Prediction of Remaining Useful Life of Lithium-ion Battery Based on Improved Auxiliary Particle Filter

In order to effectively predict the remaining useful life of lithium-ion batteries, particle filter algorithm is introduced in this paper. However, the standard particle filter algorithm is difficult to ensure the accuracy of battery life prediction due to its weight degradation, particle exhaustion and other problems. In this paper, a method based on the improved auxiliary particle filter algorithm and the double exponential capacity degradation model to predict the remaining useful life of lithium-ion batteries is proposed. Based on the standard particle filter, the algorithm introduces an auxiliary variable and performs two weighting operations to make the particle weight change more stable. Then, using the nonlinear mapping ability of BP neural network, the particle weights are split and adjusted to improve the particle diversity. The experimental results show that the improved algorithm is more reliable than the auxiliary particle filter, and the estimated relative error is smaller, that is, the remaining useful life of lithium-ion battery can be predicted more accurately.