liming-chang/LiBMap-46800-NCM811-Graphite
LiBMap-46800-NCM811-Graphite A Design–Performance Dataset for Cylindrical NCM811/Graphite Lithium-Ion Cells Developed jointly by Tsinghua University and Electroder Ltd. For dataset enquiries: Chang Liming Email: clm24@mails.tsinghua.edu.cn Contents Overview — scope, provenance, and the present public subset Design Space — five varied porous-electrode design parameters Performance Simulations — protocols and termination conditions Data Tables and… See the full description on the dataset page: https://huggingface.co/datasets/liming-chang/LiBMap-46800-NCM811-Graphite.
<style> h2 { font-size: 1.485rem !important; font-weight: 700 !important; }
h2:first-of-type + ol { list-style-type: upper-roman; }
h2:not(:first-of-type)::before { display: inline-block; margin-right: 0.28em; }
h2:nth-of-type(2)::before { content: "I. "; } h2:nth-of-type(3)::before { content: "II. "; } h2:nth-of-type(4)::before { content: "III. "; } h2:nth-of-type(5)::before { content: "IV. "; } h2:nth-of-type(6)::before { content: "V. "; } h2:nth-of-type(7)::before { content: "VI. "; } h2:nth-of-type(8)::before { content: "VII. "; } </style>
<h1 align="center">LiBMap-46800-NCM811-Graphite</h1>
<p align="center"><strong>A Design–Performance Dataset for Cylindrical NCM811/Graphite Lithium-Ion Cells</strong></p>
<p align="center"> Developed jointly by Tsinghua University and Electroder Ltd.<br> For dataset enquiries: Chang Liming Email: <a href="mailto:clm24@mails.tsinghua.edu.cn">clm24@mails.tsinghua.edu.cn</a> </p>
Contents
- [Overview](#overview) — scope, provenance, and the present public subset
- [Design Space](#design-space) — five varied porous-electrode design parameters
- [Performance Simulations](#performance-simulations) — protocols and termination conditions
- [Data Tables and Relationships](#data-tables-and-relationships) — schemas, cardinalities, and join keys
- [Metric and Scene Catalogue](#complete-metric-and-scene-catalogue) — all 28 metrics and exact scene identifiers
- [Format and Loading](#format-and-loading) — Parquet layout and loading examples
- [Licence](#licence) — CC BY-NC 4.0 terms
Overview
To address the scarcity of high-quality design–performance data for lithium-ion batteries, we developed what is, to our knowledge, the first large-scale, high-quality dataset designed for battery engineering. The dataset integrates cell format, geometry, material system, kinetic parameters, and porous-electrode design parameters within a unified design space. It also establishes a systematic performance framework based on representative traction-battery and energy-storage applications. Data were generated through material-balance calculations and electrothermal performance simulations on the ElectroderDIGITAL® platform. Large-scale parallel computing produced performance data for approximately 770,000 cell designs.
Each design record contains more than 100 design-related fields describing the cell format, geometry, material formulation, porous electrodes, mechanical components, and thermodynamic and kinetic properties. For the complete LiBMap database, performance was characterised using eight categories comprising 29 metrics: discharge capacity, discharge energy, peak discharge power, continuous discharge power, recharge power, fast-charge time, direct-current resistance, and energy efficiency. The dataset also retains complete electrothermal response data from the simulations, including electrode potentials, terminal voltage, polarisation decomposition, heat generation, and temperature rise. It therefore provides a lithium-ion battery design–performance resource with comprehensive design information, broad performance coverage, and detailed process data.
The present public release is a subset of the complete database. It covers a cylindrical 46800 cell with a single-crystal NCM811 cathode and a graphite anode. This subset contains seven performance categories and 28 metrics; energy efficiency is not included in the present release. It systematically describes relationships between five porous-electrode design parameters and cell performance, comprising 72,000 complete cell designs and their associated detailed performance data.
Design Space
<div align="center">
</div>
Each parameter was sampled at equally spaced values over its specified range, including both endpoints. The Cartesian product of the five discrete parameter sequences produced the complete design space:
20 × 20 × 20 × 3 × 3 = 72,000 distinct cell designs
Performance Simulations
Capacity and energy. The cell was discharged from full charge at constant current under a specified temperature and C-rate. Discharge continued until the lower voltage cutoff was reached. Capacity and energy were obtained from the same simulation.
