An Ensemble Hybrid Model with Outlier Detection for Prediction of Lithium-ion Battery Remaining Useful Life

Lithium-ion batteries are widely used in many electronic devices, in order to ensure the safe and stable operation of these devices, it is necessary to know when the battery will reach the end of life (EOL).This paper introduces an ensemble hybrid model with outlier identification to predict the remaining useful life(RUL) of lithium-ion battery. A model based on unscented Kalman filter(UKF) preliminarily predicts the battery capacity which is used as a health indicator to track the degradation process of lithium-ion batteries, at the same time, multiple back propagation neural network sub-models predict the error evolution in parallel, and the two are added to complete the error correction. After marking and eliminating anomaly points by the isolated forest algorithm, all the predicted outputs are merged through a simple averaging strategy to obtaine the final lithium-ion battery RUL . Finally, the feasibility of the proposed method is verified by experiments.