In view of the hysteresis characteristics of flat OCV-SOC of LiFePO4 battery, the OCV estimated by using the traditional Equivalent Circuit Model (ECM) has the problem of low accuracy under the charge-discharge switching condition, So hysteresis modeling of battery is proposed. In order to highlight the necessity of considering hysteretic characteristics of LiFePO4 batteries, three battery models were compared to evaluate their complexity, accuracy and applicability. The results show that the first-order RC model is only suitable for pure charge or discharge conditions without considering the influence of hysteresis. The first-order RC hysteresis model adds a hysteresis on the basis of the first-order RC model. Although the influence of hysteresis characteristics is considered, the hysteresis is greatly affected by parameter identification and the OCV estimation fluctuates. The Preisach model has good accuracy for charging and discharging switching conditions, but the training data and time cost are high. Therefore, practical applications can be selected according to different application scenarios. Under NEDC charging and discharging conditions, state of charge (SOC) estimation is carried out for different models combined with algorithms and the estimation error is within 5%, among which the Presisach error is within 3%.