Prediction of SOC Values for Sodium Batteries by Using the GRU and BiGRU Network Models

To enhance the predicted accuracy of state of charge (SOC) values, this paper proposes the gated recurrent unit (GRU) network model and bidirectional gated recirculation unit (BiGRU) network model to predict the SOC values of sodium batteries. The dynamic characteristics of sodium batteries were measured under two operation conditions (DST and FUDS) to collecting the training data and test data. Here, the values of voltage and current are the input parameters and the SOC values are the output results. The comparison results indicate that the BiGRU network model is more accurate than the GRU network model. The RMSE value and MAE value can be decreased to 32.28% and 17.77% by using the BiGRU network model. Therefore, The BiGRU network model can improve the prediction accuracy of SOC values for sodium batteries.