Unsupervised domain-adaptive Transformer for remaining useful life prediction of proton exchange membrane fuel cells under dynamic ripple-current ageing
Description
This dataset contains the processed data and reproducibility files used for a steady-to-dynamic transfer study of proton exchange membrane fuel cell remaining useful life prediction. The processed records are derived from the public FC1/FC2 ageing data in the IEEE 2014 Data Challenge archive. FC1 is used as the labelled steady-state source trajectory, and FC2 is used as the dynamic ripple-current target trajectory. The upload includes the processed FC1 table (1155 rows), the processed FC2 table (1021 rows), window metadata (348 windows), the window tensor file (X shape (348, 48, 30)), model-comparison metrics, ablation metrics, source-dataset metadata, and the experiment script. The files are intended to support reproducibility of the associated Energy Reports manuscript. The raw FC1/FC2 data should be cited through the original Loughborough University dataset DOI, https://doi.org/10.17028/rd.lboro.3518141.v1. The present upload provides the processed tables, windowed arrays, metrics, and code used in the manuscript workflow.
Files
Steps to reproduce
Install the Python packages listed in requirements.txt. The raw FC1/FC2 ageing records should be obtained from the original Loughborough University dataset at https://doi.org/10.17028/rd.lboro.3518141.v1. The processed FC1 and FC2 tables, window metadata, window tensor file, model metrics, and ablation metrics are included in this upload. To regenerate the processed files and reported metrics, place the original IEEE 2014 Data Challenge archive in the data location expected by code/fclab2022_uda_experiment.py and run that script in a Python environment with the listed dependencies.