Rotary Bed Reactor for Chemical-Looping Combustion with Carbon

Article pubs.acs.org/EF

Rotary Bed Reactor for Chemical-Looping Combustion with Carbon Capture. Part 1: Reactor Design and Model Development Zhenlong Zhao, Tianjiao Chen, and Ahmed F. Ghoniem* Department of Mechanical Engineering, Massachusetts Institute of Technology (MIT), 77 Massachusetts Avenue, Cambridge, Massachusetts 02139-4307, United States ABSTRACT: Chemical-looping combustion (CLC) is a novel and promising technology for power generation with inherent CO2 capture. Currently, almost all of the research has been focused on developing CLC-based interconnected fluidized-bed reactors. In this two-part series, a new rotary reactor concept for gas-fueled CLC is proposed and analyzed. In part 1, the detailed configuration of the rotary reactor is described. In the reactor, a solid wheel rotates between the fuel and air streams at the reactor inlet and exit. Two purging sectors are used to avoid the mixing between the fuel stream and the air stream. The rotary wheel consists of a large number of channels with copper oxide coated on the inner surface of the channels. The support material is boron nitride, which has high specific heat and thermal conductivity. Gas flows through the reactor at elevated pressure, and it is heated to a high temperature by fuel combustion. Typical design parameters for a thermal capacity of 1 MW have been proposed, and a simplified model is developed to predict the performances of the reactor. The potential drawbacks of the rotary reactor are also discussed. while it is looping between the reactors. Since first proposed by Richter and Knoche in 1983,1 the selection of the OC has been acknowledged as one of the most important aspects in CLC. Most of the work on CLC thus far has been focused on the development and investigation of OCs configured in particle form using fixed- or fluidized-bed reactors. Some of the most commonly investigated OCs are nickel-,3−7 copper-,3,7−11 iron-,3,12−14 and manganese-based3,7,15,16 OCs. Research on the CLC reactor design has almost exclusively concentrated on an interconnected fluidized-bed reactor with OC particles circulated throughout the reactors.17−24 In this system, both reactors are fluidized beds. OC particles are fluidized and pneumatically transported continuously between the fuel reactor and the air reactor. A cyclone in the top and a loop seal in the bottom are used to separate the OCs from gas streams. The reactor system based on a fluidized-bed design has advantages, such as perfect particle mixing, homogeneous temperature distribution, and smooth, liquid-like particle flow inside the reactor with continuous automatically controlled operations.25 Major challenges of this design are related to the particle circulation process:17,26 (1) extra energy is needed to fluidize the beds, and hence, the pressure drop throughout the reactor is usually high; (2) an efficient cyclone is critical to the recovery of OCs from the flue stream; (3) fine particles need to be removed from the flue stream before entering the gas turbine in a combined cycle (CC); (4) agglomeration may happen at high operating temperatures; and (5) particle collisions would impair the lifetime of the reactors. Besides, nitrogen and CO2 leakage can occur, which leads to reduction of the capture efficiency or the need for extra separation downstream to purify the CO2 stream. These issues are more

1. INTRODUCTION It has been widely acknowledged that emissions of greenhouse gases are a primary contributor to global warming, and CO2 is the most prevalent of these gas emissions. One possible approach to restrict anthropogenic CO2 emissions, apart from improving conversion and use efficiency and expanding the use of alternative sustainable energy, is carbon capture and sequestration (CCS). Thus far, extensive research focus has been placed on three general processes for capturing CO2 from combustion in power plants: pre-combustion capture, postcombustion capture, and oxy-combustion. One of the key issues that limits the applications of CCS approaches is the large energy penalty during the separation process, which renders CCS inefficient. Recently, a new approach for CO2 capture has been investigated. This approach was named “chemical-looping combustion (CLC)” by Richter and Knoche1,2 and belongs to oxy-fuel combustion. In CLC, two steps of combustion are involved (as seen in Figure 1). Fuel is oxidized by metal oxide

Figure 1. General scheme of a CLC system.

in a fuel reactor to generate CO2 and water steam. The reduced metal oxide is then regenerated by air in an air reactor. The flue gas from the fuel reactor only contains CO2 and H2O, where CO2 can be readily captured after steam condensation. During this two-step CLC process, the looping metal oxide adsorbs oxygen in the air reactor and releases oxygen in the fuel reactor; that is, it acts as an “oxygen carrier” (OC) transporting oxygen © 2012 American Chemical Society

Received: August 28, 2012 Revised: November 5, 2012 Published: November 13, 2012 327

dx.doi.org/10.1021/ef3014103 | Energy Fuels 2013, 27, 327−343

Energy & Fuels

Article

severe at elevated pressures. Although a great variety of successful industrial applications using fluidized-bed reactors have been built, the complexity of the multiphase, multi-scale reactive flow makes it difficult to design, optimize, and scale up the reactors. Alternative designs, such as the moving-bed reactor,27−29 the fixed packed-bed reactor,26,30,31 or the rotating packed-bed reactor,32,33 have also been proposed and investigated. Fan and co-workers27−29 suggested using a moving bed for the fuel reactor. In comparison to fluidized-bed reactors, the movingbed reactor has the advantages that the mixing of the gas (or solid) phase along the moving direction is small such that the dilution effect of the product on the incoming fuel is limited. Similar technical difficulties still exist in the particle circulation process as in the fluidized-bed design. To avoid circulation issues, Noorman and co-workers26,30,31 proposed the packed-bed reactor design. In this design, the OC particles are packed into the reactor and are alternately exposed to reducing and oxidizing conditions via periodic switching of the gas feed streams. Two reactors in parallel are used alternately to ensure a continuous high-temperature gas stream supply to the downstream gas turbine.17 The main advantages of the packed-bed reactor are that the separation of gas and particles is intrinsically avoided, that the reactor design can be much more compact, and that they allow for better use of the oxygen. Besides, because the operation is in stationary beds, no extra energy is needed for circulation. The use of such reactors, however, requires the effective control of large volumes of gas under high temperature and high pressure, in which the flow must be continually initiated and terminated. One potential challenge is that heating and cooling the packed particles may cause a large temperature fluctuation within the reactor, as observed by Noorman et al.30 Besides, fuel slip may occur during the switching period, which leads to safety issues. Dahl et al.32,33 designed a rotating reactor. The rotating reactor is an extension of the packed-bed reactor. An annulus packed bed containing OC particles rotates when fuel and air streams are introduced radially outward through the reactor. Inert gas (in this case, steam) is fed between air and fuel sectors, and separation walls on the outer and inner walls are used to avoid mixing. The advantages of the rotating bed reactor include the compactness of design with continuous operations, the limited energy for circulation, and the feasibility of scale-up and commercialization. The main challenge for this design is to avoid the gas leakage and dilution between fuel and air streams, which, at the moment, are unavoidable.17 In this work, we investigate whether CLC can be conducted in reactor concepts based on rotary-bed reactor technology with microchannels rotating between the reducing and oxidizing environments. The main advantages of this concept are that the separation of the gas and particles is intrinsically avoided and that the operation is continuous, while cyclical, it is stationary. Moreover, the reactor design is compact and easy to scale-up. The idea of a rotary chemical-looping reactor system is based on two innovative applications of rotary technology. Rotary desiccant wheel: A desiccant wheel matrix consists of a large number of microchannels with walls constructed of supporting material coated or impregnated with desiccant material. The desiccant material adsorbs water vapor when moist air passes through the process air side, and it is regenerated when it desorbs water, while heated air flows through the regeneration side.34

