The gradually increased amount of battery aging data has enabled rapid development of data-driven and machine learning algorithms in battery state assessment and lifetime prediction. Despite deep commitment and broad excitement, significant gaps still exist which hinders a thorough comparison and rapid iterative of the prediction algorithm. First, the formats across many public data sources are inconsistent; second, a majority of the algorithms are close source and lack of reproducibility; Last, the definition of evaluation metrices varies under different prediction scenarios, which pose challenges to compare different algorithms and develop novel one.
If you do not have a Python environment locally, you can click on the Colab badge below and run the experimental example using the code we shared.
Datasets available for battery RUL prediction tasks
| Data Source |
Chemistry of cathode |
Nominal capacity and end of life(EOL) |
Degration Characteristics |
|---|---|---|---|
| NASA | NCA | 2Ah/1.4Ah | Linear, Capacity recover |
| CALCE | LCO | 1.1Ah/0.88Ah | Linear, Have aging knee point |
