An Adaptive Method for Automotive Traction Battery Pack SOC Estimation Based on In pack Cell Uniformity Condition

Abstract:

The current state of charge (SOC) prediction methods for the traction battery pack (TBP) do not take into consideration of the cell uniformity problem which can not be neglected in TBP consisting of dozens or thousands of battery cells with their own characteristics. A new approach for online TBP SOC adjustment is proposed, which combines the tranditional and adaptive network based fuzzy inference system (ANFIS) methods. Fuzzy inference system (FIS) is used to adjust the traditional SOC estimation results in the pack in running time. Since the ANFIS is introduced, the training stage of the FIS can be completed offline; the trained knowledge base is appropriate for online application in an embedded system with acceptable computation complexity. The model structure, training method and verification process are introduced, and the verification result shows good generalization ability of the trained FIS.