---
pretty_name: batgrad
license: cc-by-4.0
tags:
  - battery
  - timeseries
  - parquet
  - foundation-models
---

# batgrad datasets

Normalized battery time-series data for the
[`batgrad`](https://github.com/marplan/batgrad) template.

Included dataset IDs:

- `pozzato-2022`
- `synthetic-pozzato-2022-m50t`

The repository preserves the canonical
`type=.../dataset=.../source=normalized/...` paths expected by `batgrad`. Download
both datasets with:

```sh
uv run scripts/hf_assets.py download \
  --dataset pozzato-2022 synthetic-pozzato-2022-m50t
```

## Provenance

The published-data portion is a modified derivative of Gabriele Pozzato,
Anirudh Allam, and Simona Onori, "Lithium-ion battery aging dataset based on
electric vehicle real-driving profiles," Data in Brief 41, 107995, 2022.
[doi:10.1016/j.dib.2022.107995](https://doi.org/10.1016/j.dib.2022.107995)

The files are not the original dataset. Processing by `batgrad` includes
canonical mapping, type conversion, validation, derived features,
protocol-specific sharding, normalization, and resampling/downsampling. The
original authors are not responsible for these modifications.

- [Dataset overview](https://osf.io/qsabn/overview?view_only=2a03b6c78ef14922a3e244f3d549de78)
- [Raw data (Dropbox)](https://www.dropbox.com/scl/fo/3ss0age6ggfcm67okldhw/h?rlkey=tnczvb82gukfe2n4gol2uyo7x&dl=0)
- Original dataset license: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)

The synthetic portion contains randomized profiles generated with PyBaMM using
a modified `OKane2022` parameter set. Particle cracking was disabled, and the
LLI/LAM parameters were increased.