Enhanced Physics-Based Models for State Estimation of Li-Ion Batteries

Battery models and state estimation algorithms are a key components of todays advanced Battery Management Systems (BMS). Thereby, the battery models are used to estimate non-measurable states in the battery to ensure safety and availability while prolonging its life. This paper focuses on pseudo-2D physics-based battery models namely the Doyle-Fuller-Newman (DFN) model and Single Particle Model (SPM) that are capable to represent battery internal electrochemical states, that are vital for high precision simulation of the battery behavior. A three-step DFN model parameter identification procedure including QR decomposition with column pivoting, microstructure analysis and model optimization is proposed and applied on a commercial 18650 lithium-ion battery. The DFN model is validated with drive cycles as they occur in Electric Vehicles (EV) revealing a RMSE smaller than 18mV on average over the full SOC range. In the end, the DFN model is used to validate a state-space implementation of a SPM with electrolyte dynamics, which can be implemented on an embedded system to estimate battery states in real-time