Final Report Summary - KEEPEMALIVE (Knowledge to Enhance the Endurance of PEM fuel cells by Accelerated LIfetime Verification Experiments)
Executive Summary:
In the KeePEMalive project a comprehensive test program on fuel cell degradation has been carried out, accumulating more than 25 000 accelerated test hours on single cells. Performance data were analysed and compared to results from a Danish field test project (at Lolland) in which 32 fuel cell systems for combined heat and power are being deployed in households to identify detrimental operating conditions. By improving cell materials and optimising the system operation conditions the lifetime of these fuel cell systems has been increased from ½ to 2 years.
There has been five (5) focus areas of activities in the project:
1) Material development and selection
Development of PFSA based membrane materials has been a key activity in the project. In total 5 generations of membrane materials have been synthesized and characterized with respect to key properties. The initial batch production was successfully scaled up to continuous membrane series production in the range of 2000 m2/year.
Partner CRNS showed that by introducing Ce-based radical scavengers the stability of the Membrane was further enhanced. Manufacturing of MEAs and stack assembly was done in-house by IRD.
The most stable and best performing fuel cells were obtained using reinforced membrane and a cathode catalyst with improved catalyst dispersion on an oxidant resistant support.
2) Fuel cell characterisation during operation
To quantify the degradation of the fuel cells at a reasonable speed, a set of Accelerated Stress Test (AST) protocols were developed and applied, exacerbating selected stressing conditions relevant for real life µCHP application, including:
• Continuous operation, • Reformate operation, • Dead end operation, • Start-stop, shut-down, • Fuel Starvation, • Electric load cycling
Distinct sets of experiments were run for each of these stressing conditions by systematically varying operating conditions such as fuel gas humidity, cell temperature and cell current.
After conditioning the fuel cell for 24 hours, the cell performance was mapped every 24 hour to identify changes as input for degradation rate quantification.
Fuel cells performance was assessed using: • Electrochemical Impedance Spectroscopy, • Hydrogen cross-over & Cyclic Voltammetry, • Carbon corrosion by NDIR, • Membrane degradation by Fluorine emission rates, • Segmented cell to reveal localised degradation
3) Identification of material changes
The following materials were subject to assessment: - Catalyst materials (SINTEF), - Catalyst layer surface & cross section (CNRS), - Membrane materials (CNRS, Fumatech)
and these MEA components' properties including e.g.,: - Mechanical strength, membrane (Yuong modulus), - Morphology, Microstructure, - Catalyst agglomeration and migration by (SEM), - Electrode ElectroChemical Surface Area (ECSA), - Membrane Conductivity, Ohmic resistance, - Water-uptake, hydration number
4) Quantification of degradation rates
Changes in performance with time were assessed and the degradation rate was calculated. The used fuel cell materials subject to stressing conditions were compared to pristine materials revealing corresponding material changes.
Statistical analyses has clearly shown that degradation depends on several factors and varies along the operation curve as illustrated below for load cycling over 144 hours AST.
5)Enhanced insight/understanding
Based on studies of more than 60 cells and assessment of corresponding material changes for some of these, the cause and effect relationships were identified.
The complexity of processes taking place in PEM fuel cells were verified by statistical analysis showing that factors causing degradation are inter-related.
The enhanced insight provided a sound basis for selection of improved materials for next generation MEAs to be tested.
Project Context and Objectives:
The main objective of KEEPEMALIVE's was to establish improved understanding of degradation and failure mechanisms, accelerated stress test protocols, sensitivity matrix and lifetime prediction models for LT PEMFC to enable a lifetime of 40 000 h at realistic operation conditions for stationary systems, in compliance with performance and costs targets.
As 40 000 hours' testing is not possible to achieve within the 3 ½ years' duration, the aim throughout the project was to find relevant procedures to highly accelerate the degradation rate observed in µCHP applications and limit the testing-time required for estimating the lifetime of PEMFCs for stationary applications. Hence, an accelerated stress test (AST) experimental program was set up. Because the degree of acceleration is strongly dependant on the operating conditions, the program was divided into two phases.
The topic "degradation of fuel cells" is highly complex in its nature. Therefore, a thorough assessment of available literature on the relationships between operation conditions and lifetime issues was carried out leading to definition of an ambitious initial AST program. To realise true collaboration between the involved European research laboratories and industry partners, certain ASTs were intentionally shared between laboratories.
Conditions causing degradation were identified and their relevance for stationary µCHP applications was verified. In total 6 stressors were selected to account for the most typical conditions encountered during real life operation of µCHP units. The stressors were:
1. Continuous operation: at nominal current density with long term exposure to liquid water and high voltage (relevant for winter period operation)
2. Reformate operation: fuel quality and reformer issues
3. Dead-end operation: Purging of air causing elevated pressure in the anode chamber
4. Fuel starvation: typically encountered during transients
5. Start/stop cycling: Hydrogen/air fronts in fuel channel (relevant for summer period operation)
6. Electric load cycling: (spring/fall period profile)
Operation variables such as Temperature, Relative Humidity (RH), cell voltage and current density were selected as controllable factors for the ASTs. From our best aggregated knowledge, adequate levels of these factors were selected to accelerate the degradation process roughly 100-fold.
During the initial phase of the AST program (Month 1-25), baseline experiments were carried out at the available laboratories (7 in total) in the consortium, to reveal the inter-laboratory variance. Unfortunately, it turned out that the variation in performance for identical cells tested at different laboratories were substantial, as documented during the 2nd progress meeting (M13) and in D 5.2. These discrepancies are assumed to be linked to differences in test equipment hardware configurations (e.g. flow fields) and varying active cell area used at the various laboratories (5 to 25 cm2).
Moreover, based on available protocols, experience and recommendations in literature, a break-in procedure was selected and validated to ensure that maximum performance was reached prior to executing the ASTs. The best available membrane materials were selected based on feedback from cell and stack tests. As new membrane materials were developed, however, we found that the originally verified break-in procedure (for first and second generation membrane materials) was not adequate for the latter generation of membranes (reinforced) (for details, see D 2.2). Thus, revision of the break-in protocol was needed.
Acknowledging the complexity of these relationships, a systematic approach using statistically designed experiments (factorial design) was used, as described in the DoW. To get maximum information out of the factorial designed experiments (for details, see D 5.1) complete sets of so-called 23 experiments (3 factors, each at 2 levels = 8 different operation conditions) were planned at each partner's test facilities. During the execution of those set of experiments, however, it turned out that many of the selected sets of operation parameters led to instable operation (e.g. flooding or dehydration) and several of the 6 distinct AST protocols were not executable (experimental set point values defined in D 1.2). The foreseen statistical analyses to reveal the factor's effect on degradation were, hence, not possible due to incompleteness of the full sets of eight experiments.
Difficulties encountered during the initial iterations of statistical analysis of the data from 37 single cell tests carried out in phase one of the AST program (Month 1-25), lead to the conclusion that only qualitative relationships between factors and degradation could be identified (for details, see D 5.2). As a consequence Stack tests were postponed until reliable single cell characterization could be demonstrated.
During the revision of the initial AST protocols (Months 26-29) four corrective actions were taken (details on D 1.3) in order to secure more reliable results in the phase 2 of the AST program (Months 30-42):
1. The number of AST protocols was reduced from 6 to 4, to allow for more replicate experiments and reduce the standard deviation, and thereby, increase the statistical significance of the results from the ASTs.
2. No sharing of ASTs between laboratories, due to the high inter-laboratory variation.
3. Parameter verification experiments were introduced prior to execution of the revised ASTs to verify that all combinations of parameters lead to stable single cell operation.
4. The duration of the ASTs was reduced from 400 to 200 hours, to allow for more cells to be tested (combination of operation conditions) and be able to complete full sets of factorial experiments with replicates.
This way, a more robust approach was taken during phase 2 of the AST program (Months 30-42), with the aim of getting more reliable and quantitative results, and thereby facilitates the establishment of a lifetime prediction model (WP5). Despite these corrective actions and the immense effort put down by all project partners, the statistical evaluation of the data in WP 5 is not revealing reliable and significant relationships making it possible to generate a Life Prediction Model, as depicted in the Description of Work. However, we have gained a lot of experiences from our substantial experimental AST-program. Dissemination of these Key Achievements is realised in a merged public Deliverable, representing D5.3 Life Prediction Model and D1.5 Design Guide, entitled: "Experiences from a comprehensive PEMFC degradation study as basis for Lifetime Prediction Modelling and a Corresponding Design Guide for PEMFC degradation studies."
Recommendations for researchers planning to execute degradation studies on PEMFCs:
The list below is by no means exhaustive, but should provide valuable hints to avoid pitfalls commonly encountered during execution of such studies.
1. Acknowledge that fuel cell operation and degradation mechanisms are complex
Fuel cells operation include a complex set of inter-related processes which altogether enable these energy converters to convert chemical energy directly into electricity. For low temperature proton exchange membrane fuel cells (PEMFCs), e.g. water plays a key role in membrane proton conductivity and without external humidification of anode gases or sufficient back diffusion of water from the cathode, the ohmic loss will increase dramatically. It is a prerequisite to acknowledge the complexity and reflect it in the scientific approach to be able to reveal the nature of degradation.
2. Consider the use of statistical design of experimental program
Various statistical tools for experimental design are available, and the use of such tools is highly recommended. However, one should consider the tool's adequacy for the given task and pay attention to the inherent assumptions these design tools are based on, and take these into account when interpreting the data.
3. Carefully select the stress factors to mimic your fuel cell application
Stressing factors adequate for your fuel cell application (e.g. µ-CHP, automotive etc.) should be carefully selected based on the typical real life operation profiles. It might be necessary to down-select and try to mimic more than one stress factor by identifying similarities and merging some originally defined stressing conditions.
4. Execute adequate baseline experiments as benchmark for AST program
Carry out baseline experiments with benchmark materials as reference for your AST program every time a precursor or manufacturing technique is changed. An AST protocol will to a certain degree always reflect the specific application and materials in use, although the ultimate goal is a set of generic protocols for wide application.
5. Establish a set of End of Test (EoT) criteria for each AST protocol
In some cases the AST protocol applied causes a steadily decreasing performance, whereas other ASTs eventually cause certain cell failure. The duration of an Accelerated Stress Test until the cell meets your EoT criteria will highly depend on the aggressiveness of the conditions these cells are subject to. EoT criteria may be cell voltage (at a given current density), hydrogen crossover current, cell failure (pin holes) etc.
