Operating characteristics of a molten carbonate fuel-cell power-generation system

INTERNATIONAL JOURNAL OF ENERGY RESEARCH

Int. J. Energy Res. 23: 1331}1344 (1999)

OPERATING CHARACTERISTICS OF A MOLTEN CARBONATEFUEL-CELL POWER-GENERATION SYSTEM

WEI HE*

Process Technology, Technology Centre, Norsk Hydro ASA, Oslo N-0246, Norway

SUMMARY

This paper investigates the operating characteristics of a molten carbonate fuel-cell power-generation system by usingdynamic simulations. The implementation of the system model and the evaluation of the system model performance arepresented and the model e!ectiveness for improving system operation and design is demonstrated. Copyright ( 1999John Wiley & Sons, Ltd.

KEY WORDS: molten carbonate fuel}cell system; dynamic simulation; operating characteristic

1. INTRODUCTION

A fuel cell is an electrochemical device that converts the chemical energy of a reaction into electrical energyand heat, without combustion as an intermediate step. A molten carbonate fuel cell (MCFC) power-generation system provides a new and promising option for the e$cient conversion of fossil fuels to electricitybecause of its high e$ciency (40}60%, based on the low heating value of the fuel) and low negativeenvironmental impact. In order to use accessible resources and to satisfy the load demand of consumer, aswell as to achieve high e$ciency, a fuel-cell power-generation system requires chemical, thermal andelectrical integration. Accordingly, a fuel-cell system often consists of the following four major sub-systems:

(a) fuel processing (converting accessible fuel to appropriate inputs for fuel cell),(b) fuel cell power-generation (producing DC power and heat),(c) power conversion (converting electricity from DC to required AC power),(d) heat recovery (utilizing the heat from fuel cell).

The prediction of the power output characteristics from a complete MCFC system is not straightforwardbecause the power output depends not only on the stack characteristics but also on the gases from the fuelprocessing sub-system and from the heat recovery sub-system. Furthermore, the integration between thefuel-cell stack and the reformer may result in dynamic interactions under load-following operation modes.

A numerical tool is useful to predict the system output power characteristics under load-following modes,to identify potential operating problems and to compare alternative system designs. Since few generalsimulation models are available to analyse the dynamic characteristics of MCFC systems, we have developeda system simulation model based on physical laws. This study is a part of the research project entitled&&Analysis of molten carbonate fuel-cell power-generation systems by using dynamic simulations'' (He, 1998).This paper describes the implementation of the system model, evaluation of the system model performanceand demonstration of the model e!ectiveness for improving system operation and design.

*Correspondence to: Wei He, Process Technology, Technology Centre, Norsk Hydro ASA, Oslo, N-0246, Norway.

CCC 0363}907X/99/151331}14$17.50 Received 26 November 1998Copyright ( 1999 John Wiley & Sons, Ltd. Accepted 26 April 1999

2. MODEL IMPLEMENTATION

The system model comprises a connected network of operational units, which emulate the process #owdiagram. The operational units are inter-connected by the mass and heat streams, and they are individuallyparameterized with design and operating data to activate them.

The software package SPEEDUP (1996) has been selected for the implementation of the fuel cell systemmodel. The structure of a SPEEDUP software package is shown in Figure 1. It consists of: (i) SPEEDUPExecutive Translator (which turns user problem description "le into a data "le which SPEEDUP solver canunderstand and obey), (ii) SPEEDUP Solver (which compiles and links the user library, mathematicalsolution library and physical property library), and (iii) output.

Setting up a fuel-cell system simulation through SPEEDUP code is ful"lled through both the user problemdescription and the user library. The user library includes nine types of newly developed component modelsand these are: (1) fuel cell stack, (2) reformer, (3) steam generator, (4) water separator, (5) mass-transportequipment (e.g. blower, compressor, expander and pump), (6) heat exchanger, (7) control valves, (8) splitter,and (9) mixer. These component models are described by papers (He, 1997, 1998; He and Chen, 1998).

The operational units are integrated into the system model and the implementation of the connection ofthese units is shown in Figure 2 which is based on the system diagram (He, 1998) (Q1, Q2, Q3 and Q4 are theheat transfer quantity in the steam generator, fuel preheater, heat exchanger and air preheater, respectively).

