Performance of a Natural Gas Solid Oxide Fuel Cell System With and Without Carbon Capture

The solid oxide fuel cell (SOFC) technology through its nearly reversible electrochemical conversion of chemical potential into electric power has the potential for significantly higher electric efficiency power systems relative to conventional Carnot-cycle-based heat engines. In addition, the oxy-combustion of syngas in a sealed SOFC system renders itself readily available for carbon capture and sequestration (CCS) with the requirement of only a small oxy-combustor downstream of the fuel cell to combust the fuel that is not utilized electrochemically. (Electrochemical utilization of fuels varies typically between 75% and 90% for current fuel cell technology due to practical considerations.) The heat rejected by the fuel cell system can be recovered further in a combination of Brayton and Rankine cycles, depending on whether the fuel cell system is operating at elevated or atmospheric pressures.

Accordingly, the United States (US) Department of Energy (DOE) National Energy Technology Laboratory (NETL) has been pursuing the development of the SOFC technology to enable future power generation systems that are consistent with the cornerstones of the DOE mission—energy security, energy reliability, and environmental sustainability. The US DOE fuel cell program is currently focused on the development of low-cost, highly efficient, and reliable fossil-fuel-based SOFC power systems [1]. NETL’s SOFC technology development roadmap shown in Fig. 1 is aligned with near-term market opportunities in the distributed generation sector to validate and advance the technology while paving the way for utility-scale natural gas (NG)- and coal-derived synthesis gas-fueled applications via progressively larger system demonstrations [2].

National Energy Technology Laboratory is continually carrying out analytical evaluations of the SOFC systems to aid in prioritizing technological advances along research pathways to the realization of utility-scale SOFC systems, which is the ultimate objective of the US DOE fuel cell program (Fig. 1). NETL’s previous pathway studies [3,4] are based on simplified representations of the SOFC performance at only a specific design point. While a variety of detailed computational models can be employed to predict the electrical performance along with the thermal and flow fields of SOFCs, these models are generally computationally intensive for use in system-level thermodynamic and material balance models typically used in techno-economic analyses. Through a partnership with the Pacific Northwest National Laboratory (PNNL), NETL has developed a reduced order model (ROM) that is based on detailed computational models of the SOFC stack. This model, which is referred to as the PNNL-ROM model here, is not computationally intensive, by design, and allows for effective use in system-level models. The PNNL-ROM model, in addition to the capability to predict the stack performance over the entire range of operating conditions, is designed to provide other desirable information about the state of the stack, such as internal temperature gradients, which can be critical constraints in system optimization studies.

National Energy Technology Laboratory is utilizing the PNNL-ROM tool to refine the earlier pathway studies. The performance results of NG fuel cell (NGFC) systems, which form the first of the series of these evaluations, are presented and discussed in this study. Specifically, the performance of an atmospheric NGFC system was analyzed under different operating configurations and conditions. The impact of advances in underlying SOFC technology on electrical performance was also explored.

A block flow diagram of the NGFC system with carbon capture is shown in Fig. 2. Desulfurized NG with less than 100 ppbv of sulfur is sent to an oxygen (O₂)-driven auto-thermal reformer (ATR) where it is converted to syngas, which fuels the SOFC system, subsequent to expansion to the desired SOFC operating pressure. Part or all of the desulfurized NG may bypass the ATR to increase the methane content of the syngas entering the SOFC, which offers cost and performance advantages. Complete internal reformation (IR; internal to the module), while eliminating the need for specialized process equipment (ATR), utilizes heat generated in the stack directly for the endothermic reformation reaction and reduces the airflow rate needed to maintain a desired stack temperature gradient resulting in higher process efficiency. To prevent cracking and deleterious carbon formation, a pre-reformer, which converts the higher hydrocarbons into methane, is generally included before completing the reformation internal to the stack. The anode off-gas recirculation rate is maintained at a suitable oxygen-to-carbon (OTC) ratio to avoid carbon formation anywhere in the fuel path. The oxidant is supplied to the cathode by an air blower, and heat exchangers on both the anode and cathode side are appropriately designed to keep the desired temperature gradient across the stack. A cathode-gas recirculator is also included to modulate the air-side heat exchanger size. The heat from burning the electrochemically unutilized fuel to completion in an O₂-driven (oxy-) combustor is transferred to a steam bottoming cycle. Waste heat from the cathode exhaust is transferred to the steam cycle after satisfying the steam requirements of the cryogenic air separation unit (ASU), which is also used to supply the O₂ to the ATR and the combustor, to enable efficient capture of carbon dioxide (CO₂). The cooled anode exhaust from the heat recovery steam generator (HRSG) is further purified to pipeline specifications in a CO₂ purification unit (CPU). For systems without carbon capture, shown in Fig. 3, the O₂ supply to the reformer and the combustor may be substituted with air.

