The modified multi-innovation adaptive EKF algorithm for identifying battery SOC

For a second-order battery circuit model, a modified multi-innovation adaptive extended Kalman filter (MMI-AEKF) algorithm is investigated to accurately estimate the state of charge (SOC). Firstly, a process/measure noise estimation based adaptive extended Kalman filter (EKF) algorithm is presented to adjust the statistical characteristics of noise online. Then, a multi-innovation extended Kalman filter (MI-AEKF) algorithm is studied by expanding the single innovation into an innovation vector containing present/past innovations to increase the input information. Furthermore, based on the exponential sequence principle, add the attenuation memory factors to weight all the present/past innovations to distinguish different influences. Finally, build the test bench to collect data of constant current experiment for parameter identification and data of the urban dynamometer driving schedule (UDDS) and Los Angeles 92 (LA-92) for algorithm validation. Simulation results show that the investigated algorithm has excellent accuracy and robustness in SOC estimation, and the RMSE is less than 0.1% under the reasonable parameters.