Research package

Battery Cycle-Life Analyzer

How can capacity-fade models be selected and end-of-life projections bounded transparently?

Open repository ↗Pinned starting point ↗

Engineering approach

Linear, power-law and logarithmic curve fits selected using the latest held-out observations before a full-data refit.

Chronological holdout, empirical model comparison, bounded EOL/RUL and residual-bootstrap diagnostics.

Evidence inspector

Source reference
8d230473ada30cfd856414f35fcc9d12be5a01bb · current-main snapshot
Reviewed at
2026-09-16T20:13:45Z
Review scope
Pinned repository documentation reviewed; upstream scientific tests not rerun here.
Published checks
The pinned source provides tests and a reproducible synthetic CSV workflow. An opt-in Oxford measured-current example has separate data licensing and stated approximation limits.
Next action
Run the supplied CSV example, inspect the holdout score and distinguish bounded projections from measured lifetime.

Source snapshot · 2026-09-16
Pinned source 8d23047Published checks ↗

What this does not establish

Empirical curve fits, not electrochemical or safety models. Bundled LFP/NMC data are synthetic; an EOL outside the supported window remains unavailable.

Evidence associated with a tagged release or historical CI run does not automatically validate later commits. Review the exact source, environment and assumptions used for any reproduced result.

Read the shared evidence vocabulary →