A novel integrated SOC-SOH estimation framework for whole-life-cycle lithium-ion batteries

Accurate estimating the state of health (SOH) and state of charge (SOC) is crucial for ensuring the reliable and safe operation of lithium-ion batteries. Traditional methods for the joint estimation of SOC and SOH typically rely on separate models, resulting in a decoupling of their relationship. Moreover, the current convolutional and recurrent-based deep models overlook the inherent connection between local features and global temporal features. These limitations not only hinder the extraction of combined feature information relevant to SOC and SOH during the charging process, but also increase computational complexity and diminish estimation accuracy. To solve these problems, this study proposes a novel SOC-SOH Estimation Framework (SSEF). The framework achieves parameter sharing by segmented training, effectively accounting for the intrinsic coupling between SOC and SOH. This enables a unified joint estimation of the two variables, leading to a substantial enhancement in efficiency. Additionally, a novel charging encoder that alternates between Temporal Convolutional Network and Bidirectional Gated Recurrent Unit is designed. It captures local temporal information and long-term dependencies related to SOC and SOH during charging. SSEF enables precise SOC and SOH estimation for whole-life-cycle lithium-ion batteries, enhancing accuracy and efficiency compared to prevalent methods.