Interpretable Battery Aging without Extra Tests via Neural-Assisted Physics-based Modelling

arXiv cs.LG / 4/3/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes IBAM, an interpretable battery aging model that produces a 2-D “aging fingerprint” instead of relying on the single scalar state of health (SoH).
IBAM requires no extra diagnostic tests, using only routine battery management system logs to capture interpretable voltage polarization loss across charge/discharge curves and tail loss near end-of-discharge.
The framework combines a fractional-order physics-based equivalent circuit model with a two-stage least-squares approach to extract per-cycle fingerprints, then aligns these fingerprints with the SoH axis using physics-guided regression.
SoH per cycle is estimated with a bidirectional gated recurrent unit (BiGRU) using customized multi-channel voltage features, integrating neural prediction with physical constraints for better fidelity.
Across batteries with short, medium, and long lifespans, IBAM is reported to provide superior physics-model fidelity at different aging stages and to reveal degradation mechanism patterns that can inform battery health assessment and control decisions.

Abstract

State of health (SoH) is widely used for battery management, but it is a single scalar and offers limited interpretability. Two batteries with similar SoH can exhibit very different degradation behaviors and the lack of interpretability hinders optimal battery operation. In this paper, we propose IBAM for interpretable battery aging modelling with a neural-assisted physics-based framework. IBAM outputs a 2-D aging fingerprint without extra diagnostic tests and uses only routine logs from the battery management system. The fingerprint offers great interpretability by capturing a battery's curve-wide polarization voltage loss and the tail loss near the end-of-discharge. IBAM first creates a physics-based battery model based on a fractional-order equivalent circuit model, and then extracts per-cycle fingerprints from the model using a two-stage least-squares method. IBAM further anchors fingerprints on the SoH axis with physics-guided regression, where the per-cycle SoH is estimated via a bidirectional gated recurrent unit with customized multi-channel voltage features. Across batteries with short-, medium-, and long-lifespans, IBAM consistently yields the best physics model fidelity at different aging stages, and provides clear interpretations of degradation mechanisms and fingerprint patterns about batteries of different lifespans. The resulting fingerprints support interpretable battery health assessment and can inform battery control choices.