# SKIPP'D: Sky Images and Photovoltaic Power Generation Dataset

## Facts

- Provider: Environmental Assessment and Optimization Group, Stanford University (Yuhao Nie, Adam Brandt and others)
- License: CC-BY-4.0 (https://purl.stanford.edu/dj417rh1007)
- Size: 14 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/solarbench/SKIPPD
- Access: public, readable without an account
- DOI: https://doi.org/10.25740/dj417rh1007
- Category: Energy
- Source: https://doi.org/10.1016/j.solener.2023.03.043
- Page: https://scigantic.com/archive/f28f4269-7f3e-53ee-a95c-8ad70d68f342
- Keywords: Energy, benchmark, Parquet, Hugging Face

## About

SKIPP'D (SKy Images and Photovoltaic Power Generation Dataset), a benchmark for short-term solar forecasting compiled by the Environmental Assessment and Optimization group at Stanford University. It pairs ground-based all-sky camera images of the sky over Stanford with time-aligned power output from a nearby photovoltaic array, so models can learn how passing clouds drive short-term PV fluctuations; the authors released it to support deep learning for image-based solar nowcasting and forecasting. This Hugging Face copy comes from the authors' SolarBench project. Listing the repository (measured 2026-11-14) shows ground/ with five train Parquet shards of about 0.42 to 0.47 GB and a test shard, a ground_no_sky_img variant without images, and ERA5 and GFS folders with numerical weather inputs; it uses about 9.4 GB. Uses include sky-image solar nowcasting, PV power forecasting, cloud motion modelling, generative sky image prediction and benchmarking forecasting models that fuse imagery with weather data. Load with Hugging Face datasets, pandas or pyarrow and decode images with PIL. The Stanford Digital Repository record licenses the benchmark data CC BY 4.0; cite Nie et al., Solar Energy 255 (2023).

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("solarbench/SKIPPD")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `ERA5/test-00000-of-00001.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/f28f4269-7f3e-53ee-a95c-8ad70d68f342
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/f28f4269-7f3e-53ee-a95c-8ad70d68f342
- Over MCP: `get_schema_card` with id `f28f4269-7f3e-53ee-a95c-8ad70d68f342` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Energy datasets](https://scigantic.com/topics/energy)
