Index - Final Report. Report compiled by: Ming Chen

Transcription

1 Index - Final Report 1. Final report Project details Executive summary Project results Overview and milestones WP1 Cell development and testing WP2 Interconnects and seals WP3 Stack development and testing WP4 SOEC system WP5 Energy system modeling Dissemination of the project results Environmental benefits of the project Utilization of project results Project conclusion and perspective 66 Report compiled by: Ming Chen Contributors: Department of Energy Conversion and Storage, Technical University of Denmark (DTU Energy): Ming Chen, Xiufu Sun, Anne Hauch, Karen Brodersen, Benoit Charlas, Sebastian Molin, Karsten Agersted, Vinothini Venkatachalam, Wolff-Ragnar Kiebach, Yi-Lin Liu, Janet Jonna Bentzen, Christopher R. Graves, Jacob R. Bowen, Jean-Claude Njodzefon, Søren Højgaard Jensen, Mogens Bjerg Mogensen, Peter Vang Hendriksen Haldor Topsøe A/S (HTAS): Peter Blennow, Rainer Küngas, Claus Friis Pedersen Department of Development and Planning, Aalborg University (AAU): Iva Ridjan, Brian Vad Mathiesen 1

2 1. Final report 1.1 Project details Project title Solid Oxide Electrolysis for Grid Balancing Project identification Energinet.dk project no Name of the programme which has funded the project ForskEL (ForskVE, ForskNG or ForskEL) Name and address of the enterprises/institution responsible for the project Department of Energy Conversion and Storage (DTU Energy), Technical University of Denmark, Risø Campus, Frederiksborgvej 399, 4000 Roskilde, Denmark CVR (central business register) DK Date for submission

3 1.2 Executive summary Denmark s ambitious plan to rapidly increase the fraction of renewable energy supply towards 100 % over the coming 35 years will lead to huge changes in the electricity grid and a need for large-scale energy storage due to the intermittent nature of wind and solar power. Solid oxide electrolysis cells (SOECs) are a promising technology for energy storage and synthetic fuel production and it has a unique potential for grid regulation in the Danish power system. The purpose of this 2-year project was to improve performance and durability of SOEC cells and stacks targeting applications specifically for regulating the future Danish power system with a high amount of fluctuating renewable energies, and at the same time enhance the cost competitiveness and environmental friendliness of the SOEC technology, The four project partners (DTU Energy, Topsoe Fuel Cells A/S (TOFC) 1, Haldor Topsøe A/S (HTAS), Aalborg University) have been working closely on critical research issues for the upscale to MW level of the technology foreseen over the next 5-10 years. These include: 1) studies of cell performance, degradation processes and measures to improve the life time, 2) improved methodologies for studies and measures to improve the lifetime, 3) a focused effort on making the cells stronger to allow upscale in size, and 4) development of an upscalable cell manufacture technique multi-layer tape casting (MTC). In addition, key components for closely heat and flow integrated SOEC systems of kw scale were developed. Progress on all these critical issues is necessary for SOECs to become an active part of future Danish electricity system available from around In the previous project ForskEL Development of SOEC Cells and Stacks, long-term electrolysis operation was demonstrated on so-called 2.5G-Standard SOEC cells. This type of cell has a Ni/YSZ (YSZ: (Y 2 O 3 ) x (ZrO 2 ) 1-x ) support, a Ni/YSZ active fuel electrode, an YSZ electrolyte and a LSCF/CGO (LSCF: La 1-x Sr x Co 1-y Fe y O 3- ; CGO: Ce 1-x Gd x O 2- ) oxygen electrode with a CGO barrier layer at the interface between the electrolyte and the oxygen electrode. The 2.5G-Standard cells represent the state-of-the-art cell technology at DTU Energy and are representative of technology used at HTAS. At the beginning of this project (ForskEL Solid Oxide Electrolysis for Grid Balancing ), a 6-month cell test was completed on 2.5G-Standard cells, demonstrating a degradation rate of 22 mv/1000h (or 1.7%/1000h) for steam electrolysis at 800 o C and -1 A/cm 2. This represents the starting point for the cell development work in this project. In WP1 of this project extensive efforts were devoted to development and testing of SOEC cells with improved performance, durability, and mechanical strength. For 2.5G-Standard cells, we further investigated their performance and durability as functions of current density, feed gas composition, and extent of conversion for steam and/or CO 2. The main test conditions were chosen as 800 o C and -1 A/cm 2, oxygen fed to the oxygen electrode and different gas compositions fed to the fuel electrode with a flow corresponding to a steam/co 2 conversion of 56 %. The influence of gas composition was studied by varying the fuel electrode inlet gas composition from 90 % H 2 O + 10 % H 2 (i.e. steam electrolysis) to 45 % CO % H 2 O + 5 % CO + 5 % H 2 (co-electrolysis), and further to 90 % CO % CO (CO 2 electrolysis). The effect of current density or conversion was studied by varying the current density from zero up to A/cm 2, or conversion from 28 % to 56 %, respectively. Based on the cell test results, a safe operation window of current density up to -1 A/cm 2, fuel conversion up to 60 %, and CO 2 +H 2 O content in the feed gas composition up to 90 % was proposed. When operated at -1 A/cm 2, a steady state degradation rate of <30 mv/1000 h (or 2.5 %/1000 h) is achievable. 1 Topsoe Fuel Cells A/S was closed on October 31, 2014 and its tasks for this project were taken over by Haldor Topsøe A/S hereafter. 3

4 To be able to support the development work in this project and also push forward the competence edge, detailed understanding of degradation mechanisms is indispensable. In this project, advanced post-mortem characterization methodologies were developed and employed to detect electrolysis operation induced changes from micro-/nanometer to macro scale. For example, we further developed thin-film based electron backscatter diffraction (EBSD) and applied it to detect microstructure/composition/crystallography changes at the micro- and nano-meter scale in SOEC samples. Another effort was devoted to investigate impurity distribution inside a 2000 h-tested SOEC stack using secondary ion mass spectroscopy (SIMS). Unlike previously reported results from single cell testing, in this stack, no increase of Si after electrolysis operation was observed. In combination with post-mortem results on other long-term tested SOEC stacks, it was concluded that Si poisoning of the Ni/YSZ electrode is not one of the dominant degradation issues at the stack level under the electrolysis conditions explored so far. The used sealing glass, which was considered to be one of the major sources for Si poisoning in single cell testing, can be used as sealant for SOEC stacks. The focus of the cell development work in this project is on optimizing halfcells made by multi-layer tape casting (MTC) in order to improve performance, durability and mechanical strength. A MTC halfcell consists of support, active Ni/YSZ electrode, YSZ electrolyte, and CGO barrier layer with all layers co-sintered in one go. The cells were further applied with LSC/CGO oxygen electrode by screen printing. The full cells are named as 2.XG. The optimization was done via varying the Ni content and/or particle size distribution. With the most successful cells (2.XG, 2014 version), a long-term degradation rate of only %/1000 h was demonstrated at 800 o C and -1 A/cm 2 for steam electrolysis. This very low degradation rate is significantly better than any results so far reported in literature and ensures a cell lifetime of more than five years. In addition, we have improved mechanical strength of MTC support and have successfully enlarged the cell footprint from 13X13 cm 2 to 19X19 cm 2. At a footprint of 13X13 cm 2, the Weibull modulus of MTC support is significantly higher (1.5 or 2 times) than that of 2.5G-Standard support. When enlarged to a footprint of 19x19 cm 2, the Weibull modulus of MTC support is reduced, but still similar to that of 2.5G-Standard support at a footprint of 13X13 cm 2. The decrease in Weibull modulus (when increasing the footprint from 13X13 cm 2 to 19X19 cm 2 ) is mainly due to inhomogeneity introduced when up-scaling the cell fabrication process, which will be minimized in further process optimization. WP2 of this project deals with interconnects (ICs). Currently coated Crofer 22 APU is employed as IC material in SOEC stacks. It s corrosion behaviour in SOEC test conditions is of concern. We investigated corrosion of various uncoated and coated ICs when in contact with Ni or when exposed to 100 % CO 2. We found that diffusion of Ni into ICs slows down the oxidation kinetics, but causes phase transformation and introduces mechanical stress which calls for further study. For corrosion in pure CO 2, no sign of carbide formation or carbon diffusion into the alloy was detected in the present study. Co coated ICs showed poor corrosion resistance. Further improvement of corrosion resistance was achieved via developing the MnCo spinel ((MnCo) 3 O 4 ) coating by the electrophoretic deposition (EPD) technique. We demonstrated MnCo spinel coated ICs with a corrosion rate constant of 3-6 X g 2 /(cm 4 S) at 800 o C in air, in comparison with X g 2 /(cm 4 S) obtained on ICs with either no coating or Co coating. The EPD coating process was successfully up-scaled. Full scale ICs coated with EPD coating were tested at the stack level, showing promising results. In WP3, extensive stack development and testing was carried out. HTAS Delta-type stacks, which were state-of-the-art SOEC stacks at the beginning of this project, were selected for evaluating stack reliability under conditions relevant to Danish electricity balancing schemes. Two 8-cell stacks were operated according to the 100 % wind power profile defined in WP4 of this project, with a total period exceeding 3000 h. The tests demonstrate that a reversible 4

5 SOC stack can be operated in a stable manner with a real-world relevant operating profile which demands many changes in the operating point and reversals between electrolysis mode and fuel cell mode. By the end of the tests, both stacks experienced certain degree of damage caused at least partly due to testing issues. To promote application of SOC stacks for grid balancing, maintaining stable contact among various stack components is one of the main issues which should be addressed in the near future. To map out safe operation window for HTAS Delta-type stacks, three stack tests were carried out with a purpose of exploring the stack performance under three of the selected extreme conditions: high current density, high degree of steam conversion (into hydrogen), and high steam content. Detailed post-mortem analyses (PMA) were carried out on the here-tested stacks plus two tested in other projects. Based on the above stack tests and PMA results, it was concluded that HTAS Delta-type stacks can be operated rather safely at o C with electrolysis current density up to A/cm 2 and fuel conversion up to 60%, with an expected life-time of exceeding 9000 h. Re-distribution of Ni in the active Ni/YSZ electrode and contacting on the oxygen side were identified as two of the most critical issues which shall be addressed in order to further improve SOEC stack lifetime and robustness. An important achievement in WP3 of this project is on high pressure SOEC stack testing. This is a continuation of previous launched activities on high pressure testing both at the cell level and at the stack level. In this project, we have successfully completed construction of the high pressure stack test rig. In addition, an 8-cell Delta-type stack was tested under elevated pressure up to 20 bar. The durability test was carried out at 10 bar for close to 200 h. Several incidents, caused mainly by instrument technical issues, happened in-between, which eventually led to crack of cells and shutdown of the test. However, the current test proves feasibility of high pressure SOEC stack testing and improved cell performance at elevated pressure. The stack development work in this project involved testing of improved stack components (ICs with new coating and new cell contact layers) at the stack level and testing new stack design for electrolysis purpose. The new stack components show promising results and are recommended for further development. At the beginning of this project all SOEC work on stacks was performed on the Delta design. During the project a transition to the TSP-1 type stack design was done. The TSP-1 design shows several advantages as compared to the Delta design, such as increased number of cells per stack, increased cell active area, improved gas flow distribution and stack manufacturability etc. Significant efforts were devoted to design and testing TSP-1 stacks (including upgrading the test facilities at HTAS and DTU Energy for testing TSP-1 stacks). Within the project period, 6 TSP-1 stacks were tested for various purposes, with a total testing period exceeding 5100 h. The test results confirm stable electrolysis performance and good durability of the TSP-1 stacks. The focus of WP4 was on system level. The first task was from an SOEC system point of view, to identify SOEC cell and stack test conditions relevant for grid balancing schemes. The grid balancing schemes include both short-term and long-term grid balancing, and with a grid relevant to Denmark, which means the focus is on integrating intermittent wind or solar power. We investigated whether solid oxide cells (SOCs) could provide full energy balancing by reversible (electrolysis and fuel cell modes) operation with a single device. Two scenarios were considered: The first is based on wind power supply and the second is based on solar power supply. In both scenarios, we assume an energy storage scheme wherein the intermittent renewable energy provides 100 % of the needed energy for the region including ondemand electricity as well as hydrocarbon fuels for transportation produced by electrolysis. The analyses were converted into SOC stack operating profiles and employed in actual stack tests carried out in WP3 of this project. In addition to the above system analyses, two key 5

6 components (heat exchangers and electrical heaters) for closely heat and flow integrated SOEC systems were developed. The heat exchangers were successfully tested. Good agreement between the measured and model predicted performance was achieved. Testing of the newly developed heaters was unfortunately delayed due to test setup issues. Further development and testing is considered highly relevant and will be combined with similar activities in the forthcoming project. The last work package of this project (WP5) focused on energy system modelling. Different aspects of integrating SOEC into various energy systems were investigated. We first carried out a comparison between distributing produced syngas (CO+H 2 ) through natural gas grid versus direct conversion of syngas to liquid fuel and distribution in the liquid form. According to the analysis, distributing syngas would need a special pipeline network which should be potentially avoided. On the other hand, transportation of final fuel is well established. The second analysis was to find out the potential for using combined capacities of SOEC and SOFC for grid balancing purposes. SOEC mode is used for production of liquid/gaseous fuels for the transport sector, while the potential of SOFC operation was based on substitution of combined cycle gas turbine (CCGT) in combined heat and power (CHP) plants as they both convert fuels to electricity. It was found that, depending on the wind production ( history data), the combined SOEC/SOFC can substitute between 73.7 % (2008) and 96 % (2014) of CCGT capacity. On average combined SOEC/SOFC can substitute ~84 % of CCGT capacity. Furthermore, we conducted an analysis on using steam electrolysis for fuel production versus co-electrolysis. The results indicate that there are no decisive differences between these two pathways for the same fuel outputs. Finally, analysis of using SOEC in different energy systems (a 100 % renewable Danish 2050 system and two types of German energy systems with or without nuclear energy) was done in order to determine potential of using electrolysis and what utilization capacities are necessary in order to supply fuel demand. Optimum electrolysis utilization ranges were identified for all the three systems. As compared to the previous projects, significant progresses have been achieved in this project with respect to improving performance and durability of SOEC cells, stack components, and stacks under grid balancing related conditions. At the cell level, we have developed cells which are capable of electrolysis operation up to -1 A/cm 2 with a long-term degradation rate of only %/1000 h, ensuring a cell life-time of more than five years. At the stack level, we have demonstrated that reversible SOC stacks can be operated in a stable manner with a real-world relevant operating profile for grid balancing purpose which demands many changes in the operating point and reversals between electrolysis mode and fuel cell mode. We proved feasibility of high pressure SOEC stack testing and improved cell performance at elevated pressure. In addition, a new stack design (TSP-1) has been developed and tested, which according to test data is superior to the old design (Delta) on several aspects. Extensive long-term stack tests confirm stable electrolysis performance and good durability of the TSP-1 stacks. In Roadmap SOEC Steam Electrolysis and Co-electrolysis formulated by Danish Partnership for Hydrogen and Fuel cells 2, the target for improved cell durability is set as <1.25 %/1000 h at 800 o C and -1 A/cm 2, while the target is <0.75 %/1000 h at 750 o C and A/cm 2. The results obtained in this project are in line with the SOEC strategy and roadmap and have reached the target. The outcome of this project provides guidelines for the future focus of development in order to achieve increased performance, better durability, and more cost-efficient solid oxide electrolysis cells and stacks. 2 Roadmap SOEC Electrolysis and Co-electrolysis, Danish Partnership for Hydrogen and Fuel Cells, September