Rotary regenerative heat exchanger: The rotary heat exchanger transfers heat between a high-pressure, low-temperature feed gas and a low-pressure, high-temperature flue gas. As the wheel rotates between the cold and hot streams, only heat is transferred through the solid matrix. An effective sealing system, which could avoid the mixing between the low-pressure flue gas and the high-pressure inlet air, is needed to reach a high efficiency.35−38 1.1. Rotary Bed Reactor Concept. The rotary bed reactor combines the basic elements of a rotary desiccant wheel and a rotary regenerative heat exchanger. The OC, rather than the desiccant material, is incorporated (either coated or impregnated) on the surface of the spinning wheel matrix in a way similar to that in the desiccant wheel. As the reactor rotates, oxygen is adsorbed, while the channels pass through the air section, temporarily stored in the OC and then released to oxidize the fuel. During the cyclic operation, the solid wheel also behaves as a heat exchanger to transfer the reaction heat to the flowing gas. Because each microchannel is alternately exposed to reducing and oxidizing conditions via rotation, the combustion process is readily divided into reduction and oxidation sections. The direct contact between the fuel stream and the air stream is intrinsically avoided. Effective sealing systems can be adapted from the rotary regenerative heat exchanger to avoid the fuel leakage and the air dilution. One distinction though is the high operating temperature of the rotary reactor. Special attention must be paid to temperature fluctuation and the differential thermal distortion associated with the reactor. Besides, carbon deposition on the channel surface inhibits the redox reactions, lowers the carbon capture efficiency, and even leads to the channel blockage, and it must be avoided in cyclic operation. Detailed discussion on the thermal distortion as well as the carbon deposition is presented in section 4. A similar idea of using a reactor with microchannels for CLC was briefly suggested before. Pavone and co-workers39,40 discussed the possibility of using a chemical-looping redox (reduction−oxidation) chamber within a Brayton cycle. The redox chamber consists of a large number of channels. Each channel is 0.5 m long and 0.5 mm in radius, with 50 μm of OCs wash-coated on the surface and 25 μm of support matrix between adjacent channels.39,40 When the injections at the inlet are manually switched or the redox chamber is continuously rotated, each channel experiences the identical sequence of reduction and oxidation. The redox performance of one channel was simulated using a commercial computational fluid dynamics (CFD) package for the initial several cycles. Unburned fuel was observed at the outlet. For a carbon capture efficiency of 90%, the energy penalty was 15% of the power output without CO2 capture. Owing to the limited amount of the solid phase in the channel, large temperature fluctuations (>500 °C) were observed in the OC wash-coat, which could cause severe thermal stresses within the reactor. Little discussion was given regarding the reactor design, cyclic performance, or operating conditions. Cichanowicz and Muzio41 extended the rotary idea to a Rankine cycle using solid fuel with a counter-current flow pattern in the reactor with a focus on the system-level design. Coal particles enter the chamber channel and undergo a series of complex reactions, e.g., devolatilization, pyrolysis, and char oxidation, generating syngas, which is then oxidized by the OCs along the channel. However, a major barrier in this design is that the ash particles tend to stick to surfaces, which may lead 328


