Theory of battery ageing in a lithium-ion battery: Capacity fade, nonlinear ageing and lifetime prediction

Selcuk Atalay, Muhammad Sheikh, Alessandro Mariani, Yu Merla, Ed Bower, W. Dhammika Widanage
2020 Journal of Power Sources  
A battery model capable of predicting SEI and Li plating induced aging is developed. • Mass transport of EC and DMC molecules within anode is considered. • The model predicts the lifetime based on actual drive-profiles with fast charging. • The importance of multi-layered SEI and its effect on porosity change is explained. • The model is able to capture the transition from linear to nonlinear aging. A R T I C L E I N F O Keywords: Battery ageing Capacity fade Nonlinear aging Drive-cycle life
more » ... etime prediction Lithium-ion battery A B S T R A C T Forecasting the lifetime of Li-ion batteries is a critical challenge that limits the integration of battery electric vehicles (BEVs) into the automotive market. Cycle-life performance of Li-ion batteries is intrinsically linked to the fundamental understanding of ageing mechanisms. In contrast to most previous studies which utilise empirical trends (low real-time information) or rough simplifications on mathematical models to predict the lifetime of a Li-ion battery, we deployed a novel ageing formulation that includes heterogeneous dual-layer solid electrolyte interphase (SEI) and lithium-plating ageing mechanisms with porosity evaluation. The proposed model is parameterized and optimized for mass transport and ageing parameters based on fresh and an aged cell and validated against our experimental results. We show that our advanced ageing mechanisms can accurately calculate experimentally observed cell voltage and capacity fade with respect to cycling number and can predict future fade for new operating scenarios based on constant-current and a dynamic power profile cycling experimental data consisting of high discharge C-rates and fast-charging periods. Our model is able to capture the linear and nonlinear (knee-point) capacity fade characteristics with a high accuracy of 98% goodness-of-fit-error and we compared our model performance with well-accepted existing model in literature.
doi:10.1016/j.jpowsour.2020.229026 fatcat:5szvln4mhjervlhry7mntjmdju