7 1.3 Project results Overview and milestones The project was structured into five technical work packages (WPs), within which a number of work tasks (WTs) are defined. WP1 and WP2 focused on cell and stack components and were led by DTU Energy. In WP1, performance and durability of the current generation SOEC cells (2.5G-Standard) was investigated under grid balancing related test conditions. Based on the acquired knowledge on life-time limiting factors and degradation mechanisms, counteracting measures were proposed and more reliable and larger SOEC cells for up-scaling were developed. WP2 aimed to further improve corrosion resistance of coated interconnects (ICs) employed in SOEC stacks. The developed SOEC cells and stack components from WP1 and WP2 were assembled into SOEC stacks and tested in WP3. In addition, a new stack design was developed and tested for electrolysis operation. WP4 focused on SOEC system level, where two key system components (heat exchanger and heater) were developed and tested. WP3 and WP4 were led by Haldor Topsøe A/S. WP5 dealt with analysis on energy system level and was led by Aalborg University. The five WPs are listed below, with further splitting into work tasks (WTs): WP1 Cell development and testing WT1.1 Safe operation window for current generation SOEC cells WT1.2 Advanced characterization and modelling methodologies WT1.3 Optimized SOEC cells with enhanced durability WT1.4 Optimized SOEC cells with improved mechanical strength at larger scale WT1.5 Next generation SOEC cells WP2 Interconnects and seals WT2.1 IC corrosion under extreme conditions WT2.2 Seals with reduced impurity emission WP3 Stack development and testing WT3.1 Safe operation window for current generation SOEC stacks WT3.2 SOEC stack testing of improved cell and stack components WP4 SOEC system WT4.1 Grid balancing related test conditions WT4.2 Key components for heat and flow closely integrated SOEC systems WP5 Energy system modelling Listed below are the milestones that were set up in the project plan and their status at the completion of the project: M1.1 Safe operation window (in terms of current density, temperature of operation, cycling rate, degree of conversion, etc.) for current generation SOEC cells mapped out and reported. [Milestone fulfilled. We investigated performance and durability of 2.5G-Standard cells as functions of current density, feed gas composition, and extent of conversion for steam and/or CO 2. Based on the cell test results, a safe operation window of current density up to -1 A/cm 2, fuel conversion up to 60 %, and CO 2 +H 2 O content in the feed gas composition up to 90 % was proposed. When operated at -1 A/cm 2, a steady state degradation rate of <30 mv/1000 h (or 2.5 %/1000 h) is achievable.] M1.2 Analysis of impurity distribution inside long-term tested SOEC stacks evaluated. [Milestone fulfilled. Impurity (especially Si) distribution inside a long-term tested stack was analysed using time-of-flight secondary ion mass spectrometry (TOF-SIMS). Unlike previously reported results from single cell testing, in this stack, no increase of Si 7

8 during electrolysis operation was observed. In combination with post-mortem results on long-term tested SOEC stacks, it was concluded that Si poisoning of the Ni/YSZ electrode is not one of the dominant degradation issues at the stack level under the electrolysis conditions explored so far. The used sealing glass, which was considered to be one of the major sources for Si poisoning in single cell testing, can be used as sealant for SOEC stacks.] M1.3 Application of thin film based EBSD to an SOEC electrode microstructure with a spatial resolution of 10 nm or less demonstrated. [Milestone fulfilled. Transmission Kikuchi Diffraction (TKD) investigations were carried out on two samples: an YSZ porous backbone structure impregnated with Ni and CGO nano-particles and a Ni/YSZ SOEC cathode, with 10 nm spatial resolution demonstrated.] M1.4 Optimised SOEC cells tested at 1.25 A/cm 2 over a period exceeding 1000 hours with a degradation rate less than 1%/1000h. [Milestone partially fulfilled. We have developed SOEC cells with enhanced performance and improved durability for electrolysis operation. With the most successful cells (2.XG, 2014 version), a long-term degradation rate of only %/1000 h was demonstrated at 800 o C and -1 A/cm 2 for steam electrolysis. When tested at A/cm 2, the same cell underwent heavy degradation during the 1000 h test, with a degradation rate much higher than the one specified in the milestone. However, the achieved very low degradation rate at -1 A/cm 2 ( %/1000 h) is significantly better than any results so far reported in literature and ensures a cell life-time of more than five years. In Roadmap SOEC Steam Electrolysis and Co-electrolysis formulated by Danish Partnership for Hydrogen and Fuel cells 3, the target for improved cell durability is set as <1.25 %/1000 h at 800 o C and -1 A/cm 2, while the target is <0.75 %/1000 h at 750 o C and A/cm 2. The results obtained in this project are in line with the SOEC strategy and roadmap and have reached the target.] M1.5 An SOEC cell with a foot print of 18X18 cm 2 with improved Weibull modulus by 15% (as compared to 2.5 G cells of the 2012 generation) demonstrated. [Milestone partially fulfilled. We have improved mechanical strength of MTC support and have successfully enlarged the cell footprint from 13X13 cm 2 to 19X19 cm 2. At a footprint of 13X13 cm 2, the Weibull modulus of MTC support is significantly higher (1.5 or 2 times) than that of 2.5G-Standard support. When enlarged to a footprint of 19x19 cm 2, the Weibull modulus of MTC support is reduced, but still similar to that of 2.5G- Standard support at a footprint of 13X13 cm 2. The decrease in Weibull modulus (when increasing the footprint from 13X13 cm 2 to 19X19 cm 2 ) is mainly due to inhomogeneity introduced when up-scaling the cell fabrication process, which can be minimized in further process optimization.] M1.6 New electrode materials/architecture tested at 2 A/cm 2 over a period of 500 hours with a degradation rate of half that of the current generation SOECs demonstrated. [Milestone fulfilled. We have demonstrated a CGO fuel electrode with both initial electrode resistance and long-term durability significantly better than the state of the art Ni/YSZ electrode. The degradation rate at 1.9 A/cm 2 was reduced by a factor of 18.] 3 Roadmap SOEC Electrolysis and Co-electrolysis, Danish Partnership for Hydrogen and Fuel Cells, September

9 M2.1 Oxidation tests of current generation coated ICs or coated alloys under at least two of the typical extreme operating conditions (defined in WT4.1) carried out for periods exceeding 1000 hours. [Milestone fulfilled. Two scenarios were investigated: Ni diffusion induced IC phase transformation and corrosion in 100 % CO 2. Diffusion of Ni into ICs slows down the oxidation kinetics. However, inter-diffusion may cause phase transformation and mechanical stresses which calls for further study. For corrosion in pure CO 2, no sign of carbide formation or carbon diffusion into the alloy was detected in the present study, which leads to a tentative conclusion that pure CO 2 may be relatively safe with regard to the fuel side interconnect corrosion. This could be due to the fact that the formed dense oxide scale totally blocks inward carbon diffusion.] M2.2 ASR measurements of current generation coated ICs or coated alloys under no current and under current density above 1 A/cm 2 for periods exceeding 1000 hours completed and influence of current on IC corrosion clarified. [Milestone fulfilled. ASR measurements of uncoated and coated Crofer 22 APU at 0.5 and 2 A/cm 2 have been carried out for periods exceeding 1000 h. The results do not point to a strong influence of the current density on ASR degradation.] M2.3 Coated ICs or coated alloys tested in SOEC stack-relevant environments for >1000h with a degradation rate of 10% better than current generation coated ICs demonstrated. [Milestone fulfilled. We demonstrated MnCo spinel coated ICs (by EPD) with a corrosion rate constant of 3-6 X g 2 /(cm 4 S) at 800 o C in air, in comparison with X g 2 /(cm 4 S) obtained on ICs with either no coating or Co coating.] M2.4 A 50% reduction in the Si content (as compared to the one with standard sealant) found in the active Ni/YSZ electrode of a cell with alternative sealant tested for >1000 hours demonstrated. [Milestone not fulfilled. During the project course, we have found out that unlike single cell testing, Si poisoning of the Ni/YSZ electrode is not one of the dominant degradation issues at the stack level under the electrolysis conditions explored so far. In combination with post-mortem results on long-term tested SOEC stacks, we concluded that the used sealing glass, which was considered to be one of the major sources for Si poisoning in single cell testing, can be used as sealant for SOEC stacks. The originally planned tasks for M2.4 therefore became less relevant to actual needs of SOEC technology development at its current stage. The resources set aside for this task (3 manmonths) was therefore reallocated to the more development task (high pressure SOEC stack testing) in this project.] M3.1 Two stack tests of current generation SOEC stacks under conditions relevant for grid balancing (defined in WT4.1, one for short-term balancing and the other for long-term) for a period exceeding 1000 hours with detailed post-mortem characterisations on cells and stack components completed. [Milestone fulfilled. Two stacks (K-695 and K-696) were tested according to the 100 % wind power profile defined in WT4.1 with a total period exceeding 3000 h.] M3.2 Three stack tests of current generation SOEC stacks under extreme operating conditions (defined in WT4.1) with detailed post-mortem characterisations on cells and stack components carried out. [Milestone fulfilled. Three stacks (K-730, K-744, and K-753) were tested under extreme operating conditions (high current density, high conversion, high steam content). PMA analyses were carried out on some of the here-tested stacks plus two 9

10 stacks tested in other projects. Loss of Ni in the active Ni/YSZ electrode and contacting on the oxygen side were identified as two of the most critical issues.] M3.3 Safe operation window for the current generation SOEC stacks mapped out. [Milestone fulfilled. HTAS Delta-type stacks can be operated rather safely at o C with electrolysis current density up to A/cm 2 and fuel conversion up to 60%, with an expected life-time of exceeding 9000 h.] M3.4 An SOEC stack containing more than 40 cells with a foot print of 12X12 cm 2 under electrolysis operation for 500 hours demonstrated. [Milestone fulfilled. A 75-cell TSP-1 stack (Q-691) was tested for CO 2 electrolysis over a period exceeding 500 hours.] M3.5 One stack test at 10 bar for a period exceeding 500 hours carried out. [Milestone partially fulfilled. Within the project period, we have successfully completed construction of the high pressure stack test rig. In addition, an 8-cell Delta-type stack (K-740) was tested under elevated pressure up to 20 bar. The durability test was carried out at 10 bar for close to 200 h. Several incidents, caused mainly by instrument technical issues, happened in-between, which eventually led to crack of cells and shutdown of the test. The current test proves feasibility of high pressure SOEC stack testing and improved cell performance at elevated pressure.] M3.6 Three stack tests of SOEC stacks composed of improved cells or ICs or impregnated with nanoparticles carried out for a period exceeding 500 hours. [Milestone fulfilled. In addition to Q-691 tested for M3.4, four more stack tests (K-751, X-030, U-001, Q-675) were carried out for M3.6 with a total testing period exceeding 2000 h. Among these four stacks, one Delta stack was tested for screening different IC coatings, and the other three stacks for testing either the new stack design (TSP-1) and/or new contact layers. The developed MnCo spinel coating showed promising results and is recommended for further development. Stable electrolysis operation was demonstrated on the new stack design TSP-1.] M3.7 One SOEC stack test under conditions relevant for grid balancing (defined in WT4.1) for a period exceeding 3000 hours with an ASR of less than 0.45 Ω cm 2 and an average degradation rate of less than 1.25%/1000 h at 1 A/cm 2 demonstrated. [Milestone not fulfilled. Two identical 50-cell TSP-1 stacks (X-076 and X-078) were tested for a period exceeding 1800 h. Due to adoption of the new stack design (TSP- 1), the requested test conditions in M3.7 are not technically feasible at the current stage. A more moderate and realistic test strategy for conducting the long term test in M3.7 was therefore chosen. The current test confirms stable electrolysis performance and good durability of TSP-1 stacks. Further development towards realizing the technical requirements specified in M3.7 will be pursued in the coming project.] M4.1 Grid balancing related test conditions (including extreme operation conditions) and test profile defined. [Milestone fulfilled. Analyses on grid balancing schemes based on integrating either wind or solar power and where solid oxide cells provide full energy balancing (by reversible electrolysis and fuel cell operation) have been carried out. The analyses have been converted into SOC stack operating profiles and employed in actual stack tests.] M4.2 Performance of a 14x14x3 cm 3 heat exchanger tested. 10

11 [Milestone fulfilled. Heat exchangers with the specified dimension were designed and successfully tested. Good agreement between the measured and COMSOL model predicted performance was achieved.] M4.3 Performance of a 14x14x3 cm 3 electrical heater tested. [Milestone not fulfilled. Planar electrical heaters with compact and production friendly mechanical design were produced. Due to technical issues with the test setup, it was unfortunately not possible to test the heaters and fulfil the milestone within the project period. Further development and testing of the electrical heaters is considered highly relevant and will be combined with similar activities in the project ForskEL ] M5.1 Results of the energy system analyses and assessment of economic consequences of applications of SOEC in different types of European energy systems compared with other types of electrolysis reported. [Milestone fulfilled. Different aspects of integrating SOEC into various energy systems were investigated. Firstly, a comparison between distributing produced syngas through natural gas grid versus direct conversion of syngas to liquid fuel and distribution in the liquid form was carried out. Secondly, potential for using combined capacities of SOEC and SOFC for grid balancing purposes was investigated. Furthermore, analysis on using steam electrolysis for fuel production versus co-electrolysis was conducted. The results indicate that there are no decisive differences between these two pathways for the same fuel outputs. Finally, analysis of using SOEC in different energy systems (a 100 % renewable Danish 2050 system and two types of German energy systems with or without nuclear energy) was done in order to determine potential of using electrolysis and what utilization capacities are necessary in order to supply fuel demand. Optimum electrolysis utilization ranges were identified for all the three systems.] The main technical results within each WP, including accounts of the work related to fulfilling the project milestones, are described in the following sections (Sections ). 11

12 1.3.2 WP1 Cell development and testing WP1 involved investigating the current generation SOEC cells under grid balancing related test conditions, exploring life-time limiting factors and mapping out a safe operation window, continuing to further understanding of degradation mechanisms, and based on the acquired knowledge identifying counter-acting measures and developing more reliable and larger SOEC cells for up-scaling. WP1 was undertaking mainly by DTU Energy, in close collaboration with HTAS. The activities in WP1 are divided into five work tasks (WTs) as listed below WT1.1 Safe operation window for current generation SOEC cells The purpose of WT1.1 was to investigate operation parameters on performance and durability of the current generation (so-called 2.5G-Standard) SOEC cells. The 2.5G-Standard cells consist of a Ni/YSZ support, a Ni/YSZ active fuel electrode, an YSZ electrolyte and a LSCF/CGO (LSCF: La 1-x Sr x Co 1-y Fe y O 3- ; CGO: Ce 1-x Gd x O 2- ) oxygen electrode with a CGO barrier layer at the interface between the electrolyte and the oxygen electrode. Three operation parameters were selected for the study in this project: current density, gas composition, extent of conversion for steam and/or CO 2. The main test conditions were chosen as 800 o C and -1 A/cm 2, oxygen fed to the oxygen electrode and different gas compositions fed to the fuel electrode with a flow corresponding to a steam/co 2 conversion of 56 %. The influence of gas composition was studied by varying the fuel electrode inlet gas composition from 90 % H 2 O + 10 % H 2 (i.e. steam electrolysis) to 45 % CO % H 2 O + 5 % CO + 5 % H 2 (coelectrolysis), and further to 90 % CO % CO (CO 2 electrolysis). The effect of current density or conversion was studied by varying the current density from zero up to A/cm 2, or conversion from 28 % to 56 %, respectively. A number of cell tests were repeated in order to verify reproducibility of the cell test results. The test detail is given in Table 1. Table 1: Overview of tested 2.5G-Standard cells in the project. Test no. Fuel gas composition Current density, A/cm 2 Conversion, % 15test37 H 2O/H 2 (90/10) test45 H 2O/H 2 (90/10) test63 a H 2O/H 2 (90/10) test139 H 2O/H 2 (90/10) test48 H 2O/H 2 (90/10) test44 H 2O/H 2 (90/10) Testing period, hour 3test123 CO 2/CO (90/10) test50 CO 2/CO (90/10) >1000 b 15test46 CO 2/CO (90/10) test119 CO 2/CO (90/10) test47 CO 2/CO (90/10) test88 CO 2/H 2O/CO/H 2 (45/45/5/5) test89 CO 2/H 2O/CO/H 2 (32.5/32.1/17.1/18.3) test91 CO 2/H 2O/CO/H 2 (45/45/5/5) Note: a) 4test63 was launched and reported in the project ForskEL and completed in the current project. b) 15test50 is still continuing by the time when this report is submitted. c) For steam electrolysis, due to limitation in steam supply, the gas flow to the Ni/YSZ electrode (H 2O+H 2 mixture with H 2O/H 2 = 90/10) was kept constant at 13.4 l/h. The conversion of steam into H 2 was therefore varied according to the applied current density. The evolution of cell voltages with time for steam electrolysis, CO 2 electrolysis, and coelectrolysis of steam and CO 2 are presented in Figures 1, 3, and 4, respectively. For steam electrolysis, the 2.5G-Standard cells are capable of electrolysis operation at current density up to -1 A/cm 2, while at A/cm 2 the cells show accelerated degradation. Impedance and post-mortem analyses indicate that degradation happened mainly at the Ni/YSZ electrode for 12

