Optimal Experimental Design for Fuel Cell Model Parameter Identification – IEEE Access Reproducibility Release

Overview

This release contains the source code, configuration files, and data required to reproduce the results of the article:

L. M. Winkler, M. A. Danzer, H. Palm,
“Optimal Experimental Design for Fuel Cell Model Parameter Identification,” IEEE Access, 2026.
[DOI inserted once available]

The code implements the proposed optimal experimental design (OED) workflow for parameter identification of a PEM fuel cell model, including Fisher information–based design and comparison against Latin Hypercube Sampling (LHS) strategies.

Contents

• Core OED framework implementation (PEMFC model, FIM computation, virtual experiments, and design optimization)
• Simulation and evaluation scripts to generate the main figures and tables in the paper
• Example configuration files for the zero-dimensional PEMFC model with six unknown parameters
• Processed data / result files used for evaluation

Reproducibility

The following Jupyter notebooks in examples/ reproduce the key steps and results of the IEEE Access article:

• 06_calculate_LH_estimation_demo.ipynb
  Creates an initial Latin Hypercube Sampling (LHS) design and performs the corresponding parameter calibration for the PEM fuel cell model.

• 07_calculate_pi_designs_demo.ipynb
  Performs the parameter-individual optimal experimental design (pi‑OED), and visualizes the resulting experiments and their Fisher Information.

• 08_calculate_pi_estimation_demo.ipynb
  Calibrates the model using the pi‑OED experiments and visualizes the resulting parameter distributions.

• 09_estimate_LH_progression_demo.ipynb
  Evaluates and visualizes the practical identifiability of the parameter set for different LHS configurations (progressive LHS refinement).

• 10_calculate_measurement_error_demo.ipynb
  Computes the measurement error statistics and distributions from a real experimental dataset.

• 11_plot_results_demo.ipynb
  Aggregates the simulation and estimation results from the previous notebooks and reproduces the main results and figures reported in the publication in a single notebook.

The real experimental data used to estimate the measurement error distribution is provided in:

• data/stats/2025-11-19_BZ1508_m282_V2_023C_TP.csv

Some optimization routines used within the OED workflow have been modularized into a separate support library for fuel cell optimization. The corresponding release can be found at oed_fcs. Please refer to the README.md for detailed setup instructions, dependencies, and step‑by‑step commands.

Versioning

This release (v1.0.0) corresponds exactly to the code used to generate the results reported in the IEEE Access article above.

Future updates, improvements, or extensions will be released under subsequent version tags (e.g. v1.1.0, v2.0.0) and may not reproduce the published results bit‑for‑bit.

Full Changelog: https://github.com/BartgeierXC/FuelCellOED/commits/v1.0.0-ieee-access