6. Tune the Break-in procedure to ensure maximum performance at BoT
Activation of fuel cells is highly dependent on the materials in use and the pre-conditioning these have been through during manufacturing and MEA assembly. Some materials need longer break-in to reach maximum performance than others. The Break-in procedure should, hence, be tuned in each case to avoid intermediate peaks in performance during the AST execution.
7. Run parameter verification experiments prior to start ASTs
Stable operation of the fuel cell is a pre-requisite for being able to interpret performance data and determine degradation rates. The parameter span (variable space of the factors (e.g. lower and higher temperature)) should be carefully tuned to assure stable operation. A wide parameter span is beneficial to reveal a significant effect of that factor. However, the larger the span, the required linearity of effects for the factorial designed experiments at 2 levels (Section 2) may be lost.
8. Establish the variance and the statistical significance of your results
Inter-laboratory variance is generally large, primarily related to hardware differences. Splitting the experiments in an AST protocol between laboratories is, therefore, not recommended unless satisfactory variance can be verified. As an integral part of the interpretation effort, a systematic and regular assessment of the variance and statistical significance of the results should be carried out. Only when this is in place, the significance of each factor under study may be verified.
9. Compare to real life experiments for acceleration factor determination
Depending on the application and the corresponding targeted lifetime, an Acceleration Factor (AF) in the range of 10-200 is recommended, corresponding to ASTs with durations of up to 500 hours to enable efficient screening of new materials and assembling procedures. By linking up to already executed or running real life fuel cell field tests, or initiate real life experiments as part of your project, the AF may be determined. In case this is not possible, an reference experiment should be initiated as early as possible, and run continuously in parallel to the AST–program.
10. Distinguish Reversible from Irreversible contribution to performance decay
Performance decay of fuel cells is composed by two terms, the reversible performance decay which may be recovered by applying selected procedures (e.g. voltage cycling to oxidize and thereby remove CO on the catalyst surface) or changing the operating conditions (e.g. increase gas stochiometries to counteract electrode flooding) and the irreversible performance degradation (e.g. loss of electrochemical surface area (ECSA) of the catalyst or loss in membrane conductivity) which may not be recovered unless the cell components are replaced. When interpreting data from degradation studies, it is important to be aware of this and treat the data correspondingly.
11. Caution should be taken when extrapolating data for lifetime prediction
The degradation rate typically changes significantly over the lifetime of a fuel cell under real life operation as well as throughout the duration of an accelerated stress test. Moreover, there are degradation mechanisms which typically result in certain cell failure (e.g. pin-hole formation), and such factors should be carefully examined and included in cell and stack lifetime prediction. Especially for lifetime prediction of stacks and systems, it is a pre-requisite that the probability of single cell failure is included.
Project Results:
This section provides a more detailed summary of the work and progress in each Work Package.
Work Package 1 From Real-life operation to experimental programme
Work-package objectives
The overall objective of WP1 was to provide guidance for the experimental design, and revisions of the presently available AST, based on assessment of typical real operating conditions, experiences from field tests, as well as experimental results obtained in WP2 & WP3 and the subsequent statistical analysis and modelling results from WP5.
The background knowledge on real-life operation from Denmark (pure hydrogen operation) and France (reformate operation) has been compiled. The initial and revised AST protocols have been defined based on this compilation. The AST-protocols are designed to be capable of mapping out the known durability limitations and the gained improvements one-by-one within the defined test-protocol. The test protocols outlined comprises both single cell and stack tests.
The main WP1 outcome was four (4) validated and detailed AST protocols, addressing real-life durability challenges and revealing figures for the expected lifetime from short 200 hours' tests. The AST protocols comprised guidance for the following:
• Single cells operated with H2/air
• Single cells operated with reformate/air
• Stacks operated with H2/air
• Stacks operated with reformate/air
Summary of significant results
The consortium has throughout the project been continuously up-dated with the results obtained in the Danish µCHP field test, where three (3) units have been demonstrated for a full year and 29 additional units have been installed and started-up. The objective was to provide the partners with sufficient background knowledge on real-life operation to design an optimal experimental work programme. Recommendations were focused on the relevant number of start/stop cycles, relevant number of hours for durable continuous operation, information on possible abnormal events, temperature, definition of lifetime etc. Selected MEAs from the two stacks that has been field tested for a full year were given to CNRS-partner for post mortem analysis. Initial and revised AST-protocols were applied for the experimental AST work performed in the present project and the results from these AST tests have been interpreted in WP5.
The project results have inspired IRD to exchange some MEA precursors and further optimize the operational conditions in the Danish on-going field test. As it has been reported, this has significantly increased the MEA durability i.e. the degradation rate was decreased from 20 µV/h to 4 µV/h.
The experience obtained in the two field tests reported has pin-pointed the following topics as essential to enhance the LT PEM durability and lifetime:
• Improve the MEA and in particular the membrane with respect to mechanical properties
• Improve the catalyst support durability in order to minimize the possibility of carbon corrosion in case of e.g. fuel starvation
• Define optimized start-stop & idle mode strategies
• Develop operational strategies for dead-end dry hydrogen operation
• Improve the membrane and catalyst tolerance towards fuel impurities to avoid expensive filters (e.g. ammonia, NOX, soot, salt and other pollutants that may occur in uncontrolled amount).
Main issues assessed and findings upon the AST protocol revision:
• Number of experiments were reduced to ensure higher quality results
• The AST duration was reduced to 200 hours allowing for more cell tests
• The use of fractional factorial experimental design may lead to difficulties with result interpretations (confounding)
• Due to difficulties in interpretation of AST results on single cells and challenges encountered with delays in stable membrane deliveries, stack tests were postponed and limited to those most relevant for actual system operating conditions
Work Package 2 Accelerated Stress Tests on stacks and single cells
Work-package objectives
The objective of this WP was to obtain quantitative data on the impact of stressing conditions by application of AST protocols proposed in WP1 to stacks and single cells.
Summary of significant results from Stack tests
Standard IRD STP0012 stack hardware as described in the D4.2 report was used. The manufacture of 4 stacks with improved MEAs supplied to the project partners is described in the D4.6 report “Fuel cell stacks with improved materials“. The planned stack AST protocols have been outlined in the D1.2 KeePEMAlive delivery report.
Each 10 cell short stack was subject to a specific AST protocol during Month 40-43. Some experiments with hydrogen/air were finalized successfully over the targeted of 1000 hour duration, totalling around 3000 stack test hours. For reformate operation, however, problems were encountered, and stable operation was not obtain within the timeframe of the project.
Continuous operation
The cell performances of IRD stack @ BoT varied between 0.65V and 0.69V @0.4 A/cm² (i.e. 0.26-0.28 W/cm²), corresponding to a RSD of 1.9 % which is significantly lower than the RSD of 6,4 % found for single cells.
Although the relative high OCV cell values @ BoT demonstrated well activated cells, their improvement with time suggested that the break-in procedure of the
stack should be optimized.
The decay rate of the cell stack was higher at low power densities (viz. 0.05-0.2 A/cm²). The main cause of this finding might have been the drying condition.
The worse cell performances were revealed by the cells located to the H2 inlet cooling outlet (Z flow) while the cell located in the middle of the stack was the best one. The better and more stable water balance management offered in the centre of the stack would permit higher and more stable performances of the cells.
Start/Stop protocol
Under start/stop AST protocol; it was evident that the stack became more sensible to water flooding at high current densities. At BoT and 215 h, it was possible to reach 0.8 A/cm² and from 430 h until EoT (594 h), it was impossible. On the other hand, overall performances were stable, even slightly increased was observed over the time.
Comparing the performance of the individual cells, it recognized that cell 9 is the best operating cell (second cell from H2 inlet) while cell 2 the less performing (second cell from the H2 outlet).
In general, the stack voltage degraded ca. 0.9 µV/h @ 0.3 A/cm2 during the first 100 hours of operation. However, after this initial degradation in performance no degradation was observed along the start-stop test for current densities lower that 0.6 A/cm2.
Dead end operation
The dead-end stack test was a straightforward experiment under normal operating conditions as in the Danish demonstration project, where the initial degradation rate was evaluated. The test was not a real AST but served as comparison to the single cell test and the other stack tests in particular to the continuous stack test with pure hydrogen.
Four (4) start/stops due to characterization were planned during the 1000 test hours. However, different unintended events occurred and the total number of start/stops were therefore 14. The unintended stop/starts did not have any major influence on the results. The BoL cell performance was rather uniform. The difference between the cell performances were rather constant for the first 850 hours and slightly increased after this time.
The overall cell voltage decay rate calculated as linear regression on the average cell voltage in operation was 52 µV/h corresponding to 52 mV/1000 hours. However, the voltage decay was not uniform for the cells, the observed decay pattern for the individual cells most of all reflected a higher degradation at the stack ended possible due to temperature issues.
Summary of significant results from Single Cells
The comprehensive AST experimental program comprised more than 20 000 hours of single cell testing, divided on around 60 single cell tests each with duration of 200 – 400 hours. A selection of results is summarised below.
The main findings from AST protocols applied to single cells are summarized as follow:
• Under continuous operation, combination of high temperature, high RH and high current density leads to a rapid failure of the f-940Rf-based cell, while at low temperature, low RH and low current density operation no failure was observed within the 200 h of the test.
• Anode electrode experiences strong degrading effects in single cells tested under reformate operation. There was evidenced that the anode overpotential is the factor dominating the anode degradation, which included Ru dissolution.
• In Fuel Starvation experiments the combination of high temperature (85 °C) and high relative humidity (80%) led to a very short lifetime and high performance decay which is mainly caused by carbon corrosion. High CO and CO2 concentrations in the anode off-gas were measured at these operating conditions. In addition high current density (400 mA/cm²) accelerates the degradation of the MEA.
The formation of pinholes has been proven by hydrogen crossover measurements using a segmented cell. In general, the pinholes had a tendency to appear at areas where the membrane has been previously damaged, also showing a certain cross-over of hydrogen prior to execution of the fuel starvation protocol.
• From performance decay calculated from single cell tested under the load cycling AST protocols, high settings were the most detrimental operation condition. It means that elevated temperature (85 °C) combined with high frequency cycling (6 cycles/hour) at mode 1 (0.2 – 0.6 A/cm2) significantly accelerated the degradation of the cell.
• Comparison between baseline F-940-based cell studied during phase 1 of the project and improved MEAs (reinforced F-940) has revealed a real improvement of cell stability under equivalent ASTs.
The complete results from experimental work following AST protocols have been described and summarized in more details in the deliverables D2.3 (stacks) and D2.4 (single cells).
