Prediction of Lithium‐Ion Battery Remaining Useful Life via Empirical Mode Decomposition‐ Autoregressive Integrated Moving Average and Regularized Particle Filter Algorithm

Abstract

This article focuses on improving the prediction accuracy of lithium‐ion battery's remaining useful life (RUL) by combining empirical mode decomposition (EMD), autoregressive integrated moving average (ARIMA), and regularized particle filter (RPF). First, to obtain detailed information about the battery capacity degradation, the monitored capacity data are decoupled by EMD. Second, the long‐term predicted model is constructed by ARIMA for the decoupled components. Finally, the long‐term prediction results are utilized as the measurement equation of the RPF prediction framework. In the proposed novel hybrid framework combining EMD‐ARIMA and RPF, the capacity prediction values are corrected and updated during the iteration of the regularized particle filter. With the estimated capacity data of every cycle, it can be detected whether the batteries reach their service life threshold. To validate the performance of the aforementioned method, the comparative experiments which are based on the NASA Prognostic Center of Excellence battery data sets are performed. According to the analysis results, higher prediction accuracy has been obtained with the proposed method in the lithium‐ion battery RUL, compared with the other three methods.