STTEWS: A sequential-transformer thermal early warning system for lithium-ion battery safety

Marui Li,Chaoyu Dong,Binyu Xiong,Yunfei Mu,Xiaodan Yu,Qian Xiao,Hongjie Jia

DOI: https://doi.org/10.1016/j.apenergy.2022.119965

IF: 11.2

2022-12-15

Applied Energy

Abstract:The internal reactions of lithium-ion batteries are susceptible to temperature, which makes the temperature of significant impact on their safety and performance. Therefore, it is very important to predict the temperature trend of lithium-ion batteries and implement thermal early warning. In order to solve this thermal concern of lithium-ion batteries, this paper designed a sequential-transformer thermal early warning system (STTEWS). First, a new allied temporal convolution-recurrent diagnosis network (TCRDN) is constructed by combining LSTM and temporal convolution network (TCN) using an adaptive boosting algorithm. Then, a complete transformer thermal diagnosis network (TTDN) is established, which fuses the important information from lithium-ion battery thermal images and integrates the prediction results from TCRDN to achieve an accurate early warning function. TTDN combines state-of-the-art time series transformer and vision transformer. TCRDN and TTDN constitute the complete STTEWS. Experiments show that the accuracy and F1 score of STTEWS for thermal diagnosis on multiple datasets both exceed 95%.

energy & fuels,engineering, chemical