A Fusion Framework with Nonlinear Degradation Improvement for Remaining Useful Life Estimation of Lithium-ion Batteries

Fusion prognostic framework for lithium-ion battery remaining useful life (RUL) estimation has become a hot spot. Especially, the cycle life prediction has been conducted widely, for which many prognostic methods have been proposed correspondingly. However, many fusion frameworks which can achieve high precision are accompanied with high computing complexity and high time consumption which makes these methods low real-time performance. Either, some widely used prediction models with low complexity are weak to handle the nonlinear degradation features. To solve these problems, a fusion framework is proposed combining the model-based extended kalman filter (EKF) and the data-driven improved nonlinear scale degradation parameter based autoregressive (NSDP-AR) models. The proposed approach takes advantage of the state tracking ability of EKF algorithm to define the specific state transition model for the battery sample. Meanwhile, NSDP-AR model which contains the degradation features of each period is to promote the universality of the ND-AR (Nonlinear Degradation Autoregressive) model. NSDP-AR model is used to obtain the long term trend prediction results which are adopted as the observation data. Finally, a combination is made to realize the RUL prediction under the kalman filter (KF) system, which is an improvement to meet the practical applications. Experimental results with the battery test data from NASA PCoE and CALCE show that the fusion prognostic framework can predict the lithium-ion battery RUL with high efficiency and accuracy.