Unbiased Model Identification and State of Energy Estimation of Lithium-Ion Battery

The accurate estimation of state of energy (SOE) is critical for the optimized utilization of lithium-ion battery (LIB). Despite the wide use of model-based observers, their performance can be declined largely by the noise corruption in real applications. This paper focuses on noise effect compensation and SOE estimation for LIB. A method combining the bias compensating recursive least squares (BCRLS) and Frisch scheme is exploited to compensate the noise effect and eliminate the identification bias. The unbiased model identification is further integrated with an observer based on the unscented Kalman Filter (UKF) to estimate the SOE in real time. Simulation and experimental results suggest that the proposed method effectively attenuates the identification bias caused by noise corruption and provides more reliable SOE estimation. Comparison with existing methods are also performed to verify its superiority in terms of the accuracy and the robustness to noises.