Lithium-ion battery end of life prediction based on the decelerating aging point

Lithium-ion batteries behave nonlinear degradation under multiple loading factors, e.g., current, temperature, voltage windows, and their combination, throughout the whole life cycle, which brings difficulties to the state estimation and prediction. Based on the battery capacity degradation curves, dual knee points {P1, P2}, the decelerating aging point (P1), and the accelerating aging point (P2), are defined for the degradation evaluation. The Kneedle method and the Bacon-Watts model are improved to adapt the {P1, P2} identification on six public datasets including LFP batteries, NCA batteries, NMC batteries, and LCO batteries. By investigating the relationship between the dual knee points and the battery end of life (EoL) which is the cycle when the capacity degradation to 80 % of the nominal capacity, it is found that the P1 cycle (N1) and P1 capacity retention (Q1) are strongly related to the EoL with early acquisition than the P2. A method involving the combination of N1 and Q1 is used for the lithium-ion battery EoL prediction based on a stepwise linear regression model, showing a maximum 8.7 % mean absolute percentage error compared to other benchmark methods, which provides new views for feature engineering for the battery state estimation and prediction of lithium-ion batteries.