State of charge estimation for Lithium Ion cells: Design of experiments, nonlinear identification and fuzzy observer design

• A unified framework for model generation and SoC observer design is presented.
• The purely data driven approach is suitable for any type of battery chemistry.
• Model based DoE ensures proper excitation while the SoC operating range is covered.
• The fuzzy observer reduces the computational complexity of the nonlinear filter.
• The proposed concepts are validated by means of a Lithium Ion power cell.

A new systematic approach to state of charge (SoC) observer design for battery cells is presented. It is based on a purely data driven model and a nonlinear observer constructed from it. As a key novelty, a unified and generic framework for model generation (i.e. design of experiments, nonlinear model structure) and observer parametrisation is presented. An integral part of SoC observers in hybrid electrical vehicles is a dynamic battery model which describes the nonlinear system behaviour of the cell terminal voltage. In order to enable the application of the proposed concepts for any type of battery chemistry, a data based modelling approach using the architecture of local model networks (LMNs) is proposed. As an important prerequisite, optimal model based experiment design ensures proper excitation of the system dynamics while the desired SoC operating range is covered. For SoC estimation, an augmented state space representation of the LMN is derived and the nonlinear observer design is presented. In particular, the use of a fuzzy observer is beneficial in combination with LMNs since the local observers are time-invariant which greatly reduces the complexity of the global estimator. The proposed concepts are validated experimentally by means of a Lithium Ion power cell.