SOH Prediction for Lithium-Ion Batteries Based on SSABP-MLR

Accurate prediction of the state of health (SOH) is important for ensuring the safe operation of lithium-ion batteries and minimizing maintenance expenses. In practical applications, direct measurement of SOH is challenging. In the paper, the NASA lithium-ion battery capacity decay dataset is used. The multi-dimensional health indicators (HI) for charging and discharging as model inputs are selected. The Sparrow Search Algorithm (SSA) optimized backpropagation neural network (BPNN) combined with the Multiple Linear Regression Model (MLR) as a combined learning model (SSABP-MLR) for prediction is proposed. The results show that RMSE and MAE are below 0.60% and 0.51%, respectively. The generalization error (GE) remains below 0.23% in the model of the generalization test. Compared with the traditional models such as BPNN, MLR, and SSA-BP, the SSABP-MLR model demonstrates superior prediction accuracy, minimal error, and outstanding generalization performance. This demonstrates the capability of the proposed method in the paper to meet the demand for SOH prediction in lithium-ion batteries.