3. MODEL PERFORMANCE

3.1. Assumptions and setting-up cases

Assumptions. For simplicity, assumptions were made when establishing each component model and theseassumptions are still applicable to the simulation of the system in its entirety. In addition, the followingassumptions were made for simulation of the whole system performance.

(i) Fixed ratio of steam to natural gas S/N (also steam to carbon ratio). The transient changes in the steamgenerator are not considered. This is partly justi"ed by the low sensitivity of the reformer to S/Ndisturbances (when S/N ratio is high, e.g. higher than 2)5). However, the temperature in#uence due tothe steam generation under transient condition is not considered.

Figure 1. Structure of SPEEDUP software package

1332 W. HE

Int. J. Energy Res. 23: 1331}1344 (1999)Copyright ( 1999 John Wiley & Sons, Ltd.

Figure 2. Implementation of the system diagram in SPEEDUP code

Table 1. Main features of the test cases

Objectives Test cases

System performance under steady-state conditions (i) System e$ciency throughout the load range

System load-following capability (ii) Output power responses to a #5% step current change(under high fuel utilization)

(iii) Output power responses to a #5% step current change(under low fuel utilization)

(iv) Output power responses to a #5% step fuel #ow change

Dynamic interactions between components underload-following operation modes

(v) Dynamic interactions between the fuel cell stack andreformer under load-up operation mode

(vi) Dynamic interactions between the fuel cell stack andreformer under load-down operation mode

Safe operation of component under load-followingoperation modes

(vii) Pressure di!erence of fuel cell response to a 10% stepcurrent change

Model evaluation (viii) Compare the calculated system response with the mea-sured results from a 100 kW system

(ii) Fixed ratio of excess air to combustible gas for the reformer burner. The air from the air blower isproportional to the fuel for the reformer burner (e.g. "xed excess air ratio).

(iii) Open-loop conditions. Since this study emphasizes the inherent system characteristics before thecontrol study, the system properties of the open-loop conditions will be analysed. The change of theinput variables is implemented by adjusting the system external operating conditions.

(iv) Single variable change. The study analyses the system responses to a single variable. That is, withoutthe speci"cations, only the input variable mentioned changes, the other input variables retain theiroriginal values.

Setting-up cases. To evaluate the model performance, eight cases listed in Table 1 were selected.

MOLTEN CARBONATE FUEL-CELL POWER-GENERATION SYSTEM 1333

Copyright ( 1999 John Wiley & Sons, Ltd. Int. J. Energy Res. 23: 1331}1344 (1999)

Figure 3. System e$ciency in whole load range

3.2. System ezciencies

To evaluate the system performance under steady-state conditions, the system e$ciency has beencalculated throughout the load range. The system is designed to operate in a load ranging from 30 to 100%(Ooue and Yasue, 1996). It is assumed that the system operational range can stretch 10% beyond itsmaximum and minimum limits (with the same system and the component con"gurations). The system modelwas used to simulate the load level between 20 and 110% of design load.

Figure 3 shows the calculated overall system e$ciency. The system e$ciency is within the range of40}50% throughout the load range. The system reaches its highest e$ciency at partial load instead of at fullload. This is caused by the fact that the fuel cell stack has a higher e$ciency at a lower power output (Blomenand Mugerwa, 1993), but most other components in the system still have lower e$ciency at a lower poweroutput. When the load is higher than 0)45 MW, the overall system e$ciency decreases because the fuel celle$ciency decrease is more signi"cant than the e$ciency increases of the other components.

3.3. System load-following capability

The system power output can be controlled by changing fuel cell current output or fuel and air inputs. Thecurrent output of fuel cell is manipulated to accommodate fast small load variations. When there are largevariations in load, the #ow rates of fuel and air are changed to maintain the fuel and oxidant utilizations offuel cell within the allowable ranges. This section shows: (i) the system responses to a step change of therequired current (with "xed input gas #ow), and (ii) the system responses to a step change of the requiredcurrent and the same step change of the input fuel #ow (with approximate "xed fuel utilization in the fuelcell).

