Uncertainty-aware physics-guided digital twins for lithium-ion batteries: toward integrated health, safety, and fast-charging management

Lithium-ion battery management is shifting from scalar state estimation to integrated supervision of health, safety, and fast-charging capability. Existing reviews have addressed battery informatics, probabilistic prognostics, physics-guided learning, or digital-twin concepts as largely separate topics, but the connections among observability, mechanistic state representation, uncertainty quantification, and decision-oriented control remains underdeveloped. This Review reorganizes the field around uncertainty-aware physics-guided digital twins and argues that health loss, safety risk, and fast-charge limitations originate from coupled electrochemical, thermal, mechanical, and interfacial processes rather than from independent objectives. The discussion first identifies the shared physicochemical basis of degradation, with fast charging treated as a stringent validation case for a useful twin. The literature is then structured around four coupled layers: observability and data, physics-based and hybrid models, uncertainty-aware inference, and decision-facing control. On this basis, the Review examines health management beyond scalar state of health, plating-aware safety diagnosis, risk-constrained fast charging, and deployment from cell to pack and fleet levels. Finally, benchmarking, standardization, semantic data governance, and battery-passport requirements are discussed together with future opportunities in multimodal sensing, knowledge-enhanced twins, and autonomous closed-loop optimization. The central conclusion is that a deployable battery digital twin must be physics-grounded, uncertainty-calibrated, and explicitly designed to support decisions.