Energy & Fuels

Article

chamber, reacts with the OC as it passes through the rotary bed, and leaves the system from the top exit chamber. The heat generated from the exothermic reactions is used to heat the passing gas to a high temperature, which is ultimately used to drive turbines to generate electricity. The solid wheel temporarily stores the heat of reaction and releases it to the flow. The rotary bed, powered by the driving motor, rotates at a constant speed, while the chambers remain stationary. The entire reactor is surrounded by insulating walls that could sustain high temperatures and high pressures (as shown in Figure 2b). For clarity, the insulating walls of the reactor wheel are not shown in Figure 2a. Figure 2b shows the cross-sectional view of the reactor. The rotary bed matrix consists of an array of identical long and narrow channels. A typical channel size is several millimeters wide, depending upon the cell density. The OCs are coated or impregnated onto the inner surfaces of each channel. Two streams are admitted into the spinning channels from the feed side and leave into two different zones divided by insulating walls in the exit chamber. As the channel passes through the fuel zone, the gaseous fuel stream flows into the channel and reacts with the active metal oxide to generate CO2 and H2O. As the same channel passes through the air zone, air flows into the channel to fully regenerate the OC back to its original state. The gas streams in the fuel and air zones are at the same pressure. The chemical energy from the continuous redox (reduction and oxidation) reactions is temporarily stored in the solid phase and then transferred to the bulk flow by convection through the rotary matrix, which behaves in a similar way to the rotary heat exchanger. The center of the rotary bed is a small hollow channel, through which a cylindrical bearing is inserted to support the reactor construction and actuation. The design is not limited to the co-current flow pattern. For example, a counter-current flow pattern with fuel (or air) flowing from the top chamber to the bottom chamber can be an alternative option. Figure 3a shows the bottom view of the gas feed chamber. The feed chamber is divided into four sectors: a fuel sector (θfuel), an air sector (θair), a fuel-purging sector (θfuel_purge), and an air-purging sector (θair_purge). The fuel zone is divided into fuel and fuel-purging sectors, while the air zone is divided into air and air-purging sectors. Fuel gas or air is fed into the fuel or air sector, respectively, while steam is used as a “sweeping” gas in the purging sectors to flush the reactor and, hence, avoid gas carry-over between sectors. The four sectors at the bottom chamber are separated by insulating walls, which remain stationary during operation. Figure 3b shows the isometric projection of the reactor wireframe. The top exit chamber consists of a fuel sector and an air sector. While one slug of feed gas passes through a channel, the reactor bed spins continuously. Accordingly, flue gas exits at the same radial location of the wheel at a slightly different angle. The two purging sectors act as “buffer zones” to account for this angle mismatch. Therefore, as one channel spins out of the fuel (or air) sector and enters the following purging sector, a feed steam continuously flushes the residual fuel (or unburned air) into the same zone at the top without diluting the other zone. Owing to rotation, some steam from the fuel (or air) purging sector may also end up entering the following air (or fuel) sector. Figure 4 shows the gas flow pattern through the reactor. The majority of the feed gas enters the reactor from the feeding chamber, flows through the different channels, and leaves the reactor from the exit chamber. The pressure drop along the channel is attributed to the skin friction, which is generally small for a laminar channel flow. As a result, the pressure differences between different sectors are expected to be small, and hence, the pressure-driven gas leakage is limited. However, some gas leakage may occur because of the spinning motion of the reactor. For instance, as shown in panels a and b of Figure 4, some gas may flow through the gap between the insulating walls and the peripheral reactor surface and leave the reactor without being reacted. Some gas may bypass from the fuel zone to the air zone through the gap between the insulating separation walls and the rotary reactor, as shown in panels c and d of Figure 4. Sealing systems similar to those used in the rotary regenerative heat exchanger can be used to reduce the gas leakage (as seen in Figure 4). Peripheral seals will trap and force the flow into centrifugal motion and, hence, restrict the gas

to blockage and surface erosion. In addition, the burnout time needed for solid fuel is generally much longer than that for gaseous fuel, such that a highly sophisticated post-possessing system is needed to remove unburned char particles. The objective of this two-part series is to propose the rotary design for CLC, assess the feasibility of continuous operation, and test the cyclic stationary performance with respect to fuel conversion and CO2 separation. In part 1, a detailed description will be presented on the reactor design, sector arrangement, material construction, functionality, and operating conditions of the rotary bed reactor. A reduced complexity model is formulated, enabling the assessment of the performances of the rotary reactor based on the cyclic behaviors of each channel. Potential challenges of the rotary reactor operation are also discussed in details. Part 2 (10.1021/ef301411g) presents the base-case results from the model as well as the results of the sensitivity analysis to identify input parameters and variables of importance. Future work will demonstrate the use of the rotary design for system-level analysis as well as experimental investigations.

2. DESIGN 2.1. Reactor Design and Specification. The rotary bed CLC reactor consists of a rotary bed matrix, a driving motor, and two stationary gas chambers located at the top and bottom of the wheel (inlet or outlet), as shown in Figure 2a. Pressurized feed gas, both fuel and air, flows in a co-current pattern from the bottom feeding

Figure 2. Schematic diagram of the rotary CLC system design with a (a) front view and (b) cross-section view. 329


Energy & Fuels

Article

or release oxygen from the OC material, release or absorb energy with the reactor matrix, and leave the reactor with varying flow velocities and concentrations. Given constant inlet conditions during operation, as shown in Figure 6a, it is expected that, after a number of cycles, the reactor will gradually converge to a periodic stationary state: the physical and chemical processes within one cycle will go back to the original states after one cycle. The compositions at the exit of the reactor at the stationary state are shown in Figure 6b. Therefore, the sum of a large number of transient flue streams exiting from the fuel sector (or air sector) will mix well to give one steady-state flue stream. The steady separate streams from the fuel and air sectors can then be used to drive the gas turbine, as shown in Figure 7, and CO2 can be easily separated after water condensation. The following analysis will mainly focus on this periodic state behavior. 2.2. Design Criteria. The rotary reactor design should satisfy a number of criteria, including complete fuel conversion, carbon separation, and operational stability. High fuel conversion and high carbon separation efficiency are the most fundamental requirements for CCS. Additionally, operational stability ensures a steady periodic performance and, thus, a long operational lifetime and low operating cost for the power generation. The operational stability includes thermal, mechanical, and chemical stabilities: (1) temperature variation with time should be limited such that the thermal expansion and distortion associated with the temperature gradient are limited; (2) solid support material must be mechanically robust such that it is effectively resistant to agglomeration, sintering, or cracking during operation; and (3) high redox reactivity of the OC should be maintained for many cycles, while minimal reaction should occur with the support material. The discussion of the reactor design parameters and operating conditions is primarily based on these three design criteria. A variety of parameters can be specified to achieve the above design criteria. These design parameters and operating conditions can be categorized into three parts: (1) material selection: OC, supporting material, surface treatment, and coating, shown in Table 1; (2) reactor configuration: shape, size of channel, arrangement of sector, and seals, shown in Table 2; and (3) operating condition: temperature, pressure, rotational velocity, and feed velocity, shown in Table 3. These parameters are highly coupled, and they are all closely related to the performances of the reactor. For instance, a highly reactive OC material, an effectively coated porous layer, or simply a long enough channel with a small cross-sectional area all lead to high redox reactivity and, thus, high fuel conversion. The CO2 separation efficiency is directly determined by the purging gas velocity through the channel, the residence time in each zone, and the size and porosity of the channel. Furthermore, the operational stability is closely related to almost all design parameters: the OC, preparation method, size of channel, temperature, pressure, and gas flow rate, which all affect the reactivity and influence the heat balance within the reactor. The selection of support material and the geometry of the channel determines the thermal inertia, mechanical strength, and chemical regenerability of the reactor. The arrangement of the purging sector and the sealing systems is closely related to the fuel−air separation and, thus, the safety of the reactor. The reactor design is complex and involves the comprehensive consideration of all of the foregoing aspects. Tables 1−3 and Figure 8 show typical design parameters for the reactor considered in this work. Copper is used as the OC, owing to its high reactivity and low tendency for carbon deposition, high oxygen capacity, relatively low cost, and limited environmental impact.42 Boron nitride (BN) is used as the support material for both bulk and porous layers, owing to its key properties, which result from its crystalline structure, such as high thermal conductivity, low thermal expansion, effective thermal shock resistance, and high chemical stability in an oxidizing environment. The bulk dense support layer (260 μm) is pretreated by surface etching and then coated with copper nitrite by wet impregnation to form a thin porous OC layer (50 μm). The thickness of the porous OC layer is the same as that used in the design proposed by Pavone,39 and it is also related to OC kinetics experiments conducted by Garcia-Labiano et al.,43 as discussed in

