{"id":"d6ftbf8pyj","doi":{"id":"10.17632/d6ftbf8pyj.3","status":"allocated","prefix":"10.17632"},"name":"Derived data and code for target supervision budgets and update timing in PEM fuel cell voltage forecasting","description":"This dataset contains derived data, research code, model weights, frozen protocol records, figure and table source data, and reproducibility audits supporting the manuscript \"Target supervision budgets and update timing in PEM fuel cell voltage forecasting.\"\n\nThe protocol evaluates future hourly cell voltage at a primary forecast horizon of 24 h and auxiliary horizons of 6 h and 48 h. It compares initial target supervision and subsequent update policies under recorded source-model selection and target-response release rules. The endpoint is operating voltage, not remaining useful life.\n\nFC1 supplies source training and validation data, FC2 is a retrospective target record, and D06 is one external durability record. D06's 17 source files belong to one test of a 10-cell stack. The preserved baseline contains 546 protocol-condition/model score jobs and 99,660 scoreable point records. Methods, seeds, forecast origins and source files are not independent physical experiments.\n\nThis update retains all files from repository version 2 unchanged and adds the October 2026 revision analyses in 10_REVISION_20261018: paired-error and calendar-block sensitivity results for 24 H24 paired runs (4,356 paired forecast-origin rows), their 48 sealed input score records and portable scripts, five current main figures and three supplementary figures with plotting inputs and source code, an editable Visio panel, and current manuscript table exports. The additional analyses reuse existing predictions; they contain no new physical measurements, model training or prediction runs. The release includes a new full-package SHA-256 manifest and a current manuscript-to-data map. Historical metadata and figure numbering in directories 01-09 describe the preserved baseline; start with README_CURRENT_RELEASE.md for the updated package.\n\nThe raw FC1/FC2 and D06 measurements are not redistributed. Obtain them from their original repositories under the applicable terms:\nhttps://doi.org/10.17028/rd.lboro.3518141.v1\nhttps://doi.org/10.5281/zenodo.20715542\n\nMissing or ineligible hours remain identifiable in coverage records. Legacy-protocol results in 07_V1_SUPPLEMENTARY_RESULTS are separate from the main protocol and should not be pooled with it. Full model reruns require the original measurements and recorded scientific environment; verification of the added paired analysis uses only the derived files included here.","version":3,"contributors":[{"profile_id":"485718f5-1c44-4fdd-9672-1c2e4e75d381","first_name":"Zhipeng","last_name":"Xu"}],"files":[{"filename":"PEMFC_voltage_forecasting_data_code_update_20261018.zip","id":"d2d22103-cab2-411a-a591-a2ee55314f50","content_details":{"id":"a23e1428-1e13-41c7-a74b-9abe04226427","sha256_hash":"c0bf4cbc98b7587de06bc2ac15968008a5648c5a3ac20b039bc57c0bd46d7997","content_type":"application/zip","size":49606536,"created_date":"2026-10-18T00:59:48.802Z","download_url":"https://data.mendeley.com/public-files/datasets/d6ftbf8pyj/files/d2d22103-cab2-411a-a591-a2ee55314f50/file_downloaded","view_url":"https://data.mendeley.com/public-files/datasets/d6ftbf8pyj/files/d2d22103-cab2-411a-a591-a2ee55314f50/file_viewed","download_expiry_time":"2126-10-30T20:37:10.230287084Z"},"metrics":{"downloads":0,"previews":0},"size":49606536,"status":"COMPLETED"}],"versions":[{"version":3,"available":true,"publish_date":"2026-10-20T10:12:05.201Z"},{"version":2,"available":true,"publish_date":"2026-09-17T12:47:36.718Z"},{"version":1,"available":true,"publish_date":"2026-06-30T04:08:53.745Z"}],"articles":[],"categories":[{"id":"data.elsevier.com/vocabulary/OmniScience/Concept-184830434","label":"Chemical