Work Package 3 Ex-situ characterisation of materials
Work-package objectives
The objectives of WP3 were
• to characterise MEA and stack components at beginning-of-life, end-of-test and end-of life
• to develop ex situ accelerated ageing tests for individual MEA component
• to relate operation conditions to ex situ characterization observations
WP3 was dedicated to the ex-situ characterisation of complete membrane electrode assemblies (MEAs) and their components, in particular the electrocatalysts and membrane materials. This characterisation was performed on pristine materials and MEAs before use in a fuel cell (BoL characterisation), and after ageing (EoT or EoL characterisation) – this ageing having been induced in-situ in the fuel cell, and ex-situ, using accelerated ageing protocols.
Furthermore, WP3 also focused its work on to characterize improved membranes incorporating radical scavengers. Examination by scanning electron microscopy with element mapping of single cell MEAs at end of test and end of life was performed. Likewise, MEAs from stacks subjected to stressing conditions were also examined allowing for identification of the regions of the MEA which were most deteriorated during operation (inlet, central section, and outlet). All in all, this are integral parts to support the main objective of the KeePEMalice project, i.e. to increase the understanding of the degradation mechanisms.
Summary of significant results
• Three types of electrocatalyst from three different suppliers have been characterised at BoL for their metal particle size and particle size distribution. Ex-situ accelerated ageing has been carried out by voltage cycling and under simulated stop/start regime. Catalyst layers have been developed using the various candidate catalysts, and corresponding complete MEAs have been observed in cross-section using SEM. These results and observations have led to a final choice of catalyst material for the anode and cathode of the MEAs.
• Pristine perfluorosulfonic acid membrane materials and forms protected against chemical degradation have been submitted to accelerated ageing ex-situ, using liquid and vapour phase Fenton testing. The membranes have been characterised using vibrational and nuclear magnetic resonance spectroscopies to determine the degradation pathways and how these are modified by the presence of radical scavenger ions. Optimised improved membranes and their components have been characterised by scanning and transmission electron microscopy
• Examination by scanning electron microscopy with element mapping of end of test and end of life single cell MEAs and used MEAs from stacks after application of stress conditions allowed identification of the regions of the MEA most subject to modification. As one example, fuel starvation was shown to lead to significant loss of ruthenium from the anode (and transfer to the cathode) in the region of the gas outlet, compared with the beginning of life MEA.
• Electrochemical characterisation of catalysts clearly demonstrated better performances for Tanaka Kikinzoku Kogyo (improved material) over Cabot catalyst (baseline material), independently of the stress test applied. Thus, this catalyst was used as MEA component for manufacture of MEAs to be used in second phase of WP2 in situ aging tests.
• Aged MEAs clearly showed variations in composition relatively to fresh assembly along the cross-section depending on the different locations examined (outlet or inlet).
• Ex-situ bipolar plate corrosion appears negligible, even in electrolyte containing fluoride ions. This result allowed the re-use of stack bipolar plates of WP2 between the several tests.
• Application of ex-situ ageing protocols to pristine PFSA membranes, and to membranes modified by incorporation of radical scavengers. The effect of radical scavenger ions on reducing fluoride emission was characterised in-situ (WP4) and ex-situ using 19F MAS NMR spectroscopy to follow local structural changes to the perfluorinated polymer backbone and side-chain.
• Incorporation of Ce and Mn ions decreased the chemical degradation imposed by Fenton reagent (Fe/H2O2). In each case the spectrum of the Ce ion exchanged membrane was similar in peak height and width to pristine Nafion-212, in the contrast with other degraded samples. Moreover the biggest tendency to enlarged peak width was observed for the peak at 118 ppm (side chain SCF2 group), which indicated side chain scission as being a major consequence of oxidative and free radical attack. These results could be directly related to the observations made in-situ on the lower fluoride emission rate from MEAs containing Mn and Ce partially exchanged membranes.
• A range of complementary techniques were used to characterise nano-particulate cerium oxide prepared by three different approaches, in order to determine whether the synthesis method affects the proportion of Ce(III) in the material, and to select the method reproducibly providing mono-disperse CeO2. The final composite membranes comprising a surface layer of nano-fibre PFSA/CeO2 were characterised for their proton conductivity and observed using SEM/EDX analysis to confirm that the cerium oxide distribution was satisfactorily distributed across the membrane surface.
• SEM/EDX has been used to analysis MEAs that had been submitted to AST in WP2. The results on MEAs having undergone fuel starvation stress were particularly noteworthy, since clear correlation has been found between the applied stress conditions at high RH and Ru depletion at the hydrogen outlet region of the anode catalyst layer, and Ru depletion at both the inlet and outlet regions at low RH. The anode catalyst layer composition at the inlet was therefore a sensitive indicator of the combined effects of fuel starvation and the degree of hydration of the MEA on anode catalyst corrosion. Under these conditions, cell reversal effects leaded to high anode potential. Cell reversal resulted in the production of oxygen instead of hydrogen oxidation at the anode through the oxidation of water. Ru is unstable at these potentials, and dissolves into water either fed or produced at the anode (cell reversal), and is carried into the exit water stream or could diffuse into the membrane. Indeed, although Ru could not be detected by EDX analysis across the membrane thickness, it wasx detected in the catalyst layer at the cathode (in the pristine MEA only Pt is detected at the cathode), although there was no clear relation between the amount of Ru detected and the applied AST conditions in terms of applied current density, temperature or RH; instead, a possible correlation appeared between the number of starvation/recovery cycles, and the amount of Ru in the cathode catalyst layer.
• Furthermore, there was clear correlation between Ru depletion and the thickness of the anode catalyst layer. The layer thickness was ca. 5 µm in pristine MEAs and in the anode catalyst layers of MEAs with a Ru/Pt ratio of 1, and it decreased progressively as Ru was leached. These observations provided input to WP5 and were entirely consistent with those described from ex-situ analysis of the cyclic voltammograms on start/stop cycling (see deliverable report D3.5).
• The ex-situ electrochemical characterisation of improved materials showed that PtCo/C catalyst presented both higher electrochemical stability and ORR activity as cathode material in comparison to the state-of-the-art Pt/C sample.
Work Package 4 Preparation and improvement of MEAs and components
Work-package objectives:
• To provide state of the art membranes, electrodes, MEAs and stacks for WP2.
• To develop and provide improved individual materials, MEAs and stacks with higher durability.
In terms of material development activities, the main objective of WP4 was to manufacture an improved membrane material by continuous production, and assure its compatibility with the improved electrode materials in use, as well as to select the best suited catalyst material for MEA fabrication, based on the initial results.
Summary of significant results
• New reinforced chemically stabilised membrane that demonstrated significantly higher durability compared plain membrane’s film without stabilisation (i.e. w/o radical scavengers) was developed by Fuma-Tech. The stabilisation of membrane was based on reinforcement, which has impact on mechanical properties. The chemical stabilisation was based on interaction of inorganic-organic composite with PFSA matrix and which yielded homogeneous distribution of stabilisation through the whole volume of membrane.
• With the aim of producing reinforced chemically stabilized membrane in a mass production, the knowledge about stabilization of mechanical and chemical nature of PFSA membrane were combined. After comprehensive search, a unique combination of solvent mixtures was identified and membrane called as F-940rf and later as F-940rfs was mastered and produced as a roll. This strategy demonstrated the production of a volume of at least 20 m2 as a roll.
• Preparation of PFSA membranes incorporating a gradient of radical scavenging oxide by electro-spinning, giving a composite CeO2/PFSA nano-fibre mat was achieved. These electrospun mats were the means of enabling the CeO2 radical scavenger to be specifically incorporated sited in close proximity to one or other catalyst layer, rather than distributed throughout the membrane. The incorporation of an additive specifically sited at a predetermined location within the membrane was complementary to the reinforcements incorporated in KEEPEMALIVE membranes that improved the membrane mechanical strength. This route provided a further means of increasing durability of the membrane and complete MEA. The release of fluoride from the MEA on fuel cell operation was used as a gauge of success of this approach. MEAs incorporating modified membranes showed a strongly reduced fluoride emission rate (FER). Thus, the resulting membranes increased its chemical stability and showed significantly improved lifetime under accelerated stress testing. In this way, the MEA lifetime was increased from <200 h to >800 h by incorporation of this radical scavenging interlayer under OCV conditions. These results validated the approach developed in the project by CNRS and provided for this group a key result that will be used in its further research and development.
• Four (4) 10-cell stacks with improved MEAs have been manufactured. These stacks were equipped with a novel MEA type where two new precursors have been implemented. The novel precursors implemented were the reinforced membrane (F-940rf) developed by Fuma-Tech and the very stable Tanaka cathode catalyst (TEC10F60TPM). The novel MEAs required development of a new prolonged initialisation (break-in) procedure. However, the MEAs performed well once adequately conditioned.
• Adopting of latest generation of FuMA-Tech’s membrane leads to improving of lifetime of MEA’s to level that is between 10 000 – 20 000 operation hours at conditions governing stationary fuel cells.
Work Package 5 Design and evaluation of experiments
Work-package objectives
• Identify critical operating conditions for PEM fuel cell stacks, based on screening of single cells, actual testing of stacks, field tests, modelling and experimental assessment of local conditions in a stack
• Based on statistical analysis of data, in combination with fundamental knowledge of degradation mechanisms, propose new accelerated stress tests (AST) for PEM fuel cells
• Develop lifetime prediction models based on theoretical and phenomenological models, as well as the results of the statistical analysis.
The activities of this WP were directed towards developing tools for the interpretation of experimental data, and thus contribute both to an increased understanding of the degradation phenomena in fuel cells, and to propose AST protocols closer to realistic operating conditions.
Summary of significant results
• Experimental results from the AST protocols conducted during the comprehensive AST program were analysed statistically using the Yates algorithm to obtain main effects and interactions and their contributions to the degradation rates. Effects of temperature, relative humidity and current density on the degradation rate were evaluated for the "Continuous operation" protocol, the "Load Cycling" protocol and the "Fuel starvation" protocol.
• Fuel Starvation for single cells is the most comprehensively studied AST protocol in this project. Initially, the results from a fractional factorial design (23-1) were statistically evaluated and reported in D5.2. The analysis revealed that second order effects were of a significant magnitude. Hence, the pre-requisites for use of fractional factorial design were not fulfilled and this resulted in confounding of main and second order interactions. To be able to distinguish main effects from interactions, a full 23 experiment was subsequently performed in phase II of the AST program. The results were supplemented by post mortem ex-situ characterization of the MEAs in WP3, and reported in D3.6.
