence time was calculated based on the movement of the silica tracer between the annulus and spout. Different values of U_sp/U_mf and U_bg/U_mf were tested to determine the impact on material circulation rate. Mass flow controllers were used to provide constant gas flow to the spout and annulus sections of the reactor, thereby providing a means to control the gas velocities U_sp and U_bg.

The purpose of these experiments was to determine if spouting occurred at different U_sp/U_mf as used in the cold flow experiments. The bed differential pressure was used to obtain a nonvisual confirmation of spouting in the reactor.

The purpose of the hot flow reactor setup was to compare the operation of a spouted fluid bed to that of a bubbling bed using gaseous fuel species of CO and H₂. Both CO and H₂ were fed at 5.0% by volume of the annulus gas flow. The bed was filled with 2300 g of oxygen carrier. The gas flow rates were adjusted for the change in operating temperature from ambient to about 1123 K (1562 °F), such that U_sp/U_mf between the hot and cold flow experiments was constant. The results from the cold flow tests were used to find the optimum ratio between spout and background velocities.

The pressure drop over the material bed was compared for the spouted bed and the bubbling bed. The differential pressure data were compared where the temperature and gas flow rates of the individual tests were the same. Spouted fluid bed operation should provide lower pressure drop if the same gas flow rate is used in comparison to a bubbling bed. Bubbling bed and spouted bed simulations were also set up to compare the pressure differences between the two reactors. For the hot flow modeling, 2D axisymmetric simulations were set up using the Gidaspow drag law [23]. A structured grid of 29,760 cells was used.

The temperatures at the top and bottom sections of the annulus were continuously monitored during the oxidation/reduction cycling. The bottom 5.08 cm (2.0 in.) of the reactor was not heated and only insulated with ceramic wool. Material circulation in the reactor would create a more uniform temperature between the top and bottom parts of the material bed and therefore provide an indication as to whether the material underwent better mixing in the spout fluid bed or bubbling bed reactor.

Computational modeling was used to initially test several spouted fluid bed reactor configurations. The simulation results were compared via the use of a fractional factorial design. Based on the fractional factorial design, the following factors affected the material circulation rate the most:

• i.

  Factor B: Draft tube height from bottom

• ii.

  Factor C: Draft tube internal diameter

• iii.

  Factor E: Spout velocity/Minimum fluidization velocity, U_sp/U_mf

High-level values of Factors B and E and low-level value of Factor C were shown to increase material circulation rate. Based on the results, a low, an intermediate, and a high mass circulation configuration were chosen to compare with the actual results from the experimentation. The simulations were set up to ascertain a basic configuration for the cold flow reactor as well as to determine the importance of the various factors. With the experiments, the average particle size of the oxygen carrier was less than that used in the simulations. This necessitated higher U_sp–U_mf ratios to induce material circulation, whereas U_sp/U_mf = 95–142 was sufficient for the larger particles, U_sp/U_mf = 160–290 was required for the finer material. The reactor geometries of the low (U_sp/U_mf = 160), intermediate (U_sp/U_mf = 225), and high U_sp/U_mf = 290) material circulation rate cases formed the basis of the cold flow reactor setup. Because the material bed height had a lesser effect on the circulation rate, it was decided to use a constant bed height of 25.4 cm (10.0 in.) and a draft tube length of 25.4 cm (10.0 in.) for the cold flow laboratory testing.

The material residence times in the annulus during one circulation pass for the cold flow reactor experiment are indicated in Fig. 3. The MFIX simulation results (with the same average particle size as that used in the experiments) are also illustrated in Fig. 3. Similar trends were observed between the experimental and simulation results. The discrepancy between the solids annular residence time for the experiment and simulation, at the lowest spout condition (U_sp/U_mf = 161), could likely be attributed to particle size and form. In the simulations, a single particle size was used, whereas in the experiment, the material had an average particle size of 110 μm. Interlocking of different particle could possibly have caused enough resistance to impede material movement in the experiment. In the simulation, the particles were able to move more freely without interlocking due to different particle shapes and sizes; thus, the solids annular residence time was lower compared to the experimental result. The experimental results validated the importance of U_sp/U_mf on the material circulation rate. The MFIX simulations also represented accurate visual representations of the actual cold flow results where intermittent, but constant frequency spouting occurred. The intermittent spouting showed that a higher spout velocity would be required to attain a spouting fluidization regime. The spouting behavior from the experimental setup is depicted in Fig. 4 and could be described as follows:

• i.

  Material builds up in the draft tube

• ii.

  Material increases and starts to exit draft tube

• iii.

  Material uniformly exits draft tube

• iv.