Systems models were developed under the Aspen Plus^® (Aspen) platform to simulate the NGFC process configurations. Performance and process limits were based upon published reports, information obtained from vendors and users of the technology, performance data from design/build utility projects, and/or best engineering judgment as described in the NGFC pathway study [3]. The plants are fueled by NG (composition shown in Table 1) and are assumed to be located at a generic Midwestern site operating at International Standards Organization (ISO) ambient conditions [3].

The SOFC simulations represent the expected operating conditions and performance capabilities of planar fuel cell technology, having split cathode and anode off-gas steams, and operating at atmospheric-pressure. Several assumptions were applied to estimate the performance of the NGFC power island. These assumptions were obtained from SOFC test data and internal vendor reports. The salient process components are described next. The reader is referred to the NGFC pathway study [3] for a detailed description of the components and the assumptions.

The PNNL-ROM model was used to predict the SOFC performance along with the cell thermal fields for a given airflow and fuel flow. The PNNL-ROM model is based on the response surface methodology [5] with detailed SOFC stack model results to create a computationally efficient ROM for the stack that retains desirable information about its internal state. The response surface approach attempts to describe the behavior of a dependent or “response” variable of interest based on the independent or “explanatory” variables that are inputs to a given process or model. For the SOFC, the input variables could constitute operational parameters such as fuel/oxidant compositions, temperatures, flowrates, and utilizations. Response variables could include electrical performance characteristics such as stack voltage, power output, and efficiency. Peak cell/stack temperature, cell temperature gradient, and maximum local current density are among other response variables that may be of interest due to their influence on-cell/stack structural stability and performance degradation.

The developed numerical process for ROM generation is shown schematically by the flowchart in Fig. 4. The number of different model input variables and their range of values are assigned based on expected SOFC operating conditions. This defines the multi-dimensional design space for the problem, and random sampling is performed to define sets of parameters. Each sample set defines a modeling case to be run. The detailed stack model then runs each of the defined cases and collects the results. Based on the expected usage of the ROM, the output parameters of interest are also defined. The corresponding values for these parameters are then extracted from the case results, and regression is performed to a selected fitting function. The regression results provide a mathematical relationship describing the predicted response surface for each response variable as a function of the input variables. These response surfaces are then exported in a suitable format for integrating the ROM into the system model. This ROM then provides a computationally efficient predictor of a given stack’s performance without integrating the entire detailed model directly into the system model and adding further nonlinear convergence iterations to that solution. Further details of the PNNL-ROM model, the regression methodology, and its validation can be found in Ref. [6].

Although the ROM procedure is applicable to comprehensive 3D models, the stack is presently modeled using PNNL’s SOFC-multi-physics (MP) software [7] that solves for the steady-state cell distributions of species, temperature, and current density. The underlying 1D channel cell model represents a 550 cm² active area anode-supported cell with metal interconnects in a counterflow configuration. The model assumes the water-gas shift reaction is in equilibrium but uses a first-order kinetic expression for the slower on-cell steam reforming of methane with a Ni/YSZ anode [7].

The SOFC direct current (DC) voltage is converted to alternating current (AC) power in an inverter with an assumed efficiency of 97% in the present study.

The SOFC stack inlet anode gas composition can induce the formation of solid carbon deposits, which can disrupt the normal performance of the stack [8]. Anode gas recirculation is used to control the anode gas inlet conditions to maintain an atomic OTC ratio greater than 2.1, which is a generally used criterion to prevent carbon deposition anywhere in the SOFC fuel flow domain [9]. Anode gas recirculation is accomplished using hot gas blowers and serves to keep the fuel utilization within the stack below acceptable limits from a flow distribution perspective.

