Abstract
Accurate estimation of lithium-ion batteries' state of health (SOH) is critical for ensuring the efficient, safe, and long-lasting operation of battery systems. To address the challenge of insufficient feature information extraction and SOH estimation under complex operating conditions, a SOH estimation method based on multisource feature extraction and the Sparrow Search Algorithm-Long Short-term Memory Network (SSA-LSTM) is proposed. First, the initial health feature set is constructed by fusing the empirical, statistical and mechanistic dimensions of features. Second, to mitigate the impact of feature redundancy, the features are evaluated and ranked based on both correlation and importance, thereby optimizing the balance between feature quality and quantity. Finally, the obtained optimal feature set is used as input to the SSA-LSTM algorithm, which constructs a SOH estimation algorithm for accurate battery SOH estimation. Experimental results demonstrate that the proposed feature selection method successfully identifies the optimal feature set. Compared with other estimation algorithms, the SSA-LSTM algorithm outperforms in all evaluation indexes, with the maximum root-mean-square error (RMSE) and mean absolute percentage error (MAPE) of the estimation results reaching 0.73% and 0.53%, respectively, across various test cases.