Short-Term Capacity Estimation and Long-Term Remaining Useful Life Prediction of Lithium-Ion Batteries Based on a Data-Driven Method

Abstract

In this work, the characteristic data of the peaks of the incremental capacity (IC) curves, the constant-current (CC) charging time, and their neighborhoods were determined during the CC charging phases of a battery. These data were transformed into an aging characteristic series and input into a long short-term memory (LSTM) recurrent neural network to achieve an accurate short-term capacity estimate. A method was then developed to predict the long-term remaining useful life (RUL). Specifically, a double exponential empirical model (DEEM) was employed to describe the fade trend of the battery capacity. The DEEM model parameters were initialized based on the offline nonlinear least squares (NLS) method. The particle filter (PF) algorithm was then used to update the DEEM model parameters and predict the RUL based on the short-term capacity estimated by the LSTM network. The experimental results revealed that the proposed method could effectively overcome the phenomenon of lithium-ion battery capacity regeneration and inconsistency. In addition, this method could realize the RUL prediction of the continuous prediction start point (SP). Under the same prediction SP setting, the proposed method outperformed other prediction methods.