Attention‐Enhanced PSO‐LSTM for Cross‐Temperature Lithium‐Ion Battery Health Estimation Using Multidimensional Operational Data

Due to their high energy density, long cycle life, and high efficiency, lithium‐ion batteries have become one of the most widely used electrochemical energy‐storage technologies. This work addresses the challenge of extracting informative internal indicators from operational data and accurately estimating the state of health (SOH) of lithium‐ion batteries. Based on multidimensional operating measurements, we construct a set of health factors that are strongly correlated with capacity from four perspectives: time, charged energy, power, and temperature. Building on these factors, a SOH estimation model is developed by integrating particle swarm optimization (PSO) with a long short‐term memory (LSTM) network and an attention mechanism (PSO‐LSTM‐Attention). The model is trained and validated on NASA's public aging datasets, and its performance is evaluated using root mean square error (RMSE) and mean absolute percentage error (MAPE). Across the tested cells (B0005, B0006, B0007, B0029, and B0030) and operating conditions (24°C–43°C), the proposed method achieves accurate capacity estimation, with all RMSE values below 1% and average errors around 0.5%. These results demonstrate that the proposed framework can effectively exploit multidimensional operational data for SOH estimation and has the potential to support battery operation optimization and safety management in practical applications.