A State-of-Health Estimation Method for Lithium-Ion Batteries Based on Electrochemical Impedance Spectroscopy and a Hybrid BiLSTM-Attention Model
Abstract
To address the complex degradation process and strong temporal dependence in lithium-ion battery state-of-health (SOH) estimation, this paper proposes an SOH estimation method based on electrochemical impedance spectroscopy (EIS) features. First, an equivalent circuit model is employed to fit the EIS data, from which physically interpretable impedance-derived aging features are extracted and normalized. Subsequently, a bidirectional long short-term memory network is used to capture the temporal characteristics of the battery aging process. A multi-head self-attention mechanism is further introduced to model global dependencies among different cycling instants, thereby enhancing the representation of critical degradation information. The proposed method is validated using a publicly available lithium-ion battery EIS dataset and is compared with long short-term memory (LSTM) and bidirectional long short-term memory (BiLSTM) models. The results indicate that the proposed method can effectively track SOH evolution under different degradation patterns and achieve improved estimation accuracy. This method provides an effective approach for online EIS-based SOH estimation of lithium-ion batteries.