% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@INPROCEEDINGS{Kuppa:1044516,
author = {Kuppa, Raman Ashoke and Hammacher, Linus and Karyofylli,
Violeta and Jodat, Eva and Karl, André and Eichel,
Rüdiger-A.},
title = {{D}ata-driven approach for modeling and sensitivity
analysis of a {P}roton-exchange membrane water electrolyzer},
reportid = {FZJ-2025-03253},
year = {2025},
note = {Acknowledgment: Financial support was provided by the
German Federal Ministry of Education and Research (BMBF)
within the H2Giga project DERIEL (grant number 03HY122C)},
abstract = {Proton exchange membrane electrolytic cells (PEMEC) have
emerged as a promising technology to produce environment
friendly green hydrogen. These are complex systems which
involve a highly non-linear interplay between mass transfer,
fluid flow and electrochemical reactions. To improve the
commercial viability of PEMECs, performance optimization of
these systems using data-driven models has emerged as a
promising approach [1]. In this work, we develop data-driven
surrogate models using synthetic data obtained from a
physics-based one-dimensional numerical model of PEMEC [2].
The predictive performance of three unique machine learning
algorithms [3] (support vector regression (SVR), extreme
gradient boosting (XGB) and artificial neural networks
(ANN)) were evaluated and their ability to capture the
inherent non-linearities of PEMEC were compared. Important
transport properties of the system such as operating
temperature, anode catalyst and porous transport layer
thickness, electronic conductivity and membrane thickness
were selected as the input features for developing the
data-driven models. Both XGB and ANN models showed better
performance in predicting the cell current density when
compared to SVR model. The ANN model was deployed to conduct
a parametric analysis to investigate the effect of operating
conditions and transport properties. Furthermore, an
explainable artificial intelligence technique known as SHAP
(Shapely Additive Explanations) was employed to list out the
important parameters influencing the cell current density.
The SHAP analysis highlight that the influence of membrane
thickness is higher than the electronic conductivity for
supported catalyst layers, and conversely, for unsupported
catalyst layers.},
month = {Mar},
date = {2025-03-11},
organization = {21st Symposium on Modeling and
Experimental Validation, Karlsruhe
(Germany), 11 Mar 2025 - 12 Mar 2025},
subtyp = {After Call},
cin = {IET-1},
cid = {I:(DE-Juel1)IET-1-20110218},
pnm = {1231 - Electrochemistry for Hydrogen (POF4-123) / HITEC -
Helmholtz Interdisciplinary Doctoral Training in Energy and
Climate Research (HITEC) (HITEC-20170406)},
pid = {G:(DE-HGF)POF4-1231 / G:(DE-Juel1)HITEC-20170406},
typ = {PUB:(DE-HGF)6},
url = {https://juser.fz-juelich.de/record/1044516},
}