ARCHITECTURE OF PLANAR SOFC STACKS WITH PARALLEL-CONNECTED CELLS

Transcription

1 ARCHITECTURE OF PLANAR SOFC STACKS WITH PARALLEL-CONNECTED CELLS L. DÖRRER *1, R. BRANDENBURGER *1, CHR. ARGIRUSIS *1,5, G. BORCHARDT *1, H. STAGGE *2, H.-P. BECK *2, CHR. SCHMID *3, J. HAMJE *3, V. WESLING *3, A. LINDERMEIR *4 *1 TU Clausthal, Institut für Metallurgie, Robert-Koch-Str. 42, D Clausthal-Zellerfeld, Germany. *2 TU Clausthal, Institut für Elektrische Energietechnik, Leibnizstrasse 28, D Clausthal-Zellerfeld, Germany. *3 TU Clausthal, Institut für Schweißtechnik und Trennende Fertigungsverfahren, Agricolastraße 2, D Clausthal-Zellerfeld, Germany *4 CUTEC-Institut GmbH, Leibnizstraße , D Clausthal-Zellerfeld, Germany. 5 School of Chemical Engineering, National Technical University of Athens, Zografos Campus, Athens, Hellas ABSTRACT To enhance the power generated by fuel cells one has to connect single fuel cells together. A comparison between different possibilities to connect cells together, especially for unequal cell parameters, has been performed with a simulation. The simplified simulation shows that in the case of unequal single cell resistances of a given stack the power generated from a parallel connection is larger than from a serial one. Additionally it is easier to avoid overloading of single cells in a larger setup in a parallel connection. To decrease the problems with inhomogeneous temperature distribution a special concept of stacked cells in a parallel configuration is proposed. This concept additionally offers new possibilities for the construction of fuel cell systems. Experiments with a 2 cell setup could be reproduced by the simple simulation. 1. INTRODUCTION A fuel cell is an electrochemical reactor which converts chemical into electrical energy. The generated voltage for a single solid oxide fuel cell is in the range of 0.7 to 1.3 V (depending on fuel and electrical current). Operation with a lower voltage is possible but often leads to a long-term degradation especially if the oxygen partial pressure and/or the temperature are unevenly distributed across the cell. The maximum current depends on intrinsic parameters and the area of the MEA (membrane electrolyte assembly) and is given by the maximum current density. Two commonly used methods to increase the electrical power of a fuel cell stack are the enlargement of the cell area to increase output current and the serial connection of cells to increase the output voltage of the stack. Additionally the thickness of the interconnector plates is reduced as far as possible to increase the power/weight ratio of the whole system. The serial connection of cells has, especially at high power densities, two major drawbacks. The performance of the stack is reduced to negligible levels in case of a single cell failure and poor heat transfer leads to an undesired temperature gradient along the cells especially for large area cells. Some architectures were developed to overcome these drawbacks. As an example a parallel connection of 4 cells in a planar configuration is used in [1]. These quadcell-modules were stacked in series. Due to this architecture the redundancy is increased and a defect in one of the cells does not have as strong an effect on the stack as it would have in a mere series connection of cells. Even in this in-plane geometry the paths for heat transport are long and the poor thermal conductivity of the interconnectors is the limiting factor to achieve a homogeneous temperature distribution. Alternatives are an increased airflow for enhanced convective heat transfer, heat-consumption by an endothermic reaction or another kind of heat sink nearby the cells [2]. The architecture presented in this work uses cells with comparatively small areas (16 cm 2 ). These cells are electrically connected in parallel and stacked together with every second MEA turned upside down. The small lateral size in conjunction with the relatively large outer surface of the stack leads to a small temperature gradient even in the case of low thermal conductivity of the interconnectors. The advantages and drawbacks of such stack architectures will be shown and discussed. 1 corresponding author: lars.doerrer@tu-clausthal.de