Discharge power. Constant-power discharge was simulated at a specified state of charge and ambient temperature. The search identified powers for either 60 s continuous discharge or 10 s peak discharge under the maximum-temperature and lower-voltage-cutoff constraints.
Recharge power. Constant-power charging was simulated for 5 s from 90% SOC. The maximum charging power was searched under the lithium-plating boundary and upper voltage cutoff.
Fast charging. The cell was charged from 20% to 80% SOC through 12 constant-current stages. Each stage covered a 5% SOC interval, and the maximum current that avoided lithium plating was searched independently for each stage.
Direct-current resistance. Direct-current resistance was calculated from the voltage response during a constant-current discharge pulse.
Four termination conditions were applied:
<div align="center">
</div>
Data Tables and Relationships
<div align="center">
</div>
1. cell_designs: Cell-Design Table
Each row contains one porous-electrode parameter combination and the complete cell design obtained through material-balance calculations.
<div align="center">
</div>
cell_designs.mod_data.result.cell.capacity is the reference design capacity in Ah. All C-rates in the dataset are defined relative to this capacity.
2. performance_results: Performance-Result Table
Each row represents one simulation task for a cell design under a specified operating condition. It records the requested metric, completion status, and final performance result.
<div align="center">
</div>
The principal result fields are:
<div align="center">
</div>
The three power-result categories also contain performance_results.result.reached. A value of true means that the duration of the saved simulation reached the target duration. A value of false means that the saved duration was shorter than the target, and the corresponding power is retained as an approximate result. The reached flag differs from performance_results.status: normal workflow completion does not necessarily mean that the target duration was reached.
3. simulation_traces: Detailed Simulation-Trace Table
Each row stores the detailed configuration, process data, and outputs for one completed performance task.
<div align="center">
</div>
4. Relationships Among the Three Tables
<div align="center">
</div>
The 24 tasks for each design correspond to seven categories containing 28 predefined performance metrics. Discharge capacity and energy under the same condition are obtained from one simulation. The number of tasks is therefore four fewer than the number of metrics. Failed fast-charge tasks have no final result, so not every design has all 28 result values.
Complete Metric and Scene Catalogue
Each scene identifier below is the exact string stored in performance_results.scene.
Discharge Capacity
<div align="center">
</div>
Discharge Energy
<div align="center">
</div>
Continuous Discharge Power
<div align="center">
</div>
Peak Discharge Power
<div align="center">
</div>
Recharge Power
<div align="center">
</div>
Fast-Charge Time
<div align="center">
</div>
Direct-Current Resistance
<div align="center">
</div>
Format and Loading
Each table is packaged as sharded, ZSTD-compressed Parquet files. Nested objects are retained as complete JSON strings, preserving their original names, ordering, and numeric text. Some strings embedded in these JSON objects retain their original Chinese values; they have not been translated or rewritten. The control columns status, order, and sub_order were converted to int32 only after exact string round-trip validation. All identifiers remain strings. The literal \N, empty strings, and JSON null values remain distinct. SHA-256 digests for every released file are listed in `CHECKSUMS.sha256`.
The following example reads one cell-design record and parses its nested JSON fields:
import json
from pathlib import Path
import pyarrow.parquet as pq
release = Path("/path/to/LiBMap-46800-NCM811-Graphite")
part = next(iter(sorted((release / "data/cell_designs").glob("part-*.parquet"))))
design = next(pq.ParquetFile(part).iter_batches(batch_size=1)).to_pylist()[0]
variables = json.loads(design["variables"])
mod_data = json.loads(design["mod_data"])
reference_capacity_ah = mod_data["result"]["cell"]["capacity"]Each configuration can also be loaded separately with the Hugging Face datasets library:
from datasets import load_dataset
repo_id = "liming-chang/LiBMap-46800-NCM811-Graphite"
designs = load_dataset(repo_id, "cell_designs", split="train", streaming=True)The train split represents the complete table. It is not a predefined machine-learning training set. This dataset provides no benchmark training, validation, or test split. For downstream splitting, all records with the same project_id should remain in the same partition. This grouping prevents information from other operating conditions of the same design from leaking into evaluation data. Additional data are required to assess generalisation to other material systems or cell formats.
Licence
This dataset is licensed under the Creative Commons Attribution-NonCommercial 4.0 International licence (CC BY-NC 4.0). It may be used for research, education, and other non-commercial purposes. Commercial use is prohibited.