13 current densities up to -1 A/cm 2, while at A/cm 2 both Ni/YSZ and LSCF/CGO electrodes degraded. Voltage, mv o C, H 2 O/H 2 = 90/ A/cm 2-1 A/cm A/cm Time, hour Figure 1: Evolution of cell voltages with time under galvanostatic (constant current) conditions. All the cells were supplied with 13.4 l/h H 2O+H 2 mixture (H 2O/H 2 = 90/10) to the Ni/YSZ electrode and 50 l/h O 2 to the LSCF/CGO oxygen electrode, respectively. Figure 2 plots the predicted cell lifetime of 2.5G-Standard as a function of electrolysis current density for steam electrolysis. Cell lifetime is predicted by extrapolating the data from Figure 1 up to an end-of-life cell voltage of 1.5 V. A lifetime of h (1.5 years) and h (2.3 years) is predicted, for continuous electrolysis operation at -1 and A/cm 2, respectively. To achieve a lifetime of 5 years for continuous electrolysis operation, the electrolysis current density has to be below 0.45 A/cm 2. In the current project, a 6-month long-term cell test was completed for steam electrolysis at 800 o C and -1 A/cm 2 (4test63), demonstrating a steady state degradation rate of 22 mv/1000 h (or 1.7 %/1000 h). This is however still too high for industrial applications, which require a degradation rate of less than 0.5 %/1000 h (corresponding to a life time of 5-10 years). To improve the cell durability at -1 A/cm 2, the main focus shall be on improving the Ni/YSZ electrode (which has been realized in the current project, see Section of this report), while for even higher current density both Ni/YSZ and LSCF/CGO electrodes need to be improved. Life time, hour A/cm Current density, A/cm 2 Figure 2: Prediction of cell lifetime (for steam electrolysis) versus current density for the 2.5G-Standard cells tested in the current project. Similar degradation behavior was seen when the 2.5G-Standard cells were tested for CO 2 electrolysis (Figure 3). The cells show stable operation at electrolysis current density up to -1 A/cm 2, though the degradation rate is slightly higher than that for steam electrolysis. Much higher degradation was seen when the cells were tested for co-electrolysis of steam and CO 2 as compared to either steam electrolysis or CO 2 electrolysis, all at 800 o C and -1 A/cm 2 with 56 % conversion. Post-mortem SEM analyses on tested cells indicate that the higher degradation observed in cells tested for co-electrolysis might be to large degree due to impurity (Si, Al etc.) poisoning of the Ni/YSZ electrode. These impurities may come from gas supply, 13

14 evaporation of cell materials or test set-up components employed in single cell testing. It is speculated that evaporation of the impurities is enhanced in CO 2 +H 2 O mixture (as compared to H 2 /H 2 O or CO/CO 2 mixture). These impurities are transported via the gas stream to the Ni/YSZ electrode and further deposited at the active sites. The impurity deposition reaction requires a reduced P(O 2 ) which is enabled by large cathodic polarization. Reducing conversion of steam+co 2 decreases the Ni/YSZ cathode polarization and therefore prevents the impurity deposition reactions, even though the amount of evaporated impurities in the gas stream is the same as in the case of high conversion. The above speculation has been realized and confirmed in three cell tests shown in Figure 4. It has to be pointed out that impurity poisoning of the Ni/YSZ electrode is a complex phenomenon and that it depends on many operating parameters, such as glass seal, feed gas composition and purity, cathodic polarization etc. Further discussion on this topic is given in the next section (Section ). Voltage, mv o C, CO 2 /CO=90/ A/cm 2-1 A/cm A/cm Time, hour Figure 3: Evolution of cell voltages with time under galvanostatic (constant current) conditions. All the cells were supplied with 90 % CO % CO to the Ni/YSZ electrode (with a flow corresponding to 56 % conversion of CO 2 into CO) and pure O 2 to the LSCF/CGO oxygen electrode, respectively o C, -1 A/cm 2 Voltage, mv H 2 O/CO 2 /CO/H 2 =45/45/5/5, 28% H 2 O/CO 2 /CO/H 2 =32/32/18/18, 39% H 2 O/CO 2 /CO/H 2 =45/45/5/5, 56% Time, hour Figure 4: Evolution of cell voltages with time for co-electrolysis of steam and CO 2 at 800 o C and -1 A/cm 2. The cells were supplied with different gas compositions (as specified in the figure) to the Ni/YSZ electrode and pure O 2 to the LSCF/CGO oxygen electrode, respectively. Based on the cell tests and post-mortem analysis results, the following can be concluded on 2.5G-Standard cells: - The 2.5G-Standard cells are capable of electrolysis operation up to -1 A/cm 2 (at 800 o C) for electrolysis of steam, CO 2, or co-electrolysis of steam and CO 2. - When operated at -1 A/cm 2, a steady state degradation rate of <30 mv/1000 h (or 2.5 %/1000 h) is achievable. - The 2.5G-Standard cells show similar long-term durability for steam electrolysis, coelectrolysis, or CO 2 electrolysis, when the gasses are all cleaned using an in-house developed gas cleaning method. - In case of pure gas, conversion does not show a big influence up to 60 % at current density up to -1 A/cm 2. On the other hand, if the impurity level is high, reducing the conversion has been proved to be effective in reducing cell long-term degradation. 14

15 WT1.2 Advanced characterization and modelling methodologies An important impurity for the SOEC performance is Si, which tends to accumulate in the electrochemically active part of the cell. As reported in the previous project ForskEL , the Si level found in un-tested cells is much lower as compared to the values found after testing, and no enrichment was found during reduction or the initial performance evaluation. It can be concluded that the Si impurities must originate from one or more external sources, most likely from the employed sealing glass (containing 55 wt.% SiO 2 ). Extensive studies have been carried out on impurity distribution in cells exposed to long-term single cell testing, while the situation in SOEC stacks remains unclear. The first part of WT1.2 was devoted to investigate impurity distribution inside long-term tested SOEC stacks. Stack K-672, which was tested for steam electrolysis over a period of 2000 h (the test was reported in the previous project ForskEL ), was selected for the study. The stack was tested at -0.7 A/cm 2 for most of the period, with 150 hours at -1 A/cm 2. Except for the top and bottom cells, all the other nine cells showed similar performance and exhibited negligible degradation for the period at -0.7 A/cm 2, but large degradation when the current density was increased to -1 A/cm 2. To detect impurities at a concentration level of less than 1000 pm, time-of-flight secondary ion mass spectrometry (TOF-SIMS) is a powerful tool. In Stack K-672, three out of eleven cells were selected and cross-sections at 38 different locations in each cell were analyzed. An illustration of these locations is shown in Figure 5. This results in 114 measuring points in total. In this way, possible impurity gradients in the stack, e.g. caused by impurities carried in the fuel gas stream, can be identified. Figure 5: Overview image of the analyzed area. Red squares mark analyzed regions in Cells 1, 5 and 9, white numbers below refer to distance (in mm) from the first analyzed spot. Gas inlet is found on the left side, gas outlet on the right. A Cr concentration gradient is found from the gas inlet to the gas outlet. Figure 6 illustrates the Cr profile through the cross sections of cell at the gas inlet, in the middle of the cell and at the gas outlet. While a high level of Cr is found near the gas inlet, the concentration in the middle of the cell and near the gas outlet is significantly lower. Interestingly, also a Cr gradient from the outer Ni/YSZ support to the YSZ electrolyte is found. The highest Cr concentrations are measured in the electrochemically active part of the Ni/YSZ close to the dense YSZ electrolyte as shown in the line-scans in Figure 6. A more detailed analysis of the Cr gradient in Cell 1 is presented in Figure 7. For all the 38 investigated locations, the integrated Cr and Si signal intensity in the first 50 µm of the Ni/YSZ cathode is plotted. A gradient from the gas inlet to the gas outlet is clearly observed for Cr, but not for Si. Cr enrichment is mainly found throughout the first quarter of the cell, with the highest Cr concentration being found in the first 10 mm of the active part of the cell close to the gas inlet. No Cr enrichment is found at the area close to the gas outlet. Besides, a Cr gradient from bottom to top of the stack is found. 15

16 LSCF-CGO electrode + contact layer Ni-YSZ electrode Ni-YSZ support gas inlet (5 mm) gas inlet (10 mm) center of cell (40 mm) gas outlet (73 mm) Cr intensity [au] Distance from interface electrolyte/ni-ysz cathode [ m] Figure 6: Cross section intensity profile of Cr in the cell close to the gas inlet (black), 10 mm from the gas inlet (blue), in the centre of the cell (green) and close to the gas outlet (red). 200 Gas inlet Gas outlet norm. Cr intensity [au] 100 sealing area Cr concentration Si concentration sealing area Distance from gas inlet [mm] Figure 7: Integrated Cr (black) and Si (blue) counts in the active Ni-YSZ electrode for Cell 1. Since the Cr level found in the sealing area of the cells is much lower as compared to the level found in cell areas exposed to the fuel gas, it can be concluded that the Cr impurities must originate from one or more external sources. One possible external source of Cr impurities could be the non-coated steel pipes which connects the gas supply lines with the stack. However, other Cr sources are possible and must be considered as well. The effect of Cr enrichment on cell/stack electrochemical performance needs further evaluation. Unlike previously reported results from single cell testing, in Stack K-672, no increase of Si during electrolysis operation in any part of the cell was observed. Same type of sealing glass is used as sealants for both single cell testing and SOEC stacks, though the amount of glass seal differs. The sealing area in SOEC stacks is much smaller as compared to that in cell tests. This may actually help prevent Si contamination of the Ni/YSZ electrode to a large extent. In addition, most of the single cell tests carried out in the project were at -1 A/cm 2 or 16

17 above, while most of the stack tests were at A/cm 2 or below. A difference in the cathodic polarization is therefore expected. This probably contributes also to the difference observed on Si poisoning of Ni/YSZ between single cell testing and stack testing. In combination with post-mortem results on long-term tested SOEC stacks (see Section of this report), it can be tentatively concluded that Si poisoning of the Ni/YSZ electrode is not one of the dominant degradation issues at the stack level under the electrolysis conditions explored so far. The used sealing glass, which was considered to be one of the major sources for Si poisoning in single cell testing, can be used as sealant for SOEC stacks. The second part of WT1.2 was devoted to develop advanced post-mortem characterization tools: thin-film based electron backscatter diffraction (EBSD), which could be used to analyze various degradation phenomena induced by electrolysis operation under high current and to detect microstructure/composition/crystallography change at the micro- and nano-meter scale. Transmission Kikuchi Diffraction (TKD) is an adaptation of EBSD that is applied to the surfaces of polished samples in the scanning electron microscope (SEM). EBSD is classically limited to a spatial resolution of approximately 50 nm. Applying EBSD to a transmission electron microscope (TEM) sample (i.e the TKD technique) has been shown to significantly improve the spatial resolution on idea metallic samples as the sample is only ~100 nm in thickness. The novel aspect of this work is to apply the technique on less ideal samples such as SOECs. In this work, a EBSD detector was installed on a recently acquired high resolution Zeiss Merlin SEM. The EBSD system was then calibrated for TKD investigations as part of the project on an yttria stabilized zirconia (YSZ) porous backbone structure impregnated with nickel and gadolinia doped ceria (CGO) nano-particles and on a Ni/YSZ SOEC cathode. Figure 8 shows the TKD results on the first sample. The left image shows a so-called pattern quality image where bright regions correspond to high quality diffraction patterns and vice versa. This image provides a good qualitative view of both the microstructure and the data quality. Here the size and shape of the individual YSZ crystals can be easily distinguished (ranging approximately from 200 nm to 1 µm in diameter). Also many small crystals surrounding the YSZ backbone are also clearly visible which correspond to the impregnated Ni and CGO nanoparticles. These are typically less than 100 nm in diameter. The right image in Figure 8 is color-coded to show the crystal orientation relative to the direction perpendicular to the image. The orientation information is merged with the pattern quality and the points where no orientation information is available are transparent. Here the individual YSZ crystals are easily discernible by their various colors. It can be seen that in the regions of the crystal boundaries and other places there are several missing data points. In the regions around the nanoparticles the data quality is poorer; however there are a large number of nanoparticles where orientation determination was successful. In the present data it is not possible to uniquely identify individual particles as YSZ, Ni or CGO, as they all share similar crystal structure. The insets in Figure 8 however, clearly show that a resolution of 10 nm or better is obtainable in this type of sample. This can be seen by the fact that there are no missing data points on either side of the black lines that indicate internal boundaries within the Ni nanoparticles. The black lines indicate special boundaries that are unique to Ni. Based on Figure 8, TKD on the Merlin SEM meets the technical requirements specified in Milestone M M1.3 Applicaton of thin film based EBSD to an SOEC electrode microstructure with a spatial resolution of 10 nm or less demonstrated. 17

18 Figure 8: Transmission Kikuchi diffraction data obtained by applying the electron backscatter diffraction (EBSD) technique to a transmission electron microscope sample. The sample is an yttria stabilised zirconia (YSZ) porous backbone structure impregnated with nickel and gadolinia doped ceria (CGO) nano-particles. The left image is a diffraction pattern quality map and the right image is a color coding of crystallographic orientation (raw data). The insets show two Ni particles containing internal grain boundaries resolved with 10 nm spatial resolution. See text for detailed description. Inset magnification 4x WT1.3 Optimized SOEC cells with enhanced durability The objective of WT1.3 was to develop SOEC cells with enhanced performance and improved durability for electrolysis operation. It continued the efforts launched in the project ForskEL on Ni/YSZ supported SOEC cells and use various measures to further improve the durability at current density equal to or above 1 A/cm 2. The main focus is on optimizing MTC (multi-layer tape casting) halfcells via varying the Ni content or particle size distribution. A MTC halfcell consists of support, active Ni/YSZ electrode, YSZ electrolyte, and CGO barrier layer with all layers co-sintered in one-go. The cells were further applied with LSC/CGO oxygen electrode by screen printing. The full cells are named as 2.XG. In standard MTC halfcells, the active Ni/YSZ electrode has a Ni/YSZ ratio of 40/60 (volume). The optimization work on the Ni content was carried out by varying the Ni/YSZ volume ratio from 40/60 (standard), to 45/55, and further to 50/50, while keeping the oxygen electrode side unchanged. The cells were tested for steam electrolysis mode for 1000 hours. All the long-term tests were carried out at 800 o C and -1 A/cm 2, oxygen fed to the oxygen electrode and 90 % H 2 O in H 2 fed to the fuel electrode with a flow corresponding to a steam utilization of 56 %. Figure 9 presents development of cell voltage, ohmic resistance (R s ) and polarization resistances (R p ) for the three tests. The three cells show similar trend for the development of R s over time. The main difference lies in the development of R p : the lower Ni-content in the hydrogen electrode, the faster the cells resistance due to the charge transfer resistance at the triple phase boundary (TPB), i.e. the R p(ni,tpb), increases in start of the test. Basically, it reaches a stable level of hydrogen electrode performance after approximately 200 h of electrolysis test; whereas the cell with the most Ni (14t140) reach a stable level for R p(ni,tpb) after approximately 600 h of galvanostatic electrolysis testing. 18

19 Figure 9: Development of cell voltage, ohmic resistance (Rs) and polarization resistance (Rp) for 14t132 (Ni/YSZ = 40/60), 14t134 (Ni/YSZ = 45/55) and 14t140 (Ni/YSZ = 50/50) during electrolysis testing at 800 oc, -1 A/cm2, oxygen fed to the oxygen electrode, 90 % H2O in H2 fed to the fuel electrode with a flow corresponding to a steam utilization of 56 %. Post-mortem analyses via SEM of the long-term tested MTC cells show some general trends and representative SEM images for the cell with a Ni/YSZ ratio of 45/55 is given as example in Figure 10. (a) 5 µm (b) Long-term tested SOEC (c) 10 µm 5 µm Reference SOEC (d) Long-term tested SOEC 10 µm Reference SOEC Figure 10: SEM images of electrolyte-ni/ysz electrode interface of the (a) long-term tested 45/55-cell and (b) reference 45/55-cell. Arrows in (a) indicate microstructural changes in the electrode of the long-term tested SOEC. Lowvoltage in-lens SEM images of the same interface for the (c) long-term tested 45/55-cell and (d) a reference 45/55-cell. Percolating Ni particles appear bright in the low-voltage in-lens SEM images. The thickness of the active Ni/YSZ electrode is marked by the red arrows in (c) and (d). 19

