Systematic parameter identification of a control-oriented electrochemical battery model and its application for state of charge estimation at various operating conditions

Electrochemical models based on first principles have shown great potential to accurately predict both the terminal voltage and internal states of lithium ion batteries. However, such models usually require significant computational time and contain a large number of parameters that describe the physical and electrochemical properties of the battery. In this paper, a systematic methodology is presented to generate control-oriented electrochemical models and identify those electrochemical parameters. The solid-phase governing partial differential equations are reduced by the volume average method, and a Galerkin projection is applied to the liquid-phase partial differential equation for model order reduction. Then, a two-step parameter identification strategy based on particle swarm optimization is proposed to obtain 26 model parameter values. Extensive calibration and validation results based on 18 sets of experimental data show that the reduced-order model with identified parameters agrees very well with experimental data at a wide range of operating conditions, covering steady-state discharge, relaxation, different driving cycles and ambient temperatures. A cubature Kalman Filter is then designed to demonstrate the capabilities of the resultant model in estimating battery state of charge. Results against all datasets show that the estimation error is less than +/- 1% for most of the 18 conditions.