Temperature Prediction of Lithium-Ion Cells Using Transformers and Deep Learning Models

The accurate prediction of batter temperature is essential for maintaining the safety, longevity, and performance of Panasonic 18650PF lithium-ion cells, especially when used in high-demand appliances like electric vehicles and energy storage systems. This study mainly explores Transformers and Deep Learning Models which can be used to predict the temperature of the Panasonic 18650PF Lithium-ion-cells using the Electrochemical Impedance Spectroscopy (EIS) data gathered at different ambient temperatures. The four main Deep Learning models used are Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), Bidirectional LSTM (Bi-LSTM), and Minimal Gated Unit (MGU) are mainly evaluated for their predictive accuracy and computational efficiency in this paper. Each model's performance is measured using standard metrics ad from the results we see that MGU and transformers perform the best when compared to others. From the results we can see that transformer alone cannot be the best model when compared to other models, hence we need to ensemble other models to get the efficient balance between speed and accuracy. This work establishes a benchmark for machine learningdriven temperature prediction in battery management systems, offering insights into how advanced recurrent neural networks can enhance proactive thermal management.