P-064

Mengxi Wen

mengxi.wen@univ-lorraine.fr

Stéphane Raël, François Lapicque, Caroline Bonnet, Melika Hinaje

Université de Lorraine, CNRS, LRGP, 54000 Nancy, France

Electrochemical reduced model of lithium-ion batteries dedicated to observer design for embarked applications

Lithium-ion batteries (LIB) offer advantageous storage characteristics while they require a battery management system (BMS) to prevent damage from extreme operating conditions and to extend battery life. To estimate battery internal states, BMS generally uses an observer associated with a control-oriented mathematical model of the battery dynamics. Thus, it is crucial to develop a battery model for LIB, combining accuracy for reliable estimation and computational efficiency for real-time embedded applications. The single particle model (SPM) is a widely used electrochemical model for real-time state estimation. However, it is limited at high current rates due to significant intrinsic inaccuracies: (1) inaccuracies in the electrode open-circuit potential (OCP) evaluation due to low-order reduction of solid-phase lithium diffusion; (2) inaccuracies in the electrolyte overvoltage evaluation due to the lack of electrolyte dynamics.

Therefore, we propose enhancements to the original SPM while preserving its low-order formulations, physical meaning, and suitability for observer design. Two key contributions are presented. First, the normalized method used provides an accurate and low-order approximation of the solid diffusion partial differential equation (PDE), enabling precise electrode OCP estimation through direct approximation of the transcendental transfer function associated with solid diffusion PDE (DASPM). Second, electrolyte dynamics has been included using a finite difference method with low-order discretization, thereby accounting for electrolyte overpotentials and concentration gradients. 

The proposed overall model was compared to experimental data, the original SPM and the DASPM at the same model order. Tests on a 40Ah LTO-NMC cell on the market showed significant improvements in cell voltage accuracy in two applied current profiles emulating suburban transport (NEDC up to 2C and WLTC2 up to 2.8C). Compared to the experimental data, the root mean square (RMS) errors from the overall model were 6.9 mV and 5.7 mV, respectively. Moreover, compared to the original SPM, the improvement in solid diffusion dynamics reduced the RMS errors by approximately 17% and 29%, respectively, in the two profiles. In addition, inclusion of electrolyte dynamics allowed further RMS errors reduction by 22% and 33% for NEDC and WLTC2 profiles.