TY  - JOUR
AU  - Kong, Chun
AU  - Li, Zhe
AU  - Zhao, Lance
PY  - 2026
DA  - 2026/09/02
TI  - A Hybrid Unet-MLP Neural Network Enabling Ultra-Fast and High-Accuracy Prediction for Low-Gradient Temperature Field in Lithium-ion Batteries
JO  - Journal of Carbon Neutrality
T2  - Journal of Carbon Neutrality
JF  - Journal of Carbon Neutrality
VL  - 1
IS  - 2
SP  - 83
EP  - 96
DO  - 10.62762/JCN.2026.807406
UR  - https://www.icck.org/article/abs/JCN.2026.807406
KW  - temperature field prediction
KW  - lithium-ion batteries
KW  - hybrid neural network
KW  - 3D thermal model
AB  - Accurate prediction of the internal temperature field in batteries is of critical importance for ensuring safe operation, extending battery lifespan, and ultimately accelerating the global transition to electric vehicles (EVs). Unsafe thermal conditions not only degrade battery performance but also hinder EV adoption. However, conventional 3D thermal models involve complex computations that struggle to meet the fast computational speed requirements of Battery Management Systems (BMS), and traditional neural network models are not accurate enough for temperature field prediction, especially for low-gradient fields. To achieve precise and rapid prediction of the internal battery temperature field, this paper proposes a hybrid neural network combining Unet and MLP architecture (HNNUM) as a green energy storage technology assurance tool to enable safe, durable battery operation. In this framework, the Unet module captures the relative magnitude relationships of temperatures across different locations in the temperature field, while the MLP module learns the temperature field range information. Finally, an inverse MinMax method integrates these two modules to output the predicted internal battery temperature field. The results demonstrate that the proposed HNNUM can achieve accurate prediction of the internal battery temperature field with only 5.16e-4 K RMSE, and takes only 16.59s to complete the prediction of 3000 temperature fields. The source code is available at https://github.com/kon9chun/hnnum.
SN  - pending
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Kong2026A,
  author = {Chun Kong and Zhe Li and Lance Zhao},
  title = {A Hybrid Unet-MLP Neural Network Enabling Ultra-Fast and High-Accuracy Prediction for Low-Gradient Temperature Field in Lithium-ion Batteries},
  journal = {Journal of Carbon Neutrality},
  year = {2026},
  volume = {1},
  number = {2},
  pages = {83-96},
  doi = {10.62762/JCN.2026.807406},
  url = {https://www.icck.org/article/abs/JCN.2026.807406},
  abstract = {Accurate prediction of the internal temperature field in batteries is of critical importance for ensuring safe operation, extending battery lifespan, and ultimately accelerating the global transition to electric vehicles (EVs). Unsafe thermal conditions not only degrade battery performance but also hinder EV adoption. However, conventional 3D thermal models involve complex computations that struggle to meet the fast computational speed requirements of Battery Management Systems (BMS), and traditional neural network models are not accurate enough for temperature field prediction, especially for low-gradient fields. To achieve precise and rapid prediction of the internal battery temperature field, this paper proposes a hybrid neural network combining Unet and MLP architecture (HNNUM) as a green energy storage technology assurance tool to enable safe, durable battery operation. In this framework, the Unet module captures the relative magnitude relationships of temperatures across different locations in the temperature field, while the MLP module learns the temperature field range information. Finally, an inverse MinMax method integrates these two modules to output the predicted internal battery temperature field. The results demonstrate that the proposed HNNUM can achieve accurate prediction of the internal battery temperature field with only 5.16e-4 K RMSE, and takes only 16.59s to complete the prediction of 3000 temperature fields. The source code is available at https://github.com/kon9chun/hnnum.},
  keywords = {temperature field prediction, lithium-ion batteries, hybrid neural network, 3D thermal model},
  issn = {pending},
  publisher = {Institute of Central Computation and Knowledge}
}