Figure 3. Schematic drawing of the rotary CLC system design with a (a) bottom view and an (b) isometric projection of the wireframe. bypass from the inlet to the outlet side. Radial seals with small clearance will restrict the gas leakage rate between stationary insulating walls and the moving reactor. The rotary bed matrix consists of a large number of channels. Each channel consists of an inner gas passage and solid support material coated with the OC, as shown in Figure 5a. Gas flows through the passages and reacts with the OC on the inner surface. The OC is coated or impregnated onto a porous layer of the solid support, as seen in Figure 5b. High porosity of the OC layer enhances the surface area between the solid and gas species and, hence, increases the heterogeneous surface reaction rates. The binder material in the porous layer acts as an oxygen-permeable material that helps enhance the gas permeability, improve the pore structure, sustain the thermal and mechanical stresses, and hence, maintain the physical and chemical stabilities of the OC after repeated cycles. In addition, as shown in Figure 5b, a bulk support layer is bonded to the porous layer. This bulk support layer is made of highly conductive materials with high heat capacity, which can effectively store the heat produced in the exothermic reaction, transfer it within the reactor, and heat the flowing gas. The use of the bulk support layer is critical to the temperature distribution of the reactor. Besides, the bulk layer helps avoid the gas mixing between the adjacent channels. Note that the support material in the porous layer and the bulk layer is not necessarily the same. As one channel travels through the fuel and air zones, following a sequence of fuel, fuel purging, air, and air purging, the active OC on the matrix surface continuously releases oxygen to oxidize the fuel and adsorbs oxygen from air. Typical profiles of gas species for two consecutive cycles at the inlet and outlet of one channel are shown schematically in Figure 6. The thermal and chemical states in one channel undergo a transient process: gases enter the channel, adsorb 330


Energy & Fuels

Article

Figure 4. Schematic drawing of the gas flow pattern through reactor and gas leakage through radial and peripheral seals. (a and b) Peripheral seals prevent leakage between the insulating walls and the peripheral surface of the reactor. (c and d) Radial seals restrict leakage between the insulating walls and the top and bottom surfaces of the reactor. section 3.3.3. The mass of the bulk dense support layer (δs − δoc)ρs is one order of magnitude larger than that of the OC layer (1 − εs)δocρoc to minimize the temperature fluctuation and ensure the thermal stability. The active OC content is around 10% (by weight) in the porous layer, similar to that used by Garcia-Labiano et al.43 The channel of the reactor is square-shaped, with a width of 2.0 mm, and hence, the flow through the channel is generally laminar. The channel width is larger than that used by Pavone.39 The choice of the channel cross-section is based on the consideration of reactivity and periodic stability: a smaller size leads to higher reaction rates but a higher convective heat-transfer rate and, hence, a larger local temperature fluctuation. The entire reactor is pressurized at 10 atm. The selection of the operating pressure is based on the experiences in a non-ideal Brayton cycle, with consideration of the impact of the compressor and turbine efficiencies on the cycle efficiency and the specific work.44 The operating temperature ranges from 823 to 1300 K. The fuel is diluted with recirculated CO2 to lower the operating temperature below the melting temperature of the OC. The feed velocities are selected to ensure sufficient residence time for complete fuel conversion in the channel. The size of the reactor (diameter and height) is configured to ensure complete fuel conversion, with a capacity of 1 MWth. The

angles of the sectors and the rotational velocity are selected to provide enough residence time for complete OC regeneration and complete purging of residual fuel (or air) and also to avoid the complete reduction of the OC (full use of oxygen capacity) in the fuel sector. The reactor design is not limited by the materials or the preparation method described above. Any of the materials or preparation methods described by Adanez et al.17 are potential candidates for this rotary design, as well as other materials studied by numerous other investigators, which include, for example, in reduced metal form, Fe, Cu, Mn, Co, and Ni. The support material can be any material conventionally used as ceramic insulators: Al2O3, yttria (Y2O3)stabilized zirconia (YSZ), TiO2, and BN. The preparation method includes, for example, wet impregnation, dry impregnation, deposition precipitation, and wash coating. Surface treatment methods, such as surface etching, can be used to improve the surface porosity and the OC loading. In addition, the channels can be formed in a variety of geometries or sizes, such as, grid type, honeycomb geometry shapes, plate types, any series of corrugated shapes, or any type of geometry that presents a high specific surface area. Besides, the reactor can be operated at different pressures, temperatures, flow velocities, etc. The preliminary study in this two-part series is based on the parameters described in Tables 1−3 and Figure 8. The performance of the reactor 331


