Rapid empirical battery electromotive-force and overpotential modelling using input–output linear parameter-varying methods

In this paper, battery overpotential model identification approaches based on local and global Linear Parameter-Varying (LPV) input–output models are developed. Key features such as model structure, number of local models, and type and order of basis functions are considered. The LPVcore toolbox (Boef, 2021) has been used to solve the global identification problems. Furthermore, an iterative scheme is proposed which identifies a complete empirical battery model, i.e., both the ElectroMotive Force (EMF), also known as open-circuit voltage, and the overpotential model. This is achieved by iteratively obtaining an EMF realisation by (1) subtracting the modelled overpotential from a measured terminal voltage resulting from Constant-Current (CC) (dis)charging, and (2) using this EMF to calculate the overpotential from dynamic (dis)charging data and identifying an overpotential model using the LPV methods. This approach results in an empirical battery model with a precision similar (around 4 mV root-mean-square error in the range between 100% and 20% SoC) to models identified through a common cascaded approach in which the EMF is obtained separately from, e.g., pulse-(dis)charge data, but requires less measurement data resulting in a reduction factor in the order of 7 to 35 in terms of required experiment time. •Develops local and global linear parameter-varying modelling approaches.•Proposes an iterative scheme to rapidly identify a complete empirical battery model.•Model includes both electromotive-force and overpotential dynamics.•Total required measurement time is reduced by a factor 7 to 35.