Research on health state estimation methods of lithium-ion battery for small sample data

Abstract

The SOH (State of Health) of lithium-ion batteries is an important feature that characterizes the reliable operation of energy storage systems. In this study, according to the IC (Incremental Capacity) curve and data-driven method, an SOH estimation method based on WPCA (Weight Principal Component Analysis) data dimensionality reduction and SW-TSVR (Structural Weighted Twin Support Vector Regression) was proposed to address the problem of model over-fitting due to redundant information present in the extraction of aging features of lithium-ion batteries. In addition, this study attempted to address issues in model underfitting caused by the small number of aging samples for small sample data as well as the difficulty in extracting effective aging features. Accordingly, WPCA was able to fully excavate aging characteristics from a limited battery aging sample and greatly reduce the interference of redundant information on the accuracy of the model, thereby improving the accuracy of SOH estimation. Using SW-TSVR, the structural and weight information of battery aging samples were shown to be fully integrated into TSVR, for which a strong model generalization ability was constructed for small samples. Finally, 750 aging cycles were conducted with 0.5C–1C current to verify the above method. This study’s findings demonstrated that the proposed method was able to achieve accurate SOH estimation in cases of small data samples.

Keywords

• Lithium-ion battery
• State of Health
• Incremental Capacity
• Weight Principal Component Analysis
• Structural Weighted Twin Support Vector Regression