Processed aligned multimodal photovoltaic dataset derived from Stanford SKIPP'D (2017–2019)

This record contains a processed and aligned multimodal photovoltaic forecasting dataset derived from the Stanford SKIPP'D benchmark dataset (2017–2019). The original Stanford benchmark provides processed all-sky image data, photovoltaic (PV) power data, and timestamp files for model development and test sets. In this derived release, we retain the original benchmark structure and further align additional meteorological variables to the original timestamps in order to support multimodal photovoltaic forecasting research. The meteorological variables were obtained from Open-Meteo and aligned to the original benchmark timestamps. This record includes:(1) a ZIP archive containing the processed aligned HDF5 dataset,(2) timestamp index files for the train/validation and test subsets, and(3) supporting documentation for reuse. The HDF5 file contains two top-level groups, `trainval` and `test`. Each group includes 20 datasets: 1 image variable (`images_log`), 1 photovoltaic power variable (`pv_log`), and 18 aligned meteorological variables from Open-Meteo. This release should be regarded as an aligned and augmented derivative dataset rather than an independently collected original dataset. Original benchmark source:https://purl.stanford.edu/dj417rh1007https://github.com/yuhao-nie/Stanford-solar-forecasting-dataset Meteorological data source:https://open-meteo.com/