The anode side heat exchanger heats the incoming mixture of fuel (syngas) and recirculated anode off-gas before entering the pre-reformer using the hot anode off-gas exhaust. The heat exchange was defined by assuming an approach temperature of 15 °C between the exhaust of the hot stream and the outlet stream of the heat exchanger.

At temperatures greater than ∼500 °C, instead of following a reaction pathway to form syngas, ethane, and higher hydrocarbons tend to crack and form carbon [9]. A catalytic pre-reformer is included to adiabatically reform the higher hydrocarbons to syngas using the sensible heat in the stream exiting the anode heat exchanger.

A portion of the vitiated cathode off gas is recirculated via a high-temperature blower to mix with the process air heated by the SOFC cathode off gas exhaust through a recuperator. The heat exchanger effectiveness is varied depending on the recirculation rate to obtain the desired air inlet temperature into the stack.

The NG expander, shown in Figs. 2 and 3, expands the NG from its high-pressure condition down to the operating pressure of the reformer unit. NG is heated using the high-temperature exhaust syngas stream from the ATR to maximize the power that can be extracted from the expansion while cooling the syngas stream being piped to the SOFC module.

The anode off-gas stream is combusted with O₂ for the cases with carbon capture, while a portion of the SOFC cathode exhaust is utilized for combustion in cases without carbon capture. In both cases, the hot stream from the combustor exchanges heat in a HRSG to produce steam, which generates power in a subcritical steam bottoming cycle after satisfying process steam requirements. The subcritical steam cycle varies greatly in its steam conditions and capacity in the study cases, providing a relatively small proportion of the total plant generation output. In some cases, the heat recovery temperature available is relatively low and results in poor steam superheat conditions. Rather than performing detailed designs for each of these unique steam bottoming cycles, the steam cycle performance was estimated using a nominal efficiency of 38.1%, which was based on simulations of steam cycles in select pathway cases [4]. This approach has been shown to result in efficiencies that could be ±1.5 percentage points lower or higher than fully simulated steam cycle model values depending on whether the SOFC is operating at atmospheric or elevated pressures, respectively. The auxiliaries for the steam cycle were scaled from the Bituminous Baseline studies [10] based on the steam cycle power output.

The NG is assumed to be desulfurized from its assumed 5 ppmv total sulfur content to 10 ppbv total sulfur using the low-temperature TDA Research Inc. SulfaTrap^TM sorbent before it is introduced to the plant [11]. The ASU, ATR, oxy-combustor, and CPU form the other major components of the balance of plant.

A cryogenic ASU is utilized to supply O₂ at 99.5% purity to the ATR and the oxy-combustor. The parasitic load of the ASU is modeled to be ∼236 kWh/ton of O₂ in accordance with values used in the Bituminous Baseline studies [10].

The NG is reformed with steam and oxidant in an ATR to generate a high-heating value syngas, which is considered to be an effective method to convert NG into a syngas [12,13]. In the present process representation (Figs. 2 and 3), the percentage of external reformation of the NG can be modified by diverting a desired portion of the inlet NG into the ATR to facilitate sensitivity studies.

The anode off-gas is combusted using 99.5% oxygen from the ASU in an advanced oxy-combustor with excess O₂ limited to 1 mol%. The combusted anode gas consists of CO₂, water vapor, excess O₂, and minor trace contaminants (sulfur species and oxides of nitrogen).

The cooled combusted anode gas after heat exchange in the HRSG is sent to the CPU for compression, drying, and purification. In the CPU, the CO₂ is purified cryogenically by liquefaction followed by flash separation and distillation to yield an O₂ content of 10 ppm in the CO₂ product stream in accordance with the specifications for pipeline transport and for potential use in enhanced oil recovery (EOR) operations or for storage in saline formations [14]. The cooling for the liquefaction is accomplished by an auto-refrigeration process that takes advantage of the Joule–Thomson effect to minimize the parasitic load associated with the cryogenic CO₂ purification process [15].

The input parameters and the ranges used to generate the PNNL-ROM model for the present analysis are listed in Table 2. The global ROM input parameters such as the current density, the internal reformation fraction, the system fuel utilization, the OTC ratio, and the cathode exhaust recirculation rates were prescribed. The other ROM input parameter values such as the stack fuel inlet temperature, stack air inlet temperature, and the initial stack airflow were dynamically extracted from relevant material streams from the NGFC Aspen model flowsheet into a calculator block embedded within the model. The ROM-generated response surface regression coefficients were accessed by the calculator block through an external data file. The output parameter values corresponding to the extracted input values were computed in the calculator block using the response surface equations along with regression coefficients.

