A Remaining Useful Life Indirect Prediction Method for Lithium-Ion Batteries Based on SA-DBN

Abstract

To solve the issue that the battery capacity cannot be directly measured in practice, a more practical indirect remaining useful life (RUL) prediction method is proposed. First, the battery state characteristic parameters are analyzed and the time interval of equal discharge voltage drop (TIE-DVD) is selected as the indirect health factor. Second, the degradation relationship model is established by using back propagation (BP) neural network. Then, the deep belief network (DBN) model is used to establish the indirect health factor prediction model. Finally, to verify the adaptability of the proposed method to different types of batteries with different aging levels, both datasets from our own laboratory and the datasets from NASA Ames Research Center are used for experimental validations. The comparative experiments demonstrate that the proposed RUL prediction method is simple, accurate, and practical.