2 2. BACKGROUND In general there are two possibilities to connect EMF-sources (electromotive force) like batteries or fuel cells together: Serially or in parallel. Figure 1 shows the two possible electrical connections of fuel cells (the single cell is shown as a battery symbol). Fuel Fuel Air Figure 1 Air Sketch of the possible electrical connections of fuel cells: serial (left) and parallel (right) For a combination of identical single cells in a serial connection the current through each cell is the same and the voltages of the single cells simply add up while in a parallel connection the voltage is identical and the currents add up. In these two ideal cases the maximum power is equal for both variants. To minimize the losses caused by connecting wires and converting to user friendly voltages it is more convenient to choose the serial connection as it is state of the art. In reality the parameters of the cells are not identical and one has to compare the performance of connecting schemes again. To evaluate the connection schemes one has to consider some practical aspects. In a practical system it is easy to measure the current and the voltage at the terminals but it is a complex matter to measure all single voltages and currents separately. Therefore the distribution of these parameters is often not known. One critical aspect to avoid degradation of the fuel cell is the stability of the anode. A typical material for SOFC anodes is a Ni/YSZ cermet. As nickel tends to oxidize at higher oxygen partial pressure and as the resulting NiO has a larger volume than metallic Ni the resulting mechanical stress can destroy the electrode. The equilibrium partial pressure of the oxidation reaction is given in tables, e.g. in [3]. For given conditions one can calculate an equivalent open circuit voltage (OCV) by the Nernst equation (1): OCV = RT / 4F * ln { p O2,C / p O 2,A }. (1) R = 8,31 J / ( K mol ) and F = C / mol are the gas constant and the Faraday constant, T is the temperature given in Kelvin and p O2 is the actual oxygen partial pressure at the cathode (C) and anode (A), respectively. A larger oxygen partial pressure at the anode leads to a lower OCV. At 850 C with air as oxidant the minimum OCV for a stable anode is 680 mv. The electrical characteristics of a fuel cell can be described by the OCV and the losses during the operation. Different losses or overpotentials influence the operation of the fuel cell. The most important ones are the ohmic losses, the activation overpotentials on anode or cathode, the diffusion overpotentials and the gas conversion losses. These contributions result in a non linear current-voltage characteristic, especially for small and very large currents. In the intermediate range lies the typical operating range. These current voltage curves can be described for simplification with a constant resistance, the area specific resistance. A more detailed description of the influence of the different losses on the performance of a fuel cell is given for instance in [4], [5]. 3. CONCEPT OF THE PARALLEL STACK ARCHITECTURE If a number of single fuel cells are connected in parallel in an in plane geometry the result will be a thin structure with a large area. Such devices are difficult to handle especially from the viewpoint of homogeneous temperature distribution. To get a small footprint in a parallel connection one has to stack the single cells. This is easily possible if each second cell is flipped. The principle sketch for this configuration is shown in figure 2.

3 Air Figure 2 Fuel Principle sketch of a fuel cell stack with an electrically parallel configuration In this geometry one has to distinguish two types of interconnectors, one at the anode side and the second at the cathode side. The anode interconnector is only exposed to the fuel. The cathode interconnector and the housing at the cathode side are only exposed to air. This offers new possibilities for the choice of the material. The housing at the anode side is exposed to fuel and air. Because of this the requirements for this material are the same as in state-of-the-art serial stacks. Due to this special geometry all anode interconnectors, anode housings and fuel gas tubes are electrically connected. Therefore there are no gas tight insulating seals necessary, all connections can be welded and the gas tubes can be used as electrical connections. 4. SIMULATIONS In this chapter we will show the results of two different types of simulations. One is a numerical simulation of the location-dependent partial current and voltage in a single cell while the second simulation yields the averaged cell voltages and currents for parallel as well as serial cells. Figure 3 Simulated terminal voltage, minimum Nernst voltage and fuel utilization (FU) for an electrolyte supported cell During operation of a fuel cell the concentration of the fuel along the flow path decreases and the gas conversion losses increase. This leads to a location dependent oxygen partial pressure and, as a