While the ROM provided a variety of output parameters, the cell voltage, the cell/stack maximum temperature (T_max), and the difference, cell ΔT = T_max − T_min (where T_min represents the cell minimum temperature) formed the primary variables of interest to the present calculations. The cell ΔT is a measure of the cell thermal gradient,¹ which generally corresponds directly to the thermal stresses in the cell and is used as a surrogate for the structural state of the cell in the present study (assuming other mechanical and thermal constraints remain the same). Both the T_max and cell ΔT accordingly served as constraints. For the stack fuel and air inlet temperatures extracted at each Aspen iteration loop, the calculator block iterated over a range of stack oxidant utilization fractions to determine the minimum value of the cathode airflow rate to meet the specified constraints on T_max and ΔT. The resulting cathode airflow rate and the corresponding cell voltage values were used to update the relevant aspects of the Aspen flow sheet for the next outer Aspen iteration loop. The ROM-integrated NGFC model took less than a minute to converge in most cases.

The SOA case from the NGFC pathway studies [3] served as a reference plant for the present analysis. The prescriptions for the reference plant with a rated power of 550 MWe were generally based on current SOFC technology and are shown in Table 3. Although the fuel utilization is high, an OTC value of 2.6 was used to ensure that the stack fuel consumption was 80%. The performance of the reference plant for cases with and without carbon capture is shown in Table 4. SOFC accounts for nearly 85% of the gross power generated as shown in Fig. 5.

Without the utilization of the waste heat in the steam bottoming cycle, the higher heating value (HHV) efficiency decreases by nearly 7 percentage points. The CPU and the ASU account for ∼58% of the auxiliary loads, as shown in Fig. 6, in the case of the NGFC plant with capture, which imposes an HHV efficiency penalty of 5.8 percentage points relative to an NGFC plant without CCS (see Table 4).

As is well recognized, the internal reformation fraction has a considerable influence on the system efficiency. The thermochemical recuperation effect and the associated reduction in the required air stoichs of complete internal reformation, as shown in Fig. 7, have the potential to increase the system efficiency by over 15 percentage points. Note that the cell ΔT is maintained at 100 °C in these cases to eliminate the confounding effects of chilling due to internal reformation often cited as an adverse impact on-cell thermal stresses. Also shown in the figure is the corresponding variation in the cathode heat exchanger overall conductance, UA (overall heat transfer coefficient *surface area), a quantity that is generally used as a measure of the size (and cost) of a heat exchanger. The cathode heat exchanger size for the 100% IR case is nearly 25% of the heat exchanger size required for the complete external reformation case (0% IR).

Operating at higher current densities while decreasing the overall system efficiency can result in higher power density. Accordingly, the number of SOFCs required to meet the given power rating of 550 MWe decreases with increase in operating current density as shown in Fig. 8. The optimal operating point can be discerned as a trade-off between the plant capital cost ($/kWe) as determined by the number of SOFCs and the operating costs based on plant efficiency. Increase in performance degradation with current density should also be factored in optimizing the overall cost of electricity.

More and more of the fuel is being converted to power at the higher electrochemical conversion efficiencies as the fuel utilization is increased. The SOFC power to steam cycle power ratio shown in Fig. 9 reflects this effect, which contributes to the increase in plant efficiency with increases in fuel utilization. However, practical fuel distribution variations limit this value to generally less than 90% for conventional designs.

The effect of the cell ΔT constraint is shown in Fig. 10. The decrease in the required air stoichs generally results in an overall increase in plant efficiency as the cell ΔT constraint is increased despite parts of the cell operating at low temperatures. Accordingly, the cathode exchanger UA also decreases resulting in perhaps a lower cost for the component. However, the cell thermal stresses increase as the cell ΔT is increased, which could adversely impact system reliability. This is an important factor to consider while optimizing the operating point of the system.