  Material moves outward and fills the annular space

The spouting behavior was also observed from the pressure data. The pressure drop over the material bed was continuously monitored and the data are depicted in Fig. 5. A clear distinction was observed between the pressure drop fluctuations for low and high frequency intermittent spouting. The higher the spout velocity, the greater the fluctuations in pressure drop as the material circulation rate increased. A fast Fourier transform (FFT) was used to express the time-based pressure signal in the frequency domain. The cold and hot flow experimental data were not sampled at a sufficient time interval to identify distinct frequencies for the spouted and bubbling bed experiments. From the pressure data, the greater fluctuations for the spouted bed could be observed, but the sampling rate was too low to discern distinct frequency differences between the spouted and bubbling bed experiments. Future experiments will be conducted at higher sampling rates to investigate the bubbling and slug frequencies. The pressure fluctuations in the simulations were examined to see and characterize the differences between the spouted and bubbling bed reactors. The visual representation of the simulated spouted fluid bed configuration, with U_sp/U_mf = 225, is illustrated in Fig. 6.

The solid volume fractions increase as the color spectrum changes from red to blue as indicated by the accompanying side scales. The same change in residence time as a function of spout velocity was observed in the simulation and experimental results. As the spout velocity was increased beyond U_sp/U_mf = 225, the material residence time did not change significantly. The observation indicates that higher spout velocities eventually restrict the flow of material from the annulus to the spout thereby decreasing the material circulation rate. The material residence time could be controlled between 40 and 170 s depending on U_sp/U_mf. The difference between intermittent spouting and other spout flow regimes is discussed in Ref. [24]. Lower spout velocities result in a stationary bed and higher spout velocities result in spouting with aeration or fluidization. The slug velocity was not measured since the experimental setup was a cylindrical bed for which the material residence time was easier to quantify. In the experiment, the material residence time decreased as U_sp/U_mf was increased from 160 to 290. However, the change in residence time for U_sp/U_mf = 225 to 290 was less compared to that for the 160–225 case. The decreased rate of change in the residence time indicates that an upper operating limit exists. The spout velocity would likely have to be increased tremendously to attain spouting with aeration/fluidization, thereby negating the lower gas flow advantage offered by a spouted bed reactor.

Four reactor configurations (A, B, C, and D) were investigated. The configurations were:

• A.

  SB operation including draft tube—The reactor was operated at a condition of U_sp/U_mf = 225 and U_bg/U_mf = 0.8.

• B.

  Simulated bubbling bed with draft tube (Pseudo bubbling bed (PBB))—In this scenario, the draft tube was left inside the reactor. The gas flow rates were altered to attain similar gas velocities in the draft tube and annulus sections, representing a bubbling fluidization regime. For Configuration B, the reactor was operated at a condition of 30% greater gas flow compared to Configuration A. Configuration B had the same total gas flow as the spouted fluid bed of Configuration A.

• C.

  Bubbling bed (BB) without draft tube—The draft tube was completely removed from the reactor and therefore represented an actual bubbling bed. The gas flow rate for the bubbling bed was the same as that for the spouted fluid bed in Configuration A.

• D.

  Simulated spouted fluid bed without draft tube (Pseudo spouted fluid bed (PSB))—The PSB was operated in similar mode to Configuration A.

The hot flow reacting condition experiments were conducted over a period of three days. The oxygen carrier was always kept in the reactor and each day the material was heated from room temperature to ∼1123 K (1562 °F) under an inert atmosphere of nitrogen. At the set point temperature, the material was oxidized until breakthrough occurred. The material was then subjected to reducing conditions. The molar flow rates of CO and H₂ were kept similar to compare the conversion between configurations with different overall gas flow rates. The results from the experiments are summarized in Fig. 7. Over the three-day testing period, the reactivity of the material increased. The increase in reactivity was expected based on experience from related projects involving the use of the specific oxygen carrier. As the material became more porous with time, the conversion of both CO and H₂ increased. The CO is more pore diffusion limited than H₂ and therefore the conversion of H₂ was always greater [25]. The H₂ conversion in all the tests was similar, whereas greater differences could be observed in the CO conversion data. On day 1, the comparison between the spouted fluid bed and the pseudo bubbling bed at the higher flow rate suggests that the higher flow rate likely contributed to the lower CO conversion for the bubbling bed case. On day 2, the flow rates were similar for the spouted fluid bed and pseudo bubbling bed cases. The conversion of CO was again lower for the pseudo bubbling bed configuration. The pseudo bubbling bed configuration necessitated the use of a higher CO₂ concentration to increase the annular flow. The higher CO₂ concentration could have affected the CO conversion. The kinetics in the spouted fluid bed could also have been more favorable and therefore increased the CO conversion. On day 3, the gas flow rate and CO₂ concentrations were similar for the bubbling bed and pseudo spouted fluid bed reactors. The CO conversion in the pseudo spouted fluid bed was similar or better than the bubbling bed.

The aforementioned results all indicate that the spouted and pseudo spouted fluid bed reactors could be operated to achieve equivalent or even better fuel conversions compared to the bubbling and pseudo bubbling bed reactors.

