Related Experiment Video
Updated: Jun 2, 2026
10:36
Optimization of An Air-Based Heat Management System for Dusty Particulate Matter-Covered Lithium-Ion Battery Packs
Published on: November 3, 2023
Low-Temperature Prediction in Commercial Lithium-Ion Batteries during Dynamic Usage via Enhanced Physics-Informed
Eric L Pereira1, Davi M Soares1
1Department of Electrical and Computer Engineering, Wichita State University, 1845 Fairmount Street, Wichita, Kansas 67260, United States.
ACS Omega
|June 1, 2026
Summary
Accurate temperature prediction for lithium-ion batteries (LIBs) is vital for safety, especially at low temperatures. An enhanced physics-informed neural network (PINN) model accurately predicts LIB surface temperature, improving safety and performance in cold conditions.
Area of Science:
- Materials Science
- Electrochemistry
- Computational Science
Background:
- Real-time temperature monitoring of lithium-ion batteries (LIBs) is critical for operational safety across diverse conditions.
- Accurate temperature prediction, particularly at low temperatures, enables early detection of hazardous states like lithium plating.
- Proactive intervention by battery management systems is necessary to prevent damage and enhance performance in cold environments.
Purpose of the Study:
- To develop an enhanced physics-informed neural network (PINN) model for accurate prediction of LIB surface temperature.
- To incorporate thermal laws and entropy effects, including Joule heating, entropic heat, and radiative losses, into the model's loss function.
- To focus on the understudied low-temperature cycling range for improved battery safety and performance.
Main Methods:
- An enhanced PINN model was developed, integrating thermal laws and entropy effects.
- Joule heating, reversible entropic heat, and radiative losses were incorporated into the loss function.
- Adaptive weighting, a charge-consistency term, and unit consistency were employed to enhance prediction stability and accuracy.
Main Results:
- The proposed model achieved mean absolute errors as low as 0.07 °C for LFP cells at 5 °C.
- High accuracy was demonstrated across three commercial cell chemistries (LFP, LCO, NMC) at various temperatures (5 °C, 25 °C, 45 °C).
- The model achieved this performance using only 30% of the dataset for training and evaluation within the same chemistry, showing robustness and agreement with thermocouple measurements.
Conclusions:
- The enhanced PINN model provides robust and accurate real-time temperature prediction for LIBs.
- The model's ability to predict temperature accurately, especially at low temperatures, significantly enhances battery safety and performance.
- This approach supports improved battery management systems, leading to safer and more efficient operation of lithium-ion batteries in diverse environmental conditions.