Energy & Fuels

Article

individual channel experiences the same sequence of events: reduction in the fuel zone, purging of the fuel stream, regeneration in the air zone, followed by purging of the leftover air stream. All channels have the same residence time in each sector. Thus, in the preliminary analysis, we can reasonably neglect the radial variations among different channels in the reactor and only focus on one channel while changing inlet conditions. The channel in the rotary reactor typically has a very high height/equivalent diameter ratio (H/d = 500, as seen in Table 2). Therefore, using a one-dimensional plug-flow model with heat and mass transfer between the solid and gas flow, we should be able to characterize the physical and chemical processes within the channel. The model includes both the gas phase and the solid phase, as shown in Figure 9. The flow inside the channel is laminar (Red = ρud/μg ∼ 100). The pressure drop is caused by laminar skin friction [ΔP ∼ 1 kPa; see part 2 (10.1021/ef301411g)], which is small compared to the operating pressure. Therefore, it is reasonable to assume a constant thermodynamic pressure along the channel. Because of the relatively low temperatures and the small channel size, radiative heat transfer between the solid and flow gas is generally negligible.45 As each channel travels between the fuel and air sectors, the local solid temperature changes with time. Therefore, one may expect heat transfer between neighboring channels in the peripheral direction. However, the thermal resistance of the solid layer is small (Biot number = hgsδs/ks ∼ 10−4), and the local solid temperature fluctuation with time is small [within 20 K; see part 2 (10.1021/ef301411g)]. Therefore, the solid phase can be treated as a transient fin with uniform temperature distribution in the horizontal direction (as shown in Figure 9).45 As shown in Figure 9, the numerical model is composed of three parts, corresponding to the fluid flow, mass transfer, and energy transfer. As the gas flows through the channel, the fuel is heated by convection while reacting with the OC to generate CO2 and steam, which further increases the flow velocity. The mass diffusion of the fuel into the OC layer and CO2 and H2O out of the same layer and the convective heat transfer between the solid and gas both depend upon the local flow. Therefore, all three parts, i.e., the fluid flow, mass transfer, and energy transfer, in the reactor are solved simultaneously. 3.2. Governing Equations. The conservation of the gas species i in the bulk flow and porous OC layer can be expressed as (Figure 10a)

Figure 5. Schematic layout of the (a) individual channel structure and (b) OC coated on the surface.

⎡ ∂(Xb, iC b) ∂(ubXb, iC b) ⎤ ⎥ = −PcJi + A p⎢ ∂t ∂z ⎦ ⎣

(1)

⎡ ∂(εsXs, iCs) ∂(εsusXs, iCs) ⎤ ⎥ = PcJi + Pcωg, i + Aoc⎢ ∂t ∂z ⎦ ⎣

(2)

2

where Ap is the cross-section area (m ) of the channel. Aoc is cross-section area (m2) of the porous OC layer. Pc is the channel perimeter (m). Cb and Cs are the local molar densities (mol/m3) in the bulk flow and porous layer, respectively. Xs,i and Xb,i are the mole fractions of the gas species i in the porous OC layer and bulk flow, respectively. ub and us are the bulk velocities (m/s) in the plug flow and porous layer, respectively. εs is the volume fraction of the gas in the porous layer. ωg,i is the overall molar generation rate (mol m−2 s−1) for species i in the porous OC layer. Ji is the external mass diffusion rate (mol m−2 s−1) from the bulk flow to the surface of the porous layer. The species diffusion flux is calculated on the basis of

Figure 6. Schematic profiles of the gas species concentration at the (a) inlet and (b) exit for two cycles. Each cycle includes fuel, fuel-purging, air, and air-purging sectors. is evaluated on the basis of the model discussed in the next section. The simulation results as well as the sensitivity analysis will be presented in part 2 (10.1021/ef301411g).

3. NUMERICAL MODEL 3.1. Model Description. The modeling of the rotary reactor starts with analyzing an individual channel during a full cycle. As the wheel rotates at a constant angular velocity, each 332


Energy & Fuels

Article

Figure 7. Simplified layout of the rotary CLC cycle. The red lines show the carbon flow pattern through the system. The feed purge streams are omitted in this layout for clarity.

Table 1. Properties of the OC and Support Materials Used in the Base Case OC support material density of the bulk support layer porosity of the porous OC layer conductivity of the support active CuO load volume fraction of CuO in the OC layer volume fraction of Cu in the OC layer pre-exponential factor for reduction (eq 10) pre-exponential factor for oxidation (eq 11) activation energy for reduction (eq 10) activation energy for oxidation (eq 11) pressure coefficient for reduction (eq 10)a pressure coefficient for oxidation (eq 11) reaction order for reduction (eq 10) reaction order for oxidation (eq 11)

symbol

value

CuO BN ρs

copper oxide boron nitride 3450

εs

0.57

ks

Table 2. Reactor Design and Configurations for the Base Case

unit channel width channel height reactor diameter cross-section area ratio of solid (both layers) thickness of solid (dense support and porous OC layers) thickness of the porous OC layer (CuO supported by BN) thickness of the dense support (BN) layer size of the air sector at the feeding chamber size of the fuel sector at the feeding chamber size of the air-purge sector at the feeding chamber size of the fuel-purge sector at the feeding chamber size of the air zone at the exit chamber size of the fuel zone at the exit chamber

kg/m3

Wm wt %

−1

K

−1

εCuO

600 10 0.0281

εCu

0.0161

k0,CH4

1.125 × 106

mol0.6 m−1.8 s−1

k0,O2

2.043 × 104

mol0.0 m0.0 s−1

ECH4

60

kJ mol−1

E O2

15

kJ mol−1

aCH4

0.83

aO2

0.68

nCH4

0.4

n O2

1.0

value

unit

2 1.0 1.70 0.45 260

mm m m μm

δoc

50

μm

δsupport θair θfuel θair_p

210 7π/15 π 2π/15

μm rad rad rad

θfuel_p

2π/5

rad

θair_zone θfuel_zone

3π/5 7π/5

rad rad

Table 3. Operation Conditions for the Base Casea operating pressure fuel inlet temperature air inlet temperature purging steam temperature volume fraction of fuel at the inlet cyclic period time air flow velocity fuel flow velocity steam velocity of the fuel-purging sector steam velocity of the air-purging sector

a

The coefficient is for CuO reduction with CO. There is no available data for CH4.