Due to the inability to visually observe the spouting behavior in the hot reactor experimental setup, another method had to be employed to determine if spouting occurred. The differential pressure measurement over the reactor bed was used for the determination. The intermittent spouting that was observed in the cold flow experiments could be detected by the fluctuations in the differential pressure value. The same method was employed for the hot flow experiments and the results for the bubbling bed and spouted fluid bed reactors are illustrated in Fig. 8. The spouted fluid bed reactor exhibited a slightly lower pressure drop in comparison to the bubbling bed. The spouting behavior could also be observed by the fluctuations in the differential pressure data. The observation confirmed the notion that the use of a spouted fluid bed would provide a lower pressure drop in comparison to a bubbling bed with the same gas flow rate. Figure 9 represents the frequency domain results after applying the FFT to the simulation results of the bubbling and spouted bed simulations. The sampling rate was 20 samples per second and therefore fast enough to capture distinct differences between the bubbling bed and spouted bed simulations. The dominant frequency for the spouted bed simulation was 7–8 Hz and for the bubbling bed simulation, it was between 0 and 1 Hz. In Ref. [26], the authors described the characteristic regimes in a spout-fluid bed and found that a slugging bed exhibits a broad frequency shape and power values greater than 104 Pa². Based on prior cold flow tests, the spouting regime could be classified as slugging or intermittent spouting. The broad frequency from 8 to 9 Hz in Fig. 9 likely indicates slugging. The difference in power and frequency is likely attributable to differences in the system properties such as bed geometry and particle properties. The simulations indicate that it could be possible to observe distinct differences between spouted and bubbling beds if the pressure sampling rate is at least greater than 20 samples per second.

The temperature profiles for the hot flow experiments were plotted to determine the level of material circulation within the spouted fluid bed reactor. The bottom 5.0 cm of the reactor was not heated and only insulated with ceramic wool. Material circulation in the reactor would create a more uniform temperature between the top and bottom of the material bed. Based on the temperature profile in Fig. 10, material circulation did occur for the spouted bed case since the top and bottom temperatures in the annulus differed by no more than ∼50–60 K (122–140 °F). For the bubbling bed case, the temperature profile differed significantly from that of the spouted fluid bed reactor. The temperature difference between the top and bottom annulus sections was ∼200–260 K (∼392–500 °F).

The temperature difference would have been minimized had the entire reactor been uniformly heated. However, in large-scale systems, dead zones can typically occur and cause the material to undergo sintering due to lack of material movement. In such a system, the reactor would have to be operated at a lower overall temperature to prohibit sintering and agglomeration of the bed material [27]. If better material mixing occurs, the reactor can be operated at temperatures closer to the sintering point to achieve higher fuel conversion.

The dynamic characteristics of bubbling fluidized beds could be successfully predicted using the two-fluid model in MFIX. Cold flow model results were used to design a spouted fluid bed reactor and identify flow parameters to achieve spouting. Intermittent spouting with consistent frequency was achieved in the experiments as opposed to spouting with aeration. The cold flow model results provided key information for rapid experimental design and operating envelope determination. Cold flow models proved adequate in depicting the intermittent spouting regime as well as providing valuable information pertaining to material circulation rate. The hot flow reactor was successfully designed based on the cold flow data. Reduction and oxidation cycling for chemical looping combustion was successfully conducted with the oxygen carrier and CO/H₂ as gaseous reactants. The results from the hot flow experiments indicated good conversions for both spouted fluid bed and bubbling bed configurations. The spouted fluid bed configuration exhibited significantly greater thermal uniformity from the top to the bottom of the reactor in comparison to the bubbling bed reactor. The greater material circulation in the spouted fluid bed reactor can potentially lessen the likelihood for sintering to occur. The pressure drop in the spouted fluid bed reactor was less in comparison to the bubbling bed reactor for the same overall gas flow rate.

The spouted fluid bed design illustrated the potential of the configuration to improve gas/solid contact, lower energy requirements and increase operational robustness in comparison to a bubbling bed reactor.

We acknowledge the technical comments of Dr. Bhima Sastri, Program Manager, United States Department of Energy (USDOE), to the research study. This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

• This material is based upon work supported by the United States Department of Energy (USDOE), Office of Science, Small Business Technology Transfer Research Grant under Award No. DE-SC0015204.

BB =

bubbling bed

CFB =

circulating fluidized bed

CLC =

chemical looping combustion

Hz =

Hertz

MFIX =

multiphase flow with interphase exchanges

NETL =

National Energy Technology Laboratory

OC =

oxygen carrier

PBB =

pseudo bubbling bed

PSB =

pseudo spouted bed

U_bg =

background velocity (m/s)

U_mf =

minimum fluidization velocity (m/s)

U_sp =

spout velocity (m/s)