A reference-electrode-free adaptive AC heating framework for lithium-ion batteries accounting for cold-spot temperature prediction

Low-temperature environments severely degrade the discharge capability of lithium-ion batteries (LIBs) and induce lithium plating during charging, leading to irreversible capacity loss. To address these challenges, this study proposes a reference-electrode-free adaptive AC heating framework, which is built on a fractional-order electro-thermal coupling model and integrates physics-constrained neural network (PCNN) parameter estimation, cold-spot temperature prediction, and a multi-constraint formulation module. The PCNN achieved parameter identification with a maximum MARE of 5.46% despite limited training samples. Supported by these adaptive parameters, the coupling model predicted terminal voltage and lumped temperature accurately, with a voltage root-mean-square error (RMSE) below 25 mV and a temperature error below 0.45 °C. To ensure safety under local lithium-plating risk, a cold-spot temperature prediction module was developed to estimate the minimum temperature with a 3.6% error, in combination with electrochemical analysis to establish a local lithium-plating boundary without requiring reference electrodes, which is further integrated with terminal voltage and equipment limits for excitation optimization. Experimental results demonstrated stable and efficient heating at −30 °C. Under a 4C current limit, the heating rate was increased by 6.6% compared with the existing cut-off frequency strategy. Moreover, the capacity degradation remained below 1% after 80 heating cycles even at I equip = 6C. Overall, this work provides a practical framework to determine excitation under multiple safety constraints for reference-electrode-free LIBs in extreme climates, incorporating cold-spot temperature prediction to ensure localized safety.