Dynamic battery remaining useful life estimation: An on-line data-driven approach

Performance degradation and remaining useful life (RUL) estimation for lithium-ion battery has broad and practical applications in almost all industrial fields. The model-based prognostics is so complicated, moreover, they are not suitable for on-line application since that more parameters and modeling information should be obtained in advance. An on-line data-driven battery RUL prediction approach based on Online Support Vector Regression (Online SVR) is proposed. With Online SVR algorithm, the lithium-ion battery monitoring data series can be forecasted precisely, on the other hand, an ensemble approach is adopted to realize combined prediction with multi-models containing off-line and on-line algorithms to achieve better prediction capacity. Experimental results with the NASA battery data show that the proposed method can effectively predict the RUL of lithium battery.