This record provides an EVRPTW benchmark dataset with heterogeneous charging stations. For each instance, two variants are included: (A) original time windows with added charging stations, and (B) relaxed time windows obtained by applying a distance-based due-date relaxation procedure. Charging stations are generated with minimum spatial separation and are assigned density-driven attributes: the charging rate is deterministically mapped from a normalized local density score, while price and waiting time are represented by log-normal distribution parameters. The package includes augmented instance files (.txt), station attribute files (*_stations.csv), the generation scripts (Algorithm 1 and Algorithm 2), and params.json for full reproducibility. Note on base instances:The folder homberger_1000_customer_instances/ contains the original benchmark instance files obtained from the public benchmark source and is included unchanged for convenience and reproducibility. All added charging-station data and attributes (variant_A_original_TW/, variant_B_relaxed_TW/, *_stations.csv) as well as the generation code are produced by the authors of this dataset.