Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Column name 'timestamp_s;"timestamp_origin";"sd_block_id";"v_raw_V";"i_raw_A";"p_raw_W";"t_cell_degC";"delta_q_Ah";"delta_e_Wh";"soc_est";"ocv_est_V";"meas_condition";"bms_condition";"v_state";"i_state";"t_state";"over_q";"under_q";"over_soc";"under_soc";"eol_cap";"eol_imp";"scheduler_state_dec";"scheduler_state_phase";"scheduler_state_checkup";"scheduler_state_charge";"scheduler_timeout";"scheduler_cu_pending";"scheduler_pause_pending";"scheduler_state_sub";"delta_q_chg_Ah";"delta_q_dischg_Ah";"delta_e_c...' is too long. It should be less than 500 characters.
Error code:   TooLongColumnNameError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

lg-hg2-stroebl-2024-raw

Tier 1 raw mirror of the LG INR18650HG2 cycling dataset by Stroebl, Mayer, Lienkamp (KIT, 2024), hosted under the BSEBench organization.

This dataset closes a major BSEBench V1.0 coverage gap : authentic WLTC dynamic driving telemetry on NMC cells with silicon-carbide composite anode (C-SiO blended graphite, 3-6% SiO per Obrovac 2014). Si-C anodes exhibit path-dependent hysteresis ~10% at mid-SOC that classical 2-RC ECM cannot correct — a stress-test for ensemble-based filtering approaches.

Status

🚧 Upload pending — placeholder card. This Tier 1 mirror contains the truncated WLTC subset (28 CSV files, ~7.9 GB) used in the BSEBench paper2b validation work. The full 69.4 GB raw distribution is preserved under canonical DOI 10.35097/1947 ; this subset is sufficient for BSEBench V1.1 evaluation.

What this is

LG Chem INR18650HG2 cells cycled with WLTC (Worldwide Harmonized Light-duty Vehicles Test Cycle) profiles across 12 parameter sets (P065-P076) × 3 replicates. Schema is plain CSV (pyarrow-readable). 5 dataset types : LOG (59 cols, 2 s resolution), EOC (57 cols per half-cycle), LOG_AGE (11 cols), EIS (22 cols), PULSE (19 cols).

Citation (verbatim from Scientific Data paper) :

Stroebl, F., Mayer, A. and Lienkamp, M. (2024). Open Source Cyclic Aging Dataset of 228 Commercial Lithium-Ion Battery Cells. Scientific Data, 11, 944.

Cell specifications

Property Value
Manufacturer LG Chem
Model INR18650HG2
Form factor 18650 cylindrical
Cathode chemistry NMC (typical 622 stoichiometry)
Anode chemistry C-SiO composite (3-6% SiO blended with graphite)
Nominal capacity 3.0 Ah
Total cells in full dataset 228
Cells in this truncated subset 28 (P065-P076 × 3 reps, ~36 minus a few missing)
Temperatures 0, 10, 25, 40 °C (read from pool_meas_t_degC column)
Aging duration up to 449 days
Cell ID format lg_P{param:03d}_rep{r}_S{stage:02d}

File inventory

This subset contains 28 CSVs named cell_logext_P{param}_{rep}_S{stage}_C{cell}_truncated.csv covering the 12 WLTC parameter sets with 3 replicates each.

Profile

WLTC driving (Worldwide Harmonized Light-duty Vehicles Test Cycle), repeated until SoC < 10%. Sign convention in raw : i_raw_A positive = charge ; BSEBench Tier 2 will flip to BPX-1.1 (charge=positive) — wait, BPX-1.1 IS charge=positive so no flip needed. Verify empirically at adapter time.

Si-C anode anomaly (important caveat)

Per Obrovac 2014 and McDowell 2013 : Si-containing anodes produce path-dependent hysteresis ~10% at mid-SOC, uncorrectable by classical 2-RC ECM. This dataset is therefore a stress-test for advanced filter approaches (BMA ensembles, neural state-space, particle-filter mixture models). BSEBench filter benchmarks should explicitly report Si-C-vs-graphite performance differential.

Why "raw mirror" tier (truncated)

BSEBench dual-tier strategy :

  • Tier 1 (this repository) — truncated WLTC subset (28 CSVs, 7.9 GB), SHA-256 verified
  • Tier 2 — see bsebench-org/lg-hg2-stroebl-2024 for harmonized BSEBench-canonical Parquet
  • Full raw : the canonical 69.4 GB distribution under DOI 10.35097/1947 is NOT mirrored in BSEBench V1.1 — interested users should download directly from the KIT data portal. BSEBench mirror covers the BSEBench-relevant subset only.

Original source

Karlsruhe Institute of Technology (KIT) data portal :

License

Creative Commons Attribution 4.0 International (CC-BY-4.0) per Scientific Data 2024 publication.

How to use

from huggingface_hub import snapshot_download

local_dir = snapshot_download(
    "bsebench-org/lg-hg2-stroebl-2024-raw",
    repo_type="dataset",
)
# 28 truncated CSV files of WLTC profiles

Citation

@article{stroebl2024lghg2,
  author  = {Stroebl, Fabian and Mayer, Alexander and Lienkamp, Markus},
  title   = {{Open Source Cyclic Aging Dataset of 228 Commercial Lithium-Ion Battery Cells}},
  journal = {Scientific Data},
  volume  = {11},
  pages   = {944},
  year    = {2024},
  doi     = {10.1038/s41597-024-03831-x}
}

@misc{bsebench2026,
  author = {Akir, Oussama and BSEBench Contributors},
  title  = {{BSEBench}: an open-source benchmark for battery state-estimation filters},
  year   = {2026},
  url    = {https://bsebench.org},
}

Provenance manifest

bsebench-org/bsebench-datasets/manifests/lg_hg2_stroebl_2024.yaml.

Downloads last month
47