A novel robust back propagation neural networkdual extended Kalman filter model for state-ofcharge and state-of-health co-estimation of lithiumion batteries

Abstract

To improve the real-time estimation accuracy of battery state, a novel back propagation neural network-dual extended Kalman filter (BP-DEKF) model for state-of-charge (SOC) and state-of-health (SOH) co-estimation of lithium-ion batteries is proposed by establishing a second-order equivalent circuit model (SO-ECM). Considering the coupling effect between SOC and SOH, the DEKF is designed to achieve the synergistic estimation to obtain better estimation results. To offset for the model error of the extended Kalman filter (EKF), a BP neural network is introduced for correction to further improve the SOC and SOH estimation accuracy. Under hybrid pulse power characterization (HPPC) working condition, the maximum and root-mean-square errors of SOC and SOH are 1.02%, 0.19% and 0.27%, 0.20%, respectively. The corresponding results are 1.03%, 0.62% and 0.083%, 0.057% under Beijing bus dynamic stress test (BBDST) working condition, respectively. The method put forward in this paper has high precision and robustness, which lays a theoretical foundation for battery state monitoring.