Ji = hm, i(Xb, i − X s, i)C b

symbol d H D εM δs

(3) a

where hm,i is the mass-transfer coefficient (m/s) for species i derivable from the Sherwood number. For laminar flow in the square-shape channel, the Sherwood number is 3.61 for the uniform wall flux model and 2.98 for the uniform wall concentration model.45 For this work, the Sherwood number of 3.61 is used because, in the periodic state, the reactant (or product) concentration decreases (or increases) from the inlet to the exit in a similar way as in the uniform wall flux model. The axial gas diffusion is generally much smaller than the convection terms, and thus, it is neglected in eqs 1 and 2. Equations 1 and 2 describe the gas species transport among the bulk flow, porous spaces in the solid OC layer, and

symbol

value

unit

P Tfuel Tair Tpurge

10 823 823 823 15 30 0.7 0.09 0.11 0.7

atm K K K vol % s m/s m/s m/s m/s

τ uair ufuel ufuel_p uair_p

Velocities are evaluated at the local temperature (inlet temperature).

heterogeneous reaction. The gas components are transported from the bulk toward the OC layer surface by external diffusion (Ji) and then are consumed as species i diffuses through the porous layer (ωg,i). Some of the gas species are also stored or transported by convection in the porous layer. As implied in Table 2, the gas volume in the porous OC layer (εsAoc) is one order of magnitude smaller than that in the channel (Ap). Thus, the amount of gas contained in the pores within the porous OC layer is negligible compared to that in the bulk flow. The bulk 333


Energy & Fuels

Article

which means that all of the gas that diffuses from the bulk flow into the porous layer will be completely consumed and all of the product molecules that are generated from the heterogeneous reaction in the porous layer will be transported out to the bulk flow. The effect of internal mass diffusion is described in the extrinsic reaction rate, ωg,i, and it will be discussed in section 3.3. The conservation of the OC (CuO or Cu) in the porous OC layer is described as Aoc

∂[(1 − εs)Coc, i] ∂t

= Pcωoc, i

i = CuO or Cu

(5)

where 1 − εs is the volume fraction of the solid in the porous layer. Coc,i is the CuO/Cu molar concentration (mol/m3), and ωoc,i is the extrinsic molar reaction rate (mol m−2 s−1) for the OC. The overall mass conservation equation for the bulk flow can be obtained by summing eq 1 for all species, as follows: ⎡ ∂C ∂(ubC b) ⎤ A p⎢ b + ⎥ = −Pc ∑ Ji ⎣ ∂t ∂z ⎦ i

(6)

Only n − 1 species equations (eqs 1 and 2) are solved because the mole fractions, Xi, sum to unity. The overall mass conservation, eq 6, can be rewritten in terms of the pressure, p, using the ideal gas law. The energy conservation equation for bulk flow is given by (Figure 10b) ⎡ ∂Eg ∂(ubHg) ⎤ ⎥ = −PcQ gs − + A p⎢ ⎢⎣ ∂t ∂z ⎥⎦

∑ Pchs,̂ iJi

(7)

i 3

where Eg and Hg are the energy and enthalpy (J/m ) of the bulk flow, respectively. ĥs,i is the molar enthalpy (J/mol) for species i and is evaluated at the temperature of the porous layer. Qgs is the interphase heat-transfer rate (W/m2) as a result of heat convection and can be calculated from

Figure 8. Design configuration of the reactor: (a) channels and (b) sectors.

Q gs = hgs(Tg − Ts)

(8)

where Ts and Tg are the bulk temperatures (K) of the solid (including both layers) and the flow. hgs is the convective heattransfer coefficient (W m−2 K−1) derivable from the Nusselt number. The Nusselt number used in this model is 3.61 from the uniform wall heat flux laminar heat-transfer model.45 The energy conservation equation for the solid phase (including both layers, as shown in Figure 10b) can be written as follows: As

∂Es ∂ ⎛ ∂T ⎞ = A s ⎜ks s ⎟ + PcQ gs + ∂t ∂z ⎝ ∂z ⎠

∑ Pchs,̂ iJi i

(9)

2

where As is the cross-sectional area (m ) of the solid phase. Es is the energy of the solid phase (J/m3), including both the support material and the OC. ks is the conductivity (W m−1 K−1) of the solid phase. As discussed in section 2.2, the size of the bulk layer is much larger than that of the porous layer [(1 − εs)δoc/(δs − δoc) ∼ 0.1]; thus, the thermal processes within the solid phase are mainly determined by the properties (heat capacity, density, and thermal conductivity) of the bulk layer. Equations 7 and 9 characterize the energy balance between the bulk flow and the solid phase by means of convective heat transfer, conductive heat transfer, and energy transfer associated with species transfer. The solid phase behaves as a heat

Figure 9. Modeling for the reactive flow in one channel for the (a) fuel zone and (b) air zone. T is the temperature. pfuel and pO2 are the partial pressures of fuel and oxygen, respectively.

convection velocity in the porous layer is lower than the velocity in the bulk flow because of the surface drag on the flow. Thus, the convection term is also negligible. Therefore, eq 2 can be greatly simplified 0 = PcJi + Pcωg, i (4) 334


Energy & Fuels

Article

Figure 10. Schematic drawing of the (a) species transport and (b) energy transfer in the channel.

temperatures up to 1000 °C. Thus, the gaseous reactant (or product) can easily diffuse through (or out of) the porous medium. Each grain of OC is exposed to the ambient gas concentration, and hence, the active metal oxide in each segment within the porous medium is converted in the same way as that in the plate-like USCM. Thus, Garcia-Labiano et al.43 simplified the USCM as follows:

reservoir during operation to temporarily store the energy from the heterogeneous reactions in the porous layer, conduct that heat from the hot to the cold locations, and finally, heat the bulk flow by convection. The solid phase, especially the bulk support layer, is critical to the overall energy balance in the reactor. The energy equation for the gas phase in the pore volumes in the porous OC layer is neglected because the gas temperature there is almost identical to the solid temperature. Because the pressure is assumed to be constant throughout the channel, the momentum equations are excluded from the governing equations. Therefore, eqs 1, 4−7, and 9 completely characterize the fluid flow, mass transfer, and energy transfer within the channel. The boundary conditions at the inlet, z = 0, are specified by the inlet conditions (flow rate, species concentration, and temperature) for each sector, as shown in Figure 6 and Table 3. 3.3. Reaction Kinetics. The reaction mechanism used here is based on the one-step extrinsic (mass-specific) kinetics proposed by Garcia-Labiano et al.43 The kinetics data43 were obtained by thermogravimetry (TG) at atmospheric pressure with CuO−Al2O3 as OC particles. The operating temperature in TG ranges from 823 to 1073 K. The OC is prepared by wet impregnation with 10% (by weight) active metal oxide load. The preparation method, active metal oxide load, and porosity of the OC are the same as those assumed in this study, as shown in Table 1. The diameter of OC particles is within 0.1− 0.3 mm. The reduction and oxidation reactions are given by the following equations:

voc, iki n dX = Cs, i dt ρm,̂ i

where ρ̂m,i (mol/m3) is the molar density of CuO for reduction or Cu for oxidation. voc,i is the stoichiometric coefficient. ki is the Arrhenius reaction rate constant (m(3n−3) mol(1−n) s−1). Cs,i is the gas concentration (mol/m3) of CH4 (for reduction) and O2 (for oxidation) at the porous surface, and n is the reaction order. The kinetic parameters are shown in Table 1. X is the non-dimensional OC conversion, defined as X=

Coc,CuO Coc,CuO + Coc,Cu

(13)

where Coc,CuO and Coc,Cu are the molar concentrations (mol/ m3) of copper oxide and copper, respectively. X equals zero when the OC is fully reduced and unity when fully oxidized. Figure 11 shows the OC conversion rate (dX/dt) for reduction and oxidation for a typical range of gas concentrations and operating temperatures. As seen in Figure 11, the reduction reactivity is generally lower than the oxidation rate. For reduction, as the temperature rises by 100 K, the reaction rate doubles, while for oxidation, only 10% increase is observed. However, the oxidation is more sensitive to the gas species concentration than the reduction. To apply these reaction kinetics to the rotary reactor, attention must be paid to differences in the operating conditions used in the experiments43 and those in the rotary channel design. The experiment conducted by Garcia-Labiano et al.43 is for CuO−Al2O3 spherical particles at atmospheric pressure with temperatures up to 1073 K, while the numerical model in this work is for the reactive flow over the CuO−BN plate-like layer at higher pressures and higher temperatures. The discrepancies in these aspects may have large impacts on the reactor performance. On the basis of the analysis in the following three sub-sections, it is reasonable to scale the kinetics to the current design conditions to obtain qualitatively reliable results.

4CuO(s) + CH4(g) → 4Cu(s) + CO2 (g) + 2H 2O(g) (10)

2Cu(s) + O2 (g) → 2CuO(s)

(12)

(11)

The extrinsic kinetics obtained by Garcia-Labiano et al.43 is based on the unreacted shrinking-core model (USCM).46 The USCM assumes that, as the reaction progresses, it leaves behind a layer of product that consists of reacted solid, i.e., Cu for reduction and CuO for oxidation, and the binder. As the reaction proceeds, the thickness of the product layer increases, producing a gradually shrinking unreacted core of the OC, although the thickness of the OC remains constant. Garcia-Labiano et al.43,47 further pointed out that, for CuO prepared by wet impregnation, the active metal oxide is uniformly dispersed in the porous surface of the support material. The external and internal mass-transfer resistances are less important than the chemical kinetics under operating 335


Energy & Fuels

Article

In general, the total reaction rate increases with increasing the total pressure48−52 because of the increase in the reactant partial pressure. However, Garcia-Labiano et al.10 reported that this increase in the reaction rate is lower than the expected increase because of the saturation of gas molecules in the pore cavity. This pressure-inhibition effect can be expressed as10

⎛ P ⎞−a =⎜ ⎟ k ⎝ P0 ⎠

kp

(14)

where kp is the Arrhenius reaction rate constant at pressure P and a is the pressure coefficient. k is the constant at atmospheric pressure, P0. From eq 14, it is clear that a higher operating pressure leads to a lower reaction rate constant kp. In this study, eq 14 is used to account for the pressure effect. The coefficient a for CuO is listed in Table 1. Note that the coefficient a is provided only for CO (a = 0.83), H2 (a = 0.53), and O2 (a = 0.68).10 Thus, the larger coefficient (a = 0.83) is used to give a conservative evaluation of the pressure effect on the CH4 reduction reaction. In the experiment conducted by Garcia-Labiano et al.,43 the operating temperature ranges from 823 to 1073 K, while in the current study, the temperature near the exit of the channel could be as high as 1300 K. Under high operating temperatures, the effects of mass-transfer resistance start to become significant, and hence, one may expect a lower reactivity than that described in eq 12. Generally, it is necessary to use the redox reaction kinetics specifically for this temperature range (823−1300 K). However, limited redox experimental data are available in the literature for high-temperature CLC.17 As mentioned before, for CuO prepared by wet impregnation, the active metal oxide is uniformly dispersed on the porous surface of the support material. The mass-transfer resistances are less significant than the chemical kinetics of reduction (eq 10) and oxidation (eq 11) for operating temperatures up to 1300 K.43,47 This conclusion is also validated in the next sub-section and in the simulated results shown in part 2 (10.1021/ef301411g). Therefore, it is reasonable to extend the kinetics43 to the current operating conditions. 3.3.3. Effect of the Surface Curvature. Experiments on the redox reactions of the OC43 have been conducted using spherical particle samples in TG with dp ranging from 0.1 to 0.3 mm, while in this study, the porous layer of the same material is used with essentially a plate-like geometry. The geometry of the OC may affect the heterogeneous reactions in two ways. External diffusion resistance: As shown in Figure 12, a larger curvature of the particle in traditional applications is accompanied by a higher surface area being exposed to bulk flow. This influences the external mass-transfer resistance from the bulk flow to the surface of the OC. Internal diffusion resistance: OCs with different shapes and sizes have different diffusion distances through the porous medium. This leads to different internal mass-transfer resistances within the OC. In this study, the external mass transfer is readily characterized by eq 4. However, the effect of the internal resistance is inherently integrated into the extrinsic kinetics. As mentioned before, the extrinsic kinetics obtained by GarciaLabiano et al.43 assumed limited internal mass diffusion resistance. Therefore, to apply the kinetics,43 special attention must be paid to the selection of the thickness of the porous layer to ensure that the internal mass diffusion resistance in the

Figure 11. OC conversion rate (dX/dt) as a function of the gas species concentration and operation temperature. Solid lines are for reduction, and dashed lines are for oxidation.

