State of Charge Estimation of the Lithium-Ion Power Battery Based on a Multi-Time-Scale Improved Adaptive Unscented Kalman Filter

The accurate estimation of the state of charge (SOC) in lithium-ion power batteries is crucial for ensuring battery reliability, optimizing energy management strategies, enhancing battery efficiency, and prolonging battery service life. To account for the diverse time-varying characteristics of both SOC and model parameters in lithium-ion power batteries, this article introduces a multi-time-scale improved adaptive unscented Kalman filter. An improved method for adaptive updating of the noise covariance matrix is introduced, aiming to bolster algorithm convergence, diminish filtering result oscillations, guarantee the positivity of the system process noise covariance matrix, and ensure system response speed. Then, a multi-time-scale SOC estimation approach is introduced under micro, medium, and macroscopic three different timescales. The experimental findings indicate that even though the initial SOC estimation value and the initial maximum available capacity estimation value are both wrong, the SOC estimation results maintain an accuracy level below 1.00% for mean absolute error, 1.00% for root mean square error, and 1.50% for maximum absolute error under dynamic stress test (DST) and Federal Urban Driving Schedule (FUDS) operational modes, except for the extreme temperature 0 °C, which rarely appears in actual scenes. Moreover, it can also reach the accurate terminal voltage estimation results, the reasonable maximum available capacity, and the polarization voltage estimation results. Meanwhile, it can effectively reduce the battery management system (BMS) calculation burden by using micro–medium–macroscopic three different timescales.