Data-Driven Simulation of the Electric Battery Performance for Powertrain Testing

Authors

Philipp Gesner, M.Sc., Christian Gletter, M.Sc., Dr.-Ing. Philipp Klein, Mercedes-Benz AG, Stuttgart, Germany;
Prof. Dr.-Ing. Bernard Bäker; Dipl.-Ing. Lutz Morawietz, Institute of Automotive Technology Dresden, Dresden, Germany

Summary

Efficient and cost-effective development of electric powertrains is fundamental to the transformation towards sustainable mobility. Modeling the electric energy storage is an essential tool for all related engineering tasks. This work aims to enhance given battery models, which are based on electrochemical insights into the single lithium-ion cells, with a purely data-driven approach.

A parallel hybrid model structure was created to compensate the existing simulation error with a recurrent neural network trained on extensively collected battery data. Combining those two worlds leads to a higher accuracy while still ensuring the stability of the output concerning the battery system's voltage response. Hence, it is a promising proposition for the safety-critical application on simulator devices, which replace real traction batteries on powertrain test benches.

Members Login