tem-CHAPTER 5. SIMULATIONS 5.7. SUB-CONCLUSIONS
perature. This is in direct disagreement with what is exhibited by the data. This behaviour is attributed to the fact that the diffusion losses in the model originate from gas phase diffusion only. A way to remedy this shortcoming of the model could be to change the catalyst model to make acid phase diffusion more of a limiting factor. An agglomerate model would accomplish this by increasing the length and tortuosity of the diffusion path in the acid phase
To sum up, the model is capable of fitting the behaviour of the of the polar-isation curves and the impedance spectra at one temperature, giving a seemingly reasonable estimate of the parameters governing the performance. This capab-ility could potentially be applied in analysing cells that have been subjected to degradation testing. By comparing the model fit to a data set recorded at the beginning of the test and one recorded at the end of the test, it would be possible to quantify what kind of degradation the cell has experienced. By making changes in the fixed parameters, the model might also be able suggest a path for improv-ing the MEA performance by alterimprov-ing the catalyst layer composition or similar.
The sub-par performance of the model in terms of temperature dependence will, however, have to be improved before such investigations can be performed with sufficient credibility. This will be the subject of future work.
Chapter 6
Conclusions
This chapter summarises the conclusion of the work performed through this Ph.D.
project and suggests a path to applying the results in future work.
6.1 Conclusions
The investigations conducted in this work have explored different tracks to im-prove the understanding of HTPEM fuel cells and their workings using electro-chemical impedance spectroscopy. The work can roughly be divided into two parts. The first part is the mainly experimental approach where EIS has been employed together with equivalent circuit models to analyse the effects of anode feed CO content and changes happening in the cell during break-in. The second part of the investigation describes the development of a mechanistic model of the steady state and impedance characteristics of HTPEM fuel cells and the evalu-ation of the ability of the model to fit and predict experimental polarisevalu-ation and impedance data.
In the first experimental investigation, the influence of CO content in the fuel stream of an HTPEM fuel cell using a commercial Celtecr-P2100 MEA was investigated using EIS and polarisation curves. The addition of CO was shown to have an effect on the whole impedance spectrum, challenging the common conception of rigid division of the impedance spectrum into distinct contributions.
The effects on the spectra were explained by different mechanisms. An increase in the ohmic resistance of the cell was assumed to derive from deactivation of the catalyst closest to the membrane. Lower active area and increased diffusion losses on the anode side due to deactivation of readily available catalyst were assumed to contribute to the capacitive loops in the impedance spectra.
The investigation into the galvanostatic break-in process of Celtecr-P sol-gel and Dapozolr 77 post-doped HTPEM MEAs revealed significant differences in the development of voltage and impedance spectra during the break-in period.
6.1. CONCLUSIONS CHAPTER 6. CONCLUSIONS
Comparison of the break-in of sol-gel and post-doped MEAs has not previously been published. The changes in the impedance was quantified using an equivalent circuit model. Three different models were tested, and the most simple model was chosen, since it produced the most consistent fitted values. The sol-gel MEAs exhibited the most significant changes in both voltage and impedance. For both types of MEAs, the rate of change was most pronounced in the initial stage of the break-in after which the rates would stabilise. The change rates stabilised within the first few hours for the post-doped MEAs, while the sol-gel took around 30-40 hours. In both cases, the results indicate that break-in times significantly shorter than the 100 hours recommended for Celtecr-P MEAs are feasible.
A common conclusion from the work involving equivalent circuit models relate to the models themselves. Particularly the test of three different models for the break-in study illustrated the ambiguity of equivalent circuit models. This leads to the suggestion of an equivalent circuit model with a minimum of degrees of freedom which should be able to account for all major contributions in an HTPEM fuel cell impedance spectrum.
The mechanistic modelling part of the project has, through a number of by-ways, resulted in a 1+1D finite volume based model, considering only the cathode side. The emphasis of the model is the interaction between the local phosphoric acid concentration in the catalyst layer and the electrode kinetics and transport phenomena. The dependence of oxygen solubility, diffusivity, and the cathode exchange current density on acid concentration is modelled in more detail than is usually employed in HTPEM models.
The model was fitted to a set of polarisation curves and impedance spectra assuming different combinations of fixed parameters. Different values of the re-action order and the transfer coefficient were tested. Changing the catalyst layer conductivity model was also investigated as well as the influence of the reaction mechanism. The results of fitting the individual cases reveals the complexity of the task of making an advanced fuel cell model fit impedance spectra and po-larisation curves simultaneously. It also illustrates the issues with determining criteria for a good fit, since some of the fits that are best in terms of the value of the objective function, has issues with either the behaviour of the simulated curves or with fitting parameters converging to seemingly unreasonable values.
The investigation underlines the strength of combining impedance spectra and polarisation curves, since the parameter combination that provides the best fit to the polarisation curves, provides a comparably worse fit to the impedance spectra.
This indicates that this parameter combination is not, after all, representative of the actual state of affairs. The fitted model is used to investigate the effects of the flow dynamics of the cathode gas on the impedance spectrum. The low frequency part of the impedance spectrum can be attributed to this phenomenon.
The model is reasonably capable of reproducing the effect of the DC current on the impedance spectra. The temperature dependence is, however, not sat-isfactorily reproduced by the model. The power of combining the two types of