3.3.1. Power response to current change. Withdrawing a higher current output from a fuel cell results ina higher power output within the allowable gas utilization range. However, a higher current output results ina lower reactant gas concentration which reduces the voltage output. This voltage drop phenomenon will bemore signi"cant at high gas utilization conditions. Consequently, the power output responses to currentchange under high and low gas utilization conditions will be illustrated. In addition, the Dutch requirementfor a #5% step change power demand is: half of this must be met within 5 s and the remainder within 25 s.Therefore, our simulation shows the power response in the time range of 100 s.

Figures 4 and 5 illustrate the responses of the power output versus a #5% step change of current densityunder high and low fuel utilizations. The increase in the current, results in a fast change of the gas utilization,that is, the hydrogen utilization increases from 80 to 84% in the fuel gas side). The voltage response has a fastbeginning with signi"cant voltage drop (due to the current increase) and a subsequent stage of voltagerecovery (due to gas concentration and temperature change). Based on the responses of the current andvoltage, the power response consists of a fast beginning with signi"cant power increase (due to the currentchange) and a subsequent stage (due to the voltage change). It should be noted that the voltage drop of 23 V

1334 W. HE


Figure 4. Power responses to a #5% step current density (high fuel utilization)

Figure 5. Power responses to a #5% step current density (low fuel utilization)

at high fuel utilization in Figure 4 is much higher than that of 14 V at low fuel utilization in Figure 5.Consequently, the power response under low fuel utilization is smoother than that under high fuel utilization.After 100 s, the power increase is 4% under low fuel utilization, while the power increase is only 2)5% underhigh fuel utilization.

This indicates that the current change can e!ectively manipulate the output power (with a signi"cantchange simultaneously). It should be noted that the change in the output power (4% at low fuel utilization or



Figure 6. Power responses to a #5% step change of fuel #ow

2)5% at high utilization) is smaller than the change in the current (5%), because of the side e!ect of thevoltage drop (14 V at low fuel utilization and 23 V at high utilization). The system load-following capabilityunder low fuel utilization is better than that at high fuel utilization.

3.3.2. System responses to fuel yow-rate change. In order to maintain approximately the same fuelutilizations in the fuel-cell stack, the gas #ow rate should have the same change as that of the requiredcurrent. This case shows the system response to a 5% step change in the required current and a 5% stepchange in fuel input #ow. Figure 6 shows the system responses. At the reformer outlet, the #ow rate of theprocess has a fast and large increase (due to step change of the fuel input) and then a decrease at the followingstage (due to temperature decrease), while the hydrogen concentration has a dip. The stack voltage has animmediate voltage drop due to the current step up, a fast change due the gas concentration in the stack, anda slow change due to the stack operating temperature variation.

The output power has a 4)2% increase with regard to 5% step changes of the current and fuel input. Itshould be noted that the side e!ect of the voltage drop (4 V) is much smaller than that from the single currentmanipulation in the previous case.

3.4. Interactions between the components

The fuel cell system integrates the stack and the reformer. The fuel included in the anode exhaust gas isrecycled to the reformer burner to provide the heat source for the reforming processes. There may be dynamicinteractions between the stack and the reformer. The temperature #uctuations in the reformer burner and thehydrogen used by the stack are important for safe operation of the reformer and the stack. Therefore, it isinteresting to study the interactions of the reformer burner temperature and stack fuel use under load-following operations. During load-up operation, increasing both the fuel input and current output isa frequent operation. Thus, the cases with a 5% step change of both fuel input and of current output areselected.

1336 W. HE


Figure 7. Interactions between stack and reformer under load-up operation mode

Figure 7 shows that high current output in the fuel cell stack causes high fuel utilization, which results inlow hydrogen concentration at the stack outlet. The low hydrogen concentration causes low combustionheat release and the low temperature of the #ue gas, which will reduce the reforming processes and result ineven higher gas utilization in the fuel cell. In this case, the hydrogen utilization in the stack increases fromaround 80}87)5% then falls back to around 81%. The hydrogen utilization exceeds a limit of 85% (Uchijamaand Segawa, 1991). This excessive hydrogen utilization results in a hydrogen concentration drop from 6)5 to4% at the reformer burner inlet. Then the #ue gas at the outlet of the burner has a temperature drop of 553C(1228}11733C). After the lag of feeding gas transport, both the hydrogen utilization and combustion-gastemperature reach new equilibrium values. This case indicates the potential problem of excessiveH

2utilization in the stack resulting from the dynamic interaction between the fuel cell stack and reformer

during the load-up operation mode.Figure 8 shows the dynamic interaction between the fuel cell stack and reformer under load-down

operation mode. There is a high combustion temperature in the reformer caused by the anode exhaust gaswith higher H

2concentration at lower power output.