• Despite the immense experimental effort put down by all project partners totalling in the range of 25 000 hours' testing, the statistical evaluation of the data in WP 5 is not revealing reliable and significant relationships making it possible to generate a Life Prediction Model, as depicted in the Description of Work. This is seen as a deviation from the original plan and was therefore conveyed already during the Mid-Term Review (M33). However, we have gained a lot of experiences from our substantial experimental AST-program. Dissemination of these Key Achievements was realised in a merged public Deliverable (http://www.sintef.no/keepemalive) representing D5.3 Life Prediction Model and D1.5 Design Guide, entitled: "Experiences from a comprehensive PEMFC degradation study as basis for Lifetime Prediction Modelling and a Corresponding Design Guide for PEMFC degradation studies."
• Modelling work was carried out in order to support the interpretation of experimental data along two routes: (i) application of a single cell model for assessment of local conditions along a channel/plate (JRC), in line with experimental data carried out with a segmented cell and with local reference probes and (ii) modelling of Pt degradation mechanisms in line with ex-situ catalyst characterisation (SINTEF). Regarding (i), a 3D model was constructed, including transport processes in the gas phase and the solid phase, reaction kinetics, as well as gas solubility in condensed water. The model is a two-phase model, assuming fine mist in convective flow. Results have been used for determination of temperature gradients and hot regions, as well as conditions leading to non-uniform current distribution, thus causing fuel starvation or low O2 concentration. Regarding (ii), ex-situ catalyst tests of Cabot catalysts from WP3 (cycling between 0.9 and 1.2 V) were described by a model describing chemical dissolution of PtO, in accordance with models provided in the literature , . The model provided a qualitative explanation for the dependence of the ECSA loss on cycling rate. Furthermore, the model reproduces the slightly non-linear behaviour of the ECSA loss vs. time, which is related to the size-dependent surface tension of the rate constant.
• Calculated effect using the decay rates from samples exposed to Fuel Starvation AST protocol showed that the main effects of T and RH are large and negative. The effect of j does not appear to correlate with performance decay. The impact of RH is much higher at high T than at low T (-31 vs. -3.6). It implies from the fact that T*RH is equal to RH*T, that the impact of T is higher for high RH. However, making assessment of the magnitude of T and RH main effects is difficult when large interaction is documented between the factors. It was plausible that T and RH are coupled in PEMFC. It is also important to recall that a prerequisite for the 23 design is orthogonally: that all factors are completely independent of each other.
Despite the failure in the establishment of a tool for lifetime prediction, WP5 has generated a lot of valuable information regarding assessment of lifetime of PEMFC systems. This know-how was processed and briefly summarised as follow:
1. At the end of the break-in period of a cell, the variance between cells has been calculated. All effects observed must be larger than this variance in order to be statistical significant.
2. The variance for several of the characterization techniques used in this project (i.e. cyclic voltammetry, electrochemical impedance spectroscopy, hydrogen cross-over current) has been established. Knowledge of requirement for significance of effect has been also established.
3. Some parallel experiments have been conducted in order to evaluate variance between experiments beyond beginning of test.
A systematic approach using statistically designed experiments (factorial design) was used, as described in the DoW. To get maximum information out of the factorial designed experiments, complete sets of so-called 23 experiments (3 factors, each at 2 levels = 8 different operation conditions) were planned at each partner's test facilities. During the execution of those set of experiments, however, it turned out that many of the selected sets of operation parameters led to instable operation (e.g. flooding or dehydration) and several of the 6 distinct AST protocols were not executable. The foreseen statistical analyses to reveal the factor's effect on degradation were, hence, not possible due to incompleteness of the full sets of eight experiments.
Difficulties encountered during the initial iterations of statistical analysis of the data from 37 single cell tests carried out in phase one of the AST program (Month 1-25), lead to the conclusion that only qualitative relationships between factors and degradation could be identified.
Therefore, it was concluded that during the second phase of the AST program (Month 30-42) complete sets of factorial designed experiments were to be executed.
Assessment of real stack performance data
Stack performance data from the Vestenskov field tests in Denmark was assessed statistically and results are reported in this section. Through the availability of real life stack data at sufficient resolution (1 Hz logging) and with availability of individual cell voltages, competence in evaluating large data sets has been obtained:
1. Import and storage of data in a useable format for evaluation.
2. Efficient smoothing of data, removing noise, yet retaining valuable information.
3. Extraction of subsets (i.e. transients) for evaluation of degradation rates.
4. Evaluation of population distributions.
5. Fitting of data.
Statistical evaluation of real life stack data has given evidence of a change in cell population distribution from being normal at beginning of life to become better fitted by a Weibull distribution later in stack life. This is an important finding, previously assumed in literature that could be utilised for prognostic purposes (Remaining Useful Life estimate) as well as lifetime prediction.
In order to assess the precision in data used for calculation of degradation rates and lifetime prediction, more than one measurement is required. Optimally, a population of identical stacks running under identical ambient and operational conditions should be available for statistical evaluation. This is rarely the case. For a single stack, the individual cell voltages can be utilized in order to evaluate the variance. More interesting than the average cell voltage and its corresponding standard deviation is the distribution of the population.
Stack geometry will affect the distribution of individual cell voltages. From a design point-of-view, it is of interest to evaluate individual cell degradation rates in order to verify stack design. From evaluation of individual cell voltages in the stack, deviation in the end cells were observed. Therefore, the two outermost cells at each end of the stack were excluded from further cell voltage evaluation. So, in this particular exercise, 43 cells were used to evaluate the population distribution. In order to improve distribution assessment, 1 Hz data was used where 200 seconds of individual cell voltages were used to create voltage populations. At Beginning of Life (BoL), the distribution is best described normally.
The probability density is better described by a normal distribution, symmetric about the mean value. The goodness of fit can be evaluated numerically e.g. by a log likelihood parameter. It was postulated that a check for normality in single cell voltage distribution for a running stack could say something about the state of health for the stack. A normal Chi-Square test was used to evaluate the "degree of normality"
There are many ways to estimate the performance degradation rate of a fuel cell stack. The estimates obtained are strongly dependent of which and how data are fitted. Some reversible (recoverable) degradation effects are typically encountered (e.g. flooding/dehydration/CO-poisoning), and these should preferably be separated from irreversible degradation when calculating degradation rates.
Potential Impact:
Impact
Assessment of the field data from real life operations (at Lolland) related to the KeePEMalive project has enabled the project system development partner IRD to improve the µCHP system and take the technology one step closer to fulfil the stringent requirements for long term durability. By exchanging some MEA precursors and further optimise the operational conditions in the Danish on-going field test the MEA durability was increased significantly e.g. the degradation rate was decreased five-fold from 20 to 4 µV/h, corresponding to increasing the system lifetime from the previous level of 3 500 hours to an expected 17 000 hours (~2 years). The heat and electricity demand and the related energy and emission savings from utilizing the CHP-units in Danish households have been mapped for various seasons and during the course of the project the electric system efficiency has been improved to 50%.
Dissemination activities
Dissemination activities have had a high focus in the project and the consortium acknowledges the importance of promoting the development of fuel cell technologies for sustainable and efficient utilization of hydrogen as energy carrier. Our public webpage (www.sintef.no/keepemalive) contains a complete list of dissemination activities.
During the KEEPEMALIVE project development of new materials and concepts has been the focal point. Several novel solutions have been implemented in the related field test program at Vestenskov, Denmark, at which in total 32 PEM fuel cell based µ-CHP systems have been installed in households. This again is reflected in the dissemination plan, where one paper is already published, two manuscripts are submitted and another 6 scientific papers based upon the results obtained in the project are planned for publication in the coming year, as listed below:
1 Characterisation Protocols of Stationary Type of Fuel Cells Focused on Prediction of Lifetime Tomas Klicpera The Journal of Fuel Cell Technology Vol. 11, 2013 Special Issue FCDIC Tokyo, Japan 2013 pp. 12- 31 www.fcdic.com
2 Study of the transport of radical scavenger in PFSA membranes and effect of their loading on membrane properties. M. Zaton J. Membr. Sci. Elsevier Manuscript submitted
3 Significant lifetime improvement of PFSA MEAs M. Zaton Electrochem. Comm. Manuscript submitted
4 Parametric study of the effects of cell reversal during fuel starvation on fuel cell performance loss and anode electrocatalyst degradation Victor Hacker Electrochimica Acta Elsevier Manuscript submission in 2013
5 PEMFC fuel starvation degradation evaluated by experimental design Thor A. Aarhaug Journal Power Sources Elsevier Manuscript submission in February 2014
6 Improvement of testing techniques for characterisation techniques for the localisation of MEA failures Astrid Hofer TBD
7 Stability of PEM fuel cell cathode catalyst under potential cycling conditions. An ex-situ AST characterization approach Luis Colmenares Journal of Power Sources Elsevier Manuscript submission in 2014
8 Tailoring the catalytic activity of PtRu/C catalysts by potential cycling Luis Colmenares Electrochimica Acta Elsevier Draft available, Manuscript submission in 2013
9 Experimental strategy for PEMFC durability assessment Steffen Møller-Holst International Journal of Hydrogen Energy Elsevier Manuscript submission in April 2014
Significant results from the project have continuously been disseminated to the scientific community through more than 20 oral presentations and 7 posters, some at major international conferences such as WHEC and ECS meetings as well as seminars and workshops targeted to a wider community of private and public stakeholders.
Moreover, an International Summer School on advanced studies of Polymer Electrolyte Fuel Cells, was held at TU Graz, August 26th – September 1st, 2010. The Summer School was organized in co-operation between Graz University of Technology, Austria (Dr. Viktor Hacker) and Yokohama National University, Japan with financial support of the European Commission (Seventh Framework Programme, FP7/2007-2013, Fuel Cells and Hydrogen Joint Technology Initiative under grant agreement n° 245113) and the Cost Action 543, Research and Development of Bioethanol Processing for Fuel Cells. The Summer school was attended by 50 students. A lecture was given by the Coordinator, Gaby Janssen, on the Lifetime and Durability issues in PEMFC and how KEEPEMALIVE tries to address these issues.
Last but not least, to reveal synergies and exchange high level knowledge in Europe, a joint Workshop entitled "Degradation of PEM Fuel Cells - experience exchange & discussions", was arranged at SINTEF in Oslo, Norway as a collaboration between 3 FCH JU-supported projects: Stayers, Premium Act and KeePEMalive, April 3rd and 4th 2013.
