Data and code for Model Predictive Control of PEM Electrolysis Driven by Wind and Solar Energy with Terminal Buffer Constraints and Source-Specific Uncertainty Calibration
Description
This dataset contains derived numerical outputs and deterministic Python code supporting the associated Applied Energy manuscript. It includes data quality summaries, dynamic hydrogen response model metrics, source-specific conformal calibration results, controller comparisons, full-duration replays, MPC ablation results, runtime benchmarks, publication figures, source links, file checksums, and a manuscript-to-data map. The analyses use public 1 Hz PEM electrolysis records for historical solar PV, historical wind, and simulated wind operation released through NLR/OEDI. Original row-level NLR/OEDI records are not redistributed. Official download links, license links, filenames, sizes, and SHA256 checksums are provided. The deposited outputs were generated by deterministic scripts. No research data were generated by a large language model. File description for Mendeley_Data_Upload.zip Compressed reproducibility package containing derived CSV and JSON results, deterministic Python scripts, publication figures in PDF and PNG formats, source-data links and SHA256 checksums, column dictionaries, a manuscript evidence map, environment requirements, and reproduction instructions. Original NLR/OEDI row-level records are not included. Steps to reproduce Download and extract Mendeley_Data_Upload.zip. Use Python 3.10 or later. Install the packages listed in environment/requirements.txt. From the extracted package root, run: python code/download_source_data.py Confirm that the three downloaded source files pass the recorded SHA256 checks. Run: python code/run_pipeline.py Compare the regenerated outputs with derived_data, FILE_MANIFEST.csv, and MANUSCRIPT_DATA_MAP.csv. The complete controller scans may require substantial computation time. Runtime values depend on the processor and software environment, whereas feasibility and performance results should remain numerically consistent within floating-point tolerance.