Figure 8. Interactions between stack and reformer under load-down operation mode

3.5. Safe operation of the components

During the load-following operation modes, it is also important to know whether each componentoperates within its safe constraints. The safe operational constraints of the fuel cell stack are usually identi"edas having the high priority (Sasaki et al., 1988). Consequently, the safe operation of the fuel cell stack underload-following modes is observed. The criteria for safe operation of the fuel cell stack here only apply to itspressure. In order to prevent damage to the cell, the constraint of the pressure di!erence range is around4 kPa under normal operation and 8 kPa under transient conditions (Yoshida and Inoue, 1993). Theresponses of pressure di!erence between anode and cathode, pressure of anode (P

!) and cathode (P

#) are

illustrated in Figure 9 under current !10% step change.Figure 9 shows the fuel cell pressure response under the current !10% step change. The anode pressure

(P!) decreases and the cathode pressure (P

#) increases because the !10% current step change causes less

electrochemical reaction, which decreases the gas-phase molecules at anode but increases the gas-phasemolecules at cathode. The time constant for the pressure response is of the order of a few seconds. The

1338 W. HE


Figure 9. Pressure responses to a !10% step change of the current density

response of the pressure di!erence has an overshoot above the new steady-state value. In this case, theovershoot of the pressure di!erences gives no cause for concern since it is much smaller than 8 kPa.

3.6. Model evaluation

No experimental data are available for the dynamic performance of the 1 MW system. However, a 100 kWclass MCFC test plant (Yoshida and Inoue, 1993) has been set up to establish the fundamental operationtechniques for the 1 MW class MCFC system. The 100 kW test system has a con"guration comparable tothat of the 1 MW system we studied, although it uses a dummy fuel cell stack and di!erent type of reformer.Therefore, the available measured data from the 100 kW MCFC system have been used to evaluate themodel performance. Yoshida and Inoue (1993) described the test system and the measured responses of thestack gas utilization to a ramp change of the power from 30 to 100 kW. When the load demand has a rampchange, the fuel input and the current output are manipulated to have a similar ramp change to match therequired power output.

Figure 10(a) gives the responses of the power output and H2

and O2

utilization to a ramp changes of fuelinput and current power. Figure 10(b) gives the pro"le based on the measurement. Though power output inFigure 10(a) has a small overshoot, its change matches the load demand in Figure 10(b). The computedresponses of H

2and O

2utilization qualitatively agree with the pro"les of the measured results. The model

validations should be continued, when the appropriate experimental data are available.

4. MODEL APPLICATIONS

The system model is applied for improvement of operation and design. The operation improvement is carriedout with regard to the reduction of the dynamic interactions between components. The e!ectiveness of thesystem model for system design is demonstrated by applying it for selection of control valve location.

4.1. Operation improvement

Dynamic interactions exist between the fuel cell stack and reformer during load-up operation mode. Asshown in Figure 7, the high hydrogen utilization in the fuel cell stack results in low operation temperature ofthe reformer, which reduces the reformer performance and provides the process gas with less hydrogenconcentration. The process gas with less hydrogen results in higher hydrogen utilization in the fuel cell stack.In the case shown in Figure 7, the hydrogen utilization exceeds its limit of 85%.

