The nonlinear characteristics of lithium ion battery dynamic system were analyzed and its chaotic characteristics were identified. The chaotic characteristics of lithium ion battery dynamic system were restored by phase space reconstruction technology and the time sequence of multidimensional state space was obtained. The prediction of reconstructed time sequence was conducted with the LS-SVM model and the predicted value for state of charge (SOC) was assessed. The simulation results show that the prediction method has higher prediction accuracy and better adaptability than BP neural network prediction model, which has a certain guiding significance for practical application.