List of Websites:
Project website address:
http://www.sintef.no/keepemalive
Project's coordinator :
Steffen Møller-Holst
Vice President Marketing,
Hydrogen and Fuel Cells
SINTEF
Tel: + 47 92604534
Fax: + 47 73591105
E-mail: steffenh@sintef.no
In the KeePEMalive project a comprehensive test program on fuel cell degradation has been carried out, accumulating more than 25 000 accelerated test hours on single cells. Performance data were analysed and compared to results from a Danish field test project (at Lolland) in which 32 fuel cell systems for combined heat and power are being deployed in households to identify detrimental operating conditions. By improving cell materials and optimising the system operation conditions the lifetime of these fuel cell systems has been increased from ½ to 2 years.
There has been five (5) focus areas of activities in the project:
1) Material development and selection
Development of PFSA based membrane materials has been a key activity in the project. In total 5 generations of membrane materials have been synthesized and characterized with respect to key properties. The initial batch production was successfully scaled up to continuous membrane series production in the range of 2000 m2/year.
Partner CRNS showed that by introducing Ce-based radical scavengers the stability of the Membrane was further enhanced. Manufacturing of MEAs and stack assembly was done in-house by IRD.
The most stable and best performing fuel cells were obtained using reinforced membrane and a cathode catalyst with improved catalyst dispersion on an oxidant resistant support.
2) Fuel cell characterisation during operation
To quantify the degradation of the fuel cells at a reasonable speed, a set of Accelerated Stress Test (AST) protocols were developed and applied, exacerbating selected stressing conditions relevant for real life µCHP application, including:
• Continuous operation, • Reformate operation, • Dead end operation, • Start-stop, shut-down, • Fuel Starvation, • Electric load cycling
Distinct sets of experiments were run for each of these stressing conditions by systematically varying operating conditions such as fuel gas humidity, cell temperature and cell current.
After conditioning the fuel cell for 24 hours, the cell performance was mapped every 24 hour to identify changes as input for degradation rate quantification.
Fuel cells performance was assessed using: • Electrochemical Impedance Spectroscopy, • Hydrogen cross-over & Cyclic Voltammetry, • Carbon corrosion by NDIR, • Membrane degradation by Fluorine emission rates, • Segmented cell to reveal localised degradation
3) Identification of material changes
The following materials were subject to assessment: - Catalyst materials (SINTEF), - Catalyst layer surface & cross section (CNRS), - Membrane materials (CNRS, Fumatech)
and these MEA components' properties including e.g.,: - Mechanical strength, membrane (Yuong modulus), - Morphology, Microstructure, - Catalyst agglomeration and migration by (SEM), - Electrode ElectroChemical Surface Area (ECSA), - Membrane Conductivity, Ohmic resistance, - Water-uptake, hydration number
4) Quantification of degradation rates
Changes in performance with time were assessed and the degradation rate was calculated. The used fuel cell materials subject to stressing conditions were compared to pristine materials revealing corresponding material changes.
Statistical analyses has clearly shown that degradation depends on several factors and varies along the operation curve as illustrated below for load cycling over 144 hours AST.
5)Enhanced insight/understanding
Based on studies of more than 60 cells and assessment of corresponding material changes for some of these, the cause and effect relationships were identified.
The complexity of processes taking place in PEM fuel cells were verified by statistical analysis showing that factors causing degradation are inter-related.
The enhanced insight provided a sound basis for selection of improved materials for next generation MEAs to be tested.
Project Context and Objectives:
The main objective of KEEPEMALIVE's was to establish improved understanding of degradation and failure mechanisms, accelerated stress test protocols, sensitivity matrix and lifetime prediction models for LT PEMFC to enable a lifetime of 40 000 h at realistic operation conditions for stationary systems, in compliance with performance and costs targets.
As 40 000 hours' testing is not possible to achieve within the 3 ½ years' duration, the aim throughout the project was to find relevant procedures to highly accelerate the degradation rate observed in µCHP applications and limit the testing-time required for estimating the lifetime of PEMFCs for stationary applications. Hence, an accelerated stress test (AST) experimental program was set up. Because the degree of acceleration is strongly dependant on the operating conditions, the program was divided into two phases.
The topic "degradation of fuel cells" is highly complex in its nature. Therefore, a thorough assessment of available literature on the relationships between operation conditions and lifetime issues was carried out leading to definition of an ambitious initial AST program. To realise true collaboration between the involved European research laboratories and industry partners, certain ASTs were intentionally shared between laboratories.
Conditions causing degradation were identified and their relevance for stationary µCHP applications was verified. In total 6 stressors were selected to account for the most typical conditions encountered during real life operation of µCHP units. The stressors were:
1. Continuous operation: at nominal current density with long term exposure to liquid water and high voltage (relevant for winter period operation)
2. Reformate operation: fuel quality and reformer issues
3. Dead-end operation: Purging of air causing elevated pressure in the anode chamber
4. Fuel starvation: typically encountered during transients
5. Start/stop cycling: Hydrogen/air fronts in fuel channel (relevant for summer period operation)
6. Electric load cycling: (spring/fall period profile)
Operation variables such as Temperature, Relative Humidity (RH), cell voltage and current density were selected as controllable factors for the ASTs. From our best aggregated knowledge, adequate levels of these factors were selected to accelerate the degradation process roughly 100-fold.
During the initial phase of the AST program (Month 1-25), baseline experiments were carried out at the available laboratories (7 in total) in the consortium, to reveal the inter-laboratory variance. Unfortunately, it turned out that the variation in performance for identical cells tested at different laboratories were substantial, as documented during the 2nd progress meeting (M13) and in D 5.2. These discrepancies are assumed to be linked to differences in test equipment hardware configurations (e.g. flow fields) and varying active cell area used at the various laboratories (5 to 25 cm2).
Moreover, based on available protocols, experience and recommendations in literature, a break-in procedure was selected and validated to ensure that maximum performance was reached prior to executing the ASTs. The best available membrane materials were selected based on feedback from cell and stack tests. As new membrane materials were developed, however, we found that the originally verified break-in procedure (for first and second generation membrane materials) was not adequate for the latter generation of membranes (reinforced) (for details, see D 2.2). Thus, revision of the break-in protocol was needed.
Acknowledging the complexity of these relationships, a systematic approach using statistically designed experiments (factorial design) was used, as described in the DoW. To get maximum information out of the factorial designed experiments (for details, see D 5.1) complete sets of so-called 23 experiments (3 factors, each at 2 levels = 8 different operation conditions) were planned at each partner's test facilities. During the execution of those set of experiments, however, it turned out that many of the selected sets of operation parameters led to instable operation (e.g. flooding or dehydration) and several of the 6 distinct AST protocols were not executable (experimental set point values defined in D 1.2). The foreseen statistical analyses to reveal the factor's effect on degradation were, hence, not possible due to incompleteness of the full sets of eight experiments.
Difficulties encountered during the initial iterations of statistical analysis of the data from 37 single cell tests carried out in phase one of the AST program (Month 1-25), lead to the conclusion that only qualitative relationships between factors and degradation could be identified (for details, see D 5.2). As a consequence Stack tests were postponed until reliable single cell characterization could be demonstrated.
During the revision of the initial AST protocols (Months 26-29) four corrective actions were taken (details on D 1.3) in order to secure more reliable results in the phase 2 of the AST program (Months 30-42):
1. The number of AST protocols was reduced from 6 to 4, to allow for more replicate experiments and reduce the standard deviation, and thereby, increase the statistical significance of the results from the ASTs.
2. No sharing of ASTs between laboratories, due to the high inter-laboratory variation.
3. Parameter verification experiments were introduced prior to execution of the revised ASTs to verify that all combinations of parameters lead to stable single cell operation.
4. The duration of the ASTs was reduced from 400 to 200 hours, to allow for more cells to be tested (combination of operation conditions) and be able to complete full sets of factorial experiments with replicates.
This way, a more robust approach was taken during phase 2 of the AST program (Months 30-42), with the aim of getting more reliable and quantitative results, and thereby facilitates the establishment of a lifetime prediction model (WP5). Despite these corrective actions and the immense effort put down by all project partners, the statistical evaluation of the data in WP 5 is not revealing reliable and significant relationships making it possible to generate a Life Prediction Model, as depicted in the Description of Work. However, we have gained a lot of experiences from our substantial experimental AST-program. Dissemination of these Key Achievements is realised in a merged public Deliverable, representing D5.3 Life Prediction Model and D1.5 Design Guide, entitled: "Experiences from a comprehensive PEMFC degradation study as basis for Lifetime Prediction Modelling and a Corresponding Design Guide for PEMFC degradation studies."
Recommendations for researchers planning to execute degradation studies on PEMFCs:
The list below is by no means exhaustive, but should provide valuable hints to avoid pitfalls commonly encountered during execution of such studies.
1. Acknowledge that fuel cell operation and degradation mechanisms are complex
Fuel cells operation include a complex set of inter-related processes which altogether enable these energy converters to convert chemical energy directly into electricity. For low temperature proton exchange membrane fuel cells (PEMFCs), e.g. water plays a key role in membrane proton conductivity and without external humidification of anode gases or sufficient back diffusion of water from the cathode, the ohmic loss will increase dramatically. It is a prerequisite to acknowledge the complexity and reflect it in the scientific approach to be able to reveal the nature of degradation.
2. Consider the use of statistical design of experimental program
Various statistical tools for experimental design are available, and the use of such tools is highly recommended. However, one should consider the tool's adequacy for the given task and pay attention to the inherent assumptions these design tools are based on, and take these into account when interpreting the data.
3. Carefully select the stress factors to mimic your fuel cell application
Stressing factors adequate for your fuel cell application (e.g. µ-CHP, automotive etc.) should be carefully selected based on the typical real life operation profiles. It might be necessary to down-select and try to mimic more than one stress factor by identifying similarities and merging some originally defined stressing conditions.
4. Execute adequate baseline experiments as benchmark for AST program
Carry out baseline experiments with benchmark materials as reference for your AST program every time a precursor or manufacturing technique is changed. An AST protocol will to a certain degree always reflect the specific application and materials in use, although the ultimate goal is a set of generic protocols for wide application.
5. Establish a set of End of Test (EoT) criteria for each AST protocol
In some cases the AST protocol applied causes a steadily decreasing performance, whereas other ASTs eventually cause certain cell failure. The duration of an Accelerated Stress Test until the cell meets your EoT criteria will highly depend on the aggressiveness of the conditions these cells are subject to. EoT criteria may be cell voltage (at a given current density), hydrogen crossover current, cell failure (pin holes) etc.
6. Tune the Break-in procedure to ensure maximum performance at BoT
Activation of fuel cells is highly dependent on the materials in use and the pre-conditioning these have been through during manufacturing and MEA assembly. Some materials need longer break-in to reach maximum performance than others. The Break-in procedure should, hence, be tuned in each case to avoid intermediate peaks in performance during the AST execution.
