Key points are not available for this paper at this time.
The state of health (SOH) of the lithium-ion battery (LIB) is a key parameter of the battery management system. Due to the complex internal electrochemical properties of LIBs and the uncertain external working environment, it is difficult to achieve accurate SOH determination. In this paper, we propose a new SOH estimation method using a directed acyclic graph (DAG) structure based on incremental capacity analysis and empirical mode decomposition (EMD), and finally with gated recurrent unit (GRU) for fitting. First, we combine IC curves and real features into the input feature map and use EMD to separate out high-frequency capacity regeneration fluctuations. Then, the feature maps are input into the DAG-GRU structure to fit multiple EMD decomposition functions and build SOH prediction models, which are compared with different neural network prediction models. The prediction method simplifies the prediction process, does not need to select complex health indicators as features, and has the ability to capture the fluctuations of capacity regeneration, and can fit the fluctuations in the capacity decay curve with high precision, this integrated multi-linear model takes into account accuracy and computational efficiency, reduces manual subjective operations, and uses artificial intelligence to complete most of the work, which is one of the important directions in future SOH research. The experimental results show that using the method proposed in this paper, the minimum mean square error and mean absolute error of SOH are reduced to 0.65‰ and 1.61%, respectively, and it also possesses excellent generalization ability.
Sun et al. (Tue,) studied this question.