2012 Prognostics & System Health Management Conference
(PHM-2012 Beijing)

Application of Unscented Particle Filter in Remaining Useful Life Prediction of Lithium-ion Batteries

Heng-Juan Cui1, Qiang Miao1, Wei Liang1, Zhonglai Wang1 & Michael Pecht2

1School of Mechanical, Electronic and Industrial Engineering University of Electronic Science and Technology of China Chengdu, Sichuan 611731, China

2Center for Advanced Life Cycle Engineering (CALCE)
University of Maryland
College Park, MD 20742



Accurate prediction of the remaining useful life of a faulty component is important to the health management of the system. It gives operators information about when the component should be replaced. This paper studied the remaining useful life prediction of the lithium-ion batteries. Some work has been done to solve this problem, but it still remains challengeable. Particle filter (PF) is a relatively effective method. However, the accuracy is not high. This paper introduces an improved PF algorithm - unscented particle filter (UPF) into the prediction. First, PF algorithm and UPF algorithm are described separately. Then, a degradation model is built based on the understanding of lithium-ion batteries. Finally, the model is combined with the training data and the algorithms to get the prediction results. According to the analysis results, it can be seen that UPF can predict the actual RUL with an error less than 4%.

Keywords:PF;UPF;Remaining useful life; Degradation model.

Full Text