7. Run parameter verification experiments prior to start ASTs
Stable operation of the fuel cell is a pre-requisite for being able to interpret performance data and determine degradation rates. The parameter span (variable space of the factors (e.g. lower and higher temperature)) should be carefully tuned to assure stable operation. A wide parameter span is beneficial to reveal a significant effect of that factor. However, the larger the span, the required linearity of effects for the factorial designed experiments at 2 levels (Section 2) may be lost.
8. Establish the variance and the statistical significance of your results
Inter-laboratory variance is generally large, primarily related to hardware differences. Splitting the experiments in an AST protocol between laboratories is, therefore, not recommended unless satisfactory variance can be verified. As an integral part of the interpretation effort, a systematic and regular assessment of the variance and statistical significance of the results should be carried out. Only when this is in place, the significance of each factor under study may be verified.
9. Compare to real life experiments for acceleration factor determination
Depending on the application and the corresponding targeted lifetime, an Acceleration Factor (AF) in the range of 10-200 is recommended, corresponding to ASTs with durations of up to 500 hours to enable efficient screening of new materials and assembling procedures. By linking up to already executed or running real life fuel cell field tests, or initiate real life experiments as part of your project, the AF may be determined. In case this is not possible, an reference experiment should be initiated as early as possible, and run continuously in parallel to the AST–program.
10. Distinguish Reversible from Irreversible contribution to performance decay
Performance decay of fuel cells is composed by two terms, the reversible performance decay which may be recovered by applying selected procedures (e.g. voltage cycling to oxidize and thereby remove CO on the catalyst surface) or changing the operating conditions (e.g. increase gas stochiometries to counteract electrode flooding) and the irreversible performance degradation (e.g. loss of electrochemical surface area (ECSA) of the catalyst or loss in membrane conductivity) which may not be recovered unless the cell components are replaced. When interpreting data from degradation studies, it is important to be aware of this and treat the data correspondingly.
11. Caution should be taken when extrapolating data for lifetime prediction
The degradation rate typically changes significantly over the lifetime of a fuel cell under real life operation as well as throughout the duration of an accelerated stress test. Moreover, there are degradation mechanisms which typically result in certain cell failure (e.g. pin-hole formation), and such factors should be carefully examined and included in cell and stack lifetime prediction. Especially for lifetime prediction of stacks and systems, it is a pre-requisite that the probability of single cell failure is included.
Project Results:
This section provides a more detailed summary of the work and progress in each Work Package.
Work Package 1 From Real-life operation to experimental programme
Work-package objectives
The overall objective of WP1 was to provide guidance for the experimental design, and revisions of the presently available AST, based on assessment of typical real operating conditions, experiences from field tests, as well as experimental results obtained in WP2 & WP3 and the subsequent statistical analysis and modelling results from WP5.
The background knowledge on real-life operation from Denmark (pure hydrogen operation) and France (reformate operation) has been compiled. The initial and revised AST protocols have been defined based on this compilation. The AST-protocols are designed to be capable of mapping out the known durability limitations and the gained improvements one-by-one within the defined test-protocol. The test protocols outlined comprises both single cell and stack tests.
The main WP1 outcome was four (4) validated and detailed AST protocols, addressing real-life durability challenges and revealing figures for the expected lifetime from short 200 hours' tests. The AST protocols comprised guidance for the following:
• Single cells operated with H2/air
• Single cells operated with reformate/air
• Stacks operated with H2/air
• Stacks operated with reformate/air
Summary of significant results
The consortium has throughout the project been continuously up-dated with the results obtained in the Danish µCHP field test, where three (3) units have been demonstrated for a full year and 29 additional units have been installed and started-up. The objective was to provide the partners with sufficient background knowledge on real-life operation to design an optimal experimental work programme. Recommendations were focused on the relevant number of start/stop cycles, relevant number of hours for durable continuous operation, information on possible abnormal events, temperature, definition of lifetime etc. Selected MEAs from the two stacks that has been field tested for a full year were given to CNRS-partner for post mortem analysis. Initial and revised AST-protocols were applied for the experimental AST work performed in the present project and the results from these AST tests have been interpreted in WP5.
The project results have inspired IRD to exchange some MEA precursors and further optimize the operational conditions in the Danish on-going field test. As it has been reported, this has significantly increased the MEA durability i.e. the degradation rate was decreased from 20 µV/h to 4 µV/h.
The experience obtained in the two field tests reported has pin-pointed the following topics as essential to enhance the LT PEM durability and lifetime:
• Improve the MEA and in particular the membrane with respect to mechanical properties
• Improve the catalyst support durability in order to minimize the possibility of carbon corrosion in case of e.g. fuel starvation
• Define optimized start-stop & idle mode strategies
• Develop operational strategies for dead-end dry hydrogen operation
• Improve the membrane and catalyst tolerance towards fuel impurities to avoid expensive filters (e.g. ammonia, NOX, soot, salt and other pollutants that may occur in uncontrolled amount).
Main issues assessed and findings upon the AST protocol revision:
• Number of experiments were reduced to ensure higher quality results
• The AST duration was reduced to 200 hours allowing for more cell tests
• The use of fractional factorial experimental design may lead to difficulties with result interpretations (confounding)
• Due to difficulties in interpretation of AST results on single cells and challenges encountered with delays in stable membrane deliveries, stack tests were postponed and limited to those most relevant for actual system operating conditions
Work Package 2 Accelerated Stress Tests on stacks and single cells
Work-package objectives
The objective of this WP was to obtain quantitative data on the impact of stressing conditions by application of AST protocols proposed in WP1 to stacks and single cells.
Summary of significant results from Stack tests
Standard IRD STP0012 stack hardware as described in the D4.2 report was used. The manufacture of 4 stacks with improved MEAs supplied to the project partners is described in the D4.6 report “Fuel cell stacks with improved materials“. The planned stack AST protocols have been outlined in the D1.2 KeePEMAlive delivery report.
Each 10 cell short stack was subject to a specific AST protocol during Month 40-43. Some experiments with hydrogen/air were finalized successfully over the targeted of 1000 hour duration, totalling around 3000 stack test hours. For reformate operation, however, problems were encountered, and stable operation was not obtain within the timeframe of the project.
Continuous operation
The cell performances of IRD stack @ BoT varied between 0.65V and 0.69V @0.4 A/cm² (i.e. 0.26-0.28 W/cm²), corresponding to a RSD of 1.9 % which is significantly lower than the RSD of 6,4 % found for single cells.
Although the relative high OCV cell values @ BoT demonstrated well activated cells, their improvement with time suggested that the break-in procedure of the
stack should be optimized.
The decay rate of the cell stack was higher at low power densities (viz. 0.05-0.2 A/cm²). The main cause of this finding might have been the drying condition.
The worse cell performances were revealed by the cells located to the H2 inlet cooling outlet (Z flow) while the cell located in the middle of the stack was the best one. The better and more stable water balance management offered in the centre of the stack would permit higher and more stable performances of the cells.
Start/Stop protocol
Under start/stop AST protocol; it was evident that the stack became more sensible to water flooding at high current densities. At BoT and 215 h, it was possible to reach 0.8 A/cm² and from 430 h until EoT (594 h), it was impossible. On the other hand, overall performances were stable, even slightly increased was observed over the time.
Comparing the performance of the individual cells, it recognized that cell 9 is the best operating cell (second cell from H2 inlet) while cell 2 the less performing (second cell from the H2 outlet).
In general, the stack voltage degraded ca. 0.9 µV/h @ 0.3 A/cm2 during the first 100 hours of operation. However, after this initial degradation in performance no degradation was observed along the start-stop test for current densities lower that 0.6 A/cm2.
Dead end operation
The dead-end stack test was a straightforward experiment under normal operating conditions as in the Danish demonstration project, where the initial degradation rate was evaluated. The test was not a real AST but served as comparison to the single cell test and the other stack tests in particular to the continuous stack test with pure hydrogen.
Four (4) start/stops due to characterization were planned during the 1000 test hours. However, different unintended events occurred and the total number of start/stops were therefore 14. The unintended stop/starts did not have any major influence on the results. The BoL cell performance was rather uniform. The difference between the cell performances were rather constant for the first 850 hours and slightly increased after this time.
The overall cell voltage decay rate calculated as linear regression on the average cell voltage in operation was 52 µV/h corresponding to 52 mV/1000 hours. However, the voltage decay was not uniform for the cells, the observed decay pattern for the individual cells most of all reflected a higher degradation at the stack ended possible due to temperature issues.
Summary of significant results from Single Cells
The comprehensive AST experimental program comprised more than 20 000 hours of single cell testing, divided on around 60 single cell tests each with duration of 200 – 400 hours. A selection of results is summarised below.
The main findings from AST protocols applied to single cells are summarized as follow:
• Under continuous operation, combination of high temperature, high RH and high current density leads to a rapid failure of the f-940Rf-based cell, while at low temperature, low RH and low current density operation no failure was observed within the 200 h of the test.
• Anode electrode experiences strong degrading effects in single cells tested under reformate operation. There was evidenced that the anode overpotential is the factor dominating the anode degradation, which included Ru dissolution.
• In Fuel Starvation experiments the combination of high temperature (85 °C) and high relative humidity (80%) led to a very short lifetime and high performance decay which is mainly caused by carbon corrosion. High CO and CO2 concentrations in the anode off-gas were measured at these operating conditions. In addition high current density (400 mA/cm²) accelerates the degradation of the MEA.
The formation of pinholes has been proven by hydrogen crossover measurements using a segmented cell. In general, the pinholes had a tendency to appear at areas where the membrane has been previously damaged, also showing a certain cross-over of hydrogen prior to execution of the fuel starvation protocol.
• From performance decay calculated from single cell tested under the load cycling AST protocols, high settings were the most detrimental operation condition. It means that elevated temperature (85 °C) combined with high frequency cycling (6 cycles/hour) at mode 1 (0.2 – 0.6 A/cm2) significantly accelerated the degradation of the cell.
• Comparison between baseline F-940-based cell studied during phase 1 of the project and improved MEAs (reinforced F-940) has revealed a real improvement of cell stability under equivalent ASTs.
The complete results from experimental work following AST protocols have been described and summarized in more details in the deliverables D2.3 (stacks) and D2.4 (single cells).