In order to reduce these dynamic interactions, one approach is to decrease the dependence of the reformerburner temperature on the fuel cell exhaust gas. We propose supplying the fuel gas directly to the combustor



Figure 10. (a) Calculated responses of stack output power and gas utilizations. (b) Response of gas utilization pro"les based onmeasurement

in the reformer. If a small amount of fuel gas (0.30 mol s~1) is directly supplied to the combustor for1 min from t"50 to 110 s, the temperature of the #ue gas at the burner outlet increases in comparisonwith the previous temperature pro"le as shown in Figure 11. The 50 K peak temperature increase inthe #ue gas provides more heat for reforming reaction, which is favourable to the reforming pro-cess and reduces the excessive hydrogen use in the fuel cell stack. Figure 11 also shows that thehydrogen utilization is lower than the limit of 85% under the load-up conditions. The side e!ect of thetemperature increase of the combustor is much less than #ue gas temperature increase of 50 K, whichis not very harmful in this case. This proposed operation strategy uses a small amount of fuel gasinjection (around 7% of the system fuel feeding gas) for only 1 min and thus the in#uence on the systeme$ciency is small. Therefore, this simulation model provided a valuable recommendation for systemoperation improvement.

This alternative operation strategy uses a small amount of fuel gas injection (around 7% of the system fuelfeeding gas) for only 1 min and thus the in#uence on the system e$ciency is small. So the proposed strategy isa valuable recommendation for system operation improvement.

1340 W. HE


Figure 11. E!ect of providing fuel gas to reformer combustor

Figure 12. Alternative fuel-#ow control with normal or high-temperature valves

4.2. Improvement of system design

The model can also be used to select di!erent control strategies, for example control valve location. Thegas temperature at the anode inlet is around 7003C and the fuel gas at reformer inlet is around 4503C



Figure 13. Power responses to #ow change at reformer inlet or at anode inlet

(He, 1997). There are two options for controlling the fuel #ow rate: one at the anode inlet and the other at thereformer inlet. The fuel #ow rate at the reformer inlet can be controlled by a normal temperature controlvalve, but the gas #ow at anode inlet must be controlled by a high-temperature valve. Figure 12 shows thepossible location of these two fuel-#ow control methods. In order to make the correct selection, it isimportant to predict the performance of the fuel cell system by using both methods.

Figure 13 illustrates the simulated power responses to a #10% step change in the fuel gas at the anodeinlet or at the reformer inlet, while the current simultaneously has a #10% step change. The absolute #owrate change at anode inlet is around 20}25 mol s~1, which is much larger than that at the reformer inlet(around 4}5 mol s~1). The power response to the fuel #ow control at the anode inlet (high-temperature

1342 W. HE


control value) is however faster and smoother. These results are expected, because the gas entering at theanode inlet immediately generates power, but the fuel gas entering at the reformer inlet has to be reformedbefore it enters the stack anode. It is noticed that the power response to the normal temperature control valveregulation consists of a fast power increase at the beginning, and then a dip, followed by a slow powerincrease to the new desired value. The fast power increase results from the simultaneous change of the currentand the residual fuel gas in the stack, and the dip is caused by the delay in the fuel supply from the reformer.The subsequent slower power increase is related to the slow response of the reformer. The dip is much largerunder high gas utilization than under low gas utilization.

A fuel cell system usually uses one reformer but several fuel cell stacks (Uchijama and Segawa, 1991). Thefuel-#ow control at the anode inlet needs several high-temperature control valves. The usual control designcriteria demand the use of fewer control valves and minimum high-temperature valves. Therefore, if thepower response can meet the demand, it is more favourable to control the #ow at the reformer inlet.

5. CONCLUSIONS

Firstly, this section describes the implementation of the system model in SPEEDUP code. The systemmodel comprises a connected network of modular component models, which emulate the process #owdiagram. Our approaches to the model integration and the steady state and dynamic simulations have beenpresented.

Secondly, applying the system simulation tool to investigate the 1 MW MCFC system, the calculatedsystem e$ciency, load-following capability, component interactions, and safe operation of the componentare as follows.

1. System e.ciency. The system e$ciency is in the range of 40}50% throughout the load range. Thesystem reaches it highest e$ciency at partial load instead of at full load.

2. System load-following capability with regard to current change. The manipulation of the current resultsnot only in a fast change in the power output, but it also causes a large voltage drop as a side e!ect. Fora #5% step current change, the power output increases 2)5% within a time order of 10 s, but witha voltage drop of 23 V (3)3%) at high fuel utilization (80%). The manipulation of the current is moree!ective at low gas utilization. For a same #5% step current change, the power output increases 4%within a time order of 10 s with a voltage drop of only 14 V (1)8%) at low fuel utilization (40%).

