Joint nonlinear-drift-driven Wiener process-Markov chain degradation switching model for adaptive online predicting lithium-ion battery remaining useful life

Abstract

The accurate prediction of the remaining useful life (RUL) of lithium-ion batteries is very important for battery management systems and predictive maintenance. However, lithium-ion batteries have a high degree of internal nonlinearity. There are two switching states during the operation of batteries operating, while the switching time point is also uncertain. In different switching states and random switching times, various unpredictable phenomena, such as capacity recovery or capacity decline could occur, which renders the accurate prediction of RUL challenging. To address this problem, a method for predicting the RUL was proposed in this work based on the nonlinear-drift-driven Wiener process and the Markov chain switching model. First, the nonlinear-drift-driven Wiener process was used to describe the time-varying battery degradation characteristics. The switching model was then applied to predict the future battery working state. Finally, the fuzzy system was employed to integrate the two by combining the battery degradation characteristics. The online update strategy of the model was simulated and validated, resulting in good adaptability and robustness. Two sets of real-case battery data from the National Aeronautics and Space Administration were also included during the validation process. The proposed method was systematically compared to other models in predicting the RUL of the batteries. From the acquired results, it was demonstrated that the proposed method was superior in predicting the RUL of batteries with improved accuracy and safety.