Trace · Digital battery passport

TracePass

From 2027, EU rules require every EV and industrial battery to carry a digital passport: carbon footprint, recycled content, state of health and its history, with different views for the public, for repairers and second-life operators, and for authorities. TracePass estimates the carbon footprint with an explainable ML model, issues a passport with a QR code, and records every lifecycle event in a hash-chained log that anyone can verify.

Footprint model trained on a synthetic, LCA-inspired dataset · classes A–E are illustrative

Issue a passport

Carbon footprint estimate

-
kg CO₂e per kWh ·
-
t CO₂e for this pack

Why: Shapley contributions vs. a typical reference pack

Recycled-content minimums (Art. 8)

Registry

most recent passports on this node

Why it matters

Footprint declarations will decide market access and, later, maximum allowed footprints. Most of a battery's carbon comes from cathode materials and the electricity of the cell factory, so where things are made matters as much as what. A passport is only useful if its history can be trusted after the battery has changed hands several times.

Method

  • Random forest trained on 4,800 synthetic packs from a semi-physical model: CAM intensity × sourcing grid, recycled-content credits, factory electricity × grid, anode type, density, logistics.
  • Exact Shapley values over 6 feature groups (64 coalitions) against a reference pack. They add up exactly to the prediction.
  • SHA-256 hash chain: each event commits to the previous one, so any edit breaks verification from that point on.

Limits & next steps

  • Synthetic training data: the real thing needs primary LCA data and the official calculation rules (delegated acts).
  • Grid intensities are rounded approximations; performance classes are not the official ones.
  • A hash chain shows tampering but doesn't stop it. Next: signed events per actor and anchoring of chain heads.