20 Comparing electrolyte-ni/ysz electrode interface (Figure 10a and b) - the area fraction of the pore phase is significantly larger in the long-term tested 45/55-cell compared to the reference cell. Some examples of these larger pores are pinpointed by the white arrows in (a). The larger pores are found in the 3-5 m of the Ni/YSZ electrode closest to the electrolyte. This could indicate that the larger pores that can be found in the long-term tested 45/55-cell (but not in the reference cell) are located in positions where Ni particles were initially positioned. Regarding the percolating Ni network for the long-term tested 45/55-cell and the reference cell (Figure 10c and d) - the long-term tested 45/55-cell has a significantly lower fraction of percolating Ni in the 3-5 m of the Ni/YSZ electrode closest to the YSZ electrolyte. Both the number of and area fraction of percolating Ni in this region is significantly lower for the long-term tested 45/55-cell when compared to the reference cell. Roughly speaking these observations correspond to the fact that the tested 45/55-cell now has an extra layer of porous electrolyte. This observation corresponds well to the observed significant and continuous increase in the ohmic resistance during long-term electrolysis testing. The cell test results for the three MTC cells having varying Ni/YSZ ratios and the sub-sequent post-mortem analyses via SEM led to further optimization of the Ni/YSZ electrode aiming for increased long-term stability during steam electrolysis and special efforts were made to manufacture electrode structures to minimize the observed migration of Ni from the active Ni/YSZ electrode. The most successful one turns out to be the cell having standard Ni/YSZ ratio of 40/60 but optimized microstructure in the active Ni/YSZ electrode. The cell was tested for steam electrolysis at the same condition as the earlier tests. As shown in Figure 11, the cell (2.XG, 2014 Version, the red curve) has a long-term degradation (last 1000 h) of only %/1000 h (voltage degradation), much lower than that for 2.5G-Standard (1.7 %/1000 h). 800 o C, -1 A/cm 2, H 2 O/H 2 =90/10, 56% Voltage, mv G-Standard 2.XG (2014 Version) Time, hour Figure 11: Evolution of the cell voltages with time under galvanostatic (constant current) conditions. All the cells were supplied with 90 % H 2O + 10 % H 2 to the Ni/YSZ electrode (with a flow corresponding to 56 % conversion of H 2O into H 2) and pure O 2 to the LSCF/CGO oxygen electrode, respectively. Post mortem SEM analysis of the 2000 h tested 2.XG cell (2014 Version) confirms that the structure stability of the active Ni/YSZ electrode has been improved significantly. Figure 12 shows representative SEM images of the reference cell (left, being reduced but not long-term tested) and the 2000 h tested cell (right), which provides qualitative evidence that a very satisfying percolating Ni network exists in the fuel electrode after 2000 h of electrolysis testing at -1 A/cm 2 and 90 % H 2 O in inlet gas for the structure optimized cell which fits well with the observed long-term stability observed during cell testing (Figure 11). 20

21 Reference 2000 h electrolysis test Figure 12: Representative SEM images of electrolyte-ni/ysz electrode interface of the reference cell (non-long-term tested sister cell) and the 2000 h electrolysis tested one. The images are in-lens low-voltage images in which Ni in a percolating network will appear bright. The cell (2.XG, 2014 Version) was also tested at A/cm 2 as specified in Milestone 1.4. In addition, reproduction of the cells was further carried out and the durability of the reproduced cells was evaluated at both -1 and A/cm 2. For the tests at A/cm 2, a new cell test house developed in the project was employed, which allows for increased steam supply and the steam conversion can therefore be kept unchanged (56 %). Figure 13: Development of cell voltage, ohmic resistance (R s) and polarization resistance (R p) for the tests on 2.XG cells (2014 Version). All galvanostatic tests were operated at 800 o C, oxygen fed to the oxygen electrode, 90 % H 2O in H 2 fed to the fuel electrode with a flow corresponding to a steam utilization of 56% both for the -1 A/cm 2 test and for the A/cm 2. The cells used for the blue and black marked tests are sister cells. As shown in Figure 13, when the cell was tested at A/cm 2, it underwent a heavy degradation during the 1000 h test and the long-term degradation rate based on the 500 h of test is approximately 16 %/1000 h. Even though we have a long-term degradation as low as %/1000 h at -1 A/cm 2 (which is significantly better than any results so far reported in literature), the same cell cannot withstand similar SOEC test at A/cm 2 as it led to a long-term degradation rate around 16 %/1000 h. For the long-term degradation behavior (1000 h-scale) of these 2.XG cells (2014 Version), it is ohmic resistance that is the dominant course of degradation. Furthermore, it seems that the applied LSC/CGO oxygen electrode is 21

22 not stable when operated at A/cm 2. Based on the results presented here and previously reported cell test results, we believe that the cell and/or electrode over-potential plays an important role in the degradation of the cells, especially with respect to initiating the observed long-term (almost linear) ohmic resistance (R s ) increase WT1.4 Optimized SOEC cells with improved mechanical strength at larger scale The strength of a halfcell relies mainly on its Ni/YSZ support layer. For 2.5G-Standard cells, the mechanical properties of the Ni/YSZ support have been well characterized and reported. In the previous project ForskEL , mechanical properties of the Ni/YSZ support layer from a specific batch of 2.5G-Standard (2012 version) were evaluated. In the nonreduced state, the support has a porosity of %. The measured Weibull strength and modulus are 336 MPa and 13, respectively. These results are in line with those previously reported 5. The purpose of WT1.4 was to develop large cells with same or even enhanced mechanical strength, focusing on MTC cells. MTC constitutes a nice mean to reduce the manufacturing costs of SOEC cells by reducing the number of steps necessary to produce a halfcell. However, due to thermal expansion mismatch between the layers high residual stresses may appear. In this project we first investigated the stresses in MTC cells. The mechanical strength of the MTC halfcells without (MTC-3) or with CGO barrier layer (MTC-4) was characterized at room temperature. The samples were investigated in a bi-axial flexure experiment (ball on ring: BoR) in two directions: support at the bottom or support on top. The estimated stresses calculated by finite element method, leading to values of support strength (in the case with support at the bottom) close to the previously reported values 5 (Figure 15). The Weibull modulus was however much higher (~20), which could be due to the fact that all the MTC samples come from one and unique cell. However, it shows in any case that the MTC halfcells are of good quality with homogeneous repartition of flaws. In addition, a parametric study has been pursued on the residual stresses of a MTC-4 halfcell with varying layer properties. It resulted in a major influence of thermal expansion coefficient (TEC) values and secondary of YSZ layer thickness. Figure 14: Schematic presentations of a) MTC-3 and b) MTC-4 halfcells. 5 H. L. Frandsen, T. Ramos, A. Faes, M. Pihlatie, and K. Brodersen, Optimization of the strength of SOFC anode supports, Journal of the European Ceramic Society, 32 [5] (2012). 22

23 Figure 15: a) Weibull strength and b) Weibull modulus of the MTC-3 cell (468-13) and MTC-4 cell (471-1) compared to previously reported values from literature 5. Further cell development work in WT1.4 focused on improving the mechanical strength of the Ni/YSZ support for MTC halfcells produced at DTU Energy. This was done by optimizing the fabrication process (and hence reducing the number of flaws in the produced Ni/YSZ support). A second route was to increase the alumina content in the Ni/YSZ support. The standard MTC Ni/YSZ support was produced from a mixture of high purity NiO powder, commercial 3 mol% Y 2 O 3 stabilized ZrO 2 powder (3YSZ) and 0.6 wt.% Al 2 O 3. In the second type, the 3YSZ powder was replaced by TZ3Y20AB (3YSZ with 20 wt.% Al 2 O 3 ). It corresponds to ~8.8 wt.% of Al 2 O 3 in the support. The support layers were prepared by tape casting, further subjected to sintering. The strength of the samples was then tested by 4-point bending, either at room temperature in air or at 800 C in H 2 /N 2 (9/91) atmosphere. The results are summarized in Table 2. Table 2: Weibull strength ( 0), Weibull modulus (m), effective volume and scaled strength for the various samples. Sample porosity (MPa) m V 0 (mm 3 ) (MPa) scaled to 1 mm 3 Std_RT_Ox AB_RT_Ox Std_RT_Red AB_RT_Red Std_HT_Red AB_HT_Red Note: The first part of the sample names indicates the alumina content: Std for 0.6 wt.% and 20AB for 8.8 wt.%. The second part points to the temperature of the 4-point bending measurements: RT for room temperature and HT for 800 o C. The last part indicates whether the sample is reduced ( Red ) or not ( Ox ). As compared to the Ni/YSZ support layer from 2.5G-Standard (2012 version), the MTC Ni/YSZ support exhibit both higher Weibull strength and Weibull modulus. The improvement was mainly due to improved ceramic processing. Further improvement of MTC Ni/YSZ support via increasing the alumina content was not successful. The oxidized 20AB supports were slightly stronger that the standard ones but the reduced ones were slightly weaker (both at room temperature and 800 C). This strength variation is not very significant, but it is clear that the 20AB supports are not notably stronger or weaker than standard supports. On the other hand, even though the 20AB supports are not really stronger, another interest of 20AB would be the decrease of thermal expansion coefficient of the supports which would become closer to the one of the electrolyte. As shown in Figure 16, the halfcells with 20AB support is significantly flatter as compared to the one with Std, which may be advantageous for stack production. 23

24 Figure 16: Schematic explanation of curvature origin in SOEC halfcells and picture of half cells with standard support (Std_HC) and 20AB support (20AB_HC). The enlargement of SOEC cell footprint was then done on standard Ni/YSZ support produced by tape casting. The standard support has a footprint of 13x13 cm 2. Enlarging the support size from 13x13 cm 2 to 18x18 cm 2 is not a trivial matter and many factors might induce a decrease of the Weibull modulus, which represents the homogeneity of the strength in a ceramic material or compound. The simple fact of changing from a known, standard procedure might induce some variations. Additionally, increase in the tape dimensions induces larger potential inhomogeneity along the support tape width and also a potentially larger thermal gradient inside the sintering furnace. Despite all these challenges, supports of 19x19 cm 2 (19x19_DTU_2014_tape30) were successfully produced (Figure 17) from a large green tape (30 cm in width, before sintering). The mechanical strength of sintered Ni/YSZ support was tested by 4-point bending method and the results are presented in Table 3. Figure 17: Picture of sintered Ni/YSZ support with a footprint of 13x13 cm 2 and 19x19 cm 2. No result is available on the mechanical strength of 2.5G-Standard support at a foot print of 19X19 cm 2. A comparison can then be made only between a MTC support of 19X19 cm 2 and a 2.5G-Standard support of 13X13 cm 2, which is not fair. However, based on the results presented in Table 3, we can still conclude the following: - Due to improved ceramic processing, the Weibull modulus of MTC support at a footprint of 13X13 cm 2 is significantly higher (1.5 or 2 times) than that of 2.5G-Standard support with the same footprint. - The Weibull modulus of MTC support at a footprint of 19x19 cm 2 is similar to that of 2.5G-Standard support at a footprint of 13X13 cm 2. 24

25 - For MTC support, a decrease in Weibull modulus happened when increasing the footprint from 13X13 cm 2 to 19X19 cm 2. This is mainly due to inhomogeneity introduced when up-scaling the cell fabrication process, which will be minimized in further process optimization. Table 3: Weibull parameters and main characteristics of the compared batches. Sample Total porosity, % 0, MPa Modulus Footprint, cm 2 2.5G-Standard support (2012 Version) 13.95% X13 2.5G-Standard support (2014 Version) 12.75% X13 MTC support (2014) 18.51% X13 Large MTC support (2014) 13.75% X WT1.5 Next generation SOEC cells Both 2.5G-Standard and 2.XG cells show accelerated degradation when tested at A/cm 2 or higher current density. Impedance and post-mortem analysis results indicate that neither the fuel electrode (Ni/YSZ), nor the oxygen electrode (LSCF/CGO or LSC/CGO) can withstand such high current density. Further enhancement on cell durability at current density above A/cm 2 requires exploration of new electrode/electrolyte materials or architecture. In this project, we have developed a novel Ce 0.9 Gd 0.1 O 2 δ (CGO) fuel electrode and investigated its performance and durability at high current density. CGO electrode is a mixed ionic and electronic conducting (MIEC) electrode with nonnegligible electronic conductivity. CGO is known to have a higher ionic conductivity than Ni/YSZ, better spreading the electrochemical reactions in the electrode thickness so that the over-potential is less concentrated near the electrode/electrolyte interface. With an expected overall lower over-potential than Ni/YSZ, the electrode is expected to display not only a better performance but also better durability at high electrolysis current density than the Ni/YSZ. In this work, a CGO fuel electrode in symmetric cell geometry has been investigated and 550 h of operation at -1.9 A/cm 2 demonstrated. The investigated solid oxide cell (Cell A) is a CGO electrode screen-printed on a 190 µm thick 8YSZ electrolyte in a symmetric cell geometry (i.e. CGO 8YSZ CGO) and sintered at 1200 C for 2 h. The active electrode area is ca cm 2 and ca. 15 µm thick. The reference cell (Cell B) is a fuel electrode supported full cell with a 4 x 1 cm 2 active electrode area with Ni/3YSZ support (ca. 0.3 mm), Ni/8YSZ fuel electrode (ca. 10 μm), 8YSZ electrolyte (ca. 10 μm), PVD CGO buffer layer (ca. 1 μm) and LSC/CGO oxygen electrode (ca. 30 μm). Both cells were characterized at 800 C in a fuel mixture H 2 /H 2 O of 0.5/0.5. Cell A, although intended to be operated at -2 A/cm 2, was operated in a symmetrical cell setup at -1.9 A/cm 2 instead due to maximum current limitations of the galvanostat. Cell B was tested at -1.5 A/cm 2 in a full cell test setup. The fuel flow rate for Cell B was 25 l/h against 6 L/h for Cell A maximum possible flow rate of the symmetrical cell test set-up. As shown in Figure 18, after 540 h of operation at -1.9 A/cm 2, the voltage of Cell A has only changed by 185 mv against 683 mv for that of Cell B after 348 h of operation at -1.5 A/cm 2. These values correspond to relative degradation of 11 and 56 %. Thus Cell A has degraded by a factor of 5 less than Cell B. The corresponding degradation rates are 0.34 and 1.96 mv/h or 0.02 and 0.16 %/h for Cell A and Cell B, respectively. 25

26 Figure 18: Voltage evolution of Cell A (a) and Cell B (b) as well as the corresponding OCV impedance spectra in (c) and (d) respectively. The corresponding DRTs are displayed in figures (e) and (f) respectively. To crystalize out the contributions of the fuel electrode (CGO or Ni/YSZ) to the overall ageing, CNLS fitting of the measured impedance spectra was carried out and the results are summarized in Table 4. For Cell A operated at 1.9 A/cm 2 for 540 h, the CGO fuel electrode has degraded by 67.8 %, in comparison with Cell B that is operated at -1.5 A/cm 2 solely for 348 h and the Ni/YSZ fuel electrode degraded for 1207 %. Further, with an initial contribution of 51.5 mω cm 2 for both electrodes of the symmetric cell of Cell A, the single CGO electrode would be having an ASR of 51.5/2 = 25.6 mω cm 2 against 42.9 mω cm 2 for the Ni/YSZ fuel electrode of Cell B. It also needs to be pointed out that by polarizing the symmetric cell so that the working electrode operates in SOEC mode, the counter electrode is operated in SOFC mode. As such the response obtained from both electrodes is actually an average value of SOEC operation of the working electrode and SOFC operation of the counter electrode. Table 4: CNLS fit results of reference cell (Cell B) spectra before and after 1.5 A/cm 2 load operation at 800 C and 50/50 H 2/H 2O ratio. Cell A (CGO) Cell B (Ni/YSZ) Time, h Fuel electrode resistance, mω cm 2 Time, h Fuel electrode resistance, mω cm 2 0 h h h h Degradation, mω cm Degradation, mω cm Degradation, % 67.8 Degradation, % Degradation rate, mω cm 2 /h Contribution to total ASR degradation, % 0.06 Degradation rate, mω cm 2 /h 28.8 Contribution to total ASR degradation, % In conclusion, we have successfully demonstrated a CGO fuel electrode with an initial electrode resistance of 25.6 mω cm 2 at 800 C in a 50/50 H 2 /H 2 O mixture, better than a state of the art Ni/8YSZ electrode with an initial ASR of 42.9 mω cm 2 under similar conditions of tem- 26

