Abstract: A least-squares online parameter identification method with an improved forgetting factor（IMAFFRLS） is proposed for the extended Kalman filter（EKF） method-based lithium-ion battery SOC estimation, which is vulnerable to the online identification accuracy of model parameters by the least-squares method and its improvement methods. Based on the dual-polarized equivalent circuit model, the causes of errors in the traditional forgetting factor-based least squares（FFRLS） method for identifying model parameters are analyzed, and it is noted that it is difficult to accurately track multiple model parameters varying at different rates with a single forgetting factor. By decoupling the covariance and gain matrices in the FFRLS algorithm, multiple variable forgetting factors are introduced to independently correct the estimation errors of different parameters. The adaptive variation of each forgetting factor is achieved based on the degree of input current fluctuation and the output voltage observation error in the moving interval. In addition, the two parameter identification algorithms before and after the improvement are combined with the EKF algorithm to realize the SOC estimation of Li-ion batteries. Finally, a comparative simulation based on Matlab is performed to verify the proposed model. The simulation results show that the proposed IMAFFRLS-EKF algorithm can identify the model parameters and estimate the SOC with higher accuracy than the FFRLSEKF algorithm.