3.3.1. Effect of the Support Material. In this study, BN is used as the support material because it has a much higher thermal conductivity (∼600 W m−1 K−1) than alumina (∼30 W m−1 K−1). Because of the excellent thermal and chemical stabilities, BN ceramics are commonly used as part of hightemperature equipment. The effect of the binder on the reactivity is mainly related to the regenerability of the OC. During periodic operations, the inert binder acts as a porous matrix to help improve the gas permeability, maintain the pore structure, and sustain the thermal and mechanical stresses. Thus, support material with similar properties should exhibit similar periodic performances in CLC. Alumina and BN have very similar physical characteristics: both show excellent thermal, chemical, and mechanical stabilities at high temperatures. In their crystalline form, c-BN and α-Al2O3, the hardness of these two materials makes them suitable as an abrasive medium and as a component in cutting tools. Both alumina and BN are electrical insulators but good thermal conductors. When in contact with oxygen at high temperatures, a thin passivation layer of B2O3 forms to prevent further oxidation in a way similar to Al2O3 on aluminum. Thus, it is reasonable to adapt the mechanism obtained by Garcia-Labiano et al.43 to the BN-supported OC in this work. 3.3.2. Effect of the Total Pressure and Operating Temperature. Few experiments have been conducted to investigate the OC behavior at high pressures, and a comprehensive examination of the pressure effect on reactivity in CLC is still missing.17 One approach to account for the pressure effects is to adapt the results from heterogeneous surface-catalyzed reactions (e.g., Langmuir−Hinshelwood mechanism). However, reactions in CLC are primary noncatalytic, and OCs act as sources of undiluted oxygen with relatively high quantities of active material, in contrast to heterogeneous catalysts. The above difference makes it insufficient to apply directly our knowledge of catalytic reactions to CLC. 336


Energy & Fuels

Article

ΔCs, i =

porous layer is indeed small and the active metal oxide in the porous medium is converted in the same way as that described by Garcia-Labiano et al.43 Thus, the following discussion focuses on how the gas components diffuse within the porous medium in a spherical particle (in ref 43) or a plate-like layer (in this study). As shown in Figure 12, the internal mass diffusion resistance affects the reaction in a way such that the gas concentration decreases as it diffuses through the OC. Therefore, the reactivity at the center of the OC is lower than that on the surface. However, as pointed out by Garcia-Labiano et al.,43 an OC prepared by wet impregnation exhibits a well-dispersed metal oxide distribution in the porous surface of the support material. The internal mass-transfer resistance is less important in such CLC systems, and the conversion of the OC is mainly limited by slower chemical reaction rates.17,43,47 The variation in the gas concentration within the porous medium is thus limited, and the OC is reduced or oxidized at a highly uniform rate throughout the entire particle. A simple estimation can be made to validate this conclusion. In the porous layer, the species conservation is expressed by

∂t

+ ∇(εgρg ugYg, j) = −∇(εgJg, j ) + εgrg,̇ j

−

(15)

(16)

where Da is the diffusion coefficient (m2/s) in the porous medium, Cs,i is the molar density (mol/m3), and ṙg,i is the reaction rate (mol m−3 s−1). Integrating the above equation over the cross-section for a spherical particle (Figure 12a) and a plate-like layer (Figure 12b), respectively, we can obtain the maximum concentration drop through the medium as ΔCs, i =

rg,̇ id p2 24Da

(18)

ωg, i vg, i

=−

ωoc, i voc, i

=

δocεiρm,̂ i dX voc, i dt

(19)

where vg,i is the stoichiometric coefficient for the gas species and εi is the volume fraction of CuO (or Cu) in the porous OC layer. 3.4. Numerical Implementation. Because of the highly nonlinear and stiff nature of the system of governing equations, the method of lines has been applied to transform the governing equations into a system of ordinary differential equations (ODEs) with finite volume discretization along the spatial coordinate. The resulting system of ODEs is then integrated in time using a fully implicit scheme. The code was written in MATLAB, and the temporal integration was conducted with the ode15s solver.53 The physical domain is uniformly discretized into 100 cells using the control volume formulation covering the entire channel from the inlet to the outlet. All state variables are calculated at the centers of the control volumes. All advection terms are evaluated using upwind difference approximation. Gas- and solid-phase properties are calculated dynamically as functions of the local state variables. The heat capacities, enthalpies, and conductivities of the gas mixture and the solid phase are evaluated as a function of the temperature using the values from the National Institute of Standards and Technology (NIST) property database.54 Binary diffusion coefficients are calculated according to the work by Fuller et al.55 The flow diagram for the overall simulation of the model is shown in Figure 13. On the basis of the specified design

By assuming that the transient and convection terms are negligible and the porosity, mass diffusion coefficient, and reaction rate are constant, we can simplify eq 15 as ∇(Da ∇Cs, i) + rg,̇ i = 0

for a plate‐like layer

2Da

where dp is the particle diameter (m) and δoc is the thickness of the porous layer (m). Da is the effective diffusivity coefficient. For the reduction reaction at 10 atm and 1273 K, the relative concentration drop ΔCs,i/Cs,i is less than 0.5% and the relative difference in reactivity between the core and the outer surface of the OC particles is less than 0.5%. Similar results are obtained for the oxidation reaction. This estimation validated the conclusion obtained by Garcia-Labiano et al.43 (and section 3.3.2) that the internal mass-transfer resistance is small relative to the chemical reactions (even at higher operating temperatures, as in section 3.3.2). We can select the thickness of the porous OC layer by matching the maximum concentration drop between the spherical particle case (eq 17) and the plate-like layer case (eq 18); i.e., δoc = d/2√3. Therefore, the OC particle diameter dp ranging from 0.1 to 0.3 mm corresponds to the OC layer thickness (δoc) between 29 and 87 μm. Similar results can be obtained from the USCM,46 as shown in Appendix A. Therefore, for the plate-like OC layer with a thickness lying within the above range, the maximum species concentration drop through the medium should be small (

Rotary Bed Reactor for Chemical-Looping Combustion with Carbon

Recommend Documents