27 perature and fuel composition. Furthermore, despite operating at 1.9 A/cm 2, the CGO fuel electrode had only degraded by 67.8 % against 1207 % of the Ni/YSZ electrode at 1.5 A/cm 2. This reveals a factor of 18 better durability of the CGO electrode. In this project, six milestones were set up in the project plan for WP1 Cell development and testing: M1.1 Safe operation window (in terms of current density, temperature of operation, cycling rate, degree of conversion, etc.) for current generation SOEC cells mapped out and reported. M1.2 Analysis of impurity distribution inside long-term tested SOEC stacks evaluated. M1.3 Application of thin film based EBSD to an SOEC electrode microstructure with a spatial resolution of 10 nm or less demonstrated. M1.4 Optimised SOEC cells tested at 1.25 A/cm 2 over a period exceeding 1000 hours with a degradation rate less than 1%/1000h. M1.5 An SOEC cell with a foot print of 18X18 cm 2 with improved Weibull modulus by 15% (as compared to 2.5 G cells of the 2012 generation) demonstrated. M1.6 New electrode materials/architecture tested at 2 A/cm 2 over a period of 500 hours with a degradation rate of half that of the current generation SOECs demonstrated. For Milestone 1.1, we investigated performance and durability of 2.5G-Standard cells as functions of current density, feed gas composition, and extent of conversion for steam and/or CO 2. The cells show similar long-term durability for steam electrolysis, co-electrolysis, or CO 2 electrolysis, when the gasses are all cleaned using an in-house developed gas cleaning method. Based on the cell test results, a safe operation window of current density up to - 1 A/cm 2, fuel conversion up to 60 %, and CO 2 +H 2 O content in the feed gas composition up to 90 % was proposed for the 2.5G-Standard cells. When operated at -1 A/cm 2, a steady state degradation rate of <30 mv/1000 h (or 2.5 %/1000 h) is achievable. With the above results, M1.1 is fulfilled. For Milestone 1.2, impurity (especially Si) distribution inside a long-term tested stack (Stack K-672) was analysed using TOF-SIMS. Unlike previously reported results from single cell testing, in Stack K-672, no increase of Si during electrolysis operation was observed. The difference between cell and stack test results was ascribed to different amounts of glass seal employed in testing and different cathodic polarization the cells were exposed to. In combination with post-mortem results on long-term tested SOEC stacks, it was concluded that Si poisoning of the Ni/YSZ electrode is not one of the dominant degradation issues at the stack level under the electrolysis conditions explored so far. The used sealing glass, which was considered to be one of the major sources for Si poisoning in single cell testing, can be used as sealant for SOEC stacks. Milestone 1.2 is considered fulfilled. For Milestone 1.3, TKD investigations were carried out on two samples: an YSZ porous backbone structure impregnated with Ni and CGO nano-particles and a Ni/YSZ SOEC cathode, with 10 nm spatial resolution demonstrated. This fulfils M1.3. For Milestone 1.4, we have developed SOEC cells with enhanced performance and improved durability for electrolysis operation. The main effort was devoted to optimizing MTC halfcells via varying the Ni content or particle size distribution, with a focus on manufacturing electrode structures to minimize the observed migration of Ni from the active Ni/YSZ electrode. The halfcells were further applied with LSC/CGO oxygen electrode by screen printing. With the most successful cells (2.XG, 2014 version), a long-term degradation rate of only %/1000 h was demonstrated at 800 o C and -1 A/cm 2 for steam electrolysis. Post-mortem results confirm that a very satisfying percolating Ni network exists in the fuel electrode after 2000 h testing. When tested at A/cm 2, the same cell underwent heavy degradation 27

28 during the 1000 h test. The degradation was correlated to changes in both Ni/YSZ and LSC/CGO electrodes. Even though the 2.XG cells showed much higher degradation rate at A/cm 2 than specified in Milestone 1.3, the degradation rate at -1 A/cm 2, which is as low as %/1000 h and which ensures a cell life-time of more than five years, is significantly better than any results so far reported in literature. In Roadmap SOEC Steam Electrolysis and Co-electrolysis formulated by Danish Partnership for Hydrogen and Fuel cells 6, the target for improved cell durability is set as <1.25 %/1000 h at 800 o C and -1 A/cm 2, while the target is <0.75 %/1000 h at 750 o C and A/cm 2. The results obtained in this project are in line with the SOEC strategy and roadmap and have reached the target. Based on the above, we conclude that M1.4 is partially fulfilled. For Milestone 1.5, we have improved mechanical strength of MTC support and have successfully enlarged the cell footprint from 13X13 cm 2 to 19X19 cm 2. At a footprint of 13X13 cm 2, the Weibull modulus of MTC support is significantly higher (1.5 or 2 times) than that of 2.5G- Standard support. When enlarged to a footprint of 19x19 cm 2, the Weibull modulus of MTC support is reduced, but still similar to that of 2.5G-Standard support at a footprint of 13X13 cm 2. The decrease in Weibull modulus (when increasing the footprint from 13X13 cm 2 to 19X19 cm 2 ) is mainly due to inhomogeneity introduced when up-scaling the cell fabrication process, which will be minimized in further process optimization. M1.5 is considered partially fulfilled. For Milestone 1.6, we have demonstrated a CGO fuel electrode with an initial electrode resistance of 25.6 mω cm 2 at 800 C in a 50/50 H 2 /H 2 O mixture, better than a state of the art Ni/8YSZ electrode with an initial ASR of 42.9 mω cm 2 under similar conditions of temperature and fuel composition. Furthermore, despite operating at 1.9 A/cm 2, the CGO fuel electrode had only degraded by 67.8 % against 1207 % of the Ni/YSZ electrode at 1.5 A/cm 2. This reveals a factor of 18 better durability of the CGO electrode. With these results we conclude that M1.6 is fulfilled. 6 Roadmap SOEC Electrolysis and Co-electrolysis, Danish Partnership for Hydrogen and Fuel Cells, September

29 1.3.3 WP2 Interconnects and seals WP2 focused on stack components, mainly interconnects (ICs) and seals. Originally two lines of work were planned: 1) to investigate and further improve corrosion resistance of the current generation ICs, 2) to further develop alternative seals with low Si emission WT2.1 IC corrosion under extreme conditions The first part of WT2.1 deals with IC corrosion under some extreme electrolysis operating conditions. Two scenarios were investigated: Ni diffusion induced IC phase transformation and corrosion in 100 % CO 2. To study Ni diffusion induced IC phase transformation, Crofer 22 APU alloy sheets with electroplated Ni coating were oxidized at 800 C in humidified hydrogen for periods up to 2000 hours and the corrosion resistance were evaluated according to mass gain and oxide scale growth. As shown in Figure 19, the corrosion rate obtained for the Crofer 22 APU initially plated with 13 µm thick Ni layer is almost 4 times lower than that for uncoated Crofer 22 APU. This was further confirmed by SEM cross-section images (Figure 20), where the Ni coated samples show thinner oxide scale than the uncoated one. Ni containing interconnect alloys (> 8 wt.%) crystallize in a face centered cubic (FCC) structure whereas Fe-Cr alloys normally have a body centered cubic structure (BCC). Diffusion of Cr in the FCC lattice is much slower than in the BCC lattice, which may account for slowing down the corrosion kinetics by pre-coating Ni. However, for real application other effects must be studied, such as possible mechanical stresses due to TEC mismatch between FCC and BCC structure. Corrosion rate / x10-14 g 2 cm -4 s no Ni 7 um Ni 13 um Ni Figure 19: Evaluated corrosion rates for Crofer 22 APU with and without nickel plating oxidized at 800 C in humidified hydrogen for 2000 h. Figure 20: SEM cross-section images on the oxide scale formed on steels after 1000 hour oxidation at 800 C in humidified hydrogen. Left: no coating; Middle: with 7 µm thick Ni coating; Right: with 13 µm thick Ni coating. 29

30 For the study of IC corrosion in pure and dry CO 2, the following samples were included: shaped ICs with Co coating and without coating, Crofer 22 APU and Crofer 22 H flat metal sheet without coating, Crofer 22 APU flat metal sheet coated with Ce 0.9 Gd 0.1 O 1.95 (CGO). Crofer 22 H is considered as a future alternative for Crofer 22 APU, while Ce 0.9 Gd 0.1 O 1.95 is chosen as one of the candidate materials for the fuel side coating. The samples were oxidized at 800 C in flowing CO 2 gas for periods up to 1000 h. The corrosion behaviour was evaluated by measuring mass gain and oxide scale thickness. The corrosion rate constants evaluated from mass gain are presented in Figure 21. Among the five types of samples, CGO coated Crofer 22 APU possesses the lowest corrosion rate constant. The corrosion rates of Co-coated and uncoated ICs are about five times and twice of that of its base alloy (uncoated Crofer 22 APU), respectively. This indicates that the IC shaping process may have a negative influence on alloy corrosion in CO 2. Cobalt coating on the fuel side of ICs seems to have a negative influence on corrosion resistance in CO 2, while the CGO coating instead has a positive effect. 800 o C in CO 2 CGO-coated Crofer22APU Sample Uncoated Crofer22H Uncoated Crofer22APU Co-coated IC uncoated IC Corrosion rate / x10-14 g 2 cm -4 s -1 Figure 21: Evaluated corrosion rates during oxidation in CO 2 at 800 o C. The corrosion rate constant for Co-coated IC was corrected for oxidation of Co into Co 3O 4. The cross-sectional view of the oxidized samples is presented in Figure 22. For shaped ICs (both Co-coated and uncoated), the oxide scale shows very poor adherence to the steel. It spalls off during SEM sample preparation and destroys the structure. The SEM analysis was then focused on flat samples. As shown in Figure 22, the oxide scale was well adhered to the base alloy. The oxide scale on uncoated Crofer 22 APU and Crofer 22 H shows similar thickness, while the CGO-coated Crofer 22 APU has slightly thinner oxide scale. This is in agreement with the results from the measured mass gain. In the present study, no sign of carbide formation or carbon diffusion into the alloy was detected, which leads to a tentative conclusion that pure CO 2 may be relatively safe with regard to the fuel side interconnect corrosion. This could be due to the fact that the formed dense oxide scale totally blocks inward carbon diffusion. Increasing carbon activity in the gas atmosphere or introducing carbon deposition via catalysts such as Ni could be considered in future studies. Figure 22: High magnification SEM images of the oxide scale on alloys after 1000 hours oxidation in pure CO 2 at 800 C. Left: uncoated Crofer 22 APU, middle: uncoated Crofer 22 H, right: CGO-coated Crofer 22 APU. The second part of WT2.1 was devoted to investigate effect of current density on IC corrosion. Area specific resistance (ASR) of Crofer 22 APU with different types of coatings was measured at 800 o C in air at either 0.5 or 2 A/cm 2. As shown in Figure 23, increasing the 30

31 current density from 0.5 to 2 A/cm 2 did not cause any significant change on ASR degradation of coated Crofer 22 APU samples. 0,05 0,04 MnCo spinel coated-2 A/cm 2 MnCo spinel coated-0.5 A/cm 2 0,05 0,04 Y+MnCo spinel coated-2 A/cm 2 Y+MnCo spinel coated-0.5 A/cm 2 0,05 0,04 Co coated-2 A/cm 2 Co coated-0.5 A/cm 2 ASR, Ohms.cm 2 0,03 0,02 ASR, Ohms.cm 2 0,03 0,02 ASR, Ohms.cm 2 0,03 0,02 0,01 0,01 0,01 0,00 0,00 0, Time, hours Time, hours Time, hours Figure 23: ASR plots of Crofer 22 APU with different coatings measured at 800 o C in air under either 0.5 or 2 A/cm 2. In the previous project ForskEL , the MnCo spinel ((MnCo) 3 O 4 ) coating developed by the electrophoretic deposition (EPD) technique showed promising results both at the component level and at the stack level. The coating development work was continued in the present project. The last part of WT2.1 was to scale up the electrophoretic deposition (EPD) coating process and produce full scale coated ICs to be tested at the stack level. In this project, the EPD coating process was further optimized and adapted to a new powder composition (Co 1.5 Mn 1.5 O 4 MCO15, as the supplier no longer offers the old composition Co 2 Mn 1 O 4 MCO20). Oxidation of full scale ICs with different coatings was carried out at 800 o C in air for periods up to 2000 hours. As shown in Figure 24, the samples with EPD coating (MCO15 and MCO20) show considerably lower mass gain than uncoated samples. In addition, a new coating method, electrolytic deposition (ELD), was also tried, and showed promising results. The samples with ELD yttria (Y 2 O 3 ) coating showed similar mass gain as compared to those EPD coated samples. On the other hand, the samples with Co coating (on both sides) showed high mass gain due to initial oxidation of Co to Co 3 O 4 and then a mass increase similar to those of the uncoated samples, proving that the Co coating is less corrosion protective. As an outcome of the coating development work in this project, 30 coated ICs with two different types of EPD coatings (MCO15 and MCO20) were produced. An experimental stack (Stack K- 751), which incorporates these coated ICs plus Co coated ones, were produced and tested, showing promising results of EPD coatings. The results are presented in Section of this report. Mass gain,mg/cm Uncoated MCO15 MCO20 ELD_Y_1m ELD_Y_3m Co_IC Corrosion rate, g 2 cm -4 s -1 Oxidation time at 800 C, hour Uncoated MCO15 MCO20 ELD_Y_1m ELD_Y_3m Co_IC Figure 24: Mass gain (left) and evaluated corrosion rate constants (right) of full scale ICs with different coatings oxidized at 800 C in air. Uncoated and Co-IC represent ICs without coating or with Co coating, respectively. The IC samples with EPD MCO coating (MCO15 and MCO20) were coated on one side only. The corrosion rate constant for Cocoated IC was corrected for oxidation of Co into Co 3O 4. 31

32 WT2.2 Seals with reduced impurity emission The original purpose of WT2.2 was to develop alternative seals with either reduced SiO 2 content or with Ag-based braze. During the project course, we have found out that unlike single cell testing, Si poisoning of the Ni/YSZ electrode is not one of the dominant degradation issues at the stack level under the electrolysis conditions explored so far. The difference between cell and stack test results was ascribed to different amounts of glass seal employed in testing and different cathodic polarization which the cells were exposed to. In combination with post-mortem results on long-term tested SOEC stacks, we concluded that the used sealing glass, which was considered to be one of the major sources for Si poisoning in single cell testing, can be used as sealant for SOEC stacks. The originally planned tasks in WT2.2 therefore became less relevant to actual needs of SOEC technology development at its current stage. The resources set aside for WT2.2 (3 man-months) was therefore reallocated to the more development task (high pressure SOEC stack testing) in this project. In WP2, four milestones were set up in the project plan: M2.1 Oxidation tests of current generation coated ICs or coated alloys under at least two of the typical extreme operating conditions (defined in WT4.1) carried out for periods exceeding 1000 hours. M2.2 ASR measurements of current generation coated ICs or coated alloys under no current and under current density above 1 A/cm 2 for periods exceeding 1000 hours completed and influence of current on IC corrosion clarified. M2.3 Coated ICs or coated alloys tested in SOEC stack-relevant environments for >1000h with a degradation rate of 10% better than current generation coated ICs demonstrated. M2.4 A 50% reduction in the Si content (as compared to the one with standard sealant) found in the active Ni/YSZ electrode of a cell with alternative sealant tested for >1000 hours demonstrated. With the results presented above, we conclude that M2.1, M2.2, and M2.3 are fulfilled, while M2.4 not. 32

33 1.3.4 WP3 Stack development and testing The tasks in WP3 included investigating the current generation SOEC stacks under grid balancing related test conditions, exploring life-time limiting factors and mapping out a safe operation window, and testing improved cells and stack components at the stack level. These tasks were undertaken by both HTAS and DTU Energy. Table 5 gives an overview of stack testing in this project. In total we have carried out 13 stack tests under various conditions with a total testing period exceeding hours. The results obtained in WP3 are summarized below, grouped according to the work tasks. Table 5: Overview of the stack testing. Stack no. (number of cells) Stack type Milestone Main purpose Testing period, hour Test location K-695 (8) Delta M3.1 Grid balancing 1500 DTU K-696 (8) Delta M3.1 Grid balancing 1600 DTU K-730 (8) Delta M3.2 Extreme condition 600 DTU K-744 (8) Delta M3.2 Extreme condition 800 DTU K-753 (8) Delta M3.2 Extreme condition 600 DTU Q-691 (75) TSP-1 M3.4/M3.6 Large stack and improved stack design 500 HTAS K-740 (11) Delta M3.5 High pressure 200 DTU K-751 (15) Delta M3.6 Improved ICs 1000 DTU X-030 (75) TSP-1 M3.6 Improved stack design 500 HTAS U-001 (75) TSP-1 M3.6 Improved stack design 50 HTAS Q-675 (8) TSP-1 M3.6 Improved stack design 500 DTU X-076 (50) TSP-1 M3.7 Long-term stack test 1800 HTAS X-078 (50) TSP-1 M3.7 Long-term stack test 1800 HTAS WT3.1 Safe operation window for current generation SOEC stacks The first part of WT3.1 was devoted to evaluate reliability of the current generation SOEC stacks (HTAS Delta-type stacks) through long-term tests under conditions relevant to Danish electricity balancing schemes (defined in WT4.1, see Section of this report). In this project, we have operated two 8-cell HTAS Delta-type stacks (K-695 and K-696) according to the 100 % wind power profile defined in WT4.1. The operating profile entailed continuously (every 5 minutes) varying the power and sometimes switching between fuel-cell and electrolysis modes ( reversible operation ). The profile was generated directly from time-series data of wind power supply and electricity demand from the Danish island Ærø. The goal was a demonstration of stack operation in highly variable conditions to provide complete load-balancing and enable 100 % renewable energy for the region. The first test ran for approximately 1500 h in total, of which about 820 h was for the reversible 100 % wind power load-balancing profile. A number of issues were encountered which interrupted the profile, including issues with the testing rig and an error in the programming. The longest continuous segment of operation in the reversible profile was 170 h, during which the stack performance showed negligible degradation. The stack performance was not harmed by these interruptions, demonstrating stable performance for several hundred hours in such a continuously fluctuating operating profile, until a programming error led to a fuelside gas supply which caused complete oxidation of the Ni electrode and damaged the stack. The second test ran for approximately 1600 h in total, of which about 1200 h was for the reversible wind load-balancing profile. The issues encountered in the previous test were successfully eliminated for this test, which enabled operating the wind profile continuously for nearly the entire 1200 h. The stack data is shown in Figure 25 together with the input electricity supply and demand data and the energy-balancing simulation data which was directly 33

