Physics-based prognostics of lithium-ion battery using non-linear least squares with dynamic bounds

Abstract

Real-time health diagnostics/prognostics and predictive maintenance/control of lithium-ion (Li-ion) batteries are essential to reliable and safe battery operation. This paper presents a physics-based (or mechanistic) approach to Li-ion battery prognostics, which enables online prediction of remaining useful life (RUL) with consideration of multiple concurrent degradation mechanisms. In the proposed approach, robust online prediction of RUL is achieved by employing a non-linear least squares method with dynamic bounds that traces the evolution of individual degradation parameters. The novelty of this approach lies in its ability to incorporate mechanistic degradation analysis results into RUL predictions using nonlinear models. Results from a simulation study with eight Li-ion battery cells demonstrate that the mechanistic prognostics approach produces more accurate RUL predictions than a traditional capacity-based prognostics approach in 78 of the 80 cases considered (97.5% of the time). Additionally, it is shown that the use of dynamic bounds ensures a low level of uncertainty in the prediction throughout the entire life of a cell.