Bridging data-driven and physics-based models for lithium-ion battery state analysis and future performance assessment via active material volume fraction

Accurately estimating the current state and predicting the future performance of lithium-ion batteries are essential for their optimal operation and lifetime extension. However, the limited availability of internal degradation metrics and reliable historical data often constrains effective battery management, particularly in the second-life applications. To address this challenge, this work introduces a novel hybrid model that synergizes the adaptability of Data-Driven Models (DDMs) with the robustness of Physics-Based Models (PBMs). The key novelty lies in the use of the active material volume fraction as a coupling variable, enabling the adaptation of the PBM to the actual degradation state of the battery. The DDM is responsible for actual state estimation, predicting three key diagnostic parameters from a standard charge test: State of Health (SOH), Full Equivalent Cycles (FEC), and the active material volume fraction ( ɛ a c t , k ) of both electrodes. These parameters are then used to initialize the PBM and simulate future performance under specific operating conditions. The informed PBM then simulates degradation kinetics, under defined future operational profiles, to forecast the remaining useful life of the battery. Experimental validation was conducted using NMC811-SiGrOx cells, demonstrating that the proposed approach achieves high predictive accuracy for SOH and ɛ a c t , k , with RMSE below 0.5% and 1%, respectively, and enables the PBM to successfully forecast performance degradation over a 50-day horizon with errors below 0.6%. These results demonstrate that the proposed methodology provides a non-invasive and computationally efficient framework suitable for Digital Twin implementations, effectively integrating data-driven adaptability with physics-informed accuracy to assess battery health and lifetime across diverse capacities, geometries, and applications. • Novel Hybrid Model framework that synergizes DDM adaptability with PBM robustness. • Uses active material volume fraction to adapt PBMs to the current degradation state of the LIB. • Enables non-invasive actual state estimation and performance forecasting from charging data. • High accuracy demonstrated on NMC811-SiGr cells over a 50-day horizon. • Facilitates informed decision-making for in-service LIBs and Second-Life Batteries.