34 used to control the stack. An overview of the entire stack test is shown in Figure 26. A sudden increase in the voltage of Cell 1, indicated by a red circle in the figure, occurred at about 800 h into the test (400 h into the wind operating profile). This event also leads to a change in the stack temperature and temperature distribution. This event is believed to be a change in end-plate contacting which we have observed to randomly affect these stacks. Before this event, negligible stack degradation is observed, and then the event seems to initiate stack degradation with a linear trend (Figure 27). Figure 25: Illustration of the reversible SOC 100 % wind power energy-balancing profile and stack test data. In the top plot, the time-series wind power supply and electricity demand data for Ærø island (2010 data) is shown. The shaded regions indicate where the energy would be used directly, stored, and retrieved from storage. The middle plot shows the result of the energy balancing simulation. The bottom plot shows the data from the stack test which was operated according to the energy-balancing profile defined by the simulation. 34

35 Figure 26: Overview of the entire stack test operated with the reversible SOC 100 % wind power energy-balancing profile from 400 h onwards. The red circle indicates the end-plate contacting change event which initiated degradation of the stack. Figure 27: Stability of the stack during the wind profile segment, showing the average cell voltage only when the electrolysis current was more than A/cm 2. Before 400 h, there is negligible degradation. At just before 400 h, the voltage suddenly increases and then follows a linear trend. Near the end of the test, the voltage begins to decrease. However, this is not due to a re-activation; open-circuit voltage (OCV) measurements show that the OCV has decreased, indicating gas leakage and significant damage to the stack at that stage of the test. The dashed lines are linear fits to the corresponding segments of the data. 35

36 The two stack tests demonstrate that a reversible SOC stack can be operated in a stable manner with a real-world relevant operating profile which demands many changes in the operating point and reversals between electrolysis mode and fuel cell mode. After some time, the end-plate contacting event led to significant degradation of the stack, but that issue is specific to the stack design and affects all different kinds of stack tests. Due to that event, we cannot determine if the operating profile leads to accelerated degradation or reduces degradation. In previous studies on individual cells, we have even seen that reversible operation can lead to much lower degradation rates than constant electrolysis operation. In any case, the test successfully demonstrated application of the SOC stack for a real-worldrelevant energy-balancing profile. By the end of the tests, both stacks experienced certain degree of damage caused at least partly due to testing issues, it was therefore decided not to carry out further post-mortem analysis on these two stacks. To promote application of SOC stacks for grid balancing, maintaining stable contact among various stack components (including end-plate contacting) is one of the main issues which should be addressed in the near future. In the second part of WT3.1, three stack tests (K-730, K-744, and K-753) were carried out with a purpose of exploring the stack performance under three of the selected extreme conditions: high current density, high degree of steam conversion (into hydrogen), and high steam content. All the three stacks are HTAS Delta-type stacks. Stack K-730 was tested for investigating stack performance under high current density. This type of extreme operating profile was defined as alternating between a safe low current density (-0.5 A/cm 2 ) and a dangerous higher current density, with short cycle length at first and then increasing cycle length to find the stability limit. The actual testing profile and stack performance are presented in Figure 28. The test was carried out at 800 o C. The stack was first exposed to -0.5 A/cm 2, with 50 % H 2 O + 50 % H 2 to the fuel electrode and oxygen to the oxygen electrode, for 150 hours. After this initial stabilization period, the fuel gas composition was changed to 90 % H 2 O + 10 % H 2 with constant 60 % conversion. Periodically changing from -0.5 A/cm 2 to higher current density up to A/cm 2 did not introduce any accelerated degradation. At about 670 h, the steam generator failed which damaged the stack. The test was therefore terminated A/cm 2 stabilize i max = -0.7 A/cm 2 i max = A/cm 2 Went to zero! H 2 burner (steam generator) failed 50% H 2 O / H 2 90% H 2 O, maintaining constant 60% conversion Figure 28: Testing profile and performance of Stack K

37 Stack K-744 was tested for investigating stack performance under high degree of steam conversion (into hydrogen). The stack was exposed to 800 o C and -0.5 A/cm 2 for steam electrolysis, with 90 % H 2 O + 10 % H 2 supplying to the fuel electrode and oxygen to the oxygen electrode. The test was performed in consecutive 1-week segments in which the degree of conversion increased from 60 %, through 75 %, 90 % and ended at 95 %, while the steam concentration was held constant at 90 %. The stack performance was assessed by iv characterization between the segments. The results from running at high conversion degrees are somewhat influenced by the fact that performance of one of the eight cells (Cell 4) seemed to depart from the average stack performance already during the second segment (75 % conversion). The true cause has not been identified but the behaviour indicates that it might be an initial loss of evenly distributed contact that later developed into a hot spot, which eventually caused the cell to crack and develop a cross-over leak. The experiment was terminated when both the voltage of Cell 4 and the general stack temperature(s) started to increase rapidly during the 95 % conversion segment. Results for Stack K-744 have been summarized in Table 6, from where it is read that the almost doubling of the degradation rates seen initially until 90 % conversion tend to change at 95 % conversion although the last number may be significantly influenced by the increased stack temperature. The changes in ASR obtained at 50 % steam show a different trend, which resembles more the general temperature increase of the stack. The relatively constant rate of ASR-increase (degradation) has been calculated until the end of 90 % conversion segment and results in 0.16 mω cm 2 per hour. Table 6: Performance degradation of Stack K-744 excluding contribution from Cell 4. ASR-values refer to measurements after each segment. Sequence Voltage change, mv/1000 h ASR per cell, Ω cm 2 Initialization (60 % conversion) C 60 % conv C 75 % conv C 90 % conv C 95 % conv C Average stack temperature Stack K-753 was tested for investigating stack performance under high steam content. The stack was exposed to 800 o C and -0.5 A/cm 2 for steam electrolysis, with H 2 O+H 2 supplying to the fuel electrode and oxygen to the oxygen electrode. The experiment was performed in consecutive 1-week segments in which the steam concentration increased from 90 % through 95 % and ended at 97 %, while the degree of conversion was held constant at 60 %. The stack performance was assessed by iv characterisation between these segments. Table 7: Performance degradation of Stack K-753. ASR-values refer to measurements after each segment. Sequence Voltage change, mv/1000h ASR per cell, Ω cm 2 Initialization (90 % steam) C 90 % steam C 95 % steam C 97 % steam C Average stack temperature The results from running at high steam concentrations were not similarly influenced by malfunctions in the stack and have been summarized in Table 7. The voltage change through each sequence may again be seen as a markedly increased degradation during the first segments, which then levels off to attain more constant degradation through the last two sequences. In this case, temperature also increased during the entire run, but far from as much as in the previous stack. The average ASR of cells started out at a lower level than measured for Stack K-744, but the relative increase of the ASR was again seen as relatively 37

38 constant, which has been calculated until the end of 97 % steam segment and results in 0.21 mω cm 2 per hour. Based on the above three stack tests, it can be tentatively concluded that for HTAS Deltatype stacks, the stacks may be exposed to extreme electrolysis operation conditions if the current density is above A/cm 2, and/or the conversion above %, and/or steam content above %. Accelerated degradation may be seen under one of these extreme conditions. Due to limited project period and resources, the three tests mentioned above were carried out for <1000 h. Repeating the above tests but with extended periods is recommended in order to better define the extreme conditions. Among the above five tested stacks, K-744 and K-753 were selected for detailed PMA study. In addition, two other stacks tested by HTAS in other projects were included for PMA as well: 1) Stack K-649 tested in a bilateral collaboration with EIFER in Germany for steam electrolysis over 9000 h; 2) Stack K-706 tested for CO 2 electrolysis over 7000 h. Below some of the common features on microstructure changes of the cells and stack components after longterm electrolysis testing are summarized. One of the most significant changes found in cell microstructure after long-term electrolysis stack testing is loss of Ni in the active Ni/YSZ electrode. An example is given in Figure 29, where the cells are from Stack K-706 operated for CO 2 electrolysis over 7000 h. The image on the left side is taken from an active area, i.e. an area exposed to CO 2, high temperature and electrolysis current throughout the test. The one on the right side is taken in the part of the cell that is covered with glass, but accessible to reducing gasses during the reduction step and during operation. However, very little current is flowing through this part of the cell, as the electrodes are covered in glass. Despite being exposed to high temperature for 7000 h, the electrode structure resembles that of a fresh cell, with uniform porosity and apparently good percolation between Ni particles. For the active area (Figure 29a), image analyses tell that the porosity of the Ni-deficient region was close to 40%, which agrees well with the fact that this part of the electrode has lost approximately half of its Ni. The region directly underneath (i.e. the neighbouring support) is much more dense than it is supposed to be (15.8 % porosity vs % in the original electrode), indicating that Ni might transport from the active electrode into the neighbouring support. It should be mentioned that such phenomena (Ni re-distribution and loss of Ni) have also been observed in cells exposed to long term testing for steam electrolysis, both from single cell testing (see Section of this report) and from stack testing (for example, Stack K-649, tested for steam electrolysis for 9000 h). Figure 29: (a) (Left) Microstructure of the Ni/YSZ electrode + support exposed to CO 2. The image is taken in the part of the cell that is accessible to CO 2 during the entire test. The electrolyte is seen in the top part of the images. (b) (Right) Microstructure of the reduced Ni/YSZ electrode + support. The image is taken in the part of the cell that is covered with glass, but accessible to reducing gasses during the reduction step and during operation. The numbers inside the images are porosity of the analyzed regions. 38

39 A value of % porosity in the Ni/YSZ support (Figure 29a) implies that many of the pores within this region are closed, i.e. not percolating. Additionally, the Ni-deficient region will have low electrochemical activity, which means that the damaged Ni/YSZ layer acts as additional ohmic resistor (increasing the effective thickness of the electrolyte), and that the actual electrochemical activity is taking place in the dense region in the support. Based on the thickness and the porosity of the damaged region and the fact that it does not contain enough Ni required for percolation, the ohmic resistance of the cell in damaged areas can be estimated to be at least a factor of 2 or 3 higher than in other areas. The exact value depends on the Ni content and on the tortuosity of the remaining YSZ matrix. Interestingly, the electrode damage is in many cases localized to contact points. These observations indicate that the oxygen side contacting is insufficient in re-distributing current across the cell, and that more current is flowing through the cell in close proximity to the contact points. There may be a direct link between electrode delamination (between electrolyte and fuel electrode) and electrode damage. However, the definite cause-and-effect relationship has not yet been established. The second critical issue identified from PMA analyses is the (IC-cell) contacting. Even though the fuel side contacting remains satisfactory, delamination between oxygen electrode and IC (or other stack components) was commonly seen in several long-term tested SOEC stacks, which will likely contribute significantly to the degradation of the stacks. An improved contacting for the oxygen side with better interface adhesion can significantly improve stack robustness. Based on the above stack tests and PMA results, it can be concluded that HTAS Delta-type stacks can be operated rather safely at o C with electrolysis current density up to A/cm 2 and fuel conversion up to 60%, with an expected life-time of exceeding 9000 h (as demonstrated in testing of K-649 and K-706). Re-distribution of Ni in the active Ni/YSZ electrode and contacting on the oxygen side are two of the most critical issues which shall be addressed in order to further improve SOEC stack lifetime and robustness. WT3.1 involves also an activity on high pressure SOEC stack testing. This is a continuation of previous launched activities on high pressure SOC testing both at the cell level and at the stack level. The activities, which include both constructing high pressure test setup and actual testing, were co-funded by several projects including ForskEL Within this project, construction of a test setup (autoclave) for testing small 1 kw SOC stacks at elevated pressure was completed. A photo of this high pressure SOC stack test system is shown in Figure 30. On the left hand side of Figure 30 is the pressure vessel. Feedthroughs are used for all the gas tubing, power lines and data acquisition cables connected into the pressure vessel. The feed-throughs are fastened on the flanges that are mounted on the pressure vessel. The gas and pressure controlling systems are mounted inside the cabinet on the right side of Figure 30. The safety of the system is monitored and controlled by a programmable logic controller (PLC) located below the gas/pressure handling cabinet. The pressure vessel accommodates the furnace, stack housing and manifold, heat exchangers, H 2 O evaporator, condensation flasks, probes etc. A high pressure gas yard (not shown in Figure 30) is placed outside the building to provide gasses to the pressure test setup. 39

40 DTU Energy Figure 30: High temperature high pressure SOEC stack test system developed at DTU Energy. In addition, a short-term durability test on a 11-cell HTAS Delta-type stack (K-740) has been conducted at 10 bar. The stack was reduced in situ and afterwards experienced to initial performance characterization at 750 o C under various pressures up to 20 bar. The durability test was carried out at 10 bar, with 400 l/h H 2 +H 2 O mixture (H 2 /H 2 O = 50/50) supplied to the fuel electrode and 400 l/h air to the oxygen electrode. A plot of the stack voltage over the entire durability test period is presented in Figure 31. The test started at a current density of A/cm 2 with an initial degradation of 2.6 V/1000 h, corresponding to a cell voltage degradation of 236 mv/1000 h. Such high degradation is often seen in the initial period of electrolysis testing. An unplanned pressure and temperature cycle happened at around 60 h, which seemed to cause no harm to the stack performance. The test continued at A/cm 2 with slightly lower degradation rate, until an incident of steam starvation happened at around 90 h which caused certain damage to the stack performance. The test was then restarted at A/cm 2 and continued for another 100 h. The test was terminated at 190 h due to crack of Cell 1. Stack Voltage (Volt) A/cm A/cm 2 A/cm V/kh 1.7 V/kh Pressure and Temperature Cycle 1.2 V/kh Steam starvation Service on H 2 O pressure fluctuation damper 0.18 A/cm V/kh Stack Electrolysis Time (Hours) Figure 31: Stack voltage as a function of time for the durability test. 40

41 In total, the stack was tested for durability at 10 bar for close to 200 h. The incidents happened in-between are caused mainly by technical issues. Due to the fact that this high pressure stack test setup is the first one of its type constructed at DTU Energy, these unexpected technical issues can only be identified and solved during actual testing. The current test proves feasibility of high pressure SOEC stack testing and confirmed improved cell performance at elevated pressure (from initial performance characterization, not shown in this report) WT3.2 SOEC stack testing of improved cell and stack components The purpose of WT3.2 was to evaluate performance and durability of SOEC cells, ICs, other stack components or SOEC stacks developed in this project. Within the project period, seven stacks have been tested, among which one stack contains improved ICs, one stack with modified contacting and new stack design, and all the other five stacks with new stack design. The specification of these stacks is listed in Table 5. Stack K-751 is a 15-cell stack, in which three types of coated ICs were assembled in one stack and tested. The ICs are all based on Crofer 22 APU, but with different coatings: Co coating on both sides, Mn 1.5 Co 1.5 O 4 (MCO15) or MnCo 2 O 4 (MCO20) spinel coating on the oxygen side and no coating on the fuel side. The cells are from one batch of 2.5G-Standard cells. The stack assembly is presented in Table 8. It is seen that the cells are subjected to six different environments, which is then used as the basis for evaluation of the results. Not all the environments are represented by an equal number of cells. Table 8: Design of Stack K-751. Interconnect no. Oxygen side coating 1 Co Single repeating unit no. 2 Co 1 Co - Co 3 Mn 1.5Co 1.5O 4 2 Co - no 4 Mn 1.5Co 1.5O 4 3 Mn 1.5Co 1.5O 4 - no 5 MnCo 2O 4 4 Mn 1.5Co 1.5O 4 - no 6 MnCo 2O 4 5 MnCo 2O 4 - no 7 MnCo 2O 4 6 MnCo 2O 4 - no 8 Co 7 MnCo 2O 4 - Co 9 Co 8 Co - Co 10 MnCo 2O 4 9 Co - no 11 MnCo 2O 4 10 MnCo 2O 4 - no 12 Mn 1.5Co 1.5O 4 11 MnCo 2O 4 - no 13 Mn 1.5Co 1.5O 4 12 Mn 1.5Co 1.5O 4 - no 14 Mn 1.5Co 1.5O 4 13 Mn 1.5Co 1.5O 4 - no 15 Co 14 Mn 1.5Co 1.5O 4 - Co 16 Co 15 Co - Co Cell environment (coatings: oxygen side coating fuel side coating) After initial performance characterization, the stack was operated at 800 o C for coelectrolysis of steam and CO 2 for more than 1000 h. The durability test started at a current density of -0.5 A/cm 2 with 60 % conversion. A gradual temperature increase in the stack was detected along with the test. The current density was lowered to A/cm 2 after approximately 840 h and further to A/cm 2 after 980 h. Gas flow conditions were kept constant during the entire durability test period. The test was finally terminated at 1008 h. The evaluated cell area specific resistance (ASR) during the durability test period is presented in Figure 32, grouped according to the different IC coating environments which the cells were exposed to. 41

