Predictability Analysis of Lithium-Ion Battery Remaining Useful Life with Multiscale Entropy

Lithium-ion battery Remaining Useful Life (RUL) estimation and prediction are very important in the fields of reliability, automatic test, power sources, and electric vehicles, etc. The performance of battery RUL estimation relies on the predictability. Multiscale entropy (MSE) can be used to analyze the predictability of time series across multiple time scales. MSE is used to analyze the predictability of lithium-ion battery RUL. Results show that the predictability of battery discharge voltage decreases along with time scale and discharge cycle, and when the cycle time exceeds certain value, the predictability of discharge voltage decreases sharply. As an input of RUL prediction, the predictability of discharge voltage influences the prediction performance of RUL. For better prediction of RUL, if the MSE of discharge voltage of certain cycle changes greatly, other models or different model parameters should be considered.