3. System load-following capability with regard to changes of current and fuel input. The output power has4)2% increase with regard to 5% step changes of the current and fuel input. The side e!ect of the voltagedrop (4 V) is much smaller than that from the single current manipulation.

4. Interactions between the fuel cell stack and reformer. The dynamic simulation results indicate theoccurrence of excessive hydrogen utilization in the stack under load-up mode and a high reformer-combustion temperature under load-down mode. During a load-up operation (i.e. 5% step changein both fuel input and current output), the H

2utilization in the stack increases from around 80 to

87)5% then falls back to 80% in a time order of 100 s; so the H2

utilization exceeds the safety limitof 85%.

5. Safe operation of the components. (i) The response of the pressure di!erence in the fuel cell has anovershoot above the new steady-state value with regard to a step change of current. The overshoot isnot harmful since it is much smaller than the constraint of 8 kPa. (ii) Regulating the cathode recyclingratio controls the stack temperature. When the cathode gas recycling ratio changes from 40 to 60%, thestack operating temperature increases by about 10 K.

The simulated utilizations of H2

at fuel-cell anode and CO2

at cathode under load-ramp changeare qualitatively consistent with the experimental pro"les from the 100 kW test system reported inliterature.



Finally, the e!ectiveness of the system model for improvement of system operation and system design hasbeen illustrated by applying it to evaluate the alternative operation strategy to reduce the dynamicinteractions between components and to select the control valve location. For example:

(a) System operation. By supplying a small amount of fuel gas directly to the reformer combustor fora short period provides an e!ective way of solving the problem mentioned above of excessive hydrogenutilization during load-up operation. With a fuel injection of 7% of the system fuel feeding directly tothe reformer combustor for 1 min, the high H

2utilization can be kept within the limit of 85%.

(b) System design. The fuel-#ow control at the anode inlet is more e!ective than that at the reformer inletwith regard to the rapid, smooth load-following capability. However, the fuel-#ow control at the anodeinlet needs more high-temperature control valves.

ACKNOWLEDGEMENTS

This research was carried out at the Laboratory of Thermal Engineering, Delft University of Technology, theNetherlands. The author wishes to thank Professor R.W.J. Kou!eld for the opportunity given to perform theinvestigation.

REFERENCES

Blomen, J. M. J. and Mugerwa, N. N. (1993). &Research, development, and demonstration of molten carbonate fuel cell systems', In FuelCell Systems, Plenum Press, New York, p. 367.

He, W. (1997). &Modeling a reformer in fuel cell power-generation systems', J. Fuel Processing ¹echnol., 53, 99}113.He, W. (1998). &Analysis of Molten Carbonate Fuel-cell Power-generation Systems by using Dynamic Simulations', Ph.D Thesis, Delft

University of Technology (in preparation).He, W. and Chen, Q. (1998). &Three-dimensional simulation of a molten carbonate fuel cell stack under transient conditions', J. Power

Sources, 73(2), 182}192.Ooue, M. and Yasue, H. (1996). &Development of 1000kW-class MCFC Pilot Plant', Proc. Fuel Cell Seminar, Orlando, FL, November

17}20, pp. 51}54.Sasaki, A., Matsumoto, S. and Tanaka, T. (1988). &Dynamic characteristics of a molten carbonate fuel cell stack', Proc. 27th Conf. on

Decision and Control, TX, U.S.A., pp. 1044}1049.SPEEDUP (1996). Speed;p ;ser Manual, Version 5.5, Aspen Technology, Inc., Cambridge, MA, U.S.A.Uchijama, Y. and Segawa, T. (1991). &Subsequent results of a 1 MW class pilot plant development', Technology Research Association for

Molten Carbonate Fuel Cell Power Generation System, Tokyo, JapanYoshida, T. and Inoue, T. (1993). &Test results of 100 kW class MCFC system and control test plant', Note, Ishikawajima-Harima Heavy

Industries Co., Ltd., pp. 5

1344 W. HE


Documents

Operating characteristics of a molten carbonate fuel-cell power-generation system