Automated Machine Learning for Estimating the State of Health of Lithium Ion Batteries

To optimize the operation of Lithium-Ion-Batteries and to accelerate the move to renewable energy, it is crucial to determine the condition of a battery with high accuracies at all times. Machine Learning and other statistical algorithms show high potential to estimate the State-of-Health without the need for a complex electrochemical model. The highest efforts for these models have to be put in the process steps of data pre-processing and the optimization of model parameters. This paper presents the state estimation of battery cells with automated machine learning. Therefore, the development time and efforts can be reduced dramatically. After training and testing the data-driven models, the estimation results for the state of health with an mean absolute error 0.855% show the competitiveness of the suggested approach in comparison to conventional methods and the potential to reduce development efforts.