Temperature estimation of extreme-fast -charging pouch lithium-ion cell based on finite element model and data-driven fusion

To address the issue of range anxiety in electric vehicles, extreme-fast-charging technology has emerged as a mainstream solution. However, extreme-fast-charging can easily lead to heat accumulation in batteries, inducing rapid temperature rise and thermal runaway. Accurate temperature estimation in fast-charging scenarios is therefore critical. To address computational inefficiencies in electrothermal coupling models, a battery pack thermal surrogate model integrating three-dimensional(3D) thermal modeling and data-driven methods has been developed. This model improves computational efficiency while ensuring that battery temperature predictions remain within reasonable error margins. A cell-based 3D thermal battery pack model has been developed, with training data generated through computational simulations. The long short-term memory (LSTM)-based surrogate model with self-attention mechanisms has been trained on multi-condition charging datasets under different ambient temperatures, charging rates, initial state of charge (SOC), and temperatures. A highly accurate data-driven thermal surrogate model for the battery pack is obtained, and its accuracy is validated through comparison with the 3D thermal model. The results show that the model and data-driven temperature estimation achieves an average error of 1.66 °C and an average relative error of 3.84 %.