42 Figure 32: Evolution of cell ASR values over the durability test period for cells exposed to different IC coating environments. As shown above, the cells protected by Co coating on the oxygen side and no coating on the fuel side shows generally the best and the most reproducible performance. Adding a Co coating also on the fuel side increases the scatter and the absolute value of ASR. The cells exposed to spinel coatings on the oxygen side exhibit inferior reproducibility on ASR values. The cells exposed to MCO20 are only marginally better than those exposed to MCO15 in this respect. The cell with the best performance and durability was the one exposed to MCO15 coating on the oxygen side and no coating on the fuel side, which holds promises for developing much better long-term properties of spinel coating than the metallic cobalt coating in terms of resistance. In the start of the project all SOEC work on stacks was and had been performed on HTAS/TOFC s old stack design called the Delta design. During the project a transition to the TSP-1 type stack design was implemented. A summary of the major differences between the Delta and TSP-1 stack designs is presented in Table 9. 42

43 Table 9: Summary of the major differences between the Delta and TSP-1 stack designs. Design Shape Main feature - 50 cells. Delta - Internal gas manifolds. - Approximately 80 cm 2 active cell area. - Thick end plates. - Need for compression rods during handling. TSP cells in a stack core in a casing. - External air manifolds. - Approximately 110 cm 2 active cell area. - Improved gas flow distribution. - Improved manufacturability. - Compression free handling of stack core. Four stacks of TSP-1 design were then tested in the project to demonstrate SOEC operation on the new TSP-1 design and/or full-sized stacks. Three stacks (Q-691, X-030, and U-001) are of 75-cell stack, while the other stack Q-675 is 8-cell stack. The cells are of 2.5G- Standard with a foot print of 12X12 cm 2. After a successful conditioning and reduction, Stack Q-691 was first tested in SOFC mode for approximately 500 h (700 ºC, 25 A with natural gas as fuel) before attempting to run in SOEC mode. This was done primarily to make sure that the stack behaved as intended under known operating conditions. SOEC operation was then commenced using 750 ºC as furnace and inlet temperatures, CO 2 /H 2 (95/5) as fuel and air on the oxygen side. The stack was operated with 40 % conversion at -70 A and the stack had one thermal cycle during the electrolysis test. The results for Q-691 under electrolysis current are shown in Figure 33 below. Figure 33: Average cell voltages and stack voltage for Stack Q-691 tested for CO 2 electrolysis for approximately 500 h. 43

44 Considering that this was one of the first times a full sized TSP-1 stack has been operated in SOEC mode, the results were quite promising. The degradation between h was approximately 7 V/1000 h on stack voltage (or 97 mv/1000 h on average cell voltage). However, it is known from previous stack testing experiences that it usually takes approximately 1000 h before the initial degradation starts to flatten out. So degradation was not the immediate focus of this test. In order to verify the promising test with Q-691, it was decided to test another TSP-1 stack (X-030) at slightly different operating conditions and start up procedure. X-030 was a custom made stack equipped with single cell voltage probes on a number of cells in order to better monitor the individual voltage response from various cells in the stack during operation. Normally a stack is divided into cell groups containing 3-6 cells in each. After conditioning and reduction it was realized that there was a cell crack in the cell group containing cells C This was indicated from the voltage signal during the initial quality testing of the stack. Nevertheless, it was still deemed possible to test the stack in electrolysis mode for a short period of time to evaluate a different SOEC operating point. SOEC operation was started without prior testing in SOFC mode. Inlet gasses and furnace temperature were kept at 750 ºC. CO 2 /H 2 (95/5) was used as fuel and air on the oxygen side. The stack was operated with 30 % conversion at -50 A. The results for X-030 under electrolysis current are shown in Figure 34 below. Figure 34: Average cell voltages and stack voltage for Stack X-030 tested for CO 2 electrolysis for approximately 500h. The cell group (Cells 52-57) containing the cracked cell has been removed from the figure. The initial performance of the stack was very promising with a relatively low cell voltage compared with Q-691. However, it was quite clear that Stack X-030 had a much higher degradation rate than Q-691 during the 500 h electrolysis test. The degradation rate on stack level was 33.8 V/1000 h or 450 mv/1000 h on average cell voltage. This is an abnormal degradation and as illustrated by the single cell voltage probes in Figure 35 there was an additional accelerated degradation occurring at around 400 h testing. 44

45 Figure 35: Single cell voltage probe measurements on stack X-030. Note the accelerated degradation occurring on especially 5 cells (in red) after approximately 400 h operation. The reason for the abnormal degradation is still under investigation but the hypothesis so far is that the cell group with a cracked cell affected the other cells and/or impurities in the supplied gasses gave rise to high degradation. The positive thing with this stack test is that we were able to capture and shut down the stack with a clear difference in performance on single cells. This was later used to perform PMA analysis and investigate differences on individual cells comparing the worst cells with some better performing ones (red vs. blue in Figure 35). This analysis has so far resulted in a conclusion that a more robust and electronically conducting contact layer on the air side is needed. This is not necessarily related to the abnormal degradation but rather an observation done during PMA. Poor lateral conductivity on the air side contact layer leads to current restrictions, which particularly affects the fuel electrode leading to Ni-loss in the active electrode region. To improve oxygen side contacting, component development and testing has been performed in order to find a better candidate for a contact layer on the air side. This work has involved materials development, ink formulation, and other relevant fabrication parameters for producing a new and improved contact layer. Eight different contact layer formulations have been tested at component level. The best candidate (CL#4, having the highest conductivity) was further evaluated at the stack level. U-001 was a 75-cell TSP-1 type stack having 3 cell groups (6 cells in each group) with cells having different contact layer solutions. The conditioning and reduction of the stack went well and the stack was approved after various in-house quality controls. SOEC testing was then started and the electrolysis current was ramped up in steps to -85 A for a quick robustness control of the stack (see Figure 36). Furnace temperature and inlet gasses were held at 750 º C with CO 2 /H 2 (95/5) as fuel and air on the oxygen side. Throughout the initial test the cell groups with the new contact layer solutions had initially slightly better performance (lower average cell voltage) than the reference cell groups. Only the cell group with Cells 1-6 had better performance. However, that is expected from the temperature profile of the stack where Cells 1-6 are exposed to slightly higher temperatures giving lower ASR and cell voltage. Unfortunately, there was a power failure on the test rig and the stack experienced an unexpected shut down. When trying to restart the test the initial testing sequence was repeated (see Figure 37). During the short test point in SOFC mode there were indications of contact loss in one of the cell groups containing the new CL#4 contact layer. Later in the test when going to -50 A (SOEC mode) all 3 cell groups containing CL#4 experienced irregular and increased cell voltages. This led to eventual cell cracks when the test continued. It is clear that U-001 gave mixed results. On the positive side is the initially improved performance of the cell groups having the new contact layer. However, the clear trend that cell groups with 45

46 new contact layer made the stack fail after a restart is something that needs further attention. PMA analyses made so far indicate that there has been poor contact between the contact layer and the interconnect. This will be further investigated in the coming project (ForskEL ) to give feedback to the continued development and testing of new and improved contact layers. Nevertheless, the results strengthen the hypothesis that improving only the electronic conductivity of the contact layer is not sufficient. Improving robustness and interface quality is equally important. Figure 36: Initial testing sequence of U-001. Figure 37: Testing sequence after Stack U-001 was restarted again after power failure. To verify performance and durability of TSP-1 stacks for electrolysis operation, one more short-term stack test was carried out. Stack Q-675 is a 8-cell stack designed specifically for impedance measurement. After reduction and initial performance characterization, Q-675 was tested for CO 2 electrolysis at the same condition as for X-030. Figure 38 plots evolution of cell and stack voltages with time. After the initial stabilization period of h, the stack showed stable performance with negligible degradation until 430 h when the test had to be terminated due to test setup reconfiguration. 46

47 10.8 Voltage, V Stack voltage Cell voltage Time, hour Figure 38: Evolution of cell and stack voltages for Stack Q-675. The last stack test of this project was devoted to fulfilling Milestone M3.7. The requested test conditions in M3.7 are not technically feasible on the current TSP-1 stack design nor with the current test rigs at HTAS. We have therefore chosen a more moderate and realistic test strategy for conducting the long term test in the project. Two 50-cell TSP-1 stacks (X-076 and X-078) are currently being tested in a dedicated test rig designed for parallel testing of 2 stacks at the same time (see Figure 39). Figure 39: Schematic showing how the test rig is designed for simultaneous parallel testing of 2 stacks. The stacks were tested with inlet gasses heated to approximately 800 ºC. The stacks are tested without a furnace (adiabatic testing), which means that only the heated inlet gasses supply the heat to the stacks. The stacks are operated with CO 2 /CO (99/1) on the fuel side and air on the oxygen side at -48 A and approximately 25 % conversion. The stacks are furthermore exposed to a controlled thermal cycle once per week for robustness testing. The results achieved at the time of writing this final project report is shown in Figure 40 with the temperature evolution in Figure 41. The tests with thermo-cycles will continue, hopefully > 3000 h, for evaluating the long term stability and robustness of the stacks. 47

48 Figure 40: Average cell voltage of two 50-cell stacks tested in parallel (X X-078). Figure 41: Temperature evolution of fuel and air inlet/outlet during the test. The thermo-cycles were made down to below 100 o C. The stacks show stable performance and after an initial degradation period the cell voltages are now beginning to be more stable and degradation rate is currently around 2 %/1000 h (see Figure 42). The degradation is resulting in higher ASR of the stacks, which gives rise to ohmic heating of the cells/stack. Higher ohmic heating leads to higher outlet temperatures (and inlet temperatures due to the heat exchangers), as shown in Figure 41. During the continued test the inlet temperature was adjusted after the 11th thermo-cycle. Figure 42: Degradation rates as a function of thermal cycle for the X-076 and X-078 stacks. 48

49 In WP3, seven milestones were set up in the project plan: M3.1 Two stack tests of current generation SOEC stacks under conditions relevant for grid balancing (defined in WT4.1, one for short-term balancing and the other for long-term) for a period exceeding 1000 hours with detailed post-mortem characterisations on cells and stack components completed. M3.2 Three stack tests of current generation SOEC stacks under extreme operating conditions (defined in WT4.1) with detailed post-mortem characterisations on cells and stack components carried out. M3.3 Safe operation window for the current generation SOEC stacks mapped out. M3.4 An SOEC stack containing more than 40 cells with a foot print of 12X12 cm 2 under electrolysis operation for 500 hours demonstrated. M3.5 One stack test at 10 bar for a period exceeding 500 hours carried out. M3.6 Three stack tests of SOEC stacks composed of improved cells or ICs or impregnated with nanoparticles carried out for a period exceeding 500 hours. M3.7 One SOEC stack test under conditions relevant for grid balancing (defined in WT4.1) for a period exceeding 3000 hours with an ASR of less than 0.45 Ω cm 2 and an average degradation rate of less than 1.25%/1000 h at 1 A/cm 2 demonstrated. For M , five stack tests were carried out in the project. Among which, two stacks were tested according to the 100 % wind power profile defined in WT4.1 with a total testing period exceeding 3000 h, while the other three stacks were tested under extreme operating conditions. PMA analyses were carried out on some of the here-tested stacks plus two stacks tested in other projects. Loss of Ni in the active Ni/YSZ electrode and contacting on the oxygen side were identified as two of the most critical issues. A safe operation window for HTAS Delta-type stacks was proposed. The milestones M are fulfilled. For M3.5, within the project period, we have successfully completed construction of the high pressure stack test rig. In addition, an 8-cell Delta-type stack was tested under elevated pressure up to 20 bar. The durability test was carried out at 10 bar for close to 200 h. Several incidents, caused mainly by instrument technical issues, happened in-between, which eventually led to crack of cells and shutdown of the test. The current test proves feasibility of high pressure SOEC stack testing and improved cell performance at elevated pressure. The milestone M3.5 is considered partially fulfilled. For M3.4 and M3.6, five stack tests were carried out with a total testing period exceeding 2500 h. Among which, one Delta stack was tested for screening different IC coatings, and the other four stacks for testing either the new stack design (TSP-1) and/or new contact layers. The developed MnCo spinel coating showed promising results and is recommended for further development. Stable electrolysis operation was demonstrated on the new stack design TSP-1. With the above results, we conclude that M3.4 and M3.6 are fulfilled. For M3.7, two identical 50-cell TSP-1 stacks were tested for a period exceeding 1800 h. Due to adoption of the new stack design (TSP-1), the requested test conditions in M3.7 are not technically feasible at the current stage. A more moderate and realistic test strategy for conducting the long term test in M3.7 was chosen. M3.7 was therefore not fulfilled. However, the current test confirms stable electrolysis performance and good durability of TSP-1 stacks. Further development towards realizing the technical requirements specified in M3.7 will be pursued in the coming project. 49

50 1.3.5 WP4 SOEC system This work package involved two tasks: - from an SOEC system point of view, to identify SOEC cell and stack test conditions relevant for grid balancing schemes (WT4.1); - to develop key components (electrical heaters and heat exchangers) for closely heat and flow integrated SOEC systems (WT4.2) WT4.1 Grid balancing related test conditions The grid balancing schemes include both short-term and long-term grid balancing, and with a grid relevant to Denmark, which means the focus is on integrating intermittent wind or solar power. In the short-term (or the near-term), an SOEC would be operated only during bursts of low-cost intermittent electricity, for peak shaving. In the long-term, when wind or solar power provides 100 % of electricity supply as expected in 2050 based on the Danish energy plan, we have investigated whether solid oxide cells (SOCs) could provide full energy balancing by reversible (electrolysis and fuel cell modes) operation with a single device, as shown in Figure 43. Our analysis involves two scenarios. The first is based on wind power supply and the second is based on solar power supply. In both scenarios, we assume an energy storage scheme wherein the intermittent renewable energy provides 100 % of the needed energy for the region including on-demand electricity as well as hydrocarbon fuels for transportation produced by electrolysis. Figure 43: Illustration of full energy system balancing using a reversible fuel cell. The full energy system is here defined as one with an intermittent electricity supply, and on-demand electricity and fuel consumption. For the wind load-balancing analysis, we have analyzed time-series supply and demand data from 2010 for Danish Ærø island. As the energy consumption on the island is currently met by about 50 % wind power from its wind turbines, we scaled up the wind power supply for the future scenario. When the intermittent electricity supply exceeds electricity demand, it drives electrolysis mode in the SOC to produce synthesis gas (CO and H 2 mixture) which is converted to methanol. When the electricity demand exceeds the supply, the methanol is converted back to electricity by operating the SOC in fuel cell mode. Also, the produced methanol is taken as needed and used in the transportation sector, and some of it is converted to dimethyl ether (DME) by a catalytic dehydration process, and used in place of diesel in other vehicles, such as the large fleet of ferries operated by the island. Carbon dioxide 50

