A deep learning approach for state-of-health estimation of lithium-ion batteries based on differential thermal voltammetry and attention mechanism

Abstract

Accurate estimation of the State of Health (SOH) of lithium-ion batteries is crucial for ensuring their safe and reliable operation. Data-driven methods have shown excellent performance in estimating SOH, but obtaining high-quality and strongly correlated features remains a major challenge for these methods. Moreover, different features have varying importance in both spatial and temporal scales, and single data-driven models are unable to capture this information, leading to issues with attention dispersion. In this paper, we propose a data-driven method for SOH estimation leveraging the Bi-directional Long Short-Term Memory (Bi-LSTM) that uses the Differential Thermal Voltammetry (DTV) analysis to extract features, and incorporates attention mechanisms (AM) at both temporal and spatial scales to enable the model focusing on important information in the features. The proposed method is validated using the Oxford Battery degradation Dataset, and the results show that it achieves high accuracy and robustness in SOH estimation. The Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) are around 0.4% and 0.3%, respectively, indicating the potential for online application of the proposed method in the cyber hierarchy and interactional network (CHAIN) framework.