Multi-step Forward State of Charge Prediction Method for Lithium-Ion Batteries

Lithium-ion batteries are increasingly applied in electric vehicles. One serious problem with electric vehicle applications is range anxiety. The anxiety will be relieved if drivers know the future State of Charge (SOC) of batteries, and therefore, multistep-forward prediction of SOC matters. However, it is very difficult to obtain the future discharge current of lithium-ion batteries in the period. Instead, a concept of equivalent current is introduced and used to replace the future discharge current when SOC is calculated for simplification. Similarly, the real discharge power of lithium-ion batteries varies while the average discharge power is relatively constant during a trip. Then, the average discharge power of lithium-ion batteries is predicted to represent the future driving cycles. Next, the equivalent current is predicted based on the average discharge power of lithium-ion batteries. Finally, the future SOC is obtained based on the predicted average power by the connection of the equivalent current. The results show the maximum SOC prediction error is less than 3%. The 200-step forward SOC prediction error of the proposed method is 0.15% in cycled dynamic stress test discharge, which is less than that of RNN and long short-term memory.