Health State Estimation of Lithium-ion Batteries Based on APDE-Bi-GRU-Attention Model

To maintain safe and dependable battery operation and lower battery system maintenance costs, state of health (SOH) estimate, one of the primary BMS tasks for electric vehicles, is essential. In order to improve the SOH prediction accuracy of lithium battery, a SOH prediction method based on the improved differential evolution algorithm (APDE) and the bi-directional gated recurrent unit network (Bi-GRU-attention) combined with the attention mechanism is proposed. Firstly, the feature factors related to battery capacity are extracted from battery charging data and correlation analysis is performed, and the four with high correlation are selected as inputs of model. Then, the Attention mechanism is applied to the Bi-GRU network to form the Bi-GRU-attention network, and the key parameters of the network parameters are optimally searched by the improved differential evolution algorithm (APDE) to build the final prediction model. Experiments are conducted on NASA dataset, and the results show that the APDE-Bi-GRU-attention method has higher prediction accuracy and fit, which verifies the feasibility of the prediction method.