Effective capacity early estimation of lithium thionyl chloride batteries for autonomous underwater vehicles

Abstract

Effective capacity early estimation is one of the critical tasks for electric autonomous underwater vehicles (AUV) to guide energy allocation and ensure equipment reliability. An effective capacity estimation method for Li/SOCl2 is proposed here. Firstly, we generate a comprehensive dataset consisting of 138 Li/SOCl2 cells, where the test-matrix is designed based on the AUV's working modes. Then, an effective capacity estimation architecture is established. In this architecture, a calendar aging model (CAM) based on the storage mode is obtained; a polynomial regression model (PRM), trained by features exacted from the early discharge data, is coupled to CAM based on the unscented Kalman filter (UKF), where one-step fusion is adopted to ensure lightweight. The verification results show that the architecture significantly improves prediction accuracy compared to traditional methods only considering the single working mode and performs well with generalization ability in the new database. Moreover, the early prediction and lightweight characteristics make it the most promising candidate for AUV power estimation in engineering.