According to the principle that traditional battery state of charge (State of Charge, SOC) estimation method is based on precise mathematical model which relies on a lot of modeling assumptions and empirical parameters, therefore, SOC accuracy by the model prediction is limited. In order to improve the prediction accuracy for the battery SOC, Artificial Fish Swarm Algorithm (AFSA) is applied to predicting the SOC by optimizing Radial Basis Function (RBF) neural network, which solve the uncertainty in RBF network parameter choice. The simulation test results show that this method can easily, quickly and accurately achieve SOC prediction and has practical value. 