Reproducibility Data for Safe Voltage Control: Processed PTProsumer and Monash SwitchDin IEEE 33-Bus Scenarios

This record provides the reproducibility data accompanying the manuscript "Uncertainty-Aware Safety Framework for Multi-Agent Reinforcement Learning-Based Voltage–VAR Control in Distribution Networks under Model Mismatch". The package contains two processed three-minute IEEE 33-bus scenarios derived from publicly available datasets: PTProsumer: Derived from the Longitudinal Dataset of Net-load, PV Production and Solar Irradiation from Madeira Island, Portugal (OSF DOI: 10.17605/OSF.IO/9XS3A). Monash SwitchDin: Derived from Residential Power and Battery Data (Zenodo DOI: 10.5281/zenodo.8219786). For each scenario, the package includes the network model, active and reactive load profiles, PV generation profiles, profile mappings, fixed experimental data splits, preprocessing metadata, data quality reports, and SHA-256 checksums. The original measurements have been processed through temporal aggregation, profile selection, scaling, and IEEE 33-bus scenario construction. Detailed preprocessing procedures and experimental split definitions are provided in the accompanying metadata files. The original MAPDN time-series dataset is not redistributed in this record. Instructions for obtaining it from its official distribution are provided in MAPDN_DOWNLOAD.md. The included network model is based on the MAPDN IEEE 33-bus scenario. License information and required source attributions are provided in LICENSES.md and the corresponding LICENSE_NOTICE.txt files. This dataset is intended to facilitate reproducibility and independent verification of the voltage safety control experiments reported in the associated manuscript.