In the field of battery management system, the accurate estimation of state-of-charge (SOC) is one of basic techniques among many state estimations. However, the equivalent circuit model (ECM), which is convenient for practical applications, cannot well simulate the battery characteristics in some specific SOC ranges. Therefore, the accuracy of SOC estimation based on the traditional Kalman filtering algorithm will decrease. The accurate estimation in the entire SOC range is realized using a novel Kalman filtering algorithm based on incremental error. Based on the offline analysis of the ECM performance in different SOC ranges, the control table for noise covariance in each SOC range is obtained. The specific noise covariance is applied in the corresponding SOC range so that the entire SOC range is accurately estimated using the proposed Kalman filtering algorithm. Experimental results showed that the estimation accuracy in the range with SOC lower than 20% can be significantly improved using the Kalman filtering algorithm based on incremental error.