Investigating Identification Input Designs for Modelling Lithium-ion Batteries with Hysteresis using LPV Framework

This paper proposes a methodology to identify models for lithium-ion batteries exhibiting pronounced hysteresis, such as the lithium-iron-phosphate (LFP) batteries. The proposed battery model structure considers a multiple--input single--output linear parameter--varying (LPV) framework, wherein a second input representing maximum hysteresis overpotential is introduced along with the battery current. Furthermore, three designs for the input-current profile used for model identification are realized to investigate the extent of the hysteretic battery behaviour relevant to certain model applications. Finally, several battery models with varying model order and basis-function complexity are identified using each of the three current profiles, which are subsequently validated using a real drive-cycle dataset that includes occasional charging spikes due to regenerative braking as well as prolonged constant-current charging conditions.