In order to improve the accuracy of SOC estimation for lithium battery, a lithium battery SOC estimation method was proposed based on the combination of dynamic forgetting factor recursive least squares and improved particle filtering algorithm. The fixed forgetting factor recursive least squares method was difficult to maintain the fast convergence and recognition accuracy at the same time in battery parameter identification, a dynamic genetic factor was hence introduced and the residual difference between the identified and actual values of model was used as the variable to construct a correction formula to achieve the dynamic adjustment of forgetting factor. In order to improve the problem of particle diversity loss in particle filter (PF), the egret swarm optimization algorithm (ESOA) was used to optimize the particle filtering algorithm. The simulation results show that the estimation error of lithium battery SOC always remains within 0.3% after using the dynamic forgetting factor recursive least squares method and the improved particle filtering algorithm, with the mean absolute error and standard deviation of 0.15% and 0.17%. Compared with other algorithms, the new algorithm has better accuracy and stability.