Online Parameter Identification of Lithium Battery Model Based on Bias Compensated Least Square

Accurate lithium-ion battery modeling affects accuracy of battery state estimation, which will affect the safe operation of the electric vehicle. However, due to complex operating environments and influence of sensor noises, measured signals inevitably contain random noises, which hence influences model estimation accuracy. To solve the problem, this paper introduces a variable error model to estimate the average weighted variance of random noises and uses an augmented parameter vector to calculate a bias term, which can be compensated to estimation results of the battery model parameters. Finally, unbiased estimations of model parameters are obtained to improve accuracy of the established model. Results show that compared with traditional least square algorithm, the forgetting factor least square method based on bias compensation proposed in this paper can get higher precise. Under the Urban Dynamometer Driving Schedule (UDDS) and Hybrid Pulse Power Characterization (HPPC) test conditions, the mean absolute errors are reduced by 25 and 28%, and the root mean square errors are reduced by 25.1 and 42.7%, respectively.