Health State Estimation Method of Lithium-Ion Batteries Based on Firefly Algorithm Optimization

A prediction method for the State of Health (SoH) of lithium-ion batteries is proposed to address challenges such as complex modeling, high prediction error, difficulty in model parameter tuning, and long testing times. The method extracts indicators highly correlated with SoH degradation from charging data and utilizes Singular Value Decomposition (SVD) for denoising, improving feature relevance. A differentiated feature extraction strategy is designed based on the battery’s initial charge state, significantly reducing the testing period. The Firefly Algorithm (FA) is employed to optimize the structure and parameters of the Gated Recurrent Unit (GRU) neural network, further enhancing prediction accuracy. Validation results on the CALCE battery dataset demonstrate that the model achieves a maximum Root Mean Square Error (RMSE) of 0.0542 and a coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) of 0.9889, surpassing traditional models. When only constant current charging phase data is used as input features, the accuracy reaches 95%, with a 25% reduction in testing time. When only constant voltage charging phase data is used, the accuracy exceeds 70%.