A hybrid grey approach for battery remaining useful life prediction considering capacity regeneration

Abstract

Remaining useful life (RUL) prediction is the core of prognostic and health management. Lithium-ion batteries are a kind of consumable resource, whose RUL has a revelatory effect on its safe use and management. However, it has been found that battery degradation is not a completely degrading trend, which is ignored by traditional prediction methods. This paper proposes a hybrid grey approach to predict the RUL under multiple states. To predict the detailed degradation as real as possible, the capacity regeneration phenomenon is found to be a crucial part. A hybrid grey forecasting model is established to forecast the normal degradation and capacity regeneration process in detail. Ensemble Kalman Filter (EnKF) algorithm is applied to weaken the errors of grey forecasting models for long-term prediction. Based on single models from grey modeling and filtering, the ablation study shows the influence of different components in the hybrid approach. Then, compared with other data-driven models, the hybrid grey-EnKF demonstrates a quite satisfactory prediction of RUL for a set of batteries. Results show the grey-EnKF predictions are highly accurate with mean absolute percentage error smaller than 1%.