Engineering"},{"id":"data.elsevier.com/vocabulary/OmniScience/Concept-184953393","label":"Energy Engineering"},{"id":"data.elsevier.com/vocabulary/OmniScience/Concept-239419302","label":"Electrical Energy Storage"},{"id":"data.elsevier.com/vocabulary/OmniScience/Concept-253440139","label":"Renewable Energy"}],"institutions":[],"metrics":{"views":0,"file_downloads":0,"file_previews":0},"available":true,"method":"1. Unpack the archive and read README_CURRENT_RELEASE.md, 10_REVISION_20261018/CURRENT_MANUSCRIPT_MAP.csv and the component READMEs. Verify FULL_PACKAGE_SHA256SUMS.txt against the files. The original version 2 manifest remains in 09_RELEASE_METADATA and covers only the preserved baseline.\n\n2. For the added paired-error and temporal-weighting analysis, use a fresh copy of 10_REVISION_20261018/paired_analysis. Run python analyze_paired_errors.py prepare, then python analyze_paired_errors.py run, then python verify_paired_errors.py. Compare the four generated CSV files with expected_results. The original recorded environment is Python 3.12 with NumPy 2.3.5; timestamps and runtime metadata are not expected to match byte for byte.\n\n3. The primary analysis weights forecast origins equally. Sensitivity summaries weight eligible 168-hour calendar blocks equally, with at least three scored origins per block. Retain the distinction between these summaries. Forecast origins, windows and seeds are not independent physical replicates.\n\n4. To reproduce current plots, run python run_figures.py in a fresh copy of 10_REVISION_20261018/figure_reproduction. See its README for Windows Times New Roman and Python dependencies. The script retains the supplied Visio panel exports and recreates the quantitative panels. Current accepted PDF/SVG/PNG exports are in figures_current; rounded manuscript tables are in tables_current.\n\n5. To inspect the preserved main experiment directly, use 05_DERIVED_DATA/v2_score_metrics.csv, v2_score_points.csv and v2_score_json, together with 03_PROTOCOL and 08_AUDIT. There are 546 score jobs and 99,660 scoreable point records; the added paired analysis is a subset-based reanalysis, not additional score jobs.\n\n6. A full original pipeline rerun is a separate task requiring lawful access to the raw repositories, the frozen protocol and execution lock, local path configuration, and the environment recorded in the baseline. The published v2 methods record Python 3.13.12, NumPy 2.4.6, pandas 3.0.3, PyTorch 2.12.0, scikit-learn 1.8.0 and SciPy 1.17.1. Follow the included qualification, preprocessing, fitting, prediction and scoring modules in recorded order; do not use future target responses for source fitting or target forecast issuance. This packaging update does not rerun that pipeline.","size":49606536,"owner":{"profile_id":"485718f5-1c44-4fdd-9672-1c2e4e75d381","first_name":"Zhipeng","last_name":"Xu"},"channel":"WEB","owner_id":"485718f5-1c44-4fdd-9672-1c2e4e75d381","publish_date":"2026-10-20T10:12:05.201Z","data_licence":{"id":"01d9c749-3c4d-4431-9df3-620b2dcfe144","description":"You can share, copy and modify this dataset so long as you give appropriate credit, provide a link to the CC BY license, and indicate if changes were made, but you may not do so in a way that suggests the rights holder has endorsed you or your use of the dataset. Note that further permission may be required for any content within the dataset that is identified as belonging to a third party.","url":"http://creativecommons.org/licenses/by/4.0","category":"Creative","short_name":"CC BY 4.0","full_name":"Creative Commons Attribution 4.0 International"},"related_links":[],"funders":[],"customer_id":"7377952c-6404-4428-bbfc-77242cfc301e","modified_on":"2026-10-18T00:59:48.787Z","created_on":"2026-06-28T06:35:06.332Z","confidential":false,"links":{"view":"https://data.mendeley.com/datasets/d6ftbf8pyj"},"repository":{"id":"MENDELEY_DATA","name":"Mendeley Data"}}