.png){kind=icon}
.png){kind=icon}
Return
Temperature estimation in a lithium-ion cell using a machine learning based approach
.png){kind=icon}
PRE
AI
DOI：10.1016/j.applthermaleng.2025.126201.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Battery surface temperature- a key indicator of battery health and operating conditions, is estimated using current and voltage data with machine learning based model in the work. Battery temperature is influenced by the heat generated in operation, which is significantly affected by the state of charge (SOC), C-rate, and ambient temperature. This work focuses on developing a Time Delay Neural Network (TDNN) model to predict temperature in lithium-ion cells. The model uses cell current, voltage, and ambient temperature as input variables. Voltage serves as a predictor for SOC, as its response to current depends on SOC, thereby simplifying the model by eliminating the need for an additional SOC prediction component. The model is designed to be adaptable for real-time applications, tackling challenges such as managing non-uniformly sampled data and ensuring the repeatability of the training process. It achieves a good accuracy in temperature estimation, with an R value of 0.997 and an MSE of 0.514 (degrees C)2 for the constant current dataset tested. It generalizes well across various C-rates and ambient temperature conditions not used in the training. It adapts well for transient condition testing of variable current and temperature drive cycles as well. Further, compared to the standard NARX and LSTM networks, the TDNN model stands out for its simple structure, efficiency and ability for generalization, making it a promising candidate for real-time temperature estimation in lithium-ion cells.
Keywords:
Temperature prediction
Artificial neural network
Lithium-ion
Time-delay neural network
NARX
LSTM
Journal
IF:
6.9
Papers:
2.7W
Citations:
10.6W