Research on load state prediction model of electric vehicle lithium battery based on Kalman filter algorithm

The monitoring and protection of the power core lithium battery of electric vehicles is closely related to whether the lithium battery can output energy efficiently. To predict the charging state of lithium batteries, this study determined the online identification method, namely Recursive Least Square Method with Forcing Fact (FRLS), for model parameters based on the equivalent circuit model of lithium batteries, namely Thevenin. The overall prediction error of SVD-UKF is lower than that of Adaptive Extended Kalman Filter (AEKF) and Iterative Extended Kalman Filter (IEKF), which are about 20% and 30%, respectively. FRLS&Singular Value Decomposition-Unscented Kalman Filter (FRLS&SVD-UKF) has low prediction error of lithium battery load state under the same working condition and temperature, and the prediction error of lithium battery load state under the same working condition shows a gradually increasing trend with the increase of temperature. The FRLS&SVD-UKF joint prediction model can accurately predict the load state of lithium battery in electric vehicles in real time, and can improve the recycling performance of lithium battery.