In order to solve the challenge of accurately estimating and predicting the state of charge (SOC) of battery in new energy vehicles, this paper focuses on the actual charging and discharging needs of electric vehicles. Relying on an experimental platform, characteristic experiments were performed on lithium battery. Based on the measured data, the Thevenin equivalent circuit model was selected for modeling, and parameter identification was performed using the Least Squares Method with a forgetting factor. Aiming at the problem that Kalman filter is difficult to cope with extreme driving cycles, the SOC estimation algorithm based on extended Kalman filter is used to estimate the lithium battery SOC. Finally, the SOC estimation errors under two different driving cycles at standard temperature were systematically compared and analyzed. The results show that in two driving cycles, the estimated SOC values closely approximate the actual SOC values, and with slightly smaller estimation errors under the new european driving cycle (NEDC) condition compared to the urban dynamometer driving schedule (UDDS) condition. The root mean square error, average error and maximum error under UDDS are 0.9058, 0.7998 and 1.3406, respectively, and the error in the case of NEDC is smaller than that of UDDS, which are 0.4439, 0.3415 and 0.9570, separately.