Physics-Informed Neural Network for Online State of Health Estimation of Lithiumion Batteries

This paper presents a novel approach for estimating the state of health (SOH) of lithium-ion batteries, which addresses the challenge of being unable to measure the internal cell temperature during operation. The proposed approach, termed physics-informed neural network (PINN), integrates prior physical knowledge with measurable actual data to estimate the SOH of the batteries. To achieve this, an equivalent circuit model is established to characterize the electrical behavior characteristics of the batteries. An electric-thermal partial differential equation is also set to describe the batteries' heat generation mechanism and heat transfer process, and the batteries' instantaneous temperature field is reconstructed based on the PINN model. Finally, the online estimation of the lithium-ion batteries SOH is realized using the piecewise Arrhenius model. The simulation and experimental results show that the proposed approach achieves an average error of 0.37% in the temperature field reconstruction of the lithium-ion batteries and an average error of 0.15% in the online SOH estimation, even when the internal cell temperature cannot be measured.

Keywords: Physics-informed neural network, SOH estimation, Temperature field reconstruction, Arrhenius model, Cycle degradation.

Download PDF