Remaining Useful Life Prediction of Lithium-Ion Batteries Considering Long-Range Dependence and Capacity Regeneration

Accurate remaining useful life (RUL) prediction of lithium-ion batteries is critical for reliable operation of the system. The performance evolution of lithium-ion batteries shows long-range dependence and capacity regeneration. However, existing performance evolution models fail to properly consider the joint effect of long-range dependence and capacity regeneration, and consequently have a deficiency in mechanism interpretability, which may limit the prediction accuracy of RUL. To address this gap, this paper separately characterizes the degradation and regeneration processes of the discharge capacity, and proposes a novel discharge capacity evolution model incorporating fractional Brownian motion and the Poisson capacity regeneration process. For the estimation problem of the model parameters caused by the Poisson regeneration, we approximately transform the proposed model to one with independent increments according to the weak convergence theorem, and then develop a maximum likelihood estimation method. From a limit perspective and using the theory of total probability, we derive the distribution of RUL, and provide the point estimation of RUL. Finally, we utilize the data set of lithium-ion batteries produced by NASA to verify the effectiveness of the proposed method, and the mean absolute errors of the proposed method are declined by at least 24% compared to the existing methods.