[Back]

C. Hametner, S. Jakubek:
"Data-driven methodologies for battery state-of-charge observer design";
in: "Automotive Battery Technology", issued by: Alexander Thaler; Daniel Watzenig; Springer-Verlag, Wien, 2014, (invited), ISBN: 978-3-319-02522-3, 111 - 125.

This chapter presents a data-based approach to nonlinear observer design for battery state of charge (SoC) estimation. The SoC observer is based on a purely data-driven model in order to allow for the application of the proposed concepts to any type of battery chemistry, especially when conventional physical modelling is not easily possible. In order to cope with the complex nonlinear dynamics of the battery, an integrated workflow for experiment design, model creation and automated observer design is proposed. The nonlinear battery model is constructed using a proven training algorithm based on the architecture of local model networks (LMNs). One important advantage of LMNs is that they offer local interpretability, which enables the extraction of local linear battery impedance models for automated nonlinear observer design. The proposed concepts are validated experimentally using real measurement data from a lithium-ion power cell.

http://dx.doi.org/10.1007/978-3-319-02523-0_7