4 consequence, to a location dependent Nernst potential (see equation (1)). On the other hand the measured voltage on the cell (terminal voltage) is averaged over the whole area. To guarantee the stability of the anode it is necessary to know the lowest possible terminal voltage for which the oxygen partial pressure on the anode side is lower than the equilibrium partial pressure of the NiO/Ni redox reaction at every point of the fuel cell. We have performed numerical simulations for an electrolyte supported cell at 850 C. Other parameters were chosen similar to the typical measurement conditions. The result is shown in figure 3. The result of this numerical simulation for our experimental conditions is that the Nernst potential is always above the critical value of 680 mv if the terminal voltage is higher than 650 mv. This terminal voltage is taken as a shut-down criterion for the measurements and the simulation of the behaviour of the connected cells. Of course the shown simulation ignores possible leaks and their consequences for degradation processes. A simple simulation model in matlab simulink is used to evaluate the differences between the serial and the parallel connection of cells to increase overall power. In this model each cell is described by a voltage source with a unique OCV and a series resistance R_I representing the internal resistance of the cell. Different overpotentials are not taken into account separately for this simple model and the resistance is taken as constant regardless of the load. Also temperature dependencies are not taken into account, the model only reflects stationary behaviour of the cell. As a first step only two cells were used. The two cells can be connected in series or in parallel. They are connected to a load that draws a current from the cells. Current and voltage of the cells are calculated. For a comparison of power of the two different configurations the current drawn by the load is doubled for the parallel connection. Figure 4 shows the simulation model for a single cell. Figure 4 Cell model for the matlab simulations of a single cell Also it is possible to add an external series resistance R_E for each cell. On the one hand this is for the comparison of the simulation data to the measurements performed with the cells; on the other hand this is a resistance that will affect a practical system because it is always necessary to connect the stack to the load with a wire. This setup of the simulation is used in section 5 to compare the measurements with respect to the parallel and the (virtual) series configuration of the stack. The simulation model can also show the performance of a two-cell stack with one cell being degraded. The degradation is taken into account by increasing the series resistance of one of the cells. Resistances between the cell and the load (in the model given by R_E) are neglected for this simulation. Parameters of the model are an OCV of 1013 mv and a resistance R_I of 123 mohm while the resistance of cell 1 is increased to 300 mohm. The current is steady at 2 A for the series connection and 4 A for the parallel connection, respectively.

5 Figure 5 Power and cell voltage of a two-cell stack with a degrading cell Two conclusions can be drawn from figure 5. First the power of the stack is higher in the parallel configuration when the resistance of cell 1 increases. As a second and possibly more important effect the cell voltage of cell 1 in series configuration decreases rapidly and drops below the safe operation limit to prevent nickel oxidation as discussed before. To stay in the safe operating range the current drawn from the series configuration has to be decreased, what leads to a further power decrease compared to the parallel connection. If a single cell voltage measurement is not implemented into the stack the low cell voltage might not even be noticed. With the current not being decreased the degradation of cell 1 would even be accelerated, leading to a further increase of the resistance and eventually to a complete stack failure. The parallel configuration can avoid these effects and combines the advantages of higher power and higher failure tolerance. 5. EXPERIMENTAL INVESTIGATIONS In the experimental setup shape milled interconnectors (Crofer 22 APU) were used to measure 2 MEAs separately or both together connected in parallel. The outer parts serves as cathode connectors and air flow fields while the inner part acts as anode connector and fuel flow field. The outer connectors are not electrically connected among each other in the hot zone. To connect the electrolyte supported cells (4 x 4 cm² active area) with the interconnector we use Pt and Ni grids were inserted. The setup is shown in figure 6 as a schematic drawing and a photographic picture. Numbers in the sketch represent the following parts: 1 anode interconnector, 2 cathode interconnector, 3 MEA 2. MEA 1, located between anode interconnector and the lower cathode interconnector, is not shown. The cathode interconnector has a footprint of 70 x 70 mm².

