Core temperature estimation in a cylindrical lithium-ion cell: analysis code, identifiability study, and scored predictions

Le résumé fourni par la source

Inverse estimation of internal lithium-ion cell state from a single external temperature channel, validated against a measured core thermocouple. Three estimators are compared on identical information: a transient finite-volume inverse, an inverse physics-informed neural network, and a one-line quasi-steady algebraic relation. The algebraic relation is the most accurate at 0.141 K core RMSE against gradients peaking at 6.54 K, and the neural estimator is the worst of the three; the repository reports that negative result as its headline. A companion phase-change study finds the remaining latent buffer identifiable to 0.1 percent of capacity from one surface sensor while the melt-front position is not identifiable at any heating rate tested. Includes Cramer-Rao identifiability analysis performed before fitting, pre-registered predictions scored with misses reported, solver verification against closed-form solutions, and a leak audit.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.