DNN‐based temperature prediction of large‐scale battery pack

• Published in: Electronics letters 2023-08, Vol.59 (16), p.n/a
• Main Authors: Kim, Jiwon, Ha, Rhan
• Format: Article
• Language: English
• Subjects: Accuracy; artificial intelligence; Artificial neural networks; battery powered vehicles; Deep learning; Electric vehicles; Lithium; Lithium-ion batteries; Machine learning; Monitoring; Neural networks; Prediction models; Sensors; Temperature effects; temperature measurement
• ISSN: 0013-5194; 1350-911X
• DOI: 10.1049/ell2.12917

Temperature monitoring is critical for estimating the available capacity of Lithium‐ion batteries. In electric vehicle applications using large‐scale battery packs, monitoring individual cell temperature is challenging due to difficulties in sensor management. To address this issue, a sensor‐less battery temperature prediction technique is proposed that ensures both accuracy and rapid runtime execution using deep learning. A deep neural network‐based temperature prediction model is introduced that utilizes short sequences of battery voltage and discharge current. An adaptive sequence length strategy is then devised to ensure high accuracy and responsiveness, covering the non‐identically distributed nature of the data. The proposed technique is experimentally validated with commercial batteries, verifying its accuracy and rapid execution. Here, a sensor‐less battery temperature prediction technique is proposed that ensures both accuracy and rapid runtime execution using deep learning. A DNN model is designed that utilizes short sequences of battery information to predict temperature. Additionally, an adaptive sequence length strategy is introduced to ensure accurate predictions under various discharge conditions.