An FE-S-BiLSTM and Heatmap-Based State-of-Health Estimation Method for Lithium-Ion Batteries

Accurate battery SOH estimation can improve battery reliability and ensure its safe and efficient operation. Therefore, this study proposes a novel method for battery SOH estimation. First, a thermocouple temperature sensor monitors the battery’s operating temperature and provides feedback to the thermostat for precise temperature control during experiments. The battery’s charging voltage and current are measured using voltage transmitters and Hall current sensors, respectively, and the two are fused to obtain the battery’s charging power. Next, one-dimensional charging power data is converted into two-dimensional heatmaps using image encoding techniques. The heatmap corresponding to the first cycle is selected as the reference image, and the difference between the heatmaps of subsequent cycles and the reference image is quantified using structural similarity (SSIM). The final results serve as an indicator of battery health. Additionally, this study proposes a novel battery SOH estimation model, the Feature Enhancement-Simplification-Bidirectional Long Short-Term Memory (FE-S-BiLSTM). The feature enhancement layer in the FE-S-BiLSTM model enriches global static features through enhancement learning. Based on the model’s bidirectional long short-term memory (BiLSTM) layer and simplification layer, dynamic features in the time-space domain are double-captured. Finally, this study utilizes six batteries and designed a variety of experiments to validate the effectiveness of the proposed method. The experimental design comprises three tasks: SOH estimation based on full charge data, SOH estimation based on random SOC interval charge data, and cross-battery SOH estimation. The experimental results demonstrate that the proposed SOH estimation method for batteries exhibits significant potential for practical applications.