State of charge estimation method by using a simplified electrochemical model in deep learning framework for lithium-ion batteries

To ensure the secure and healthy usage of lithium-ion batteries, it is necessary to accurately estimate the state of charge (SOC) in battery management systems. The development of deep learning (DL) provides a new solution for battery SOC estimation. However, the directly measured physical quantities contain less useful information and have low estimation accuracy. This paper proposes a method of integrating the mechanism knowledge of the battery domain into the DL framework. Firstly, the simplified electrochemical model is utilized to obtain the mechanism-related physical variables to expand the input of the DL model. Secondly, the long short-term memory (LSTM) network is used with the Bayesian optimization, and the variables with high correlation are identified. The best SOC estimation performance can be obtained by adding all the selected highly-correlated variables to the input for training together. The results show that the proposed method can improve the SOC estimation performance with only a slight increase in computation cost. Finally, other DL models are utilized to further validate the effectiveness, to reveal the universality. These results show that the performance of the DL model can be effectively improved by using the knowledge of the battery domain.

Automated summary

In order to ensure the secure and healthy usage of lithium-ion batteries, accurately estimating the state of charge (SOC) in battery management systems is necessary. The development of deep learning (DL) provides a new solution for battery SOC estimation. This paper proposes a method that integrates mechanism knowledge of the battery domain into the DL framework, resulting in improved SOC estimation performance.