With the growing use of lithium-ion batteries, especially in electric vehicles, they have become a key energy storage solution due to their long lifespan, high energy density, and low self-discharge rate. However, long-term use leads to inevitable performance degradation and possible structural damage, resulting in reduced capacity and power output. Accurate state of health（SOH） estimation is thus essential to ensure their safety and reliability in real-world applications.

In a study published in the IEEE/ASME journal, Dr. LIN Mingqiang ’s group from the Fujian Institute of Research on the Structure of Matter, Chinese Academy of Sciences, proposed An Ensemble Convolutional Neural Network based Transfer Learning Framework for Achievable Lithium-ion Battery Health Perception in Flexible Charging Procedure.

To achieve accurate SOH estimation of lithium-ion batteries under random charging conditions, the researchers proposed an SOH estimation method that integrates transfer learning and ensemble learning. First, the random charging process is divided into multiple local voltage segments, from which two types of data, voltage and capacity, are extracted as core indicators of battery health. To address the challenge posed by feature distribution differences across different batteries, a transfer learning approach based on domain adaptive neural network (DaNN) is employed to align features between the source and target domains. This significantly reduces the model’s dependence on target battery data and lowers the amount of training data required.

Building upon this, to estimate SOH under arbitrary local voltage segments, an adaptive relevance vector machine（RVM） ensemble model optimized with a blending strategy is developed. This model utilizes multiple DaNN-based sub-models, each trained on a different voltage segment, as base learners. By leveraging their estimation capabilities on local segments, the ensemble model enhances its adaptability to diverse charging conditions.

To further improve feature weight allocation during the model fusion stage and suppress the risk of overfitting, the Pearson correlation coefficient is introduced to analyze the relationship between features from each voltage segment and SOH. This serves as a key criterion for feature fusion, thereby enhancing the stability and generalization performance of the ensemble model.

Finally, the proposed method is validated through battery aging experiments conducted under various operating conditions. Results demonstrate that even when trained solely on data from a single source battery, the method maintains high estimation accuracy across different target conditions, showcasing strong cross-condition transferability and robustness.

This study proposes an SOH estimation method that integrates transfer learning and ensemble learning, enabling accurate battery health assessment under random charging conditions and improving model generalization. Even with data from a single source domain, the method demonstrates strong cross-condition transferability.

The structure of the proposed method (Image by Dr. LIN’s group)

Contact:

Dr. LIN Mingqiang

Fujian Institute of Research on the Structure of Matter

Chinese Academy of Sciences

E-mail: kdlmq@fjirsm.ac.cn