6 Figure 6 Sketch and photo of the mounted parallel setup This mini stack was placed in a furnace to control the temperature up to 870 C. For these first experiments a H 2 /Ar/H 2 O mixture was used with varying composition as fuel and air as oxidant. The time dependent current and voltage under different loading conditions are shown in figure 7. To avoid degradation due to partial oxidation of the anode the current was kept at a level where the terminal voltage of the loaded cell was only slightly lower than 650 mv. At first only the upper cell (MEA 2) was loaded with a linearly changing current, then the lower cell. Further conditions of the shown experiment are: 850 C, fuel: 46,5 % H 2, 46,5 % Ar, 7 % H 2 O, oxidant: air. Figure 7 Measurement of the current and voltage for different loading, from 0 to 350 s only MEA 2, from 350 to 600 s only MEA 1, > 600 s both MEA in parallel

7 The maximum current density of the MEA 2 was 180 ma/cm² and the OCV was only slightly lower than the theoretical value (OCV measured = 1013 mv, OCV calc = 1017, equation (1)). The ASR of the lower cell (MEA 1) was higher and the OCV was lower than for the upper cell. For comparison of the measurements with the simulation one has to take account of the following points: In each loading path resistors are integrated which may be turned on and off. These additional resistors were taken into account for the simulation (R_E as described in the simulation model before). In this setup it is possible to measure the current in each cell separately. Also it is possible to draw current from each cell individually as well as in a parallel connection. The serial connection of the two cells could not be measured directly with this setup. We calculated the current voltage curve from the separate measurements of the single cells and call it a virtual measurement of serially connected cells. Unfortunately there was a small leakage between anode and cathode of MEA 1. Because of this there was degradation during the measurements and the ASR increased by approximately 10 %. The leak is also responsible for the differences in OCV and ASR. These facts were also taken into account for the simulation. In figure 8 the simulations are compared to the measurements for two serially connected (virtual) cells and in two parallel connected cells. Figure 8 Comparisons between measurements and simulations for two serial and parallel connected cells Two conclusions can be drawn from figure 8. The simulation and the experimental results agree well. The simplifications we have made in our simulation do not affect the results too much. Contrary to the simulation the maximum power of the measured parallel connection is less than in the virtually measured serial connection. This is probably caused by the degradation process mentioned above.

8 6. SUMMARY The state of the art connection scheme for fuel cells is the serial connection. The main advantage of this type of connection is a user friendly voltage level. In the case of similar parameters of the single ideal cells the alternative parallel connecting scheme with low terminal voltage and large current has only some constructive advantages. In the case of different parameters and/or degradation processes of the single nonideal cells this is not true any more. With our simulations we were able to show that the maximum power of parallel-connected cells is larger than in the comparable serial connection. Additionally the drawback of the serial connection, the possible overloading of single cells, can be avoided without a complex monitoring system for each single cell voltage. Our first measurements show that a setup with stacked cells in a parallel configuration is possible and that the results are comparable with the simple model we used to predict the behaviour. AKNOWLEDGEMENT This work is being funded by Landesinitiative Brennstoffzelle Niedersachsen of the Federal State of Lower Saxony. We are indebted to Dr Helmut Lessing for fruitful discussions at the beginning of our project. REFERENCES [1] Föger K., 2007, Micro-CHP Product Development, presented at Fuel Cell Seminar San Antonio [2] Liese E. A., Gemmen R. S., 2005, Performance Comparison of Internal Reforming against External Reforming in a Solid Oxide Fuel Cell Gas Turbine Hybrid System, Journal of Engineering for Gas Turbines and Power, vol. 127, pp [3] I. Barin: Thermodynamic Data of Pure Substances, 1989, VCH Verlagsgesellschaft GmbH, Weinheim, Germany [4] Fuel Cell Handbook, 2004, 7 th ed. [5] Weber A., Entwicklung von Kathodenstrukturen für die Hochtemperaturbrennstoffzelle SOFC, Dissertation, 2002, Karlsruhe (in German)