Work Package 3 Ex-situ characterisation of materials
Work-package objectives
The objectives of WP3 were
• to characterise MEA and stack components at beginning-of-life, end-of-test and end-of life
• to develop ex situ accelerated ageing tests for individual MEA component
• to relate operation conditions to ex situ characterization observations
WP3 was dedicated to the ex-situ characterisation of complete membrane electrode assemblies (MEAs) and their components, in particular the electrocatalysts and membrane materials. This characterisation was performed on pristine materials and MEAs before use in a fuel cell (BoL characterisation), and after ageing (EoT or EoL characterisation) – this ageing having been induced in-situ in the fuel cell, and ex-situ, using accelerated ageing protocols.
Furthermore, WP3 also focused its work on to characterize improved membranes incorporating radical scavengers. Examination by scanning electron microscopy with element mapping of single cell MEAs at end of test and end of life was performed. Likewise, MEAs from stacks subjected to stressing conditions were also examined allowing for identification of the regions of the MEA which were most deteriorated during operation (inlet, central section, and outlet). All in all, this are integral parts to support the main objective of the KeePEMalice project, i.e. to increase the understanding of the degradation mechanisms.
Summary of significant results
• Three types of electrocatalyst from three different suppliers have been characterised at BoL for their metal particle size and particle size distribution. Ex-situ accelerated ageing has been carried out by voltage cycling and under simulated stop/start regime. Catalyst layers have been developed using the various candidate catalysts, and corresponding complete MEAs have been observed in cross-section using SEM. These results and observations have led to a final choice of catalyst material for the anode and cathode of the MEAs.
• Pristine perfluorosulfonic acid membrane materials and forms protected against chemical degradation have been submitted to accelerated ageing ex-situ, using liquid and vapour phase Fenton testing. The membranes have been characterised using vibrational and nuclear magnetic resonance spectroscopies to determine the degradation pathways and how these are modified by the presence of radical scavenger ions. Optimised improved membranes and their components have been characterised by scanning and transmission electron microscopy
• Examination by scanning electron microscopy with element mapping of end of test and end of life single cell MEAs and used MEAs from stacks after application of stress conditions allowed identification of the regions of the MEA most subject to modification. As one example, fuel starvation was shown to lead to significant loss of ruthenium from the anode (and transfer to the cathode) in the region of the gas outlet, compared with the beginning of life MEA.
• Electrochemical characterisation of catalysts clearly demonstrated better performances for Tanaka Kikinzoku Kogyo (improved material) over Cabot catalyst (baseline material), independently of the stress test applied. Thus, this catalyst was used as MEA component for manufacture of MEAs to be used in second phase of WP2 in situ aging tests.
• Aged MEAs clearly showed variations in composition relatively to fresh assembly along the cross-section depending on the different locations examined (outlet or inlet).
• Ex-situ bipolar plate corrosion appears negligible, even in electrolyte containing fluoride ions. This result allowed the re-use of stack bipolar plates of WP2 between the several tests.
• Application of ex-situ ageing protocols to pristine PFSA membranes, and to membranes modified by incorporation of radical scavengers. The effect of radical scavenger ions on reducing fluoride emission was characterised in-situ (WP4) and ex-situ using 19F MAS NMR spectroscopy to follow local structural changes to the perfluorinated polymer backbone and side-chain.
• Incorporation of Ce and Mn ions decreased the chemical degradation imposed by Fenton reagent (Fe/H2O2). In each case the spectrum of the Ce ion exchanged membrane was similar in peak height and width to pristine Nafion-212, in the contrast with other degraded samples. Moreover the biggest tendency to enlarged peak width was observed for the peak at 118 ppm (side chain SCF2 group), which indicated side chain scission as being a major consequence of oxidative and free radical attack. These results could be directly related to the observations made in-situ on the lower fluoride emission rate from MEAs containing Mn and Ce partially exchanged membranes.
• A range of complementary techniques were used to characterise nano-particulate cerium oxide prepared by three different approaches, in order to determine whether the synthesis method affects the proportion of Ce(III) in the material, and to select the method reproducibly providing mono-disperse CeO2. The final composite membranes comprising a surface layer of nano-fibre PFSA/CeO2 were characterised for their proton conductivity and observed using SEM/EDX analysis to confirm that the cerium oxide distribution was satisfactorily distributed across the membrane surface.
• SEM/EDX has been used to analysis MEAs that had been submitted to AST in WP2. The results on MEAs having undergone fuel starvation stress were particularly noteworthy, since clear correlation has been found between the applied stress conditions at high RH and Ru depletion at the hydrogen outlet region of the anode catalyst layer, and Ru depletion at both the inlet and outlet regions at low RH. The anode catalyst layer composition at the inlet was therefore a sensitive indicator of the combined effects of fuel starvation and the degree of hydration of the MEA on anode catalyst corrosion. Under these conditions, cell reversal effects leaded to high anode potential. Cell reversal resulted in the production of oxygen instead of hydrogen oxidation at the anode through the oxidation of water. Ru is unstable at these potentials, and dissolves into water either fed or produced at the anode (cell reversal), and is carried into the exit water stream or could diffuse into the membrane. Indeed, although Ru could not be detected by EDX analysis across the membrane thickness, it wasx detected in the catalyst layer at the cathode (in the pristine MEA only Pt is detected at the cathode), although there was no clear relation between the amount of Ru detected and the applied AST conditions in terms of applied current density, temperature or RH; instead, a possible correlation appeared between the number of starvation/recovery cycles, and the amount of Ru in the cathode catalyst layer.
• Furthermore, there was clear correlation between Ru depletion and the thickness of the anode catalyst layer. The layer thickness was ca. 5 µm in pristine MEAs and in the anode catalyst layers of MEAs with a Ru/Pt ratio of 1, and it decreased progressively as Ru was leached. These observations provided input to WP5 and were entirely consistent with those described from ex-situ analysis of the cyclic voltammograms on start/stop cycling (see deliverable report D3.5).
• The ex-situ electrochemical characterisation of improved materials showed that PtCo/C catalyst presented both higher electrochemical stability and ORR activity as cathode material in comparison to the state-of-the-art Pt/C sample.
Work Package 4 Preparation and improvement of MEAs and components
Work-package objectives:
• To provide state of the art membranes, electrodes, MEAs and stacks for WP2.
• To develop and provide improved individual materials, MEAs and stacks with higher durability.
In terms of material development activities, the main objective of WP4 was to manufacture an improved membrane material by continuous production, and assure its compatibility with the improved electrode materials in use, as well as to select the best suited catalyst material for MEA fabrication, based on the initial results.
Summary of significant results
• New reinforced chemically stabilised membrane that demonstrated significantly higher durability compared plain membrane’s film without stabilisation (i.e. w/o radical scavengers) was developed by Fuma-Tech. The stabilisation of membrane was based on reinforcement, which has impact on mechanical properties. The chemical stabilisation was based on interaction of inorganic-organic composite with PFSA matrix and which yielded homogeneous distribution of stabilisation through the whole volume of membrane.
• With the aim of producing reinforced chemically stabilized membrane in a mass production, the knowledge about stabilization of mechanical and chemical nature of PFSA membrane were combined. After comprehensive search, a unique combination of solvent mixtures was identified and membrane called as F-940rf and later as F-940rfs was mastered and produced as a roll. This strategy demonstrated the production of a volume of at least 20 m2 as a roll.
• Preparation of PFSA membranes incorporating a gradient of radical scavenging oxide by electro-spinning, giving a composite CeO2/PFSA nano-fibre mat was achieved. These electrospun mats were the means of enabling the CeO2 radical scavenger to be specifically incorporated sited in close proximity to one or other catalyst layer, rather than distributed throughout the membrane. The incorporation of an additive specifically sited at a predetermined location within the membrane was complementary to the reinforcements incorporated in KEEPEMALIVE membranes that improved the membrane mechanical strength. This route provided a further means of increasing durability of the membrane and complete MEA. The release of fluoride from the MEA on fuel cell operation was used as a gauge of success of this approach. MEAs incorporating modified membranes showed a strongly reduced fluoride emission rate (FER). Thus, the resulting membranes increased its chemical stability and showed significantly improved lifetime under accelerated stress testing. In this way, the MEA lifetime was increased from <200 h to >800 h by incorporation of this radical scavenging interlayer under OCV conditions. These results validated the approach developed in the project by CNRS and provided for this group a key result that will be used in its further research and development.
• Four (4) 10-cell stacks with improved MEAs have been manufactured. These stacks were equipped with a novel MEA type where two new precursors have been implemented. The novel precursors implemented were the reinforced membrane (F-940rf) developed by Fuma-Tech and the very stable Tanaka cathode catalyst (TEC10F60TPM). The novel MEAs required development of a new prolonged initialisation (break-in) procedure. However, the MEAs performed well once adequately conditioned.
• Adopting of latest generation of FuMA-Tech’s membrane leads to improving of lifetime of MEA’s to level that is between 10 000 – 20 000 operation hours at conditions governing stationary fuel cells.
Work Package 5 Design and evaluation of experiments
Work-package objectives
• Identify critical operating conditions for PEM fuel cell stacks, based on screening of single cells, actual testing of stacks, field tests, modelling and experimental assessment of local conditions in a stack
• Based on statistical analysis of data, in combination with fundamental knowledge of degradation mechanisms, propose new accelerated stress tests (AST) for PEM fuel cells
• Develop lifetime prediction models based on theoretical and phenomenological models, as well as the results of the statistical analysis.
The activities of this WP were directed towards developing tools for the interpretation of experimental data, and thus contribute both to an increased understanding of the degradation phenomena in fuel cells, and to propose AST protocols closer to realistic operating conditions.
Summary of significant results
• Experimental results from the AST protocols conducted during the comprehensive AST program were analysed statistically using the Yates algorithm to obtain main effects and interactions and their contributions to the degradation rates. Effects of temperature, relative humidity and current density on the degradation rate were evaluated for the "Continuous operation" protocol, the "Load Cycling" protocol and the "Fuel starvation" protocol.
• Fuel Starvation for single cells is the most comprehensively studied AST protocol in this project. Initially, the results from a fractional factorial design (23-1) were statistically evaluated and reported in D5.2. The analysis revealed that second order effects were of a significant magnitude. Hence, the pre-requisites for use of fractional factorial design were not fulfilled and this resulted in confounding of main and second order interactions. To be able to distinguish main effects from interactions, a full 23 experiment was subsequently performed in phase II of the AST program. The results were supplemented by post mortem ex-situ characterization of the MEAs in WP3, and reported in D3.6.
