The State of Charge (SoC)-Open Circuit Voltage (OCV) curve and the quality of estimation algorithm are two important factors that infect the accuracy of SoC estimation for lithium-ion batteries in electric vehicles. The purpose of this study is to improve the accuracy of SoC estimation for the lithium-ion battery. The battery management system is established to monitor the state of lithium-ion batteries to ensure the safety and reliability of the battery system. Firstly, the specific experiments were designed to analyze the relationship between the SoC-OCV curve and experimental conditions (e.g., ambient temperature and current rate) and battery states (e.g., State of Health and positive materials). A series of conclusions was found and used to correct the process of SoC estimation. Secondly, by analyzing the reasons that the SoC estimation error increased in the low-capacity period and the late-stage of estimation using the extended Kalman filter (EKF), an improved estimation algorithm was proposed. In the improved estimation algorithm, the ampere-hour counting was used in the low-capacity period, and the EKF was used in the rest. The accuracy of the improved estimation algorithm was verified by two experiments. Verification results show that the improved estimation algorithm makes up for the drawback of the EKF, the estimation error in constant current discharge experiment is less than 2 %, and the estimated error under dynamic conditions is less than 3 %. Therefore, the improved estimation algorithm has a higher accuracy than the EKF for the SoC estimation and can meet the operation requirements of a lithium-ion battery. This study contributes to the improvement of the safety and reliability of battery systems in electric vehicles by improving the accuracy of SoC estimation for lithium-ion batteries.