Online estimation for parameters and state-of-charge of LiMn₂O₂ batteries with a modified adaptive Kalman filter

The accurate estimation for parameters and state-of-charge (SOC) is very important for the state monitoring and optimization in battery management system. This paper introduces an online estimation method with modified Kalman filter, where two contributions are added comparing related battery-application literature: (1) In order to improve the robustness to systematic noise brought by varying current loads and condition, an adaptive law is introduced to calibrate the noise covariance recursively. (2) To enhance the numerical stability over covariance propagating, a modified square-root procedure is derived by using sequential measurement update and the Potter algorithm. Besides, a test bench with a series-connected battery pack of 10 LiMn2O2 cells is used to evaluate the algorithm. The experimental results indicate both a better precision comparing the offline result, and the maximum error for the overall SOC is less than 2%.