Cross-Stitch Networks for Joint State of Charge and State of Health Online Estimation of Lithium-Ion Batteries

Abstract

As a superior solution to the developing demand for energy storage, lithium-ion batteries play an important role in our daily lives. To ensure their safe and efficient usage, battery management systems (BMSs) are often integrated into the battery systems. Among other critical functionalities, BMSs provide information about the key states of the batteries under usage, including state of charge (SOC) and state of health (SOH). This paper proposes a data-driven approach for the joint online estimation of SOC and SOH utilizing multi-task learning (MTL) approaches, particularly highlighting cross-stitch units and cross-stitch networks. The proposed model is able to achieve an accurate estimation of SOC and SOH in online applications with optimized information sharing and multi-scale implementation. Comprehensive results on training and testing of the model are presented. Possible improvements for future work are also discussed in the paper.

Keywords

• SOC
• SOH
• online state estimation
• deep learning
• multi-task learning