51 needed for the fuel synthesis is assumed to be captured from the atmosphere either by technical means or via biomass. The total fuel and heat consumption for the island is also used in the energy balance calculation. Therefore, the full energy consumption (electricity and fuels) is supplied by only the wind power. This can be considered a long-term scenario dominated by SOCs. This analysis of the time-series data is converted directly into an SOC stack operating profile, as shown in Figure 44. The maximum electrolysis current density is set, and the energy balance itself then sets the remaining operation parameters. The simulated operating profile has been used directly to test Stacks K-695 and K-696 (test detail described in Section of this report). Figure 44: Energy balancing of Ærø island, showing the conversion of the supply and demand data in 2010 (together with total fuel and heat consumption for the year) to current-voltage as operating profile for an SOC stack. As part of defining the grid balancing related test conditions, a statistical analysis about the simulated long-term energy balancing profile was carried out. In addition to the wind profile analysis, a solar profile using supply and demand data from southern California in US was further included. The solar supply data is from a 50 MW solar photovoltaic plant. As shown in Figure 45, the different types of intermittent renewable power supplies result in very different operating profiles for SOCs applied for load balancing. Visual comparison of the electricity supply and demand data (Figure 45a vs Figure 45d) clearly indicates the well-known stochastic vs periodic nature of wind and solar energy, respectively. Although the net demand for electricity and fuels was set to the same values for both scenarios and the energy supply scaled to meet the total demand in 1 year, the wind power supply peaks at around 45 MW whereas solar peaks at around 70 MW. This is due to the different capacity factors of wind and solar energy. Statistical analysis of the resulting SOC operating profiles for 1 year of energy balancing are shown in Figure 45b,c,e,f. The wind operating profile consists of a great number of very short segments in either fuel cell mode or electrolysis mode before reversing to the other mode, as well as just a few very long segments exceeding 100 h (Figure 45b). The short segments correspond to quickly fluctuating wind speeds and the long segments correspond to several days of high wind. The solar profile, on the other hand, consists of regular segment lengths of about 14 h in fuel cell mode and 10 h in electrolysis mode, corresponding to the daily solar cycle (Figure 45c). There are also many short segments which correspond to quick losses and regains of sunlight during the day, e.g. coverage by clouds. Another im- 51

52 portant metric is the current density ramp rate. Comparing Figure 45e and Figure 45f shows that the solar profile would operate the SOCs with much faster ramp rates. The two profiles also have some statistics in common: the number of reversals between fuel cell and electrolysis modes in the year is about 1200 for both profiles, and the ratio of the maximum current densities for electrolysis mode vs fuel cell mode is around 3. Although the present analysis was based on 100 % renewable energy supply and certain values for fuel demand (or ratio of fuel, electricity and heat demands), the methodology can be applied for other assumptions and profiles, for example in a scenario where the SOCs would not be operated reversibly but would instead be only used for fuel production in electrolysis mode. Another, especially relevant scenario would be energy balancing east/west Denmark. However, analysis of such a large region requires further development of the simulation to include economics (spot market pricing, buying/selling between countries, fuel prices etc.). Figure 45: Wind (a-c) and solar (d-f) energy supply and demand profiles (a,d) and statistical analysis of resulting SOC operating profiles (b,c,e,f). The last part of WT4.1 was the definition of extreme operating conditions to obtain operating profiles which push the limits of SOC stability. There are many possible extreme conditions that could be tested, including high current density, high current ramping rate, high steam/co 2 conversion, high steam/co 2 content in the gas supply, synchronicity of electrical and gas supply changes, and thermal cycling. Stack tests using these extreme operating profiles were carried out in WT3.1 of this project (Stacks K-730, K-744, K-753, see Section of this report). 52

53 High current density and corresponding cell voltage is known to trigger degradation mechanisms in cells and in stacks. These degradation mechanisms seem to have incubation times on the order of hours to tens or hundreds of hours. Therefore we have defined this type of extreme operating profile as alternating between a safe low current density (-0.5 A/cm 2 ) and a dangerous higher current density, with short cycle length at first and then increasing cycle length to find the stability limit, as illustrated in Figure 46. In the actual stack test (Stack K-730, test detail described in Section of this report) we performed steam electrolysis with such a profile at several different higher current densities, e.g A/cm 2, then -0.6, then -0.7 A/cm 2 etc. High current ramping rate could be needed for certain load balancing applications. However, other energy storage devices like batteries and capacitors would likely be used for very fast transients, because they do not require the synchronized gas flows of reactants as in SOC stacks. Whereas electrical supply can be adjusted rapidly, gas flows have a slower respond, and it is therefore important to ensure that the two are properly synchronized so that reactant starvation does not occur. An extreme operating profile which would test this, progressing from properly synchronized changes to reactant starvation situation, is illustrated in Figure 46. (a) (b) (c) Figure 46: Illustration of three types of extreme conditions operating profiles. High steam/co 2 content in the gas supply is another extreme condition that could occur if there is a failure in recycling of products to the inlet. This condition is also of interest because it is desirable to use the maximum steam/co 2 content that still yields stable stack operation. Similarly, high reactant conversion could be a result of a gas supply issue and is also a desirable operating point if stable. The high steam content profile is illustrated in Figure 46c. Like the high current density profile, we have defined it as stepping from safe to dangerous operating regime, and ultimately ending with 100 % steam supply, which if the cell is not polarized, would lead to gradual and destructive oxidation of the Ni in the Ni/YSZ electrode. However, if the cells are polarized before supplying 100 % steam then the polarization may sufficiently protect the Ni from oxidation. Based on the test profile defined above, two stacks (Stacks K-744 and K-753, test detail described in Section of this report) were tested under either high degree of steam conversion (into hydrogen) or high steam content. 53

54 WT4.2 Key components for closely heat and flow integrated SOEC systems The objective of WT4.2 was to develop high temperature SOEC system components (high temperature heat exchangers and electrical heaters) which are directly mechanically compatible with the SOEC stacks. Within this project, two heat exchangers with the same footprint as the stack (14 x 14 cm 2 ) have been assembled. The heat exchanger design is very similar to that of SOEC stacks. The heat exchangers were assembled and brazed in a vacuum furnace at Danfoss. A special manifold unit for gas inlet and outlet was brazed to one of the heat exchangers in a second step. A picture of this heat exchanger is shown in Figure 47. Figure 47: Picture of brazed heat exchanger with manifolds mounted for gas inlet and outlet. Figure 48: Efficiency of the planar heat exchanger as a function of flow rate and temperature. A dedicated test set-up has been established and a 10-channel plate heat exchanger has been tested at different flow rates (1-3 Nm 3 /h) and different inlet temperature ( C). Afterwards further optimization was carried out on adjusting T_(cold,out) so that the energy transferred from the hot side corresponded to the energy received by the cold side. This was done in order to compensate for the heat loss in the setup. C p * * (T hot,in T hot,out ) = C p * * (T cold,out T cold,in ) The efficiency was then increased from % (before optimization) to %, as shown in Figure

55 For electrical heaters, the task was to achieve a more compact and production friendly mechanical design and at the same time to lower heat losses from the heater surfaces. Figure 49 shows a mechanical outline of a previously developed SOEC sub-system (SOECCore) where a stack is placed on top of a manifold plate and a heater (orange cylinder to the left). By changing the heater geometry from a cylinder to a plate, the stack and manifold could be placed directly on top of the heater. Placing a stack both on top and below the heater would lead to a compact low loss configuration as sketched in Figure 50. Figure 49: Mechanical outline of a previously developed SOEC sub-system. Figure 50: Compact low loss configuration with a planar heater. Figure 51: Mechanical outline of the proposed heater design. Figure 52: CFD simulation of temperature distribution in the proposed heater design. A planar heater with a design shown in Figure 51 was developed. The heater element is a commercially available insulated heater wire, which is coiled inside a metal frame in which the stack manifold can be placed. One critical design criterion is that the maximum temperature at the surface of the heater element should not exceed 900 o C, while the stack should be heated to 750 o C. Figure 52 shows a computational fluid dynamics (CFD) simulation of temperature distribution in the heater. According to the simulation, the maximum temperature of the heater element is close to 900 o C. Further optimization has been carried out in order to investigate alternative configurations with a higher temperature margin. In addition, a prototype heater has been produced and is shown in Figure 53. Due to several technical issues with the test setup, it was not possible to test the heater within the timeframe of this project. However, further development and testing of the electrical heaters is considered highly relevant and will be combined with similar activities in the project ForskEL

56 Figure 53: Prototype of stack integrated heater. In the current project, three milestones, which are listed below, were set up in the project plan: M4.1 Grid balancing related test conditions (including extreme operation conditions) and test profile defined. M4.2 Performance of a 14x14x3 cm 3 heat exchanger tested. M4.3 Performance of a 14x14x3 cm 3 electrical heater tested. For Milestone M4.1, analyses on grid balancing schemes based on integrating either wind or solar power and where solid oxide cells provide full energy balancing (by reversible electrolysis and fuel cell operation) have been carried out. The analyses have been converted into SOC stack operating profiles and employed in actual stack tests. For Milestone 4.2, heat exchangers with the specified dimension were designed and successfully tested. Good agreement between the measured and COMSOL model predicted performance was achieved. The milestones M4.1 and M4.2 are fulfilled. For Milestone 4.3, planar electrical heaters with compact and production friendly mechanical design were produced. Due to technical issues with the test setup, it was unfortunately not possible to test the heaters and fulfill the milestone within the project period. Further development and testing of the electrical heaters is considered highly relevant and will be combined with similar activities in the project ForskEL

57 1.3.6 WP5 Energy system modeling This work package is to investigate different aspects of integrating SOEC into various energy systems. According to the project plan, the following were investigated in this project: Firstly, a comparison between distributing produced syngas through natural gas grid versus direct conversion of syngas to liquid fuel and distribution in the liquid form was carried out. Secondly, potential for using combined capacities of SOEC and SOFC for grid balancing purposes was investigated. Furthermore, analysis on using steam electrolysis for fuel production versus co-electrolysis was conducted. Finally, analysis of using SOEC in different energy systems was done in order to determine potential of using electrolysis and what utilization capacities are necessary in order to supply fuel demand. The energy system tool employed for the above study was further improved to be able to analyze different fuel pathways and a methodology was developed to determine feasible utilization capacities of electrolysis in different energy systems. WP5 was carried out mainly by Aalborg University and the results are summarized below Comparison between SOEC enabled syngas production and distribution via the natural gas grid vs. SOEC enabled syngas production combined with synthetic fuel production and liquid fuel transportation Syngas is a mixture of 2:1 of H 2 and CO. It can also contain carbon dioxide, methane, and smaller impurities such as chlorides, sulphur compounds, and heavier hydrocarbons. Transportation of syngas is possible but not in the current natural gas network in case of standard syngas properties. The problem with transporting syngas is due to both hydrogen and CO properties. Syngas has a high share of hydrogen and explosive potential. Hydrogen can cause leakages and burns with invisible flames indicating that there is a high possibility of injuries in case of an accident. In cases where hydrogen concentrations are lower than 15 %, there is a possibility to transport syngas in existing natural gas network with slight modifications and increasing the operating pressure. However, this relates exclusively to what natural gas pipes can handle, in case that the grid is connected to filling stations, gas turbines or any gas engines the percentage drops to 2%. This implies that there is a need for building a special pipeline network that can handle the syngas properties and there are existing guidelines for this purpose. As the syngas would not play the same role as natural gas in the existing system, establishing completely new network should potentially be avoided as it is a money and time consuming investment. The transportation of final fuel (in case of our analysis liquid methanol/dme or gaseous methane) is known and established. There are some limitations and adaptations in case of suggested liquid fuels. Due to the properties of methanol/dme there is a need to use either coating or new materials for storing these fuels at the filling stations. Methane can be transported in existing natural gas networks, which can be used as fuel storage. As the only fuel that is currently used in the transport sector methane seems to be a more expensive solution than suggested liquid fuels in terms of infrastructure modification for transport Combined SOEC/SOFC units for grid balancing In order to determine potential operation hours for combined SOEC/SOFC mode, it was necessary to conduct hour-by-hour analysis of the energy system. SOEC mode is used for production of liquid/gaseous fuels for the transport sector, while the potential of SOFC operation 57

58 was based on substitution of combined cycle gas turbine (CCGT) in combined heat and power (CHP) plants as they both convert fuels to electricity. It was to investigate when reversible operation of SOEC/SOFC in fuel cell mode can substitute CCGT in CHP plants and what the economic benefits for this operation can be. The maximum theoretical potential of SOFC is determined by hourly operation of SOEC and unutilized capacity in this mode. The analysis was conducted for a reference scenario of a 100 % renewable Danish energy system in 2050, in which the installed capacity of SOEC is 7,741 MW. Figure 54 Illustrates operation of SOEC during the third week of January. The unutilized SOEC capacity was determined by subtracting the hourly utilized capacity from the installed capacity. For example, in the first hour of the presented week, the unutilized capacity is 6,207 MW. To determine the maximum theoretical SOFC potential, the unutilized SOEC capacity needs to be multiplied with the SOFC efficiency. It is expected that in 2050 the SOFC efficiency can reach 60 % (LHV). The maximum SOFC operation capacity in the first hour is therefore 3,724 MW. On a yearly basis, the maximum theoretical SOFC production is 20.4 TWh. Figure 54: Theoretical maximum capacity for SOFC operation in combined SOEC/SOFC units. Next step is to determine how much CCGT capacity can be substituted by SOFC operation. It is visible from Figure 55 that in the first hour, SOFC can substitute 84 % of CCGT capacity. Due to the changes in demand and supply, SOFC cannot substitute the entire operation of CCGT. Figure 55: Substitution of CCGT by SOFC. 58

59 Even though the maximum theoretical production capacity of electricity by SOFC is 20.4 TWh per year, only 47% of this can be utilized in the system to substitute CCGT due to the hourly changes in the demand and supply (see Table 10). Interestingly, even with only 47 % of potential SOFC capacity, it is possible to substitute almost 84 % of the CCGT capacity. Table 10: Characteristics of SOFC and CCGT in reference scenario. Maximum theoretical SOFC production Utilised SOFC capacity to substitute CCGT CCGT production Remaining production of CCGT TWh/year 9.58 TWh/year % TWh/year 1.88 TWh/year A sensitivity analysis was conducted to investigate impact of different wind productions that occur in different years, while the installed wind capacity was kept the same. This type of analysis is important as the wind production can vary significantly from year to year. The nine-year period from 2005 to 2014 with historical yearly wind profiles was analyzed. The expected full load hours from the reference scenario was simulated by correcting the production output. Figure 56 illustrates changes in critical excess electricity production (CEEP), biomass consumption and total annual costs based on different wind productions during the period. As expected, biomass consumption in the system is dependent on the wind production: the higher the wind production, the lower the biomass usage and the same trend is in total annual costs. It can be concluded that a high wind production brings economic benefits in this type of energy system. Figure 56: Main system characteristics for the period with different wind production. Looking into how this is reflected on the potential of SOFC to substitute CCGT, it is visible from Figure 57 that the CHP electricity production is changing according to the wind production. The maximum CHP electricity production (15.31 TWh/year) is occurring in 2008 when the wind production was the lowest, while the minimum CHP electricity production (4.85 TWh/year) is in 2014, which was a windy year. With increase in wind production the combined SOEC/SOFC can substitute more CCGT capacity. Depending on the wind production, the combined SOEC/SOFC can substitute between 73.7 % (2008) and 96 % (2014) of CCGT capacity. On average combined SOEC/SOFC can substitute ~84 % of CCGT capacity, as the average wind production is TWh/year. 59

60 Figure 57: Sensitivity analysis with different wind years from and substituted CCGT operation with SOFC Steam electrolysis versus co-electrolysis Fuel production based on CO 2 hydrogenation and co-electrolysis were compared in order to assess which pathway is preferable. The CO 2 hydrogenation pathway is based on combining carbon dioxide from a stationary source with hydrogen from steam electrolysis to form syngas, which is further converted to methanol/dme or upgraded to methane. The efficiency of water electrolysis used in the analysis is 73 % (LHV). The co-electrolysis pathway has the same principle as the CO 2 hydrogenation pathway, but it is a combined process of water and CO 2 electrolysis. The output of co-electrolysis is syngas which is later converted to a desired fuel. The efficiency used for co-electrolysis is 77 % (LHV). The calculations are done in order to determine energy consumption per 100 PJ of fuel output. Two fuel outputs were analyzed: liquid output (methanol/dme) and gaseous output (methane). The demand is calculated based on stoichiometric equations of the reactions occurring in these processes. Additional losses are added for chemical synthesis. No bioenergy input is related to these pathways as they recycle the CO 2 emissions from the power and heat sector or industrial processes. The results indicate that there are no decisive differences between these two pathways for the same fuel outputs SOEC applications in different energy systems Three energy systems are compared in order to determine the influence of different types of energy system on utilization of SOECs. The analysed systems were a 100% renewable Danish 2050 system and two types of German energy systems: the forecasted 2050 German energy system based on smart energy system approach (without nuclear energy) and the 2050 German energy system with different levels of nuclear capacity installed. The nonnuclear German energy system is a system with a high share of intermittent renewable energy for electricity production and natural gas as a main energy source in power plants and combined heat & power. For the nuclear Germany energy system, three scenarios were analyzed in order to determine the influence of different nuclear energy shares (15%, 30% and 45%) in electricity supply on electrolysis operation for fuel production. In order to be comparative, the German scenarios have same transport demand and same levels of electricity 60