Figures 11 and 12 show the impact of the OTC ratio and the CGRF on the plant performance. The HHV efficiency was found to decrease somewhat with increase in OTC ratio mainly due to the dilution of the stack inlet fuel with the reaction products. However, the actual fuel utilization across the stack, termed the in-stack fuel utilization, as shown Fig. 11, decreases with the increase in the OTC ratio due to the concomitant increase in anode gas recirculation fraction. The lower in-stack fuel utilization values are generally preferred to mitigate adverse effects of fuel flow mal-distributions. The chosen value of 2.6 for the OTC of the reference plant results in an in-stack fuel utilization value of ∼80%, which is consistent with the limits of SOA technology. Increasing the CGRF impacts the plant efficiency favorably since it reduces the amount of fresh air required as seen by the reduction in the air stoichs noted in Fig. 12. However, the main impact of the CGRF is to reduce the cathode heat exchanger size considerably. A 50% CGRF results in a UA value that is ∼1/3 of the value with no CGRF.

Figure 13 shows the potential system efficiency benefits of enhancing the SOFC electrical performance. The US DOE fuel cell program is continually seeking to improve the SOFC electrical performance through innovations in material and manufacturing technologies. To investigate this, the ROM was re-generated with different SOFC parameters that reflect technological advances as a reduction in cell ohmic loss (enhanced or thinner electrolyte layers for example) and/or a reduction in the activation polarization losses (enhanced interlayers or new material sets for example). A 50% reduction in ohmic losses along with a 50% reduction in activation polarization losses relative to the SOA technology (over the temperature range) can result in a gain of 3.4 percentage points in system efficiency as shown in Fig. 13. The benefits were seen to be smaller with further reduction in the cell losses. An overall system-wide loss minimization will also need to be considered in further increasing the plant efficiency beyond the advanced cell performance shown in Fig. 13.

Utility-scale NGFC systems were evaluated with and without carbon capture. A ROM model developed at PNNL was integrated into the system model to predict the underlying SOFC performance while providing other detailed information about the state of the stack, such as the internal temperature gradient and the current density distribution, generally not available from simple performance models often used to represent the SOFC.

The NGFC system performance was analyzed by varying salient system parameters, including the percent of internal NG reformation (ranging from 0% to 100%), fuel utilization, and current density. The impact of advances in underlying SOFC technology on electrical performance was also explored. Imposing carbon capture was found to result in an efficiency penalty of 5.8 percentage points. The considerable advantages of complete internal reformation on the overall system performance as well as on the cathode heat exchanger size were demonstrated.

The ROM model can be easily integrated into system models and is highly effective in enabling sensitivity studies to optimize the system performance and can be a valuable tool in initial system designs.

The enabling technologies specifically for the SOFC power island such as the ability to support 100% internal reformation, high-temperature anode gas blower, and viable heat exchanger designs are key developmental priorities for the commercial distributed generation SOFC systems. The development of the ASU, advanced oxy-combustor, and CPU, which form the other key technologies for the SOFC systems with CCS, is aligned with the general carbon capture technology commercialization efforts.

The NGFC system evaluations are being extended to include costs for the various components. The research and development pathway can utilize these studies in prioritizing the process parameters of importance. These costs studies will be reported in a subsequent paper.

This work was funded by the National Energy Technology Laboratory under the Mission Execution and Strategic Analysis contract (DE-FE0025912) for site-support services, which is gratefully acknowledged. In particular, the authors would like to acknowledge the support of Shailesh Vora, Technology Manager, Fuel Cells, NETL.

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.

There are no conflicts of interest.

J =

average current density

U_f =

fuel utilization

O₂ =

oxygen

cm² =

square centimeter

gpm =

gallons per minute

kJ/kg =

kilojoules per kilogram

kWe =

kilowatts electric

kWh =

kilowatt-hour

lb/hr =

pounds per hour

lb/MWh =

pounds per megawatt hour

mA/cm² =

milliamps per square centimeter

mW/cm² =

milliwatts per square centimeter

ppbv =

parts per billion volume

ppm =

parts per million

ppmv =

parts per million volume

Btu =

British thermal unit

HX, HTX =

heat exchanger

MM =

million

MW =

megawatt

MW/K =

megawatt per Kelvin

MWe =

megawatts electric

MWh =

megawatt hour

Ni/YSZ =

Nickel/Yttria stabilized zirconia

Stoichs =

airflow rate relative to stochiometric airflow

TRL =

Technology Readiness Level

UA =

overall heat transfer coefficient*surface area