Combined Sparse Bayesian Learning Strategy for Remaining Useful Life Forecasting of Lithium-ion Battery

Due to the sparsity and uncertainty representation ability of the Sparse Bayesian Learning (SBL) algorithm, it is widely applied in the prognostics and remaining useful life (RUL) prediction. In the other hand, because of the less sample size, poor multi-step prediction performance, it is hard to obtain satisfied prognostics results for the lithium-ion battery RUL estimation. In this paper, a novel combined SBL strategy is proposed to realize lithium-ion battery RUL prediction. In the improved algorithm, the RUL multi-step prediction is achieved by combined sub-models instead of direct iterative computation. What´s more, Grey model(GM) is adopted to forecast the short-term capacity to improve the precision of each sub-model. As a result, the multi-step prediction precision with less battery sample size can be improved. The experimental results with the NASA lithium-ion battery data set indicate that the proposed method could get satisfied result compared to the basic SBL algorithm.