Estimating remaining useful life for lithium-ion batteries using kalman filter banks

In this paper, we propose a novel method based on kalman filter banks to estimate remaining useful life for industrial components. Instead of the common linear state space equation, we adopt jump Markov linear model for the proposed method. Thus, the problem that kalman filter and particle filter are not able to deal with non-Gaussian noises can be solved. Besides, proposed kalman filter banks method has no need for resampling, which is a commonly used in particle filter. We conduct a case study on Lithium-ion batteries, and find that the proposed method outperforms many existing model-based remaining useful life prediction methods, especially kalman filter and particle filter.