Passive EIS-based battery state of health estimation using internal temperature modelling under dynamic operating conditions

Reliable onboard estimation of the State of Health of lithium-ion batteries remains challenging because battery impedance is simultaneously affected by ageing, temperature, and operating conditions. Although Passive Electrochemical Impedance Spectroscopy enables impedance estimation under dynamic operating conditions without dedicated excitation signals, the resulting impedance measurements must account for temperature effects before being exploited for diagnostic purposes. This work proposes a diagnostic framework combining passive impedance measurements, internal temperature estimation, and supervised learning for battery SoH estimation. The approach relies on a thermal model identified using Thermal Impedance Spectroscopy to estimate the internal cell temperature and separate temperature-induced impedance variations from ageing-related effects. The charge-transfer resistance extracted from passive impedance measurements is then combined with the estimated temperature and State of Charge to estimate battery SoH using a supervised neural network. Unlike previous studies, which generally address passive impedance estimation, thermal modelling, or data-driven diagnosis separately, the proposed methodology integrates these components into a unified framework. The methodology is experimentally validated on lithium-ion cells under representative dynamic driving conditions and different ageing levels. The results demonstrate that coupling PEIS with internal temperature estimation improves the robustness of impedance-based diagnostic indicators, enabling accurate SoH estimation under realistic onboard operating conditions.