Lithium-ion battery state-of-charge estimation without current measurement using unscented recursive three-step filter

This study considers the state-of-charge estimation for lithium-ion battery in low-cost portable applications where the current measurement is presumed unknown. By regrading the current as an unknown input, this problem is reformulated as optimal filtering of nonlinear systems with an unknown input. A novel unscented recursive three-step filter is then proposed to simultaneously realize the SOC and current estimation. The unscented transformation is utilized to obtain the mean and covariance of the system state. Evaluation of the proposed approach is made by experiments under two different operating conditions. Experimental results show that the SOC estimation errors are within 5% bound, indicating that the proposed algorithm is feasible.