When using the extend Kalman filter (EKF) algorithm to estimate the state of charge (SOC) of lithium-battery

the problem of inaccurate initial SOC value setting and non-gaussian observation noise interference will be often encountered

which directly leads to low estimation accuracy. In order to solve the above problems

a first-order Thevenin equivalent circuit model for lithium-batteriy is established

and an extended Kalman filter algorithm of staged-transform observation noise covariance under the maximum correntropy criterion is proposed to estimate the SOC of lithium-battery. The algorithm used a small order of magnitude observation noise covariance to improve convergence speed in the initial SOC estimation stage

And the first positive and negative state transition of the observed residual first-order low-pass filter value is used as the convergence criterion. When it is judged that the estimated value has quickly converged to the allowable error range

the algorithm adaptively switches to a large order of magnitude observation noise covariance to ensure the smoothness of the subsequent estimated waveform. At the same time

the maximum correntropy criterion is introduced to iteratively and recursively correct the statistical characteristics of the observation noise in real-time

Used to reduce the impact of non gaussian observation noise on estimation accuracy. The verification results show that the proposed stragy has excellent comprehensive performance and strong robustness in SOC estimation

and has good engineering application value.