• Despite the immense experimental effort put down by all project partners totalling in the range of 25 000 hours' testing, the statistical evaluation of the data in WP 5 is not revealing reliable and significant relationships making it possible to generate a Life Prediction Model, as depicted in the Description of Work. This is seen as a deviation from the original plan and was therefore conveyed already during the Mid-Term Review (M33). However, we have gained a lot of experiences from our substantial experimental AST-program. Dissemination of these Key Achievements was realised in a merged public Deliverable (http://www.sintef.no/keepemalive) representing D5.3 Life Prediction Model and D1.5 Design Guide, entitled: "Experiences from a comprehensive PEMFC degradation study as basis for Lifetime Prediction Modelling and a Corresponding Design Guide for PEMFC degradation studies."
• Modelling work was carried out in order to support the interpretation of experimental data along two routes: (i) application of a single cell model for assessment of local conditions along a channel/plate (JRC), in line with experimental data carried out with a segmented cell and with local reference probes and (ii) modelling of Pt degradation mechanisms in line with ex-situ catalyst characterisation (SINTEF). Regarding (i), a 3D model was constructed, including transport processes in the gas phase and the solid phase, reaction kinetics, as well as gas solubility in condensed water. The model is a two-phase model, assuming fine mist in convective flow. Results have been used for determination of temperature gradients and hot regions, as well as conditions leading to non-uniform current distribution, thus causing fuel starvation or low O2 concentration. Regarding (ii), ex-situ catalyst tests of Cabot catalysts from WP3 (cycling between 0.9 and 1.2 V) were described by a model describing chemical dissolution of PtO, in accordance with models provided in the literature , . The model provided a qualitative explanation for the dependence of the ECSA loss on cycling rate. Furthermore, the model reproduces the slightly non-linear behaviour of the ECSA loss vs. time, which is related to the size-dependent surface tension of the rate constant.
• Calculated effect using the decay rates from samples exposed to Fuel Starvation AST protocol showed that the main effects of T and RH are large and negative. The effect of j does not appear to correlate with performance decay. The impact of RH is much higher at high T than at low T (-31 vs. -3.6). It implies from the fact that T*RH is equal to RH*T, that the impact of T is higher for high RH. However, making assessment of the magnitude of T and RH main effects is difficult when large interaction is documented between the factors. It was plausible that T and RH are coupled in PEMFC. It is also important to recall that a prerequisite for the 23 design is orthogonally: that all factors are completely independent of each other.
Despite the failure in the establishment of a tool for lifetime prediction, WP5 has generated a lot of valuable information regarding assessment of lifetime of PEMFC systems. This know-how was processed and briefly summarised as follow:
1. At the end of the break-in period of a cell, the variance between cells has been calculated. All effects observed must be larger than this variance in order to be statistical significant.
2. The variance for several of the characterization techniques used in this project (i.e. cyclic voltammetry, electrochemical impedance spectroscopy, hydrogen cross-over current) has been established. Knowledge of requirement for significance of effect has been also established.
3. Some parallel experiments have been conducted in order to evaluate variance between experiments beyond beginning of test.
A systematic approach using statistically designed experiments (factorial design) was used, as described in the DoW. To get maximum information out of the factorial designed experiments, complete sets of so-called 23 experiments (3 factors, each at 2 levels = 8 different operation conditions) were planned at each partner's test facilities. During the execution of those set of experiments, however, it turned out that many of the selected sets of operation parameters led to instable operation (e.g. flooding or dehydration) and several of the 6 distinct AST protocols were not executable. The foreseen statistical analyses to reveal the factor's effect on degradation were, hence, not possible due to incompleteness of the full sets of eight experiments.
Difficulties encountered during the initial iterations of statistical analysis of the data from 37 single cell tests carried out in phase one of the AST program (Month 1-25), lead to the conclusion that only qualitative relationships between factors and degradation could be identified.
Therefore, it was concluded that during the second phase of the AST program (Month 30-42) complete sets of factorial designed experiments were to be executed.
Assessment of real stack performance data
Stack performance data from the Vestenskov field tests in Denmark was assessed statistically and results are reported in this section. Through the availability of real life stack data at sufficient resolution (1 Hz logging) and with availability of individual cell voltages, competence in evaluating large data sets has been obtained:
1. Import and storage of data in a useable format for evaluation.
2. Efficient smoothing of data, removing noise, yet retaining valuable information.
3. Extraction of subsets (i.e. transients) for evaluation of degradation rates.
4. Evaluation of population distributions.
5. Fitting of data.
Statistical evaluation of real life stack data has given evidence of a change in cell population distribution from being normal at beginning of life to become better fitted by a Weibull distribution later in stack life. This is an important finding, previously assumed in literature that could be utilised for prognostic purposes (Remaining Useful Life estimate) as well as lifetime prediction.
In order to assess the precision in data used for calculation of degradation rates and lifetime prediction, more than one measurement is required. Optimally, a population of identical stacks running under identical ambient and operational conditions should be available for statistical evaluation. This is rarely the case. For a single stack, the individual cell voltages can be utilized in order to evaluate the variance. More interesting than the average cell voltage and its corresponding standard deviation is the distribution of the population.
Stack geometry will affect the distribution of individual cell voltages. From a design point-of-view, it is of interest to evaluate individual cell degradation rates in order to verify stack design. From evaluation of individual cell voltages in the stack, deviation in the end cells were observed. Therefore, the two outermost cells at each end of the stack were excluded from further cell voltage evaluation. So, in this particular exercise, 43 cells were used to evaluate the population distribution. In order to improve distribution assessment, 1 Hz data was used where 200 seconds of individual cell voltages were used to create voltage populations. At Beginning of Life (BoL), the distribution is best described normally.
The probability density is better described by a normal distribution, symmetric about the mean value. The goodness of fit can be evaluated numerically e.g. by a log likelihood parameter. It was postulated that a check for normality in single cell voltage distribution for a running stack could say something about the state of health for the stack. A normal Chi-Square test was used to evaluate the "degree of normality"
There are many ways to estimate the performance degradation rate of a fuel cell stack. The estimates obtained are strongly dependent of which and how data are fitted. Some reversible (recoverable) degradation effects are typically encountered (e.g. flooding/dehydration/CO-poisoning), and these should preferably be separated from irreversible degradation when calculating degradation rates.
Potential Impact:
Impact
Assessment of the field data from real life operations (at Lolland) related to the KeePEMalive project has enabled the project system development partner IRD to improve the µCHP system and take the technology one step closer to fulfil the stringent requirements for long term durability. By exchanging some MEA precursors and further optimise the operational conditions in the Danish on-going field test the MEA durability was increased significantly e.g. the degradation rate was decreased five-fold from 20 to 4 µV/h, corresponding to increasing the system lifetime from the previous level of 3 500 hours to an expected 17 000 hours (~2 years). The heat and electricity demand and the related energy and emission savings from utilizing the CHP-units in Danish households have been mapped for various seasons and during the course of the project the electric system efficiency has been improved to 50%.
Dissemination activities
Dissemination activities have had a high focus in the project and the consortium acknowledges the importance of promoting the development of fuel cell technologies for sustainable and efficient utilization of hydrogen as energy carrier. Our public webpage (www.sintef.no/keepemalive) contains a complete list of dissemination activities.
During the KEEPEMALIVE project development of new materials and concepts has been the focal point. Several novel solutions have been implemented in the related field test program at Vestenskov, Denmark, at which in total 32 PEM fuel cell based µ-CHP systems have been installed in households. This again is reflected in the dissemination plan, where one paper is already published, two manuscripts are submitted and another 6 scientific papers based upon the results obtained in the project are planned for publication in the coming year, as listed below:
1 Characterisation Protocols of Stationary Type of Fuel Cells Focused on Prediction of Lifetime Tomas Klicpera The Journal of Fuel Cell Technology Vol. 11, 2013 Special Issue FCDIC Tokyo, Japan 2013 pp. 12- 31 www.fcdic.com
2 Study of the transport of radical scavenger in PFSA membranes and effect of their loading on membrane properties. M. Zaton J. Membr. Sci. Elsevier Manuscript submitted
3 Significant lifetime improvement of PFSA MEAs M. Zaton Electrochem. Comm. Manuscript submitted
4 Parametric study of the effects of cell reversal during fuel starvation on fuel cell performance loss and anode electrocatalyst degradation Victor Hacker Electrochimica Acta Elsevier Manuscript submission in 2013
5 PEMFC fuel starvation degradation evaluated by experimental design Thor A. Aarhaug Journal Power Sources Elsevier Manuscript submission in February 2014
6 Improvement of testing techniques for characterisation techniques for the localisation of MEA failures Astrid Hofer TBD
7 Stability of PEM fuel cell cathode catalyst under potential cycling conditions. An ex-situ AST characterization approach Luis Colmenares Journal of Power Sources Elsevier Manuscript submission in 2014
8 Tailoring the catalytic activity of PtRu/C catalysts by potential cycling Luis Colmenares Electrochimica Acta Elsevier Draft available, Manuscript submission in 2013
9 Experimental strategy for PEMFC durability assessment Steffen Møller-Holst International Journal of Hydrogen Energy Elsevier Manuscript submission in April 2014
Significant results from the project have continuously been disseminated to the scientific community through more than 20 oral presentations and 7 posters, some at major international conferences such as WHEC and ECS meetings as well as seminars and workshops targeted to a wider community of private and public stakeholders.
Moreover, an International Summer School on advanced studies of Polymer Electrolyte Fuel Cells, was held at TU Graz, August 26th – September 1st, 2010. The Summer School was organized in co-operation between Graz University of Technology, Austria (Dr. Viktor Hacker) and Yokohama National University, Japan with financial support of the European Commission (Seventh Framework Programme, FP7/2007-2013, Fuel Cells and Hydrogen Joint Technology Initiative under grant agreement n° 245113) and the Cost Action 543, Research and Development of Bioethanol Processing for Fuel Cells. The Summer school was attended by 50 students. A lecture was given by the Coordinator, Gaby Janssen, on the Lifetime and Durability issues in PEMFC and how KEEPEMALIVE tries to address these issues.
Last but not least, to reveal synergies and exchange high level knowledge in Europe, a joint Workshop entitled "Degradation of PEM Fuel Cells - experience exchange & discussions", was arranged at SINTEF in Oslo, Norway as a collaboration between 3 FCH JU-supported projects: Stayers, Premium Act and KeePEMalive, April 3rd and 4th 2013.
List of Websites:
Project website address:
http://www.sintef.no/keepemalive
Project's coordinator :
Steffen Møller-Holst
Vice President Marketing,
Hydrogen and Fuel Cells
SINTEF
Tel: + 47 92604534
Fax: + 47 73591105
E-mail: steffenh@sintef.no
