Sparse Electrochemical Impedance Spectroscopy Measurements during Solid Oxide Cell Stack Operation Based on Power Supply

ECS Meeting Abstracts, vol. MA2025-03, pp. 290

Abstract

At Forschungszentrum Jülich a set of over 2600 electrochemical impedance spectroscopy (EIS) measurements, conducted on 15 individual solid oxide cell (SOC) stacks was consolidated. The stacks were subjected to varying operational modes and conditions, but consistently for the purpose of degradation research. Material-wise the stacks comprised similar properties utilizing fuel electrode supported cells by the commercial manufacturer Elcogen (Tallinn, Estonia) and the in-house F10-design. 1 The EIS measurements conducted during these experiments were uniformly pre-processed with a specifically designed curation procedure to compensate for recurring instabilities and parasitic inductive impedances. 2 Dominant patterns providing a low-dimensional representation of every EIS measurement could be identified by singular value decomposition of the EIS data set. 3 These information-rich patterns, referred to as eigenspectra, facilitate efficient data handling in terms of data exchange and the training of data-driven models. Furthermore, the eigenspectra can be used as a reconstruction basis U r in which every EIS measurement in the data set corresponds to a unique r -dimensional representation. 3 A straightforward advantage of this property is the usage as a filter to compensate for process-related noise during EIS by calculating the corresponding reconstruction of such an EIS measurement (cf. Fig. 1-A). Process-related noise during EIS can arise from fluctuations in the steam supply, induced by the steam generator, for instance. A less obvious advantage is the identification of a set of tailored frequencies f* from the eigenspectra which allows the reconstruction of complete EIS spectra from sparse measurement of only these selected frequencies f* (cf. Fig. 1-B). 3 The accuracy of such approaches sensitively depends on the rank r of the reconstruction basis , i.e. the number of eigenspectra considered. Common methods from the literature were used to find an optimal value for r . Depending on the measurement conditions, accurate reconstructions can be reported for r = 6...12 . Due to the theoretically separate measurement of the real and imaginary parts of the impedance, the number of frequencies required for the reconstruction of whole EIS spectra from sparse measurements could be reduced to |f*| = r/2 . In addition, the frequencies f* were found to coincide with typical ranges of distribution of relaxation times (DRT) peak-to-process-attribution, supporting the physical relevance of the identified tailored frequencies. To verify the proposed method for application in future experiments it was validated for held-out measurements conducted on an additional SOC stack with similar material properties. Thus, accurate reconstructions of EIS measurements conducted at 700°C in dry hydrogen can be reported for the sparse measurement of only three frequencies. Furthermore, the reconstruction of EIS from sparse measurements was investigated for EIS containing known faults such as temperature and feed gas flow induced drifts during the measurement as well as for measurements exhibiting a degradation of electric contact within a repeating unit of the stack and an external leakage. Though six instead of three tailored frequencies are necessary, accurate reconstructions could still be determined even under the adverse conditions. Despite being underrepresented in the data set, the reconstruction of EIS measurements conducted under load, i.e. with high DC offsets of the excitation current up to 1 Acm -2 , was found to be accurate when employing the identified three to six tailored frequencies f* , depending on the measurement conditions. These findings suggest the potential application of the developed method as a time efficient alternative for conventional EIS measurements during operation of SOC stacks including productive operation. We hypothesize that this should enable simplified EIS measurements during load operation to be carried out at significantly shorter time intervals. This would considerably facilitate data-driven modelling for degradation prediction by using the sparse EIS measurements for regularly updating the prediction model. However, it is important to note that the relevance and accuracy of the reconstructed EIS measurements are contingent on the similarity of the SOC stacks comprised in the underlying data. 3 As a technically favorable and cost-efficient solution the use of a power supply instead of dedicated EIS hardware for the implementation of EIS is proposed. Such an experimental setup has already been investigated simulatively for an alkaline electrolyzer, by utilizing the IGBT-based power converter. 4 In the present study the impedance is determined only at those frequencies f* previously identified as optimal, and the entire spectrum is reconstructed based on the described method. This approach has the potential to significantly minimize the measurement effort and enhance resilience against instationarities in temperatures and gas flows during the measurement. For this purpose, the modulation of suitable excitation signals is tested using the arbitrary waveform generator in a CAENels FAST-Bi-1K5 power supply. Together with the embedded oscilloscope three to six tailored frequencies f* are measured utilizing only the power supply unit and EIS spectra are reconstructed from this sparse measurement. The application of the proposed approach is verified for an SOC stack with in-house F10-design 1 and fuel electrode supported cells by the manufacturer Elcogen. The obtained measurements are compared to validation measurements conducted with a Zennium X potentiostat, power potentiostat PP212 and EL1002 load devices from the manufacturer Zahner Elektrik (Kronach, Germany). To date, the impedances at the tailored frequencies f* from conventionally measured EIS have been utilized for validation purposes, and the resulting reconstruction has then been compared to the conventionally measured EIS. Due to the preconditioning of these EIS measurements and the curation of parasitic inductive impedances, measurement-induced artefacts were not limiting the reconstruction capabilities. However, with the proposed experimental setup, such instabilities and parasitic inductances may now have to be taken into consideration, as the prior curation of such a sparsely measured EIS is not a viable option anymore. Consequently, the determination of the tailored frequencies f* is also carried out for the original EIS data set without prior curation. A preliminary investigation indicates a significant influence on the distribution of those frequencies f* due to curation of the underlying data set. As exemplified in Fig. 2 for a subset of 734 EIS measurements conducted at 700°C in dry H 2 both with and without prior curation, the frequency ranges yielding an optimal reconstruction quality partially deviate. This affects the proposed reconstruction method as the tailored frequencies f* are determined as the mean of the depicted distributions. Therefore, the curated and the original data set are used to determine the tailored frequencies f* and resulting reconstruction qualities for EIS measurements with and without parasitic, inductive impedances and measurement instabilities are compared. In addition, feasible extensions of the data set are proposed to facilitate high reconstruction accuracies in all scenarios. Thus, potential avenues are elaborated for integrating sparse EIS measurements at stack level and during productive operation into existing test benches, thereby ensuring a cost-effective and time-efficient diagnostic method. Acknowledgement The authors would like to thank their colleagues at For

Authors 4

1. Christian Mänken Aachen Institute of Physical Chemistry

   Forschungszentrum Jülich · RWTH Aachen University

   Affiliation as printed

   Inst. of Energy Technology (IET-1), Forschungszentrum Jülich GmbH

   Institute of Physical Chemistry, RWTH Aachen University

2. Dominik Schäfer

   Forschungszentrum Jülich

   Affiliation as printed

   Inst. of Energy Technology (IET-1), Forschungszentrum Jülich GmbH

3. Remzi Can Samsun

   Forschungszentrum Jülich

   Affiliation as printed

   Inst. of Energy Technology (IET-1), Forschungszentrum Jülich GmbH

4. Rüdiger‐A. Eichel Aachen Institute of Physical Chemistry

   Forschungszentrum Jülich · RWTH Aachen University

   Affiliation as printed

   Forschungszentrum Jülich, IET-1: Fundamental Electrochemistry

   RWTH Aachen University, Institute of Physical Chemistry