An Improved Weighting Coefficient Optimization-Particle Filtering Algorithm Based on Gaussian Degradation Model for Remaining Useful Life Prediction of Lithium-ion Batteries

Abstract

Establishing a capacity degradation model accurately and predicting the remaining useful life of lithium-ion batteries scientifically are of great significance for ensuring safety and reliability throughout the batteries’ whole life cycle. Aiming at the problems of “particle degradation” and “sample poverty” in traditional particle filtering, an improved weighting coefficient optimization - particle filtering algorithm based on a new Gaussian degradation model for the remaining useful life prediction is proposed in this research. The main idea of the algorithm is to weight the selected particles, sort them according to the particle weights, and then select the particles with relatively large weights to estimate the filtering density, thereby improving the filtering accuracy and enhancing the tracking ability. The experimental verification results under the National Aeronautics and Space Administration data show that the improved weighting coefficient optimization - particle filtering algorithm based on the Gaussian degradation model has significantly improved accuracy in predicting the remaining useful life of lithium-ion batteries. The RMSE of the B05 battery can be controlled within 1.40% and 1.17% at the prediction starting point of 40 cycles and 70 cycles respectively, and the RMSE of the B06 battery can be controlled within 2.45% and 1.93% at the prediction starting point of 40 cycles and 70 cycles respectively. It can be seen that the algorithm proposed in this study has strong traceability and convergence ability, which is important for the development of high-reliability battery management systems.